5660 lines
271 KiB
ReStructuredText
5660 lines
271 KiB
ReStructuredText
Mesa 23.1.0 Release Notes / 2023-05-10
|
||
======================================
|
||
|
||
Mesa 23.1.0 is a new development release. People who are concerned
|
||
with stability and reliability should stick with a previous release or
|
||
wait for Mesa 23.1.1.
|
||
|
||
Mesa 23.1.0 implements the OpenGL 4.6 API, but the version reported by
|
||
glGetString(GL_VERSION) or glGetIntegerv(GL_MAJOR_VERSION) /
|
||
glGetIntegerv(GL_MINOR_VERSION) depends on the particular driver being used.
|
||
Some drivers don't support all the features required in OpenGL 4.6. OpenGL
|
||
4.6 is **only** available if requested at context creation.
|
||
Compatibility contexts may report a lower version depending on each driver.
|
||
|
||
Mesa 23.1.0 implements the Vulkan 1.3 API, but the version reported by
|
||
the apiVersion property of the VkPhysicalDeviceProperties struct
|
||
depends on the particular driver being used.
|
||
|
||
SHA256 checksum
|
||
---------------
|
||
|
||
::
|
||
|
||
a9dde3c76571c4806245a05bda1cceee347c3267127e9e549e4f4e225d92e992 mesa-23.1.0.tar.xz
|
||
|
||
|
||
New features
|
||
------------
|
||
|
||
- VK_EXT_pipeline_library_group_handles on RADV
|
||
- VK_EXT_image_sliced_view_of_3d on RADV/GFX10+
|
||
- VK_KHR_map_memory2 on ANV and RADV
|
||
- fullyCoveredFragmentShaderInputVariable on RADV/GFX9+
|
||
- VK_EXT_discard_rectangles version 2 on RADV
|
||
- VK_EXT_graphics_pipeline_library on RADV
|
||
- extendedDynamicState3ColorBlendEquation on RADV
|
||
- primitiveUnderestimation on RADV/GFX9+
|
||
- VK_KHR_fragment_shading_rate on RADV/GFX11
|
||
- VK_EXT_mesh_shader on RADV/GFX11
|
||
- RGP support on RADV/GFX11
|
||
- GL_NV_alpha_to_coverage_dither_control on r600/evergreen+
|
||
|
||
|
||
Bug fixes
|
||
---------
|
||
|
||
- [radeonsi] flickering debug chunk border lines in Minecraft
|
||
- radv, radeonsi: Rogue Legacy 2 alpha-to-coverage rendering issues
|
||
- [r600, TURKS] R600: Unsupported instruction: vec1 32 ssa_1 = intrinsic image_samples (ssa_0) on spec@arb_shader_texture_image_samples@compiler@fs-image-samples.frag (23.1.0-rc4)
|
||
- vulkan/device_select: no way to select between GPUs of the same model due to bugs
|
||
- Intel/anv: Modifier problems running gamescope embedded
|
||
- radv: 7900 XTX hair flickering/rendering issues in VaM
|
||
- radv: cache crashing
|
||
- nouveau: Regression in arb_transform_feedback_overflow_query-basic from multithreading patches
|
||
- radeonsi: vaapi: \`width >2880 && width % 64 != 0` results in wrong width in h265 stream
|
||
- [regression] iris: unable to use driver as secondary GPU (primary AMD GPU)
|
||
- iris: steam doesn't render on dg2
|
||
- [llvm 16+] [microsoft-clc] opencl-c-base.h does not exist
|
||
- Vulkancts clipping / tesselation tests trigger gpu hang on DG2
|
||
- Swaped fields in picture in vlc and mythtv if hw accel is on
|
||
- WGL: Assert assigns dwThreadId variable
|
||
- nine regression with r600 (bisected)
|
||
- [ACO] [RADV] Flickering squares in some areas in The Last of Us Part 1 (with workaround)
|
||
- radv: Jedi Fallen Order flickering & blocky plants
|
||
- nouveau: NV50 (NVAC) broken in latest master
|
||
- rusticl failed to build with rust-bindgen 0.65.0
|
||
- Regression, Bisected: glsl: Delete the lower_tess_level pass breaks r600 tesselation
|
||
- vkcts-navi21-valve failing often with GCVM_L2_PROTECTION_FAULT_STATUS:0x00X00830
|
||
- Deep Rock Galactic GPU freeze (AMD, DX11 DXVK Proton)
|
||
- radv: Resident Evil 4 Chainsaw Demo GPU hang with Navi 24
|
||
- radv: Gotham Knights GPU hang with Navi 24
|
||
- aco: s_load_dword with negative soffset cause GPU hang
|
||
- piglit.spec.ext_image_dma_buf_import.ext_image_dma_buf_import crash shutting down
|
||
- overlay layer: unable to launch titles on steam
|
||
- radv/zink: spec@ext_texture_integer@multisample-fast-clear gl_ext_texture_integer
|
||
- VAAPI: Wrong H.264 playback on RX 6900 XT and RX 6700 XT (all Sienna?)
|
||
- radv: possibly not setting state dirty bits correctly
|
||
- RADV: VRS attachment not working in specific scenario
|
||
- rusticl: invalid SPIR-V kernel causes panic
|
||
- [RADV] The Last Of Us Part 1: artifacting in the menu (with workaround)
|
||
- AMD va-api outputs corrupt encoding
|
||
- !20673 regressed \`dEQP-VK.wsi.xlib.surface.query_formats`
|
||
- aco: missing dependency on generated header
|
||
- zink: spirv validation errors with spirv 1.6
|
||
- freedreno/a6xx: Assertion \`view->rsc_seqno == rsc->seqno' failed.
|
||
- iris regression in map stride after import with gen9 parts
|
||
- anv: zink ADL failures
|
||
- Vulkan loader \`vk_common_GetPhysicalDeviceFormatProperties` fails to sanitize properties bits.
|
||
- Loading a model in PrusaSlicer 2.6.0-alpha5 crashes GNOME on radeonsi
|
||
- [glx][bisected][regression]Intel HD 3000 failing to create context on applications like Unity
|
||
- v3d: dEQP-GLES3.functional.texture.specification.teximage2d_pbo.rg32f_cube and similar fails when run together with other tests
|
||
- standalone glsl compiler not getting built with mingw
|
||
- !22191 broke test-docs-mr
|
||
- mesa: index buffer leaking
|
||
- RadeonSI: null dereference in amdgpu_cs_add_buffer, potential refcount mismatch, running BeyondAllReason
|
||
- NIR can't unroll any loop from nine
|
||
- Steel Division 2 - radv/gpu hang - bisected
|
||
- Turnip: VKD3D can't run due to lack of memory property flag
|
||
- eglCreateImageKHR, error: EGL_BAD_ALLOC (0x3003), message: "createImageFromDmaBufs failed" on AMD multi-gpu with explicit format modifiers
|
||
- radv: In the game Quake II RTX appeared artifacts at fresh mesa builds
|
||
- radv: Vampire: The Masquerade - Bloodline (Unofficial Patch) regression
|
||
- radeonsi broken for gcn1 card
|
||
- libgrl.a installed but not used?
|
||
- radv: crash compiling UE5 lumen hardware RT shader
|
||
- spec@ext_transform_feedback@builtin-varyings gl_culldistance fail
|
||
- Panfrost T860 - broken system with latest mesa on gnome wayland jammy
|
||
- aco: unused vtmp_in_loop
|
||
- FTBFS: src/amd/llvm/ac_llvm_util.c:248:4: error: implicit declaration of function ‘LLVMAddIPSCCPPass’ (LLVM C interface removed upstream)
|
||
- vulkan: new generated physical_device_feature missing meson dependency
|
||
- Build broken on old-ish Python versions
|
||
- radv: Support fullyCoveredFragmentShaderInputVariable from VK_EXT_conservative_rasterization on RDNA2+
|
||
- radv,nir: dEQP-VK.ray_query.builtin.rayqueryterminate.* failures
|
||
- RFE: Use _mesa_is_foo(ctx) helpers more
|
||
- spec@ext_transform_feedback@builtin-varyings gl_culldistance fail
|
||
- ci: infinite XDG_RUNTIME_DIR spam
|
||
- ci: XDG_RUNTIME_DIR spam
|
||
- [KBL] iris failures with dEQP-GLES3.functional.texture.compressed.astc.void_extent*
|
||
- glsl compiled error when the RHS of operator \`>>` is int64_t by enabling GL_ARB_gpu_shader_int64 extension
|
||
- turnip: inline uniforms regression
|
||
- QPainter fails to render multiple shapes with a brush set since Mesa 23.0
|
||
- eglSwapBuffers blocks in wayland when it's wl_surface_frame event is stolen.
|
||
- plasmashell sometimes hangs with mesa_glthread
|
||
- pps_device.h:23:11: error: ‘uint32_t’ does not name a type
|
||
- Build fails with llvm 17: llvm/ADT/Triple.h: No such file or directory
|
||
- nir: i2f32(i2i32(x@8)) isn't being collapsed to i2f32(x)
|
||
- zink-lvp no longer running tests
|
||
- radv: Immortals Fenyx Rising: Grass Flicker on R9 380X and Steam Deck
|
||
- radv: A Plague Tale: Requiem black "flash" on 7900XTX
|
||
- 7900 XTX: Graphical corruption / artifacts in Cyberpunk
|
||
- radeonsi draws spurious values to depth buffer
|
||
- Commit ccaaf8fe04c956d9f16f98b7f7fa69a2526283bc causes GPU ring timeouts on BONAIRE
|
||
- radv: CmdCopyQueryPoolResults broken for VK_QUERY_TYPE_PRIMITIVES_GENERATED_EXT with queryCount > 1
|
||
- rusticl over llvmpipe + ffmpeg's Opencl filter = error -51
|
||
- rusticl over llvmpipe + ffmpeg's Opencl filter = error -51
|
||
- ci: Remove LAVA Gitlab section handler workaround
|
||
- System freeze when playing some h264 videos with VA-API on Rembrandt
|
||
- OpenGL crashes in X-Plane 11
|
||
- r600/TURKS: NIR Shader related errors on CLI with the game "A Hat in Time" and Gallium Nine
|
||
- agx: shifts aren't sound
|
||
- ci: build logs hidden
|
||
- fatal error: intel/dev/intel_wa.h: No such file or directory
|
||
- [Bisected] Regression: Project Zomboid renders black
|
||
- hasvk: Black pixels with 8xMSAA and fast clears on Intel(R) HD Graphics 4400 (HSW GT2)
|
||
- radv: GTA IV graphical artifacts on 7900XTX
|
||
- radv: Resident Evil Revelations 2 artifacts on 7900XTX with DCC
|
||
- radv: Prototype 2 black textures on RDNA 3 when DCC is enabled
|
||
- Mesa 23.0.0 crashes immediately with indirect rendering
|
||
- virpipe-on-gl: arb_enhanced_layouts\@matching_fp64_types crashes
|
||
- [RADV] Returnal - pistol muzzle flash fills whole screen (graphical artifact)
|
||
- ACO: dEQP-VK.binding_model.descriptor_buffer.multiple.graphics_geom_buffers1_sets3_imm_samplers hangs on NAVI10
|
||
- Build failures with recent lld
|
||
- r600,regression: Glitches on terrain with the NIR backend on Transport Fever 2
|
||
- [radeonsi] Regression with MSAA fix for Unreal / Unreal Tournament 99
|
||
- spirv: Switch Vulkan drivers to use \`deref_buffer_array_length`
|
||
- r600/TURKS: Crash of the game "A Hat in Time" with Gallium Nine and NIR path (third report)
|
||
- [gen9atom] Vulkan tests cause gpu hang: dEQP-VK.memory_model.*
|
||
- GL_SHADER_BINARY_FORMAT_SPIR_V is not added to the list of GL_SHADER_BINARY_FORMATS even if GL_ARB_gl_spirv is supported.
|
||
- mesa: "Fragmented" dynamic lights in IronWail with \`r_fsaamode 1` on
|
||
- [ANV/DG2] Vertex explosion in nvpro-samples/vk_raytracing_tutorial_KHR/ray_tracing_gltf
|
||
- CUEtools FLACCL hit assert in rusticl
|
||
- Assertion Failed on Intel HD 5500 with Linux / Mesa 22.3.1 / OpenGL
|
||
- Rise of the Tomb Raider's Ambient Occlusion pass misrenders (swimming shadows)
|
||
- vk_enum_to_str: missing VkPipelineCreateFlags
|
||
- [glsl] [spirv] ssbo unsizied array not supported ?
|
||
- Creating a vulkan physical device on an AMD GPU causes following calls to drmModeAddFB to fail with ENOENT
|
||
- Minecraft: spec related compile errors
|
||
- mesa: _mesa_glthread_upload crash
|
||
- glthread: OpenGL submission blocks while swapping buffers
|
||
- glthread: Loading a shader cache in yuzu slows down with mesa_glthread=true
|
||
- Commit ""radeonsi: enable glthread by default"" (d6fabe49cd72fb) causes a regression in gstreamer gtkglsink element
|
||
- llvmpipe: linear rasterizer / depth bug
|
||
- radv: (Using mesh shader) NIR validation failed after nir_lower_io_to_scalar_early
|
||
- panfrost Mali-G31 glamor regression
|
||
- allwinner a64: DRM_IOCTL_MODE_CREATE_DUMB failed: Cannot allocate memory after some time of apps usage
|
||
- turnip: dEQP-VK.ubo.random.all_shared_buffer.48 slow
|
||
- wine + dxvk + Rise of the Tomb Raider crashes in Soviet Installation 20% with VK_ERROR_DEVICE_LOST
|
||
- Sometimes VLC player process gets stuck in memory after closure if video output used is Auto or OpenGL
|
||
- ci: Remove LAVA Gitlab section handler workaround
|
||
- kwin_wayland crashes involving dri2_create_drawable when Plasma starts and the llvmpipe driver from Mesa 23.0-rc3 and 23.0-rc4 is used
|
||
- turnip: no ubwc fast clear for depth on a618
|
||
- anv: VK_ACCESS_2_SHADER_READ_BIT doesn't seem to be handled correctly
|
||
- Vulkan WSI flags leak into NIR, breaking build on BSDs
|
||
- Iris corruptions in zoom calls
|
||
- Sampling with aux enabled with ISL_AUX_STATE_PASS_THROUGH seems broken on Tigerlake+
|
||
- anv: incorrect task shader payload
|
||
- radv: Hi-Fi Rush incorrectly rendering face shadows with DCC on 7900 XTX
|
||
- [iris] isl_calc_min_row_pitch seems incorrect on a750
|
||
- DG2: incorrect rendering in Sascha Willems raytracing callable demo
|
||
- turnip: conditional load/store hurts some workloads
|
||
- Some blackouts / rendering issues with RADV_PERFTEST=gpl in Battlefield 1 (DX11)
|
||
- radv/zink: ACO assert with DOOM2016
|
||
- Registered special XGE not unregistered
|
||
- draw_llvm.c:788:7: error: implicit declaration of function ‘LLVMContextSetOpaquePointers’
|
||
- asahi: Optimize lower_resinfo for cube maps
|
||
- Metro Exodus hits nir validation with a driver supporting raytracing.
|
||
- ANV Gen 9.5 swapchain corruption when using newer \`VK_IMAGE_LAYOUT_ATTACHMENT_OPTIMAL` layout
|
||
- turnip: optimal bin layout
|
||
- piglit.spec.arb_shader_texture_image_samples.compiler.fs-image-samples_frag regression
|
||
- lavapipe assert fails on Windows
|
||
- zink: itoral-gl-terrain-demo rendering failure
|
||
- asahi: Implement a shader disk cache
|
||
- [ICL] Trine 4 trace causing GPU HANG
|
||
- radv: Segfault during createAccelerationStructure when the backing buffer is not bound to memory
|
||
- 7900 XTX stuck at 'compiling shaders' in Monster Hunter Rise
|
||
- radv: slow GPL (fast) link times
|
||
- libEGL warning: failed to get driver name for fd -1
|
||
- iris: Context incorrectly marked as guilty
|
||
- simple_mtx.h:34:12: fatal error: valgrind.h: No such file or directory
|
||
- [ANV] Commit 4ceaed78 causes misrendering on Cyberpunk 2077
|
||
- adding eglGetMscRateANGLE support for multiple monitors with different refresh rates
|
||
- Performance regression in Chromium WebGL when implement ANGLE_sync_control_rate with egl/x11
|
||
- ci: Ensure that the Intel/Freedreno trace pipelines only show up in relevant MR's
|
||
- anv: Performance issue with Vulkan on Wayland KWin
|
||
- Incorrect format conversion on big endian
|
||
- radv: State of Decay 2 character rendering regression
|
||
- aco_tests assembler.gfx11.vop12c_v128/gfx11 failure
|
||
- r600,regression: Loading of DOOM stuck at 0% with the NIR backend
|
||
- RADV: enabling TC-compat HTILE in GENERAL for compute queues is likely broken
|
||
- Confidential issue #8065
|
||
- VAAPI HEVC encode broken since 22.3
|
||
- GPU HANG: ecode 12:1:859ffffb (Resetting rcs0 for stopped heartbeat on rcs0) - reproducible
|
||
- zink: src/gallium/auxiliary/pipebuffer/pb_slab.c:138: Assertion failed: \`heap < slabs->num_heaps`
|
||
- [zink] Assertion \`heap < slabs->num_heaps' failed on Pascal (bisected)
|
||
- [RADV] Incorrect copies to/from compressed textures with mipmaps
|
||
- mesa_glthread=true and probably ANY id Tech 3 engine games, offroad...
|
||
- radeonsi: VRAM Leak/abnormally high usage in Minecraft mod pack
|
||
- nir/lower_blend: Bogus assert
|
||
- anv-tgl-vk: fails a multiple jobs after changing sharding
|
||
- radv CTS crashes since ebec42d799b22b7b3d06acd710f5687252446a06
|
||
- llvmpipe: dEQP-EGL programs.link failures.
|
||
- libmesa_util depends on gallium
|
||
- EGL report EGL_EXT_create_context_robustness with kms_dri drvier while can't create context with EGL_LOSE_CONTEXT_ON_RESET_EXT attribute.
|
||
- v3d: missing drm format modifier support on Raspberry Pi 4 required for mpv
|
||
- Return To Monkey Island black screen
|
||
- Return To Monkey Island black screen
|
||
- navi22 amdgpu: bo 000000002843d677 va 0x0800000400-0x08000005ff conflict with 0x0800000400-0x0800000600
|
||
- Ryzen 6800H laptop amdgpu: bo 00000000b1eb583a va 0x0800000200-0x08000003ff conflict with 0x0800000200-0x0800000400
|
||
- [RADV] [MISSED PERFORMANCE POTENTIAL] Vulkan not working when Color Depth is set to "16", but Vulkan works when Color Depth is set to "24"
|
||
- v3dv: f2f16_rtz lowering could be improved
|
||
- debug build compilation failed: inlining failed in call to ‘always_inline’ ‘src_is_ssa’: indirect function call with a yet undetermined callee
|
||
- radv: regression: broken UI rendering in Elden Ring
|
||
- radv: Missing implementation of VkImageSwapchainCreateInfoKHR and VkBindImageMemorySwapchainInfoKHR
|
||
|
||
|
||
Changes
|
||
-------
|
||
|
||
Adam Jackson (22):
|
||
|
||
- glx/dri3: Simplify protocol version tracking
|
||
- glx: Remove glx_context::screen
|
||
- glx: Remove a can't-happen NULL check
|
||
- glx: Remove support for glXGetDriverConfig for old drivers
|
||
- glx: Clean up some funny business from context bind/unbind
|
||
- glx: Reflow MakeContextCurrent a little
|
||
- glx: Check for initial "glX" first in glXGetProcAddress
|
||
- glx: Move 1.2 GLXPixmap code into glx_pbuffer.c
|
||
- glx: Inline a few single-use constant strings into their user
|
||
- glx: Fix drawable type inference in visual/fbconfig setup
|
||
- glx: Harmonize glXCreateGLXPixmap with glXCreatePixmap
|
||
- mesa: Fix extension table formatting
|
||
- mesa: Trivially advertise NV_generate_mipmap_sRGB
|
||
- wsi/x11: Make get_sorted_vk_formats handle varying channel widths
|
||
- wsi/x11: Infer the default surface format from the root window's visual
|
||
- wsi/x11: Support depth 16 visuals
|
||
- glx/dri: Use X/GLX error codes for our create_context_attribs
|
||
- dri: Validate more of the context version in validate_context_version
|
||
- glx/dri: Fix error generation for invalid GLX_RENDER_TYPE
|
||
- glx: Disable the indirect fallback in CreateContextAttribs
|
||
- glx: Fix error handling yet again in CreateContextAttribs
|
||
- mesa: Enable NV_texture_barrier in GLES2+
|
||
|
||
Adam Stylinski (2):
|
||
|
||
- glx: fix a macro being invoked with the wrong parameter name
|
||
- mesa: fix out of bounds stack access on big endian
|
||
|
||
Alan Coopersmith (1):
|
||
|
||
- util/disk_cache: Handle OS'es without d_type in struct dirent
|
||
|
||
Alejandro Piñeiro (17):
|
||
|
||
- vulkan/wsi: check if image info was already freed
|
||
- v3dv/format: remove unused v3dv_get_tex_return_size
|
||
- v3dv/pipeline: rename lower_tex_src_to_offset to lower_tex_src
|
||
- v3dv: pass alignment to v3dv_buffer_init
|
||
- v3dv/image: use 64-byte alingment for linear images if needed
|
||
- v3dv: skip two ycbcr tests
|
||
- broadcom/compiler: v3d_nir_lower_txf_ms doesn't need v3d_compile
|
||
- broadcom/compiler: treat PIPE_FORMAT_NONE as 32-bit formats for output type
|
||
- v3dv: enable shaderStorageImageReadWithoutFormat
|
||
- broadcom/compiler: fix indentation at v3d_nir_lower_image_load_store
|
||
- nir: track if var copies lowering was called
|
||
- radv: use shader_info->var_copies_lowered
|
||
- anv: use shader_info->var_copies_lowered
|
||
- v3d/v3dv: use shader_info->var_copies_lowered
|
||
- v3dv: handle ASPECT_MEMORY_PLANE aspect flags when getting plane number
|
||
- v3dv/debug: add debug option to disable TFU codepaths
|
||
- v3dv/pipeline: use pipeline depth bias enabled to fill up CFG packet
|
||
|
||
Alexandros Frantzis (2):
|
||
|
||
- egl/wayland: Fix destruction of event queue with proxies still attached.
|
||
- vulkan/wsi/wayland: Fix destruction of event queue with proxies still attached.
|
||
|
||
Alyssa Rosenzweig (351):
|
||
|
||
- nir/peephole_select: Allow load_preamble
|
||
- agx: Peephole select after opt_preamble
|
||
- asahi: Handle sampler->compare_mode
|
||
- panfrost: Don't use AFBC of sRGB luminance-alpha
|
||
- pan/bi: Fix incorrect compilation of fsat(reg.yx)
|
||
- pan/bi: Add a unit test for fsat(reg.yx)
|
||
- panfrost: Enable NV_primitive_restart on Valhall
|
||
- panfrost: Fix logic ops on Bifrost
|
||
- panfrost: Stop testing CAP_INT16
|
||
- panfrost: Remove PAN_MESA_DEBUG=deqp
|
||
- panfrost: Remove unused debug parameter
|
||
- panfrost: Fix clears with conditional rendering
|
||
- panfrost: Document render_condition_check contract
|
||
- nir: Add Midgard-specific fsin/fcos ops
|
||
- nir: Optimize vendored sin/cos the same way
|
||
- pan/mdg: Use special NIR ops for trig scaling
|
||
- pan/mdg: Scalarize LUT instructions in NIR
|
||
- pan/mdg: Remove MSGS debug
|
||
- mesa: Set info.separate_shader for ARB programs
|
||
- nir/lower_blend: Fix alpha=1 for RGBX format
|
||
- nir/lower_blend: Clamp blend factors
|
||
- nir/lower_blend: Fix SNORM logic ops
|
||
- nir/lower_blend: Avoid useless iand with logic ops
|
||
- nir/lower_blend: Don't do logic ops on pure float
|
||
- nir/lower_blend: Handle undefs in stores
|
||
- nir/lower_blend: No-op nir_color_mask if no mask
|
||
- asahi: Omit extra call to clock_gettime
|
||
- nir/opt_preamble: Treat \*size as an input
|
||
- nir/opt_preamble: Consider load_preamble as movable
|
||
- agx: Lower system values in NIR in the driver
|
||
- agx: Bump preamble_storage_size to 512
|
||
- agx: Centralize texture lowering
|
||
- asahi: Use non-UAPI specific BO create flags
|
||
- nir: Add a late texcoord replacement pass
|
||
- asahi: Run nir_lower_fragcolor during preprocessing
|
||
- asahi: Lower texcoords late
|
||
- panfrost: Implement GL_EXT_render_snorm on Bifrost+
|
||
- ail: Add layout->mipmapped_z input
|
||
- ail: Test mipmapped_z behaviour
|
||
- ail: Test 63x63 cube map
|
||
- asahi: Set layout->mipmapped_z for 3D textures
|
||
- asahi: Fix encoding of uniform size
|
||
- asahi: Strengthen agx_usc_uniform contract
|
||
- asahi/nir_lower_sysvals: Split large ranges
|
||
- asahi: Correct alignment for USC Uniform packets
|
||
- agx: Support uniform registers as LODs
|
||
- asahi: Use writeback when it looks beneficial
|
||
- asahi: Make STAGING resources linear
|
||
- asahi: Prefer blit-based texture transfer
|
||
- asahi: Implement nontrivial rasterizer discard
|
||
- asahi: DRY dirty tracking conditions
|
||
- asahi: Remove redundant tri merge disable bit
|
||
- asahi: Merge fragment control XML
|
||
- agx: Keep varyings forwarded to texture as fp32
|
||
- asahi: Don't use 16-bit inputs to 32-bit st_tile
|
||
- docs/asahi: Document clip distance varyings
|
||
- agx: Fix storing to varying arrays
|
||
- agx: Handle constant-offset in address matching
|
||
- asahi: Add XML for custom border colours
|
||
- agx/decode: Add a data parameter to stateful
|
||
- agx/decode: Handle extended samplers
|
||
- asahi: Implement custom border colours
|
||
- asahi: Fix delete_vs_state implementation
|
||
- asahi: Add compute kernel scaffolding
|
||
- asahi: Don't leak shader NIR
|
||
- asahi: Add hooks for SSBO and images
|
||
- asahi: Fake more caps for dEQP-GLES31
|
||
- asahi: Advertise seamless cube maps
|
||
- asahi: Stub out MSAA for dEQP
|
||
- asahi: Bump PIPE_CAP_MAX_TEXTURE_ARRAY_LAYERS
|
||
- asahi: Add compute batches
|
||
- asahi: Implement load_ssbo_address/get_ssbo_size
|
||
- asahi: Identify more compute-related XML
|
||
- agx: Implement compute ID intrinsics
|
||
- agx: Implement barriers
|
||
- nir/print: Extract get_location_str
|
||
- nir/print: Pretty-print I/O semantic locations
|
||
- nir/print: Pretty-print color0/1_interp
|
||
- agx: Allow uniform sources on phis
|
||
- agx: Run DCE twice
|
||
- agx: Lower uniform sources with a dedicated pass
|
||
- agx: Don't scalarize preambles in NIR
|
||
- nir/lower_clip: Only emit 1 discard
|
||
- tu,vulkan: Add common Get*OpaqueCaptureDescriptorDataEXT
|
||
- radv: Use common Get*OpaqueCaptureDescriptorDataEXT
|
||
- agx: Remove unused AGX_MAX_VARYINGS
|
||
- agx: Respect component in frag load_input
|
||
- agx: Fix AGX_MAX_CF_BINDINGS
|
||
- agx: Remove bogus gl_Position assertion
|
||
- agx: Implement load_helper_invocation
|
||
- agx: Write sample mask even with no colour output
|
||
- asahi: Submit batches that don't touch RTs
|
||
- asahi: Add XML for indirect dispatch
|
||
- asahi: Add XML for indirect draws
|
||
- asahi: Add XML for VDM memory barriers
|
||
- panvk: Take lock when tracing
|
||
- panvk: Fix varying linking
|
||
- panvk: Disable SNORM rendering
|
||
- asahi: Remove default=true on index list values
|
||
- asahi: Refactor index buffer upload for indirect
|
||
- asahi: Implement indirect draws
|
||
- panfrost: Fix some fields in v10.xml
|
||
- pan/decode: Add support for decoding CSF
|
||
- asahi: Vectorize background colour load
|
||
- panfrost: Disable CRC by default
|
||
- panfrost: Fix prim restart XML on Valhall
|
||
- nir: Augment raw_output_pan with IO_SEMANTICS+BASE
|
||
- pan/lower_framebuffer: Operate on lowered I/O
|
||
- nir/lower_blend: Don't touch store->dest
|
||
- nir/lower_blend: Don't handle gl_FragColor
|
||
- nir/lower_blend,agx,panfrost: Use lowered I/O
|
||
- asahi: Lower clip distances late
|
||
- asahi: Move agx_preprocess_nir to CSO create
|
||
- agx: Don't treat clip distances specially
|
||
- agx: Do more work in agx_preprocess_nir
|
||
- asahi: Fix rendering into mipmapped framebuffers
|
||
- agx: Lower offsets in NIR
|
||
- agx: Model and pack gathers
|
||
- agx: Implement gathers (nir_texop_tg4)
|
||
- docs/features: Sync Asahi with reality
|
||
- asahi: Advertise ARB_derivative_control
|
||
- asahi: Advertise ARB_texture_barrier
|
||
- agx: Model atomic instructions
|
||
- agx: Model local loads/stores
|
||
- agx: Disallow immediate bases to device_load
|
||
- agx: Pack global atomics
|
||
- agx: Pack local load/store instructions
|
||
- agx: Translate NIR atomics
|
||
- agx: Translate load/store_shared
|
||
- agx: Lower shared memory offsets to 16-bit
|
||
- agx: Pack local atomics
|
||
- agx: Implement b2b32
|
||
- agx: Handle group_memory_barrier
|
||
- agx: Add and use agx_nir_ssa_index helper
|
||
- agx: Handle ssa_undef as zero
|
||
- agx: Add agx_internal_format_supports_mask helper
|
||
- asahi: Implement color masks with masked stores
|
||
- asahi: Make shader-db work again
|
||
- panfrost: Use proper locations in blend shaders
|
||
- nir/lower_blend: Consume dual stores
|
||
- nir: Add nir_texop_lod_bias_agx
|
||
- asahi: Lower lod_bias_agx to uniform registers
|
||
- agx: Lower sampler LOD bias
|
||
- nir/lower_blend: Don't dereference null
|
||
- docs/feature: Mark ARB_sync as done on Asahi
|
||
- asahi/decode: Handle VDM barriers
|
||
- nir: Add nir_lower_helper_writes pass
|
||
- pan/mdg: Use nir_lower_helper_writes
|
||
- asahi: Advertise dual-source blending
|
||
- agx: Mask shifts in the backend
|
||
- agx: Fix 2D MSAA array texture register allocation
|
||
- asahi: Mark PIPE_FORMAT_NONE "supported"
|
||
- agx: Don't write sample mask from preambles
|
||
- agx: Add AGX_MESA_DEBUG=nopreamble option
|
||
- agx: Clean up after lowering address arithmetic
|
||
- agx: Factor out allows_16bit_immediate check
|
||
- agx: Inline 16-bit load/store offsets
|
||
- agx: Constify agx_print
|
||
- agx: Refactor vector creation
|
||
- agx: Use agx_emit_collect for st_tile
|
||
- agx: Don't print pre-optimization shader
|
||
- agx: Only lower int64 late
|
||
- asahi: Bump shader buffers
|
||
- asahi/meta: Use lowered I/O
|
||
- agx: Disable tri merging with side effects
|
||
- agx: Handle fragment shader side effects
|
||
- asahi: Rework system value lowering
|
||
- asahi: Wire up compute kernels
|
||
- nir/lower_tex: Add lower_index_to_offset
|
||
- pan/bi: Use lower_index_to_offset
|
||
- ir3: Use lower_index_to_offset
|
||
- nir/opt_barrier: Generalize to control barriers
|
||
- glsl/nir: Use scoped_barrier for control barrier
|
||
- pan/bi: Drop control_barrier handling
|
||
- pan/mdg: Drop control_barrier handling
|
||
- ir3: Drop non-scoped barrier handling
|
||
- gallivm: Drop non-scoped barrier handling
|
||
- agx/lower_address: Break on match
|
||
- agx/lower_address: Optimize "shift + constant"
|
||
- agx/lower_address: Handle large shifts
|
||
- agx/lower_address: Handle 8-bit load/store
|
||
- agx/lower_address: Fix handling of 64-bit immediates
|
||
- agx/lower_address: Handle 16-bit offsets
|
||
- agx: Assert that memory index is 32-bit reg
|
||
- agx: Fix clang-formatting
|
||
- agx: Pack indirect texture/sampler handles
|
||
- agx: Handle indirect texture/samplers
|
||
- asahi: Don't allow linear depth/stencil buffers
|
||
- asahi, agx: Implement dummy samplers
|
||
- asahi,agx: Implement buffer textures with gnarly NIR
|
||
- panfrost: Remove some unused definitions
|
||
- docs/panfrost: Move description of instancing
|
||
- panfrost: Don't use DECODE_FIXED16 for sample position
|
||
- panfrost: Handle fixed-point packing in GenXML
|
||
- panfrost: Add XML for framebuffer pointers
|
||
- panfrost: Use framebuffer pointer XML
|
||
- panfrost: Remove FBD tag enum from XML
|
||
- panfrost: Inline the last MALI_POSITIVE use
|
||
- panfrost: Remove MALI_POSITIVE macro
|
||
- pan/mdg: Remove reference to removed macro
|
||
- agx: Don't set lower_pack_split
|
||
- agx: Make partial DCE optional
|
||
- agx: Fix subdivision coalescing
|
||
- agx: Implement extract_[ui]16
|
||
- agx: Use nir_lower_mem_access_bit_sizes
|
||
- agx: Switch to scoped_barrier
|
||
- nir/lower_point_size: Use shader_instructions_pass
|
||
- ail: Restructure generated tests
|
||
- agx: Lower discard late
|
||
- util/prim_convert: Don't set index_bounds_valid
|
||
- pan/bi: Ignore signedness in vertex fetch
|
||
- panfrost: Identify "Base vertex offset" signedness
|
||
- panfrost: Assert that we don't see unsupported vertex formats
|
||
- panfrost: Defeature 24-bit textures
|
||
- panfrost: Handle null textures robustly
|
||
- panfrost/ci: Skip draw_buffers_indexed.random.* on Midgard
|
||
- panfrost/ci: Identify some Piglit flakes
|
||
- panfrost/ci: Add some Piglit skips
|
||
- panfrost/ci: Remove fbo-mrt-new-bind fail+flake
|
||
- panfrost: Note glDrawRangeElements underflow
|
||
- asahi: Fix occlusion query lifetime
|
||
- panfrost: Don't round up Midgard polygon list BOs
|
||
- panvk: Use vk_get_physical_device_features
|
||
- asahi: Use a dynarray for writers
|
||
- ci: Add clang-format to the amd64 container
|
||
- ci: Enforce clang-format for asahi
|
||
- gallium: Fix u_stream_outputs_for_vertices with QUADS
|
||
- nir/builder: Add nir_umod_imm helper
|
||
- blorp,anv,hasvk: Use umod_imm
|
||
- v3d,v3dv: Use udiv_imm/umod_imm
|
||
- radv: Use umod_imm
|
||
- ir3: Use umod_imm
|
||
- nir: Add Panfrost intrinsics to lower sample mask
|
||
- nir: Add Mali load_output taking converison
|
||
- panfrost: Use 0/~0 boolean for MSAA sysval
|
||
- pan/bi: Don't duplicate texture op cases
|
||
- pan/bi: Lower sample mask writes in NIR
|
||
- pan/bi: Lower load_output to make sysval explicit
|
||
- pan/bi: Allow specializing bifrost_nir_options by arch
|
||
- pan/bi: Lower gl_VertexID in NIR
|
||
- pan/bi: Remove bi_load_sysval
|
||
- pan/mdg: Use I/O semantics for MRT blend stores
|
||
- panfrost: Remove inputs->blend.rt
|
||
- panfrost: Remove unused inputs.nr_cbufs
|
||
- pan/bi: Only lower once
|
||
- pan/mdg: Only lower once
|
||
- pan/bi: Split out early preprocessing from late
|
||
- pan/mdg: Split out early preprocessing from late
|
||
- pan/lower_framebuffer: Only call for FS
|
||
- pan/lower_framebuffer: Use nir_shader_instructions_pass
|
||
- pan/blit: Lower load_sampler_lod_parameters_pan
|
||
- panfrost: Preprocess shaders in the driver
|
||
- pan/lower_framebuffer: Lower MSAA blend shaders
|
||
- panfrost: Lower clip_fs late
|
||
- panfrost: Lower texcoords late
|
||
- panfrost: Effectively lower gl_FragColor late
|
||
- panfrost: Preprocess shaders at CSO create time
|
||
- panfrost: Remove stale TODO
|
||
- panvk: Lower sysvals in NIR
|
||
- panvk: Don't use vec4 for vertex_instance_offsets
|
||
- panvk: Inline blend constants as syvals
|
||
- panfrost: Add NIR-based sysval lowering pass
|
||
- panfrost: Lower sysvals in GL
|
||
- panfrost: Move sysvals to GL driver struct
|
||
- panvk: Remove unused function
|
||
- panfrost: Move panfrost_sysvals to GL driver
|
||
- pan/bi: Export bifrost_nir_lower_load_output
|
||
- pan/bi: Call pan_nir_lower_zs_store late
|
||
- panvk: Lower blending late
|
||
- panfrost: Remove Midgard RSD fields from Bifrost
|
||
- asahi: Convert to SPDX headers
|
||
- mesa/st: Only set seamless for GLES3
|
||
- mesa/st: Normalize wrap modes for seamless cubes
|
||
- asahi: Don't lie about seamless cube maps
|
||
- panfrost: Print perf debug when flushing everything
|
||
- panfrost: Print perf debug on seqnum overflow
|
||
- panfrost: Don't redundantly call emit_const_buf
|
||
- panfrost: Mark packs as ALWAYS_INLINE
|
||
- panfrost: Don't update access with a single batch
|
||
- panfrost: Add a v9 fast path for no images
|
||
- panfrost: Clean up tiler calculations
|
||
- panfrost: Estimate vertex count for hier mask
|
||
- panfrost: Choose hierarchy masks by vertex count
|
||
- docs: Remove docs about macOS hardware drivers
|
||
- nv50,nvc0: Use u_pipe_screen_get_param_defaults
|
||
- panfrost: Always upload a workaround sampler
|
||
- pan/{mdg,bi}: Always use sampler 0 for txf
|
||
- panfrost: Unset TEXTURE_BUFFER_SAMPLERS
|
||
- gallium: Remove PIPE_CAP_TEXTURE_BUFFER_SAMPLER
|
||
- docs/gallium: Note samplers are not used for txf
|
||
- nir/print: Don't print sampler_index for txf
|
||
- asahi: Support more renderable formats
|
||
- agx: DCE even with noopt
|
||
- agx: Assert that we don't overflow registers
|
||
- agx: Constify agx_{read,write}_registers
|
||
- agx: Don't allow uniform source to local_atomic
|
||
- agx: Don't destroy usub_sat with constant
|
||
- asahi: Add perf debug for generate_mipmap
|
||
- asahi: Add perf debug for shader variants
|
||
- agx: Set loads_varying accurately
|
||
- agx: Add helper for calculating occupancy
|
||
- asahi/decode: Remove agxdecode_dump_bo
|
||
- asahi/decode: Print VDM barriers
|
||
- asahi: Set PIPE_CAP_LOAD_CONSTBUF
|
||
- agx: Coalesce more collects
|
||
- agx: Don't overallocate registers
|
||
- asahi: Honour sampler count
|
||
- asahi: Implement null textures
|
||
- asahi: Lower 1D to 2D
|
||
- asahi: Dirty track depth bias uploads
|
||
- asahi: Clamp texture buffer sizes
|
||
- agx: Tease apart some sample_mask packing magic
|
||
- agx: Rename writeout to wait_pix
|
||
- agx: Make signal_pix instructions explicit
|
||
- vulkan: Add common features2_to_features
|
||
- radv: Use vk_features2_to_features
|
||
- v3dv: Use vk_features2_to_features
|
||
- lavapipe: Use vk_features2_to_features
|
||
- pvr: Use vk_features2_to_features
|
||
- anv,hasvk: Use vk_features2_to_features
|
||
- tu: Use vk_features2_to_features
|
||
- nir: Combine if_uses with instruction uses
|
||
- nir/opt_ray_queries: Don't use list_length
|
||
- nir/opt_loop_unroll: Avoid list_length
|
||
- nir: Remove 2nd argument from nir_before_src
|
||
- nir/validate: Don't treat if-uses specially
|
||
- dxil: Avoid list_length
|
||
- nir: Reduce indirection
|
||
- nir: Factor out nir_src_rewrite_ssa helper
|
||
- nir: Use nir_src_rewrite_ssa
|
||
- dxil: Use nir_src_rewrite_ssa
|
||
- nir: Remove nir_if_rewrite_condition_ssa
|
||
- nir/repair_ssa: Refactor some use handling
|
||
- nir/validate: Only walk uses once
|
||
- mailmap: Update my e-mail
|
||
- panfrost: Symlink gallium .clang-format to common
|
||
- panfrost/winsys: Add .clang-format for winsys folder
|
||
- panfrost/winsys: Clang-format
|
||
- pan/decode: Move comment out of designated initializer
|
||
- panfrost: Re-run clang-format
|
||
- panvk: Clang-format
|
||
- ci: Run clang-format on panfrost
|
||
- mesa/st: Set uses_sample_shading when forcing per-sample
|
||
- nir/lower_blend: Set uses_fbfetch_output conservatively
|
||
- nir/lower_blend: Enable per-sample shading
|
||
- pan/bi: Lower swizzles for 8-bit CSEL
|
||
- pan/bi: Respect swizzles for more vector ops
|
||
- pan/bi: Use nir_lower_mem_access_bit_sizes
|
||
- panfrost: Allocate shared memory in OpenCL
|
||
- pan/decode: Print compute job payloads
|
||
- asahi: Fix disk cache disable with AGX_MESA_DEBUG
|
||
|
||
Amber (15):
|
||
|
||
- util/u_trace: pass utrace context to marker functions.
|
||
- freedreno: add support for markers.
|
||
- ir3, isaspec: add raw instruction to assembler/disassembler.
|
||
- ir3: support texture and sampler index with offsets
|
||
- nir: support lowering nir_intrinsic_image_samples to a constant load
|
||
- ir3: use lower_image_samples_to_one
|
||
- intel/compiler: use lower_image_samples_to_one
|
||
- freedreno: make sure depth/stencil layouts are always tiled
|
||
- freedreno: use A6XX_GRAS_SC_CNTL_SINGLE_PRIM_MODE with fb readback
|
||
- gallium: make BlendCoherent usable from gallium drivers
|
||
- freedreno: use blendcoherent to set FLUSH_PER_OVERLAP
|
||
- freedreno: check for conditional rendering in launch_grid
|
||
- nir: allow nir_lower_fb_read to support multiple render targets
|
||
- nir: Add memory coherency information to shaders.
|
||
- freedreno, nir, ir3: implement GL_EXT_shader_framebuffer_fetch
|
||
|
||
Andres Calderon Jaramillo (1):
|
||
|
||
- r600: Report multi-plane formats as unsupported
|
||
|
||
André Almeida (2):
|
||
|
||
- radv: Implement vk.check_status
|
||
- winsys/amdgpu: Fix amdgpu_cs_query_reset_state2 error log
|
||
|
||
Antonio Gomes (11):
|
||
|
||
- rusticl: Enabling reading/writing for images created from buffers
|
||
- rusticl: Enabling image fill for images created from buffers
|
||
- rusticl: Enable copy for images created from buffers
|
||
- rusticl: Enable mapImage for images created from buffers
|
||
- gallium, rusticl: Add tex2d_from_buf in image_view and sampler_view
|
||
- mesa/st, nine, nouveau: Fix uninitialized pipe_sampler_view structs
|
||
- lvmpipe/cs: Add support for 2d images created from buffers
|
||
- gallium: Add new caps PIPE_CAP_LINEAR_IMAGE_(PITCH_ALIGNMENT|BASE_ADDRESS_ALIGNMENT)
|
||
- rusticl: Implement spec for cl_khr_image2d_from_buffer
|
||
- llvmpipe: Add new caps PIPE_CAP_LINEAR_IMAGE_(PITCH_ALIGNMENT|BASE_ADDRESS_ALIGNMENT)
|
||
- iris: Add support for 2d images created from buffers
|
||
|
||
Anuj Phogat (3):
|
||
|
||
- anv: implement TES distribution mode WA 22012785325
|
||
- iris: implement TES distribution mode WA 22012785325
|
||
- intel/genxml/125: Add preferred SLM allocation size field
|
||
|
||
Asahi Lina (43):
|
||
|
||
- asahi: Split off common BO code into its own file
|
||
- asahi: Split off macOS support into its own file
|
||
- asahi: Refuse to transfer out-of-bounds mip levels
|
||
- meson: Fix Asahi build on macOS
|
||
- asahi: Fix shader key cloning overreads
|
||
- asahi: Do not use memctx for pools / meta cache
|
||
- asahi: Drop agx_device.memctx
|
||
- asahi: Only apply FS lowerings to fragment shaders
|
||
- asahi: Add BO_SHAREABLE flag
|
||
- asahi: Add readonly BO flag
|
||
- asahi: Identify USC cache invalidate
|
||
- asahi: Flush USC caches on the first draw
|
||
- asahi: Drop macOS backend
|
||
- asahi: Add nocluster,sync,stats debug flags
|
||
- asahi: Align device submission API with upcoming UAPI
|
||
- asahi: Implement Linux driver scaffolding, sans UAPI
|
||
- asahi: Add APIs for DMA-BUF sync file import/export
|
||
- asahi: Add agx_debug_fault() helper
|
||
- asahi: Add result buffer to context/batches
|
||
- asahi: Add agx_bo_mmap() calls to transfer path
|
||
- asahi: Pull device name from device struct
|
||
- asahi: Do not overread user index buffers
|
||
- asahi: Fix scissor culling check when out of bounds for FB/viewport
|
||
- asahi: Fix device fd leak in agx_close_device
|
||
- asahi: Destroy the renderonly context on screen destroy
|
||
- asahi: clang-format the world again
|
||
- asahi: Assert on TIB strides > 64
|
||
- asahi: Support importing sync objects on BO export
|
||
- asahi: Make agx_flush_resource reallocate non-shareable resources
|
||
- asahi: Extend batch tracking for explicit sync
|
||
- Revert "asahi: Advertise dual-source blending"
|
||
- asahi: Make agx_alloc_staging() take a screen instead of a context
|
||
- asahi: Enable glthread
|
||
- asahi: Locate low VA BOs correctly
|
||
- asahi: Fix style nits
|
||
- asahi: Implement valid buffer range tracking
|
||
- asahi: Make BO import path failures more robust
|
||
- asahi: Add a helper macro for debug/error messages
|
||
- asahi: Add resource debugging
|
||
- asahi: Print reasons why compression is disabled
|
||
- asahi: Fix compressed ZS support
|
||
- asahi: Flip kmsro around to allocate on the GPU
|
||
- asahi: Allow explicit non-LINEAR modifiers for scanout
|
||
|
||
Axel Davy (1):
|
||
|
||
- frontend/nine: Fix num_textures count
|
||
|
||
Bas Nieuwenhuizen (26):
|
||
|
||
- aco: Pass correct number of coords to Vega 1D LOD instruction.
|
||
- radv: Strictly limit alignment needed within a descriptor set.
|
||
- radv: Reduce descriptor pool allocation for alignment.
|
||
- radv: Set FDCC_CONTROL SAMPLE_MASK_TRACKER_WATERMARK
|
||
- radv: Shift left the tile swizzle more on GFX11.
|
||
- nir: Apply a maximum stack depth to avoid stack overflows.
|
||
- radv: Add helper to hash stages.
|
||
- radv: Hash group handles as part of RT pipeline key.
|
||
- radv: Use provided handles for switch cases in RT shaders.
|
||
- radv: Use group handles based on shader hashes.
|
||
- radv: Implement & expose VK_EXT_pipeline_library_group_handles.
|
||
- Update my mailmap aliases
|
||
- ac/surface,radv: Avoid pitch weirdness if image not used for rendertarget.
|
||
- ac/surface: Only allow stencil pitch adjustment for mipmaps.
|
||
- ac/surface,radv: Opt out of stencil adjust.
|
||
- util: Add aligned int64_t types for x86(non 64).
|
||
- util/disk_cache: Align atomic size.
|
||
- radv: Align atomic values.
|
||
- radv: Reserve space in framebuffer emission.
|
||
- radv: Reserve space in various streamout functions.
|
||
- radv: Reserve space in conditional rendering functions.
|
||
- radv: Reserve space in si_cs_emit_cache_flush.
|
||
- radv: Reserve space for updating DCC metadata.
|
||
- radv: Reserve space for fast clear related writes.
|
||
- radv: Reserve space for indirect descriptor set address writes.
|
||
- radv: Move all the dirty flags from TES binding to TCS binding.
|
||
|
||
Benjamin Cheng (1):
|
||
|
||
- radv: initialize cmd_buffer upload list earlier
|
||
|
||
Boyuan Zhang (6):
|
||
|
||
- radeonsi/vcn: check fence before destroying dpb
|
||
- radeonsi/vcn: check fence before destroying decoder
|
||
- radeonsi/vcn: validate fence handle before using it
|
||
- virgl/video: disable decoder fence
|
||
- virgl: add more formats to conv table
|
||
- frontends/va: check decoder in va surface call
|
||
|
||
Brian Paul (9):
|
||
|
||
- anv: add a third memory type for LLC configuration
|
||
- llvmpipe: do additional checks in lp_state_fs_analysis.c for linear shaders
|
||
- llvmpipe: remove debug printf spam in lp_setup_wait_empty_scene()
|
||
- gallium/xlib: call fence_finish() in XMesaSwapBuffers()
|
||
- llvmpipe: fix ps invocations query bug
|
||
- llvmpipe: rename some vars related to occlusion query and ps invocations
|
||
- llvmpipe: s/tabs/spaces/
|
||
- llvmpipe: s/unsigned/enum pipe_query_type/
|
||
- llvmpipe: clean-up llvmpipe_get_query_result()
|
||
|
||
Błażej Szczygieł (1):
|
||
|
||
- glx: Fix glXGetFBConfigFromVisualSGIX
|
||
|
||
Caio Oliveira (26):
|
||
|
||
- glsl: Account for unsized arrays in NIR linker
|
||
- hasvk: Update driver name in debug information
|
||
- intel: Add extra zeros at the end of debug identifiers
|
||
- iris, crocus: Align workaround address to 32B
|
||
- anv, hasvk: Align workaround address to 32B
|
||
- nir: Add nir_intrinsic_rotate
|
||
- nir/lower_subgroups: Add option lower_rotate_to_shuffle
|
||
- spirv: Implement SPV_KHR_subgroup_rotate
|
||
- nir: Support use_scoped_barrier in nir_lower_atomics_to_ssbo
|
||
- microsoft/compiler: Handle scoped barrier in Tess splitting
|
||
- gallivm: Fix handling of nir_intrinsic_scoped_barrier
|
||
- glsl: Implement use_scoped_barrier option for lowering memory barriers
|
||
- intel/compiler: Mark various memory barriers intrinsics unreachable
|
||
- pan/compiler: Fix handling of nir_intrinsic_scoped_barrier
|
||
- pan/midgard: Handle nir_intrinsic_scoped_barrier in Midgard compiler
|
||
- panfrost: Use NIR scoped barriers instead of memory barriers
|
||
- spirv: Don't specify nir_var_uniform or nir_var_mem_ubo in barriers
|
||
- spirv/tests: Subclass spirv_test helper to namespace the tests
|
||
- spirv/tests: Add script to generate C array from SPIR-V source
|
||
- spirv/tests: Parametrize stage in get_nir() helper
|
||
- spirv/tests: Add some basic control flow tests
|
||
- spirv: Add skip_os_break_in_debug_build option to use in unit tests
|
||
- intel/fs: Handle scoped barriers with execution scope
|
||
- intel/vec4: Handle scoped barriers with execution scope
|
||
- intel/compiler: Drop brw_nir_lower_scoped_barriers
|
||
- intel/compiler: Drop non-scoped barrier handling
|
||
|
||
Caleb Cornett (6):
|
||
|
||
- d3d12: Lower minimum supported Shader Model to 6.0
|
||
- futex: Change INT_MAX to INT32_MAX.
|
||
- util: Add #ifdefs for Xbox GDK support.
|
||
- dxil_validator: Add support for Xbox GDK.
|
||
- wgl: Add support for Xbox GDK.
|
||
- d3d12: Add support for Xbox GDK.
|
||
|
||
Charlie Birks (1):
|
||
|
||
- docs: add a few vulkan extensions supported by multiple drivers
|
||
|
||
Charmaine Lee (5):
|
||
|
||
- svga: fix resource_get_handle from resource created without SHARED bind flag
|
||
- svga: fix compatible formats for shareable surfaces
|
||
- svga: use upload buffer if texture has pending changes
|
||
- translate: do not clamp element index in generic_run
|
||
- svga: set PIPE_CAP_VERTEX_ATTRIB_ELEMENT_ALIGNED_ONLY for VGPU10 device
|
||
|
||
Chia-I Wu (38):
|
||
|
||
- turnip: replace TU_DEBUG_DONT_CARE_AS_LOAD by a bool
|
||
- turnip: make debug_flags a global variable
|
||
- freedreno: add has_implicit_modifier helper
|
||
- freedreno: support UBWC scanout
|
||
- turnip: add a comment to tu_format_for_aspect
|
||
- turnip: move a comment about FMT6_Z24_UNORM_S8_UINT_AS_R8G8B8A8
|
||
- turnip: remove tu_native_format::tile_mode
|
||
- turnip: make tu6_format_*_supported static
|
||
- turnip: let tu6_format_vtx* take pipe format
|
||
- turnip: add blit_format_texture
|
||
- turnip: add blit_format_color and blit_base_format
|
||
- turnip: handle ubwc in blit_base_format
|
||
- turnip: reorder tu6_format_*
|
||
- freedreno/registers: correct WFM bit in CP_REG_TEST
|
||
- turnip: add a comment to tu_render_pass_cond_config
|
||
- turnip: skip unnecessary CP_REG_TEST for cond load/store
|
||
- freedreno/registers: document more bits of CP_REG_TEST
|
||
- freedreno: avoid conditional ib in fd6_emit_tile
|
||
- radv: fix a hang with binning on CHIP_RENOIR
|
||
- turnip: fix a major leak with GPL LTO
|
||
- turnip: fix a null descriptor set dereference
|
||
- turnip: avoid FMT6_Z24_UNORM_S8_UINT_AS_R8G8B8A8 for event blits
|
||
- radv: add a size check in radv_create_buffer for Android
|
||
- util/log: refactor mesa_log
|
||
- util/log: allow multiple loggers
|
||
- util/log: improve logger_file newline handling
|
||
- util/log: improve logger_android
|
||
- util/log: add logger_syslog
|
||
- util/log: add support for MESA_LOG_FILE
|
||
- util/log: add logger_windbg
|
||
- mesa: add missing newlines for _mesa_debug/_mesa_log callers
|
||
- mesa: use mesa_log from output_if_debug
|
||
- anv: process utrace payloads on queue submission
|
||
- ci/radv: remove dEQP-VK.image.sample_texture.* fails/flakes
|
||
- radv: set RADEON_FLAG_GTT_WC for external mem on vram
|
||
- radv: rework radv_layout_fmask_compressed
|
||
- radv: add RADV_FMASK_COMPRESSION_PARTIAL
|
||
- radv: disable tc_compatible_cmask on GFX9 in some cases
|
||
|
||
Christian Gmeiner (1):
|
||
|
||
- etnaviv: nir: use lower_fround_even
|
||
|
||
Collabora's Gfx CI Team (3):
|
||
|
||
- Uprev Piglit to 60e7f0586bac0cfcfcb5871046e31ca2057a5117
|
||
- Uprev Piglit to 2391a83d1639a7ab7bbea02853b922878687b0e5
|
||
- Uprev Piglit to 355ad6bcb2cb3d9e030b7c6eef2b076b0dfb4d63
|
||
|
||
Connor Abbott (12):
|
||
|
||
- freedreno/a6xx: Rename CP_CSQ_IB*_STAT
|
||
- freedreno/a6xx: Add CP_ROQ_*_STAT
|
||
- freedreno/a6xx: Fix CP_ROQ_THRESHOLDS_1
|
||
- freedreno/a6xx: Fill in ROQ status registers
|
||
- freedreno/crashdec: Fix apparent off-by-one with ROQ size
|
||
- freedreno/crashdec: Add prefetch test
|
||
- tu: Fix tile_align_h on a650
|
||
- freedreno: Fix or/and'ing two BitmaskEnums
|
||
- tu: Use vk_pipeline_get_renderpass_flags()
|
||
- vk/render_pass: Support VK_EXT_fragment_density_map
|
||
- nir, spirv: Add support for VK_EXT_fragment_density_map
|
||
- tu: Don't override depth for GMEM
|
||
|
||
Constantine Shablya (12):
|
||
|
||
- anv: handle ATTACHMENT_OPTIMAL layout
|
||
- anv: use Vulkan runtime's robust buffer access
|
||
- hasvk: use Vulkan runtime's robust buffer access
|
||
- anv,hasvk: flush what UNIFORM_READ flushes on SHADER_READ
|
||
- vulkan: relocate rmv to its correct home
|
||
- vulkan: tidy up vk_physical_device_features
|
||
- vulkan: delete trailing namespace
|
||
- vulkan: add hepler for vkGetPhysicalDeviceFeatures2
|
||
- vulkan: use vk_features for vk_device::enabled_features
|
||
- anv: use vk_get_physical_device_features
|
||
- vulkan: fix building with python3.8
|
||
- vulkan: depend idep_vulkan_runtime_headers on vk_physical_device_features.h
|
||
|
||
Corentin Noël (12):
|
||
|
||
- ci/venus: Remove failure now passing
|
||
- kopper: Do not free the given screen in initScreen implementation
|
||
- ci: uprev virglrenderer
|
||
- ci/venus: Skip tests risking out of memory issues
|
||
- ci: uprev virglrenderer and crosvm
|
||
- ci: Setup XDG_RUNTIME_DIR in crosvm-init
|
||
- ci: Allow to use crosvm-runner before deqp-runner
|
||
- ci: Uprev crosvm and virglrenderer
|
||
- venus/ci: Only run one crosvm instance
|
||
- mesa: OpenGL ES 3.0 requires EXT_instanced_arrays
|
||
- glapi: Make EXT_draw_instanced functions available for GLES 2.0
|
||
- mesa: Add EXT_instanced_arrays support
|
||
|
||
Daniel Schürmann (82):
|
||
|
||
- radv: CSE ray_launch_{size|id}
|
||
- radv: rename shader_info->cs.uses_sbt -> shader_info->cs.is_rt_shader
|
||
- radv: unconditionally enable scratch for RT shaders
|
||
- radv/rt: introduce and set rt_pipeline->stack_size
|
||
- radv/rt: use dynamic_callable_stack_base also for static stack_sizes
|
||
- radv/rt: don't hash maxPipelineRayRecursionDepth
|
||
- nir: add Continue Construct to nir_loop
|
||
- nir: add assertions that loops don't have a Continue Construct
|
||
- nir: create nir_push_continue() and related helpers
|
||
- nir: add lowering for Loop Continue Constructs
|
||
- spirv: use Loop Continue Construct to emit SPIR-V loops and lower after parsing
|
||
- nir/lower_continue_constructs: special-case Continue Constructs with zero or one predecessors
|
||
- nir/lower_continue_targets: only repair SSA when necessary
|
||
- nir: simplify nir_block_cf_tree_{next|prev}
|
||
- radv/rt: rename library_pipeline->groups to library_pipeline->group_infos
|
||
- radv/rt: defer library_pipeline allocation
|
||
- radv/rt: introduce struct radv_ray_tracing_module
|
||
- radv/rt: move stack_sizes into radv_ray_tracing_module
|
||
- radv/rt: only reserve stack_sizes after rt_case insertion
|
||
- radv: expose radv_postprocess_nir()
|
||
- radv: expose radv_pipeline_capture_shaders()
|
||
- radv/rt: introduce and use radv_rt_pipeline_compile()
|
||
- radv: remove unused parameters from radv_compute_pipeline_compile()
|
||
- radv/rt: move radv_pipeline_key from rt_variables to traversal_data
|
||
- nir/gather_info: allow terminate() in non-PS
|
||
- aco: fix NIR infinite loops
|
||
- radv/rt: use terminate() when returning from raygen shaders
|
||
- aco/dominance: set immediate dominator for any BB without predecessors
|
||
- aco/value_numbering: clear hashmap between disconnected CFGs
|
||
- aco/dead_code_analysis: don't add artificial uses to p_startpgm
|
||
- aco/insert_exec_mask: allow for disconnected CFG
|
||
- aco/spill: allow for disconnected CFG
|
||
- radv/rt: place any-hit scratch vars after intersection scratch vars
|
||
- radv/rt: Fix any_hit scratch variables.
|
||
- mesa: add gl_shader_stage_is_rt()
|
||
- radv: add RT shader args
|
||
- radv: handle RT stages in radv_nir_shader_info_pass()
|
||
- radv: add RT stages to radv_get_shader_name()
|
||
- radv: add RT shader handling to radv_postprocess_config
|
||
- aco: add RT stage enums
|
||
- aco: don't set private_segment_buffer/scratch_offset on GFX9+
|
||
- aco: move rt_dynamic_callable_stack_base_amd to VGPR
|
||
- aco: implement load_ray_launch_{id|size}
|
||
- aco: create hw_init_scratch() function for p_init_scratch lowering
|
||
- aco: implement select_rt_prolog()
|
||
- radv: add radv_create_rt_prolog()
|
||
- radv: compile rt_prolog
|
||
- radv/rt: use prolog for raytracing shaders
|
||
- aco: remove aco::rt_stack variable
|
||
- radv: remove unused parameter from radv_open_rtld_binary()
|
||
- radv: separate radv_postprocess_binary_config() from radv_shader_create()
|
||
- radv: remove unnecessary copy of binary->config
|
||
- radv: inline radv_postprocess_config()
|
||
- radv: separate radv_capture_shader_executable_info() from radv_shader_create()
|
||
- radv: move gl_shader_stage from radv_binary to radv_shader_info
|
||
- radv: remove radv_create_gs_copy_shader()
|
||
- radv: refactor shader_compile()
|
||
- radv: skip pipeline caching with RADV_DEBUG=shaders
|
||
- radv: fix radv_shader_binary member fields to 32 bit.
|
||
- radv/rt: Fix VK_KHR_pipeline_executable_properties
|
||
- aco: split ps_epilog args before exporting them
|
||
- aco/ra: adjust_max_used_regs() for fixed Operands
|
||
- aco: don't use shared VGPRs for shaders consisting of multiple binaries
|
||
- radv: update PS num_vgprs in case of epilogs rather than overallocating VGPRs
|
||
- vulkan/pipeline_cache: remove vk_device from vk_pipeline_cache_object
|
||
- vulkan/pipeline_cache: Don't re-insert disk-cache hits into disk-cache
|
||
- vulkan/pipeline_cache: implement vk_pipeline_cache_create_and_insert_object()
|
||
- vulkan/pipeline_cache: use vk_pipeline_cache_create_and_insert_object() during vk_pipeline_cache_load()
|
||
- vulkan/pipeline_cache: add cache parameter to deserialize() function
|
||
- vulkan/pipeline_cache: move vk_log on failed deserialization to vk_pipeline_cache_load()
|
||
- radv: derive struct radv_shader from vk_pipeline_cache_object
|
||
- radv: unconditionally store the binary code in radv_shader
|
||
- radv: add radv_shader_serialize() and radv_shader_deserialize() functions
|
||
- radv: add struct radv_pipeline_cache_object
|
||
- radv: implement radv_shader_create_cached()
|
||
- radv: use vk_pipeline_cache
|
||
- radv: clean up pipeline-cache interface
|
||
- radv/ci: add 2 more Flakes for Navi21
|
||
- radv/rt: fix total stack size computation
|
||
- radv/rt: properly destroy radv_ray_tracing_lib_pipeline on error
|
||
- vulkan/pipeline_cache: replace raw data objects on cache insertion of real objects
|
||
- radv: add padding to radv_shader_binary_legacy
|
||
|
||
Daniel Stone (18):
|
||
|
||
- ci/fdno: Only run full tests on a limited subset of machines
|
||
- ci/radv: Skip vkCreateInstance memory-fail test
|
||
- ci/anv: Temporarily halve TGL testing load
|
||
- intel/isl: Don't scream FINISHME into logs for 3D vs. CCS
|
||
- ci/radv: Drop raven quick_shader load
|
||
- ci/fdno: Add a618 Vulkan flakes
|
||
- ci/zink: Add flake seen in the wild
|
||
- ci/radv: Lower stoney CTS load
|
||
- ci/android: Use a more aggressive timeout for the job
|
||
- ci: Actually run Piglit on LAVA
|
||
- ci: Disable Collabora LAVA farm
|
||
- Revert "ci: Disable Collabora LAVA farm"
|
||
- CI: Disable Windows runners
|
||
- CI: Disable mingw job
|
||
- ci/panfrost: Add texturesize flake seen in the wild
|
||
- CI: Disable freedreno
|
||
- ci/radeonsi: sort and dedup stoney skips
|
||
- ci/radeonsi: Skip really slow tests on stoney
|
||
|
||
Danylo Piliaiev (52):
|
||
|
||
- tu/kgsl: do not use kgsl_command_object::offset
|
||
- tu: Prevent using stale value of RB_UNKNOWN_88D0 on BLIT
|
||
- tu: Prevent using stale value of GRAS_SC_CNTL in sysmem clear
|
||
- freedreno: Document A6XX_GRAS_SC_CNTL::rotation field
|
||
- turnip: Ensure that there is no renderpass rotation in binning
|
||
- turnip: Disable draw states after dyn renderpass in all cases
|
||
- ir3: Consider dst type in ubo_vec4 to ldc lowering
|
||
- tu: Don't expose KHR_present_id,KHR_present_wait without KHR_swapchain
|
||
- turnip: Add debug option to find usage of stale reg values
|
||
- docs/freedreno: Add info about stale reg stomper dbg option
|
||
- ci/tu: Add 1/200 pass to test for stale reg usage
|
||
- ir3: Add cat5/cat7 cache related instructions
|
||
- ir3: Add cat7 sleep instruction
|
||
- freedreno/register: Define chip enum values
|
||
- util/perf: C++-proof util/perf
|
||
- util/format: Make format_table compatible with C++
|
||
- spirv: sort spirv_supported_capabilities
|
||
- vk/vk_extension_gen: Make table struct initializable in C++ on older gcc
|
||
- vk/wsi: C++-proof wsi_common_drm.h
|
||
- vk/util: remove (void \*) casts from vk_foreach_multi_draw macros
|
||
- vk/util: Generate defines to help casting structs with vk_find_struct
|
||
- freedreno/common: C++-proof freedreno_uuid.h
|
||
- ir3: C++-proofing
|
||
- tu: C++-proofing: fix offsetof with dynamic array index
|
||
- tu: C++-proofing: fix struct initializers
|
||
- tu: C++-proofing: various enum fixes
|
||
- tu: C++-proof: do not goto over variables initialization
|
||
- tu: C++-proofing: fix designator initializer order
|
||
- tu: C++-proofing: fix extension table initialization
|
||
- tu: C++-proofing: Initialize tu_reg_value in-order by pack funcs
|
||
- tu: C++-proofing: fix casting from void * fpermissive warnings
|
||
- tu: C++-proofing: ease access to global bo struct
|
||
- tu: C++-proofing: prevent taking address from rvalue
|
||
- tu: C++-proofing: cast result when extracting field from reg value
|
||
- tu: C++-proofing: misc fixes
|
||
- freedreno/msm: Rename drm_msm_gem_submit_reloc::or in C++ code
|
||
- tu: compile as C++
|
||
- vk/entry_points: Add option to generate template entrypoints
|
||
- freedreno/regs: Include assert.h in generated headers
|
||
- tu: Generate entrypoints for each gen
|
||
- turnip: add cached and cached-coherent memory types
|
||
- tu/drm: Support cached non-coherent memory
|
||
- freedreno/registers: Document new CP_EVENT_WRITE::SEQNO
|
||
- freedreno/registers: More a7xx regs
|
||
- freedreno/computerator: C++ proofing
|
||
- freedreno: C++ fixes for computerator to compile
|
||
- freedreno/computerator: Convert to C++
|
||
- freedreno: Move fd6_pack.h to common code accessible by computerator
|
||
- freedreno: Add dummy a730/a740 definition
|
||
- freedreno/computerator: Templatize a6xx backend
|
||
- freedreno/computerator: Add support for a7xx
|
||
- vulkan: Sanitize pSampleMask in CmdSetSampleMaskEXT
|
||
|
||
Dave Airlie (37):
|
||
|
||
- ci: bump vk cts to 1.3.3.1 + and a crash fix.
|
||
- vulkan/video: add common h264/h265 parameter set management code.
|
||
- vulkan/format: add a 10-bit video format
|
||
- radv: remove the status query mark it unsupported.
|
||
- radv: add new upload alloc aligned api
|
||
- ac: add name to codec info struct
|
||
- radv: adding video decode queue support
|
||
- radv: add video decoder register setup.
|
||
- radv/video: add initial frameworking.
|
||
- radv/video: add initial h264 decoder for VCN
|
||
- radv/video: add h264 support for uvd
|
||
- radv: add vcn h265 decode.
|
||
- radv/video: add h265 decode UVD support
|
||
- radv/vcn: enable dynamic dpb tier 2 for h264/h265 on navi21+
|
||
- anv: add video engine support in various places
|
||
- anv: set Y/4 tiling for video decode images
|
||
- anv: add video format features for the one supported video output format
|
||
- anv/format: handle video extensions structs by ignoring them
|
||
- intel/genxml: align some of the fields with the media driver
|
||
- intel/genxml: add missing power well control bits
|
||
- anv/image: allocate some memory for mv storage after video images.
|
||
- anv: add initial video decode support for h264.
|
||
- anv/query: add query status report
|
||
- anv: enable video decode extensions.
|
||
- anv/video: fix video memory bindings.
|
||
- crocus: disable Y tiling for render targets properly.
|
||
- crocus: switch gen4/5 tiling flags to follow suggestions.
|
||
- llvmpipe: fix compute address bits to return native pointer size.
|
||
- anv: always pick graphics queue to execute prime blits on.
|
||
- radv: add video format support to format probing.
|
||
- anv/video: fix chroma qp to be a integer value.
|
||
- anv/video: disable picture id reampping.
|
||
- anv: fix image height for field pictures.
|
||
- radv/video: fix h264 frame heights when field images are in use
|
||
- radv/video: fix used for reference flags.
|
||
- radv/video: fix h265 decoding sizes.
|
||
- radv/trace: don't attempt to emit trace on non-graphics/compute queues
|
||
|
||
David (Ming Qiang) Wu (1):
|
||
|
||
- radeonsi/vcn: add an exception of field case for h264 decoding
|
||
|
||
David Heidelberg (73):
|
||
|
||
- ci/zink: Penumbra is now fixed.
|
||
- freedreno/ci: Switch also performance a630 job to manual
|
||
- ci/anv: add multiple fails uncovered by change of sharding
|
||
- ci/intel: fully utilize asus-cx9400-volteer
|
||
- ci/piglit: explicitely define we want GLX tests
|
||
- ci: migrate from wget to curl
|
||
- ci/piglit: 2023-01-19 uprev
|
||
- ci: bump ci-fairy with session support (robust downloads)
|
||
- ci: Sir trace has small invisible change in rendering
|
||
- ci: bump Mold to the 1.10.0
|
||
- ci: uprev piglit (etag md5 checksumming support)
|
||
- ci/lavapipe: use dxvk for the traces
|
||
- ci: revert download of git cache to the wget
|
||
- ci/llvmpipe: add flake timeout for rusticl program@execute@builtin@builtin-float-sincos-1.0
|
||
- util/process_test: make the error variable static
|
||
- intel: enable -mfpmath=sse on x86
|
||
- intel: use c_see2_arg instead of explicit -msse2
|
||
- ci/freedreno: add flaking KHR-GL45.buffer_storage.map_persistent_dispatch
|
||
- meson: print c_cpp_args
|
||
- intel/vulkan: add missing dependency on generated headers
|
||
- ci/freedreno: add flaking KHR-GL45.buffer_storage.map_persistent_flush
|
||
- ci/alpine: keep the curl inside the image
|
||
- ci: alpine: install bash and coreutils for date -d
|
||
- ci: implement unified sections
|
||
- ci: make meson build and test uncollapsed
|
||
- ci: deqp-runner: drop already unused function
|
||
- ci: Retry, retry, retry... No one likes to trigger Marge more than once.
|
||
- ci/zink: add skip for the Single-GL46.enhanced_layouts.ssb_member_align_non_power_of_2
|
||
- ci/lavapipe: add recent occasional flake
|
||
- ci/freedreno: rare flake KHR-GL45.sample_variables.mask.rgba8i.samples_4.mask_3
|
||
- crocus/meson: add dependency on libintel_dev also for versioned static libraries
|
||
- ci/ci_run_n_monitor: while we usually disable many jobs, print them inline
|
||
- ci: do not exit when an error happens inside the section
|
||
- ci/lavapipe: fixes typo
|
||
- ci/zink: fixup the zink-lvp job
|
||
- ci: disable mesa-swrast runner jobs
|
||
- ci/lava: implement the priority
|
||
- ci/weston: before testing, verify that XWayland is really running
|
||
- ci/weston: add background PID
|
||
- ci: add and utilize dalboz devices
|
||
- ci/amd: move skqp and va jobs on raven from XOrg to the XWayland
|
||
- ci/panfrost: correct the job name, as it runs on gles2
|
||
- ci/lava: every LAVA job doesn't want to run gles2 deqp, drop it
|
||
- ci: build Wayland support for the amd64
|
||
- ci/iris: update apl and glk expectations, after enabling Wayland support
|
||
- ci/clover: disable the jobs
|
||
- ci/traces: disable nheko trace with zink since it flakes
|
||
- ci/freedreno: add recent occasional flakes
|
||
- ci/traces: add two skips due to flakes
|
||
- ci/intel: add dEQP-EGL.functional.wide_color.window_fp16_default_colorspace flake
|
||
- ci: distribute XDG_RUNTIME_DIR with setup-test-env script
|
||
- ci: disable weston session timeout for llvmpipe
|
||
- meson: implement quirk for the compilation under armv7 GCC with LTO
|
||
- aco: drop leftover variable
|
||
- ci: bump Alpine to 3.17 (again)
|
||
- ci/freedreno: do not build tools executables without explicitly enabling them
|
||
- freedreno/decode: fix possible overflow
|
||
- ci: rename .lava-test to .lava-test-deqp to describe it correctly
|
||
- ci: create lava-test without deqp HWCI_TEST_SCRIPT
|
||
- ci: remove deqp from lava piglit and traces runs
|
||
- ci/freedreno: split deqp from other jobs
|
||
- ci/freedreno: define Google farm specific includes
|
||
- ci/freedreno: Make traces work on LAVA caching proxy
|
||
- ci/broadcom: test occasionally fails, but typically passes
|
||
- ci: disable lima farm, currently out-of-space, needs to be fixed
|
||
- ci: implement sections for cuttlefish
|
||
- ci/v3d: add flaking spec@ext_framebuffer_blit@fbo-sys-blit
|
||
- Revert "mesa: Enable NV_texture_barrier in GLES2+"
|
||
- ci/amd: update device status
|
||
- ci/amd: raven is currently downgraded to 2 machines only, adapt
|
||
- ci/amd: add draw.dynamic_rendering flake
|
||
- ci/freedreno: fix the a530_piglit job and switch to Weston
|
||
- panvk: clear dangling pointers
|
||
|
||
David Redondo (1):
|
||
|
||
- egl/wayland: fix oob buffer access during buffer_fds clean up
|
||
|
||
David Rosca (2):
|
||
|
||
- frontents/va: Use PIPE_USAGE_STAGING for coded buffer
|
||
- frontends/va: Map VAEncCodedBufferType buffer as PIPE_MAP_READ
|
||
|
||
Dmitry Baryshkov (2):
|
||
|
||
- freedreno/a5xx: reorder GPMU registers
|
||
- freedreno/a5xx: add SP clock control register
|
||
|
||
Dmitry Osipenko (6):
|
||
|
||
- util/cache_test: Unset env vars left after Cache.List test
|
||
- util/mesa-db: Don't account header size
|
||
- util/mesa-db: Support removal of cache entries
|
||
- util/cache_test: Remove dummy cache entry added by cache_exists()
|
||
- util/mesa-db: Introduce multipart mesa-db cache
|
||
- util/disk_cache: Switch to multipart mesa-db cache
|
||
|
||
Dylan Baker (38):
|
||
|
||
- VERSION: bump to 23.1.0-devel for further development
|
||
- docs: reset new_features.txt
|
||
- meson: bump minimum required version to meson 0.59
|
||
- meson: replace has_exe_wrapper with can_run_host_binaries
|
||
- meson: replace uses of ExternalProgram.path with .full_path
|
||
- meson: drop meson < 0.54 workaround
|
||
- meson: use a feature option for dri3
|
||
- meson: use a feature option for gallium-vdpau
|
||
- meson: use a feature option for gallium-va
|
||
- meson: use a feature option for gallium-xa
|
||
- meson: use a feature option for shader_cache
|
||
- meson: use a feature option for shared-glapi
|
||
- meson: use a feature option for gles1
|
||
- meson: use a feature option for gles2
|
||
- meson: use a feature option for gbm
|
||
- meson: use a feature option for llvm
|
||
- meson: use a feature option for valgrind
|
||
- meson: use a feature option for libunwind
|
||
- meson: use a feature option for lmsensors
|
||
- meson: use a feature option for power8
|
||
- meson: use a feature option for xlib-lease
|
||
- meson: use a feature option for zstd
|
||
- meson: use a feature option for egl
|
||
- meson: use a feature option for shared-llvm
|
||
- meson: Use feature option methods for xmlconfig
|
||
- meson: remove version checks for < 0.59
|
||
- meson: use builtin support for reading version from a file
|
||
- meson: use [] instead of 'lib for !windows name_prefix
|
||
- meson: use the same workaround for setting 'lib' on windows
|
||
- meson: combine checks for linker --gc-sections support
|
||
- util: rzalloc and free hash_table_u64
|
||
- iris: consider bufmgr creation to have failed if \`dup`ing of the fd fails
|
||
- intel/mi: use 64bit constant for bitshift
|
||
- intel/dev: create a helper dependency for libintel_dev
|
||
- docs: Add calendar entries for 23.0 release.
|
||
- docs: add release notes for 23.0.0
|
||
- docs: Add sha256 sum for 23.0.0
|
||
- docs/relnotes: add 23.0.0 to relnotes.rst
|
||
|
||
Ella Stanforth (1):
|
||
|
||
- v3dv: add support for multi-planar formats, enable YCbCr
|
||
|
||
Emma Anholt (211):
|
||
|
||
- dri2: Fix exposing robustness with swkms.
|
||
- ci/llvmpipe: Drop dEQP-EGL.functional.sharing.*.link.7 flakes.
|
||
- ci/iris: Add known flakes for skqp.
|
||
- ci/iris: Generalize the 8888_pbuffer EGL known flakes and share with GLK.
|
||
- ci/zink: Add more blit conversion xfails for a618.
|
||
- freedreno: Skip CPU/GPU timestamp sync when not supported.
|
||
- ci/freedreno: Add glx-swap-event-async as a flake.
|
||
- freedreno/pps: Fix a signed/unsigned complaint.
|
||
- ci: Enable building the testing drivers with perfetto.
|
||
- ci: Add some new folks to the restricted-traces access list.
|
||
- Revert "nouveau/ci: temporary disable gk20a-gles"
|
||
- ci/virgl: Disable iris traces for now while it's unstable.
|
||
- ci: Drop windowoverlap xfails, since it's always skipped.
|
||
- ci/zink: Drop xfail for copy-sub-buffer.
|
||
- ci/zink: Drop glx-swap-copy xfails.
|
||
- ci/zink: Clear issue #7781 flakes.
|
||
- ci/freedreno: Switch the piglit job to using a deqp-runner suite.
|
||
- ci: Move PIGLIT_PLATFORM settings out of the .tomls.
|
||
- ci/piglit: Add some common piglit skips for Mesa CI's testing of glx.
|
||
- ci/piglit: Exclude swapbuffers front-readback tests with PIGLIT_PLATFORM=gbm.
|
||
- zink: Fatal error if requesting validation and we fail to load the layer.
|
||
- zink: Add missing Flat decorations on some inputs.
|
||
- zink: Fix validation failure for maxLod < minLod.
|
||
- zink: Fix up mismatches of memory model vs addressing model.
|
||
- zink: Re-emit the SpvBuiltInSampleMask access chain each load.
|
||
- ci/zink: Add coverage using the vulkan validation layer on lvp.
|
||
- ci/zink: Update TGL full-run xfails.
|
||
- ci/zink: Update radv xfails for the recent shadow fixes.
|
||
- ci/freedreno: Mark max-texture-size as a flake.
|
||
- ci: Move the performance jobs' allow_failure:true to the gl rules.
|
||
- ci: Add manual rules variations to disable irrelevant driver jobs.
|
||
- freedreno: Don't sync timestamps while perfetto isn't running.
|
||
- ci/zink: Disable Amnesia trace until the linked issue gets fixed.
|
||
- ci/zink: Move the zink-anv-tgl manual full run to custom manual deps.
|
||
- ci: Run our manual jobs during the nightly scheduled run.
|
||
- ci: Fix perf jobs blocking Marge pipelines.
|
||
- ci: Fix perf job condition.
|
||
- ci: Drop the itoral-gl-terrain demo from traces.
|
||
- tu: Mark tiling impossible if we couldn't lay out gmem in the first place.
|
||
- turnip: Optimize tile sizes to reduce the number of bins.
|
||
- tu: Only emit the conditional gmem subpass resolves when gmem is possible.
|
||
- turnip: Make the tiling-impossible case have an impossible tile layout.
|
||
- gallivm: Optimize emit_read_invocation's first-invocation loop.
|
||
- gallivm: Refactor out a shared "get the first active invocation" loop.
|
||
- gallivm: Return 0 first_active_invocation when we know that up front.
|
||
- gallivm: Use cttz instead of a loop for first_active_invocation().
|
||
- gallivm: Use first active invocation in some image/ssbo accesses.
|
||
- ci/lvp: Drop the subgroupbroadcast skips.
|
||
- llvmpipe: Enable LP_DEBUG on normal builds.
|
||
- gallivm: Enable GALLIVM_DEBUG (mostly) on non-DEBUG builds.
|
||
- gallivm: Fix the type of array nir_registers.
|
||
- gallivm: Fix codegen performance for constant-index register array stores.
|
||
- gallivm: Do the same codegen improvement for constant-index array loads.
|
||
- ci/swrast: Drop skips for tests whose perf had been fixed.
|
||
- ci/llvmpipe: Drop skip of InteractionFunctionCalls2.
|
||
- ci/freedreno: Don't forget to report flakes on a618, too.
|
||
- u_trace: Add an interface for checking trace enablement outside a context.
|
||
- zink: Add tracing of blit operations.
|
||
- ci: Disable systems in my farm that haven't recovered.
|
||
- ci/zink: Update TGL full-run xfails.
|
||
- ci/freedreno: Disable the a306_piglit_gl job.
|
||
- ci/freedreno: Update a530 manual-run xfails.
|
||
- ci/freedreno: Add an xfail for a618 VK full run.
|
||
- ci/freedreno: Update a3xx piglit_shader xfails.
|
||
- ci/nouveau: Disable the gm20b jobs entirely.
|
||
- ci/radv: Update navi21 llvm xfails.
|
||
- ci/crocus: Update HSW expectations.
|
||
- ci/freedreno: Update manual-run xfails for a530.
|
||
- Revert "freedreno/a5xx: Fix clip_mask"
|
||
- ci/radv: Add a skip for navi21-llvm for a test that consistently timeouts.
|
||
- ci/etnaviv: Drop stale xfails from gc7000.
|
||
- ci/etnaviv: Update deqp xfails for gc2000.
|
||
- egl/kopper: Add assert for no kopper in dri2_copy_region.
|
||
- egl: Add a note explaining the swapBuffers badness in dri2_x11_copy_buffers().
|
||
- egl/kopper: Use the kopper private interface for swapBuffers.
|
||
- egl/kopper: Pass ancillary invalidate flush flags down to gallium.
|
||
- ci: Add a manual full and 1/10th hasvk CTS runs.
|
||
- hasvk: Silence conformance warning in CI.
|
||
- hasvk: Fix SPIR-V warning about TF unsupported on gen7.
|
||
- anv: Fix gfx8/9 VB range > 32bits workaround detection.
|
||
- hasvk: Fix gfx8/9 VB range > 32bits workaround detection.
|
||
- glsl: Drop the (v.x + v.y + v.z + v.w) -> dot(v, 1.0) optimization.
|
||
- ci/etnaviv: Drop one more gc7000 xfail.
|
||
- ci/freedreno: Drop a530 piglit_gl coverage.
|
||
- ci/turnip: Drop the #8219 xfail.
|
||
- ci/zink+turnip: Disable flaky minetest trace.
|
||
- ci/hasvk: Add a synchronization flake.
|
||
- ci: Fix stage of etnaviv manual runs.
|
||
- ci/zink: Add a glx flake on anv
|
||
- ci/crocus: Add new tess xfails and a link to the regression bug report.
|
||
- ci/crocus: Mark unvanquished as flaky.
|
||
- anv: Skip the RT flush when doing depth-only rendering.
|
||
- anv: Skip BTI RT flush if we're doing an op that doesn't use render targets.
|
||
- glsl/opt_algebraic: Drop ~~x == x transformation.
|
||
- glsl/opt_algebraic: Drop log(exp(x)) -> x and exp(log(x)) -> x optimisations.
|
||
- glsl/opt_algebraic: Drop pow-recognizer.
|
||
- glsl/opt_algebraic: Drop abs(-x) -> abs(x) and abs(abs(x)) -> abs(x).
|
||
- glsl/opt_algebraic: Drop -(-x) -> x optimization.
|
||
- glsl/opt_algebraic: Drop f2i(trunc(x)) -> f2i(x) optimization.
|
||
- glsl/opt_algebraic: drop fsat(fadd(b2f(x), b2f(y))) -> b2f(ior(x, y)) opt.
|
||
- glsl/opt_algebraic: Drop shifts of 0 optimizations.
|
||
- glsl/opt_algebraic: Drop pow optimizations.
|
||
- glsl/opt_algebraic: Drop rcp optimizations.
|
||
- glsl/opt_algebraic: Drop and/or/xor optimizations.
|
||
- glsl/opt_algebraic: Drop fdiv(1,x) -> frcp(x) and fdiv(x,1) -> x optimizations.
|
||
- glsl/opt_algebraic: Drop add/sub with 0 optimizations.
|
||
- glsl/opt_algebraic: Drop x + -x -> 0 optimization.
|
||
- glsl/opt_algebraic: Drop csel(true/false, x, y) optimization.
|
||
- nir: Add optimization for fdot(x, 0) -> 0.
|
||
- glsl/opt_algebraic: Drop fdot 0-channel optimizations.
|
||
- glsl/opt_algebraic: Drop scalar all_eq/any_neq -> eq/neq opt.
|
||
- glsl/opt_algebraic: Drop the eq/neq add-removal optimization.
|
||
- glsl/opt_algebraic: Drop no-op pack/unpack optimization.
|
||
- glsl/opt_algebraic: Drop the flrp/ffma simplifiers.
|
||
- glsl/opt_algebraic: Drop some fmul simplifications.
|
||
- nir: Port a floor->truncate algebraic opt pattern from GLSL.
|
||
- glsl/opt_algebraic: Drop the ftrunc pattern recognizer.
|
||
- glsl/opt_algebraic: Drop the flrp recognizer.
|
||
- glsl: Remove unused as_rvalue_to_saturate().
|
||
- ci: Update traces expectations for gutting glsl opt_algebraic.
|
||
- panfrost/midgard: Fix handling of csel with a vector constant condition.
|
||
- panfrost/midgard: Drop redundant arg to emit_explicit_constant.
|
||
- glsl: Move lower_vector_insert to GLSL-to-NIR.
|
||
- nir/split_64bit_vec3_and_vec4: Handle 64-bit matrix types.
|
||
- gallivm: Return 0 for first active invocation when no invocations are active.
|
||
- gallivm: Use first_active_invocation for ubo/kernel memory loads.
|
||
- gallivm: Use first_active_invocation for scalar SSBO loads.
|
||
- gallivm: Add some notes about other invocation_0_must_be_active usages.
|
||
- ci: Add some xfail updates from VKCTS 1.3.5.0 for the manual jobs.
|
||
- ci/etnaviv: Drop the dEQP-GLES2.functional.uniform_api.random.94 xfail.
|
||
- anv+hasvk: Use driconf to disable 16-bit for zink.
|
||
- zink: Pass the cmdbuf to the end of the marker, too.
|
||
- Revert "ci: disable mesa-swrast runner jobs"
|
||
- ci: Re-enable some swrast testing using fd.o's shared runners for now.
|
||
- glsl/nir: Include early glsl-to-nir output in NIR_DEBUG=print.
|
||
- glsl_to_nir: Use a variable's constant_value if it wasn't const-propped out.
|
||
- glsl: Delete constant propagation pass.
|
||
- glsl: Delete constant folding pass.
|
||
- glsl: Delete constant-variables pass.
|
||
- ci: Update trace expectations for GLSL constant prop removal.
|
||
- ci/zink: Update TGL xfails/flakes based on the last nightly pipelines.
|
||
- ci/turnip: Extend a630 vk full timeout to 3 hours.
|
||
- ci/iris: Add skips for slow tests on APL.
|
||
- turnip: Don't push inline uniform buffer contents outside constlen.
|
||
- ci/turnip: Clear out stale xfails.
|
||
- ci/turnip: Disable dEQP-VK.image.queue_transfer.* for now.
|
||
- ci/turnip: Move some more of the 1.3.5 new xfails under links.
|
||
- glsl: Simplify vector constructors from scalars.
|
||
- glsl/lower_precision: Add a unit test that I thought we might fail at.
|
||
- glsl/lower_precision: Add a cut-down testcase for #8124
|
||
- glsl: Set the precisions of builtin function arguments and returns.
|
||
- glsl: Handle highp promotion of builtin function args in the builtins.
|
||
- glsl: Set the precision of function return value temporaries.
|
||
- glsl/lower_precision: Drop most special-casing of builtin arg precision.
|
||
- glsl: Fix the precision of atomic counter builtin function args.
|
||
- glsl/lower_precision: Add actual spec quotes for "check_parameters"
|
||
- nir/lower_mediump: Fix assertion about copy_deref lowering matching.
|
||
- ci/iris: Update more manual job xfails from the Wayland build change.
|
||
- ci/crocus: Update expectations from VK CTS 1.3.5.0.
|
||
- ci/hasvk: Update some xfails from the 8-sample fast clear disable.
|
||
- ci/etnaviv: Get the gc2000_piglit manual job mostly working.
|
||
- glsl/standalone: Pull program create/destroy out to a public function.
|
||
- glsl/standalone: Pull out a helper function for adding GLSL source shaders.
|
||
- glsl/standalone: Make all standalone contexts have NewProgram set.
|
||
- glsl: Write a new test for GLSL and NIR mediump lowering.
|
||
- ci/crocus: Fix 1.3.5.0 xfails.
|
||
- ci/etnaviv: Polish the gc2000 xfails a bit.
|
||
- ci/zink: Update the tgl manual run xfails.
|
||
- gallivm: Skip loads/stores that are definitely outside of compact vars.
|
||
- nir/lower_sysvals: Add support for un-lowered tess_level_inner/outer.
|
||
- nir_to_tgsi: Handle stores to compact outputs.
|
||
- glsl: Delete the lower_tess_level pass.
|
||
- glsl: Remove the TessLevel lowering special case from xfb.
|
||
- glsl: Drop dead prototype.
|
||
- ci/freedreno: Flake KHR-GL45.shader_image_load_store.basic-allTargets-store
|
||
- ci/broadcom: Skip another texelfetch case.
|
||
- perfetto: Add a .clang-format for the directory.
|
||
- intel/perfetto: Drop unused "pipelined" field.
|
||
- perfetto: Make a MesaRenderpassDataSource with common setup/start/stop.
|
||
- perfetto: Deduplicate clock sync packet emit from renderstage sources.
|
||
- perfetto: Move intel's cmdbuf/queue annotation code to the shared util.
|
||
- ci/zink: Drop validation exception for leaks at device destroy.
|
||
- ci/zink: Disable godot-tps-gles3 on a630.
|
||
- docs: Update Vulkan renderpass docs for !22191
|
||
- ci: Add missing dependency on doxygen sources for docs-generation jobs.
|
||
- docs: Claim less functionality for glsl_compiler.
|
||
- glsl: Move ForceGLSLAbsSqrt handling to glsl-to-nir.
|
||
- zink: Add mapping for nir_op_ldexp, but disable it for 64-bit's sake.
|
||
- glsl: Retire ldexp lowering in favor of the nir lowering flag.
|
||
- glsl/softfp64: GC the temp vars after we lower them to SSA.
|
||
- glsl/softfp64: Add fisfinite lowering.
|
||
- state_tracker: Lower frexp before lowering doubles.
|
||
- intel: Always call nir_lower_frexp.
|
||
- ir3: Move turnip's nir_lower_frexp to the shared compiler.
|
||
- nouveau: Add missing nir_opt_algebraic_late.
|
||
- nouveau: Enable frexp lowering in the backend.
|
||
- zink: Enable nir_lower_frexp.
|
||
- v3d: Lower frexp in the GL compiler like we do in Vulkan.
|
||
- agx: Enable nir_lower_frexp.
|
||
- panfrost/midgard: Enable nir_lower_frexp.
|
||
- nir_to_tgsi: Always lower frexp_exp/sig.
|
||
- glsl: Drop frontend lowering of 32-bit frexp.
|
||
- glsl: Drop PIPE_SHADER_CAP_DFRACEXP_DLDEXP_SUPPORTED.
|
||
- tgsi: Drop TGSI_OPCODE_DFRACEXP.
|
||
- ci/zink: Disable a630 portal-2-v2 due to kernel OOMs.
|
||
- etnaviv: Fix regression from if_uses change.
|
||
- blob: Don't valgrind assert for defined memory if we aren't writing.
|
||
- util/log: Fix log messages over 1024 characters.
|
||
- vulkan: Handle alignment failure in the pipeline cache.
|
||
- vulkan: Actually increment the count of objects in GetPipelineCacheData.
|
||
- ci/radeonsi: Mark glx-make-current as flaky.
|
||
|
||
EmperorPenguin18 (1):
|
||
|
||
- v3d: expose more drm formats with SAND128 modifier
|
||
|
||
Eric Engestrom (172):
|
||
|
||
- bin/ci: add gitlab_gql.py.cache to the .gitignore
|
||
- mesa/st: drop unused param
|
||
- ci/bare-metal: add more timestamps to help debugging issues
|
||
- ci: be explicit about the \`meson setup` subcommand
|
||
- docs: add release notes for 22.3.4
|
||
- docs/relnotes: add sha256sum for 22.3.4
|
||
- docs: update calendar for 22.3.4
|
||
- meson: turn android-libbacktrace into a feature option
|
||
- v3dv: mark dEQP-VK.api.command_buffers.record_many_draws_secondary_2 as flaky
|
||
- ci/android: move common config to common job
|
||
- ci/android: move virgl-specific gpu_mode to virgl-defined variables
|
||
- ci/android: move virgl-specific fails/flakes/skips lists to virgl-defined variables
|
||
- ci/android: move virgl-specific deqp suite to virgl-defined variables
|
||
- ci/android: move virgl-specific so lib name to virgl-defined variables
|
||
- ci/android: add missing line terminator at the end of the file
|
||
- docs: add release notes for 22.3.5
|
||
- docs: update calendar for 22.3.5
|
||
- panfrost: drop no-longer-needed libglsl
|
||
- gallium/u_screen.h: add missing stdint.h include
|
||
- util: avoid calling kcmp on Android
|
||
- etnaviv: use simple_mtx to avoid breaking windows in the next commit
|
||
- gallium: move etnaviv screen_lookup_or_create function to common code
|
||
- freedreno: replace custom code with u_pipe_screen_lookup_or_create()
|
||
- lima: replace custom code with u_pipe_screen_lookup_or_create()
|
||
- v3d: use u_pipe_screen_lookup_or_create() to keep track of and reuse screens
|
||
- vc4: use u_pipe_screen_lookup_or_create() to keep track of and reuse screens
|
||
- panfrost: use u_pipe_screen_lookup_or_create() to keep track of and reuse screens
|
||
- asahi: use u_pipe_screen_lookup_or_create() to keep track of and reuse screens
|
||
- u_pipe_screen_lookup_or_create: avoid re-querying the fd to have a consistent hash key
|
||
- broadcom/ci: mark test as flaky
|
||
- vk/util: keep track of extension requirements
|
||
- vk/runtime: keep track of supported instance extensions
|
||
- vk/runtime: turn vk.xml extension requirements into asserts
|
||
- meson: move float64_glsl_file one meson.build up
|
||
- meson: only build mapi when needed
|
||
- meson: only build the loader when needed
|
||
- meson: only build libglsl_util when needed
|
||
- meson: only build glsl when needed
|
||
- meson: drop \`TODO: opengl`, it's done
|
||
- ci: simplify adding & removing deqp patches
|
||
- ci: remove no-op sed
|
||
- ci: fix grouping of image tags
|
||
- ci: bump tags of deqp images
|
||
- docs: add 23.1 branchpoint & rc dates
|
||
- meson: make GLX require OpenGL
|
||
- meson/windows: only build libgl-gdi for desktop gl
|
||
- meson: allow building GLES without GL
|
||
- mesa: add _mesa_is_desktop_gl_compat() and _mesa_is_desktop_gl_core() helpers
|
||
- mesa: make use of the new _mesa_is_desktop_gl_compat() helper
|
||
- mesa: make use of the new _mesa_is_desktop_gl_core() helper
|
||
- mesa: make more use of the existing _mesa_is_gles* helpers
|
||
- mesa: add & use new _mesa_is_gles1() & _mesa_is_gles2() helpers
|
||
- mesa: make more use of the new _mesa_is_gles1() helper
|
||
- mesa: make more use of the new _mesa_is_gles2() helper
|
||
- mesa: optimize out _mesa_is_desktop_gl*() and _mesa_is_gles*() calls when not built
|
||
- ci: stop watching for changes in removed script
|
||
- meson: improve formatting of options file
|
||
- broadcom/ci: refactor a bit
|
||
- broadcom/ci: fold .vc4-rpi3-piglit:armhf into its only user
|
||
- broadcom/ci: use deqp-runner to run piglit tests
|
||
- docs/release-calendar: drop the last 22.2.x, it won't happen
|
||
- broadcom/ci: group x11 and wayland variant of the same test failing
|
||
- broadcom/ci: use weston's xwayland instead of starting X as well
|
||
- broadcom/ci: add x11- prefix to x11 EGL tests
|
||
- broadcom/ci: drop create_pixmap_surface from the fails; it passes now
|
||
- broadcom/ci: skip buffer_age.no_preserve and swap_buffers_with_damage on wayland
|
||
- broadcom/ci: add two known failures
|
||
- broadcom/ci: re-enable egl on wayland
|
||
- docs: include explicit \`setup` in instructions
|
||
- docs: add release notes for 22.3.6
|
||
- docs/relnotes: add sha256sum for 22.3.6
|
||
- docs: update calendar for 22.3.6
|
||
- v3d: update supertuxkart reference after 1c028a4d5b623e73bdf5
|
||
- docs: mention the meson summary
|
||
- docs: mention \`meson configure` and drop broken workaround script
|
||
- meson: reuse vulkan_wsi_list for defining vk_wsi_args
|
||
- meson: replace vk_wsi_args with dependencies to let meson take care of transitivity
|
||
- egl: include directly the useful vulkan header, instead of including everything
|
||
- glx: include directly the useful vulkan header, instead of including everything
|
||
- gbm: drop unnecessary vulkan dependency
|
||
- radv: split linker script for android since it requires different symbols
|
||
- glsl: align definition of _mesa_problem with the one in main/error.h
|
||
- glapi/meson: drop duplicate line in deps
|
||
- meson: allow checking for null pointers even if they're supposed to be non-null
|
||
- panfrost/ci: add EGL tests
|
||
- asahi/winsys: add .clang-format
|
||
- vk: move radv's linker symbols scripts for use in all drivers
|
||
- v3dv: add linker script to fix android symbols
|
||
- tu: add linker script to fix android symbols
|
||
- anv: add linker script to fix android symbols
|
||
- vn: add linker script to fix android symbols
|
||
- android/vk: drop unnecessary symbols
|
||
- vk: be stricter about symbols check between android and other platforms
|
||
- v3d/ci: add dEQP-GLES3.functional.texture.specification.teximage2d_pbo.*_cube flakes
|
||
- osmesa: add exported symbols check
|
||
- docs: add release notes for 22.3.7
|
||
- docs/relnotes: add sha256sum for 22.3.7
|
||
- docs: update calendar for 22.3.7
|
||
- v3dv/ci: add a test to the known failures
|
||
- meson: bump minimum version to 0.60
|
||
- meson: allow feature options to take true/false to mean enabled/disabled
|
||
- meson: inline gtest_test_protocol now that it's always 'gtest'
|
||
- v3dv: split out broadcom_shader_stage_to_gl() calls to improve readability
|
||
- ci: take valve farm offline
|
||
- ci: disable weston session timeout
|
||
- broadcom/ci: no need to skip the tests that swap buffers anymore
|
||
- ci/broadcom: move rare failure to the flakes
|
||
- ci: drop redundant .no_scheduled_pipelines-rules + .core-rules since the latter already includes it
|
||
- ci/rustfmt: simplify getting all the rust files
|
||
- ci/rustfmt: print which files are checked
|
||
- ci: group RESULT logic in a single place
|
||
- v3dv/ci: fix test name (\`,Fail` is not part of the test name)
|
||
- asahi: replace copies of .clang-format with symlinks
|
||
- asahi: fix a few typos
|
||
- v3d: fix \`dirty` bitset being too small to accept V3D_DIRTY_SSBO
|
||
- v3dv: use common GetPhysicalDeviceFeatures
|
||
- v3dv: reorder features as 1.0, 1.1, 1.2, 1.3
|
||
- v3dv: use vk_get_physical_device_features
|
||
- v3d/ci: add another depthstencil-default_fb-drawpixels-* to the flakes
|
||
- v3d/ci: group dEQP-GLES3.functional.texture.specification.teximage2d_pbo.* flakes and add another one
|
||
- ci: centralize detection of ccache in link-werror wrapper
|
||
- ci: add linker wrapper for clang
|
||
- ci: always use the -Werror wrapper
|
||
- ci: deduplicate compiler wrappers
|
||
- ci/docs: start documenting ci_run_n_monitor.py
|
||
- v3d: add link to issue investigating failure
|
||
- asahi: change create_renderonly signature to uniformize it
|
||
- etnaviv: change create_renderonly signature to uniformize it
|
||
- freedreno: change create_renderonly signature to uniformize it
|
||
- lima: change create_renderonly signature to uniformize it
|
||
- panfrost: change create_renderonly signature to uniformize it
|
||
- v3d: change create_renderonly signature to uniformize it
|
||
- vc4: change create_renderonly signature to uniformize it
|
||
- kmsro: uniformize renderonly creation
|
||
- kmsro: sort drivers alphabetically
|
||
- ci/broadcom: consolidate vc4-rpi3* jobs into a single vc4-rpi3-gl:armhf
|
||
- ci/broadcom: consolidate v3d-rpi4* jobs into a single v3d-rpi4-gl:armhf
|
||
- ci/broadcom: slightly increase coverage of vk tests
|
||
- vc4/ci: add arm64 failure to flakes as it works on armhf
|
||
- broadcom/ci: run gl jobs on arm64, just like vk
|
||
- vc4/ci: add another sync flake
|
||
- panfrost: assign the correct create_for_resource from the start
|
||
- Revert "broadcom/ci: run gl jobs on arm64, just like vk"
|
||
- v3dv/ci: mark known dEQP-VK.wsi.xlib.surface.query_formats failure
|
||
- ci/rustfmt: make sure to only check each file once
|
||
- v3d: disable GL_NV_conditional_render
|
||
- VERSION: bump for 23.1.0-rc1
|
||
- .pick_status.json: Update to 8ebc5cbe2b828f34b9bfb32c528d3514ead59798
|
||
- v3dv/ci: drop fixed failure from fails.txt
|
||
- .pick_status.json: Update to 0d7912d239dac5bf3c8b07f2a6ca467f760d6aa6
|
||
- .pick_status.json: Update to 543b6ca7c4b00c4bfff5668ba0a0643d565db201
|
||
- amd: fix buggy usage of unreachable()
|
||
- compiler: fix buggy usage of unreachable()
|
||
- pvr: fix buggy usage of unreachable()
|
||
- vk/util: fix buggy usage of unreachable()
|
||
- v3d: add flake spec@ext_framebuffer_blit@fbo-sys-sub-blit
|
||
- VERSION: bump for 23.1.0-rc2
|
||
- .pick_status.json: Update to 3017d01c9ded9c9fd097b600081b1bbe86e90fb8
|
||
- .pick_status.json: Update to a18a51a708a86f51e0a5ab031b379f65bc84fb49
|
||
- .pick_status.json: Update to c060b649c5a866f42e5df73f41c6e2809cf30e99
|
||
- ci: rework vulkan validation layer build script
|
||
- .pick_status.json: Update to 3f14fd8578549e34db2f564396f300819b2ff10f
|
||
- VERSION: bump for 23.1.0-rc3
|
||
- .pick_status.json: Update to 040aeb5a23e5cc8a71a352e55282d514dd2ab64f
|
||
- .pick_status.json: Update to 9f522ac0c65ceae11ad1a4e84ec9f32a9393a25c
|
||
- .pick_status.json: Update to efc94390f716b70ac1d5b09c6f949f938aeadcac
|
||
- VERSION: bump for 23.1.0-rc4
|
||
- .pick_status.json: Update to 6d84b34359dcbad477209adb9f9d0592c5a71bb9
|
||
- .pick_status.json: Update to cb4e4fc5de48886758a26ff19d322947b5abfcec
|
||
- dzn: fix pointer type mismatch
|
||
- .pick_status.json: Update to 57afa7c0b12d6d0c9013368853080dfea5b50d07
|
||
- .pick_status.json: Update to 31e6d15801a9904089aa2913c8eb5a31b79c7dfc
|
||
|
||
Erico Nunes (5):
|
||
|
||
- lima/ci: Add more piglit unsupported tests to skip
|
||
- Revert "CI: Lima farm is offline"
|
||
- lima: don't use resource_from_handle while creating scanout
|
||
- lima/ci: restore swap buffers egl tests
|
||
- Revert "ci: disable lima farm, currently out-of-space, needs to be fixed"
|
||
|
||
Erik Faye-Lund (54):
|
||
|
||
- zink: whitespace fixup
|
||
- zink: fix depth-clip disable cap
|
||
- zink: remove depth_clip_control_missing workaround
|
||
- radeonsi: respect smoothing_enabled
|
||
- meson: remove dupliace add_devenv call
|
||
- meson: remove deprecated osmesa-bits option
|
||
- meson: remove deprecated dri-drivers option
|
||
- meson: avoid using deprecated build_root() method
|
||
- meson: use files() instead of joining paths
|
||
- freedreno/meson: simplify script-path logic
|
||
- meson: do not reconstruct ICD paths
|
||
- anv, hasvk: remove stale TODO-files
|
||
- zink: correct companies in requirements
|
||
- zink: remove incorrect trailing comma
|
||
- meson: remove unused USE_FOO_ASM defines
|
||
- vulkan: prefer vulkan_core.h over vulkan.h
|
||
- meson: don't pass vk wsi args where they don't belong
|
||
- Revert "meson: Fix Asahi build on macOS"
|
||
- zink: prefer vulkan_core.h over vulkan.h
|
||
- zink: get rid of needless dependency
|
||
- ci: correct typo in name of linkcheck job
|
||
- docs: update link to intel optimization reference manual
|
||
- nir: add a print_internal debug-flag
|
||
- docs: implement new vk-feat role
|
||
- docs/zink: use vk-feat role for features
|
||
- docs/zink: remove some trailing spaces
|
||
- docs/zink: fixup wording of the GL 4.6 requirements
|
||
- meson: correct typo in comment
|
||
- ci: move docs-stuff out of root .gitlab-ci.yml
|
||
- docs: fixup broken envvar-role syntax
|
||
- docs: escape a few more strings
|
||
- docs: fixup broken indentation
|
||
- docs/zink: mention vk1.2 mirror-clamp feature option
|
||
- docs/zink: clean up requirements-language
|
||
- docs: move developers article to main website
|
||
- docs: remove old thanks-article
|
||
- docs: prefer http-links over ftp
|
||
- docs/freedreno: fix turnip-heading level
|
||
- docs: drop reference to modindex
|
||
- docs: move old relnotes to _extra directory
|
||
- docs: use version-number as toctree-title for relnotes
|
||
- zink: emit terminate for spir-v 1.6
|
||
- zink: use demote from spir-v 1.6 when possible
|
||
- zink: use spir-v 1.6 local-size when needed
|
||
- zink: enable spir-v 1.6 for vulkan 1.3
|
||
- docs: format code-block as ini
|
||
- docs: format code-block as toml
|
||
- docs: make code-block indents consistent
|
||
- ci: move virgl-rules after intel-rules
|
||
- virgl/ci: clean up manual rules for virgl
|
||
- ci: remove unused rules
|
||
- zink: do not use sampled-image for buffers
|
||
- nir: fix constant-folding of 64-bit fpow
|
||
- llvmpipe: fixup refactor copypasta
|
||
|
||
Faith Ekstrand (99):
|
||
|
||
- nir: Add more opcodes to nir_tex_instr_is_query()
|
||
- nir/builder: Add some texture helpers
|
||
- radv: Use the new NIR builder tex helpers for meta
|
||
- anv: Refactor Android externalFormat handling in CreateYcbcrConversion
|
||
- anv/android: Use VkFormat for externalFormat
|
||
- util/format: YUYV and UYVY have 4 8-bit channels
|
||
- vulkan/formats: Add YCbCr format information
|
||
- vulkan: Add a common vk_ycbcr_conversion struct
|
||
- anv: Use the common vk_ycbcr_conversion object
|
||
- anv: Use the YCbCr format info from common code
|
||
- nir: Add copyright and include guards to nir_vulkan.h
|
||
- anv,nir: Move the ANV YCbCr lowering pass to common code
|
||
- gallium,util: Pull u_indices and u_primconvert back into gallium
|
||
- mailmap: Remap e-mail addresses for Faith Ekstrand
|
||
- vtn: Set alignment on initial UBO/SSBO casts
|
||
- anv: Let spirv_to_nir() set UBO/SSBO base cast alignments
|
||
- hasvk: Let spirv_to_nir() set UBO/SSBO base cast alignments
|
||
- intel/compiler: Document wm_prog_key::persample_interp
|
||
- intel/nir: Lower barycentrics to per-sample in a dedicated pass
|
||
- nir: Remove nir_lower_io_force_sample_interpolation
|
||
- intel/compiler: Use SHADER_OPCODE_SEND for PI messages
|
||
- intel/fs: Return early in a couple builtin setup helpers
|
||
- intel/compiler: Convert brw_wm_aa_enable to brw_sometimes
|
||
- intel/fs: Make per-sample and coarse dispatch tri-state
|
||
- intel/compiler: Convert wm_prog_key::persample_interp to a tri-state
|
||
- intel/compiler: Convert wm_prog_key::multisample_fbo to a tri-state
|
||
- intel/fs/validate: Assert SEND [extended] descriptors are uniform
|
||
- intel/fs: Break out yet another FB write helper
|
||
- intel/fs: Rework dynamic coarse handling
|
||
- nir/deref: Preserve alignments in opt_remove_cast_cast()
|
||
- nir/from_ssa: Use more helpers in resolve_parallel_copies
|
||
- nir/from_ssa: Only re-locate values that are destinations
|
||
- nir/from_ssa: Move the loop bounds check in resolve_parallel_copy
|
||
- nir: Add a load/store bit size lowering pass
|
||
- intel/nir: Use nir_lower_mem_access_bit_sizes()
|
||
- Revert "vk/runtime: turn vk.xml extension requirements into asserts"
|
||
- Revert "vk/util: keep track of extension requirements"
|
||
- vulkan: Remove unused fields from Extension and ApiVersion
|
||
- vulkan: Improve extension parsing
|
||
- vulkan: Parse the platform in Extensions.from_xml()
|
||
- vulkan: Add a get_all_required() helper
|
||
- vulkan: Properly filter entrypoints
|
||
- vulkan: Properly filter by api in enum_to_str
|
||
- Vulkan: Properly filter structs in vk_cmd_queue_gen
|
||
- vulkan: Filter out provisional extensions
|
||
- vulkan: Move the features generator to vulkan/util
|
||
- vulkan: Properly filter structs in vk_physical_device_features
|
||
- vulkan/layers: Use PUBLIC instead of VK_LAYER_EXPORT
|
||
- vulkan/device-select-layer: Include vulkan.h
|
||
- vulkan: Update the XML and headers to 1.3.241
|
||
- nir/lower_io: Handle buffer_array_length for more address modes
|
||
- anv: Drop our manual SSBO size handling
|
||
- hasvk: Drop our manual SSBO size handling
|
||
- panvk: Drop our manual SSBO size handling
|
||
- turnip: Set spirv_options::use_deref_buffer_array_length
|
||
- lavapipe: Set spirv_options::use_deref_buffer_array_length
|
||
- v3dv: Set spirv_options::use_deref_buffer_array_length
|
||
- spirv: Always emit deref_buffer_array_length intrinsics
|
||
- nir: Check against combined alignment in nir_lower_mem_access_bit_sizes
|
||
- nir: Add mode filtering to lower_mem_access_bit_sizes
|
||
- nir: Add UBO support to nir_lower_mem_access_bit_sizes
|
||
- nir: Add a combined alignment helper
|
||
- nir: Rename align to whole_align in lower_mem_load
|
||
- nir: Rename nir_mem_access_size_align::align_mul to align
|
||
- nir: Make chunk_align_offset const in lower_mem_load()
|
||
- nir: Handle wider unaligned loads in lower_mem_access_bit_size
|
||
- intel/nir: Limit unaligned loads to vec4
|
||
- vulkan/runtime: Rename and document storage image Z range
|
||
- intel/blorp: Set array_len for 3D images properly
|
||
- isl: Set Depth to array len for 3D storage images
|
||
- intel: Use nir_lower_tex_options::lower_index_to_offset
|
||
- vulkan: Update XML and headers to 1.3.244
|
||
- vulkan: Provide wrappers for VK_EXT_map_memory2 functions
|
||
- anv: Limit memory maps to the client-allocated size
|
||
- anv: Implement VK_KHR_map_memory2
|
||
- intel/isl: Support Yf/Ys/Tile-64 in isl_surf_get_image_offset_sa
|
||
- intel/blorp: Drop the TODO file
|
||
- docs: Fix Faith's name in relnotes
|
||
- nir: Drop a bunch of Authors tags
|
||
- spirv: Drop a bunch of Authors tags
|
||
- intel: Drop some author comments and update Faith's name
|
||
- util,mesa,panfrost: Drop some author tags
|
||
- vulkan: vk_android.c should be copyright Intel
|
||
- util: Update some copyright tags
|
||
- CODEOWNERS: s/jekstrand/gfxstrand
|
||
- vulkan,anv,hasvk,radv: Add a common vk_image_usage_to_ahb_usage helper
|
||
- vulkan/android: Fix hardware buffer usage flags
|
||
- vulkan: Add an ahardware_buffer_format field to vk_image
|
||
- anv,hasvk: Set vk_image.ahardware_buffer_format
|
||
- radv: Set vk_image.ahardware_buffer_format
|
||
- vulkan,anv,hasvk,radv: Unify Android hardware buffer creation
|
||
- vulkan: Add a vk_device_memory base struct
|
||
- anv: Use the new vk_device_memory base struct
|
||
- vulkan: Record pipeline flags in the render pass
|
||
- vulkan: Plumb rendering flags through vk_graphics_pipeline_state
|
||
- anv/pipeline: Use feedback loop flags for self-dependencies
|
||
- hasvk/pipeline: Use feedback loop flags for self-dependencies
|
||
- vulkan: Drop vk_render_pass_state::\*self_dependenc*
|
||
- vulkan: Drop VkRenderingSelfDependencyInfoMESA
|
||
|
||
Felix DeGrood (10):
|
||
|
||
- intel/perf: Hide extended metrics by default
|
||
- anv: cs_stall during compute state flush on < gen12.5
|
||
- anv: only emit CFE_STATE when scratch space increases
|
||
- anv: set CFE_STATE.OverDispatchControl to default
|
||
- iris: report draw count for perfetto
|
||
- anv/blorp: support surf generation for addresses
|
||
- anv/blorp: implement anv_cmd_buffer_fill_area
|
||
- anv/blorp: add flush reasons to RT flushes
|
||
- anv: reset query pools using blorp
|
||
- anv: disable reset query pools using blorp opt on MTL
|
||
|
||
Filip Gawin (2):
|
||
|
||
- crocus: don't quantize the clear value
|
||
- nine: add fallback for D3DFMT_D16 in d3d9_to_pipe_format_checked
|
||
|
||
Francisco Jerez (11):
|
||
|
||
- intel/fs/gfx12: Ensure that prior reads have executed before barrier with acquire semantics.
|
||
- intel/disasm/gfx12+: Use helper instead of hardcoded bit access for 64-bit immediates.
|
||
- intel/disasm/gfx12+: Fix print out of non-existing condmod field with 64-bit immediate.
|
||
- intel/eu/gfx12+: Implement decoding of 64-bit immediates.
|
||
- intel/fs/gfx12+: Drop redundant handling of SHADER_OPCODE_BROADCAST in exec pipe inference.
|
||
- intel/fs: Fix src and dst types of LOAD_PAYLOAD ACP entries during copy propagation.
|
||
- intel/eu/gfx8-9: Fix execution with all channels disabled due to HW bug #220160235.
|
||
- intel/rt: Fix L3 bank performance bottlenecks due to SW stack stride alignment.
|
||
- intel/fs: Track force_writemask_all behavior of copy propagation ACP entries.
|
||
- intel/fs: Fix copy propagation dataflow analysis in presence of force_writemask_all ACP overwrites.
|
||
- intel/fs: Fix register coalesce in presence of force_writemask_all copy source writes.
|
||
|
||
Frank Binns (7):
|
||
|
||
- pvr: small cleanups
|
||
- pvr: remove start/stop transfer flags
|
||
- pvr: stop restricting the compiler to the Sascha Willems triangle demo
|
||
- pvr: remove duplicate define
|
||
- pvr: initialise size for placeholder "zeroed" shaders
|
||
- pvr: replace nop binary shader with run-time compiled shader
|
||
- pvr: fix clang-format issue
|
||
|
||
Friedrich Vock (26):
|
||
|
||
- radv/rt: Divide by the correct workgroup size
|
||
- radv/bvh: Prevent NANs when computing node cost
|
||
- radv/rmv: Also check the other pid field
|
||
- radv/rmv: Avoid more CPU unmap deadlocks
|
||
- radv/rmv: Log bo destruction before freeing it
|
||
- radv/rmv: Correct timestamp shifting
|
||
- vulkan/rmv: Use the timestamp divisor instead of a hardcoded value
|
||
- vulkan/rmv: Remove delta parameter from dump helpers
|
||
- mesa: Report GL_SHADER_BINARY_FORMAT_SPIR_V as supported
|
||
- docs: Fix formatting for RMV tracing docs
|
||
- radv: Extend hit attribute lowering for LDS
|
||
- radv: Use LDS for closest-hit hit attributes
|
||
- radv: Emit RT shader VA user SGPR
|
||
- radv/rt: Add shader config combination/postprocessing utils
|
||
- radv: Add RT shader stage names for executable properties
|
||
- aco: Swap operands for v_and_b32 in RT prolog
|
||
- radv/rt: Also adjust the SGPR count in postprocess_rt_config
|
||
- aco: Un-swap addressable VGPRs/SGPRs in RT prolog
|
||
- radv: Work around use-after-free compiler errors
|
||
- radv: Add RT stages to radv_mesa_to_rgp_shader_stages
|
||
- radv/rmv: Fix creating RT pipelines
|
||
- radv/rmv: Fix import memory
|
||
- radv/rt: Plug some memory leaks during shader creation
|
||
- radv: Don't leak the RT prolog binary
|
||
- radv: Always call si_emit_cache_flush before writing timestamps
|
||
- radv: Add driconf to always drain waves before writing timestamps
|
||
|
||
GH Cao (1):
|
||
|
||
- gallium: Add MCJIT target triplet for Windows ARM64
|
||
|
||
Ganesh Belgur Ramachandra (1):
|
||
|
||
- ac/nir: fix CDNA image lowering for array textures
|
||
|
||
Georg Lehmann (81):
|
||
|
||
- Revert "aco: Combine v_cvt_u32_f32 with insert to v_cvt_pk_u8_f32."
|
||
- aco: use s_bfm_64 for constant copies
|
||
- aco: use s_pack_ll_b32_b16 for constant copies
|
||
- aco: Improve wave64 cycle estimates.
|
||
- aco: fix imod/omod for gfx11 VOP3 opcodes
|
||
- aco: add mov/cndmask opcodes to does_fp_op_flush_denorms
|
||
- aco: don't allow output modifiers for v_cvt_pkrtz_f16_f32
|
||
- aco: allow output modifiers for ldexp_f16
|
||
- aco: don't list imod/omod support v_fmaak_f32/v_fmamk_f32
|
||
- aco: support omod/imod for v_fmac_f16
|
||
- aco: remove stale TODOs about v_interp opsel
|
||
- aco: new 16bit VOP3 opcodes can use opsel
|
||
- aco: Don't use vcmpx with DPP.
|
||
- aco: combine a ^ ~b and ~(a ^ b) to v_xnor_b32
|
||
- amd,nir: remove byte_permute_amd intrinsic
|
||
- nir: change 16bit image dest folding option to per type
|
||
- amd: don't use d16 for integer loads
|
||
- amd: d16 uses rtz conversion for 32bit float
|
||
- aco: use v_permlane(x)16_b32 for masked swizzle
|
||
- aco/gfx11: use dpp_row_xmask and dpp_row_share
|
||
- aco: use and swizzle mask in dpp quad perm
|
||
- aco/optimizer_postRA: assume all registers are untrackable in loop headers
|
||
- nir/opt_algebraic: add patterns for iand/ior of feq/fneu with 0
|
||
- aco: mark mad definition as precise if the mul/add were precise
|
||
- aco: use v_fma_mix_f32 for v_fma_f32 with 2 fp16 representable, different literals
|
||
- nir/lower_mediump: don't use fp16 for constants if the result is denormal
|
||
- aco: treat VINTERP_INREG as VALU
|
||
- aco/ir: rework IR to have one common valu instruction struct
|
||
- aco/ra: set opsel_hi to zero when converting to VOP2
|
||
- aco: validate VALU modifiers
|
||
- aco/print_ir: simplify using VALU instruction
|
||
- aco/optimizer: simplify using VALU instruction
|
||
- aco: remove VOP[123C]P? structs
|
||
- aco: add bitfield array helper classes
|
||
- aco: use bitfield array helpers for valu modifiers
|
||
- aco/assembler/gfx11: simplify 16bit VOP12C promotion to VOP3
|
||
- aco/optimizer: don't reallocate instruction when converting to VOP3
|
||
- aco: don't reallocate fma{mk,ak,_mix} instruction
|
||
- aco: copy abs/neg with assignment
|
||
- aco: use integer access for neg_lo/neg_hi
|
||
- aco: use array indexing for opsel/opsel_lo/opsel_hi
|
||
- aco: access neg/abs as int in usesModifiers
|
||
- aco: use bitfield_array for temporary neg/abs/opsel
|
||
- nir: optimize i2f(f2i(fsign))
|
||
- aco: remove duplicates from .clang-format
|
||
- amd: remove duplicate from .clang-format
|
||
- aco: don't check usesModifiers for pseudo instructions
|
||
- aco: fix p_interp_gfx11 comment
|
||
- aco: make .clang-format usable with tests
|
||
- aco/ir: fix copy paste bug in convert_to_SDWA
|
||
- aco/util: override default assignment operator for bitfield helpers
|
||
- aco: clean up to_mad_mix
|
||
- aco/ra: don't reallocate VOP3 instruction for non-vcc lane mask
|
||
- aco/vn: hash opsel for VOP12C
|
||
- aco/assembler: support VOP12C opsel
|
||
- aco: validate VOP12C opsel
|
||
- aco/to_hw_instr: use VOP1 opsel for v_mov_b16
|
||
- aco/ra: prepare for VOP12C opsel
|
||
- aco/optimizer: preserve opsel when fusing fma
|
||
- aco: handle opsel in combine_comparison_ordering
|
||
- aco: handle opsel in combine_ordering_test
|
||
- aco: handle opsel in combine_constant_comparison_ordering
|
||
- aco: update match_op3_for_vop3 for VOP12C opsel
|
||
- aco: support v_cvt_f32_f16 with opsel in combine_mad_mix
|
||
- aco: support neg(mul)/abs(mul) optimization in more cases
|
||
- aco: return true in usesModifiers for VOP12C with opsel
|
||
- aco: swap opsel when swapping VOP2/C operands
|
||
- aco/ir: copy opsel when converting to DPP
|
||
- aco: don't label mul with opsel as abs/neg
|
||
- aco/gfx11: allow opsel for VOP12C
|
||
- aco/optimizer: use opsel for VOP12C
|
||
- aco: keep label_mul/usedef/minmax in apply_extract
|
||
- aco/optimizer: remove to_SDWA
|
||
- aco: add tests for fma with opsel
|
||
- aco: add tests for dpp with opsel
|
||
- aco: add tests for swap operand with opsel
|
||
- aco: add tests for cmp ordering with opsel
|
||
- aco: add test for min/max combining with opsel
|
||
- aco/tests: run optimize.mad_mix.input_conv.modifiers on gfx11
|
||
- aco: add tests for neg(mul) with opsel
|
||
- aco/tests: add missing dependency on generated header
|
||
|
||
Gert Wollny (49):
|
||
|
||
- glsl/nir: only set uses_sample_shading when the output is a fbfetch
|
||
- nir: Add possibility to store image var offset in range_base
|
||
- nir: Add range_base to atomic_counter and an option to use it
|
||
- ntt: handle the image intrinsic range_base when translating to TGSI
|
||
- ntt: Make use of the range_base offset when translating atomics in NTT
|
||
- virgl: lower image variable offsets into the intrinsic range_base value
|
||
- virgl: Request setting the atomic offset in the range_base
|
||
- virgl: drop the separable flag for cases that can't be handled
|
||
- r600/sfn: Fix readport check
|
||
- r600/sfn: Do a bit of cleanup with the secondary read port validation
|
||
- r600/sfn: Fix opcode and result dest slot mask for variable size dot
|
||
- r600/sfn: Fix splitting of multislot alu ops
|
||
- virgl: remove unused virgl_encoder_inline_write
|
||
- r600/sfn: Use range_base for atomics and images
|
||
- r600/sfn: Work around dependency issue when splitting op to group
|
||
- r600/sfn: drop useless instr use count
|
||
- r600/sfn: Fix a typo
|
||
- r600/sfn: Silence warnings about unused parameters
|
||
- r600/sfn: Don't copy propagate indirect loads to more than one dest
|
||
- r600/sfn: Stop try scheduling in t-slot with empty related v-slot
|
||
- r600/sfn: rename texture coordinate offset for clarity
|
||
- r600/sfn: address use in group only if instr can be added
|
||
- r600/sfn: Forward setting the block ID and index
|
||
- r600: Don't start new CF for every fetch through tex clause
|
||
- r600/sfn: Fix handling of fetch through texture clause
|
||
- r600/sfn: Fix alu trans op flag setup
|
||
- r600/sfn: Fix Cayman trans from string and add test for copy prop
|
||
- vulkan/wsi: Take Xwayland into account for x11_min_image_count
|
||
- zink/kopper: Add extra swapchain images for Venus
|
||
- r600/sfn: be more conservative with channel use in multi-slot ops
|
||
- r600/sfn: Fix readport cylce map
|
||
- r600/sfn: Fix minimum required registers
|
||
- r600/sfn: Add AluGroup method to update readport validation from scratch
|
||
- r600/sfn: Split AluInstr replace_source into test and actual replace
|
||
- r600/sfn: Add method to AluGroup to replace sources
|
||
- r600/sfn: Add print method to AluReadportValidation
|
||
- r600/sfn: redirect copy propagation to alu parent group
|
||
- r600/sfn/tests: Add a test for the copy prop into a group
|
||
- r600/sfn: Fix atomic lowering
|
||
- virgl: Enable AMD_vertex_shader_(layer|viewport_index) when host supports it
|
||
- virgl: Don't try to do re-alloc or readback by transfer for blob resources
|
||
- ntt: add option to lower SSBO bindings to buffer index
|
||
- virgl: Lower binding start into buffer indices
|
||
- r600/sfn: fix container allocators
|
||
- r600/sfn: Lower tess levels to vectors in TCS
|
||
- r600/sfn: make sure f2u32 is lowered late and correctly for 64 bit floats
|
||
- r600/sfn: assign window_space_position in shader state
|
||
- r600/sfn: Ass support for image_samples
|
||
- r600/sfn: fix cube to array lowering for LOD
|
||
|
||
Giancarlo Devich (25):
|
||
|
||
- d3d12: Use varying comparison function for TESS stage key compare
|
||
- d3d12: Add unions to encompass shader key stage vars, use in hashing
|
||
- nir: Check sampler_binding is valid when lowering tex shadow
|
||
- d3d12: Don't clear d3d12_shader_key
|
||
- d3d12: Move d3d12_context_state_table_entry to d3d12_resource_state.h
|
||
- d3d12: Assign up to 16 simultaneously active contexts unique IDs
|
||
- d3d12: Track up to 16 active context resource states locally in d3d12_bo
|
||
- d3d12: Don't recompute has_flat_varyings or missing_dual_src_outputs
|
||
- d3d12: Track max varying slot, set and compare less bytes
|
||
- d3d12: Don't unnecessarily zero out gs/tcs keys
|
||
- d3d12: Don't memcmp gs/tcs keys
|
||
- d3d12: Create varying structures as necessary, reference them
|
||
- d3d12: Don't loop in update_draw_indirect_with_sysvals
|
||
- d3d12: Compare shader keys with a switch, instead of cascading if's
|
||
- d3d12: Compare shader keys with union-encompassing fields all at once
|
||
- d3d12: Compare shader key common parts with memcmp, instead of if's
|
||
- d3d12: Cache varying info to reduce compare/copy cost
|
||
- d3d12: Use memcmp for full tcs/gs variant keys
|
||
- d3d12: Track up to 16 contexts worth of pending barriers locally in bos
|
||
- d3d12: Don't unnecessarily recompute manual_depth_range
|
||
- d3d12: Use context-level sampler_state array for filling shader keys
|
||
- d3d12: Use short circuit in shader key compare; update key hash
|
||
- d3d12: Reduce gs variant key init cost; unnecessary validate gs calls
|
||
- d3d12: Unroll shader variant selection loop
|
||
- d3d12: Track up to 16 contexts worth of batch references locally in bos
|
||
|
||
Guilherme Gallo (24):
|
||
|
||
- radeonsi/ci: Update stoney test expectations
|
||
- radeonsi/ci: Skip slow traces on raven
|
||
- Revert "ci: disable Collabora's LAVA lab for maintance"
|
||
- ci/lava: Move LAVA dependencies to pip
|
||
- ci/lava: Add LavaFarm class to find LAVA farm from runner tag
|
||
- ci/lava: Fix LAVA logs issues for Collabora jobs
|
||
- ci: Upload debian-release artifact to S3
|
||
- ci: Create debian-arm64-release job
|
||
- ci: Use release builds in perf jobs
|
||
- ci: Use workflow to make CI aware of performance jobs
|
||
- ci: Reuse MESA_CI_PERFORMANCE_ENABLED in performance-rules
|
||
- ci: Handle carriage return characters in LAVA logs
|
||
- ci: Fix release build use for performance jobs
|
||
- ci/baremetal: Wrap artifact download curl with xtrace
|
||
- ci: Improve piglit-traces "no-perf" filter
|
||
- ci: Fix freedreno-rules-performance
|
||
- ci: Add piglit traces hidden jobs
|
||
- ci/freedreno: create a618-traces and perf jobs
|
||
- ci/zink: Add zink-a618 trace jobs
|
||
- ci/zink: Add zink-turnip-manual-rules
|
||
- ci/zink: Add zink-tu-a618-traces-performance job
|
||
- ci/zink: Fix zink-tu-a618-traces perf job rules
|
||
- ci/zink: Reduce zink-tu-a618-traces parallelism
|
||
- Revert "ci: disable Collabora's LAVA lab for maintance"
|
||
|
||
Hampus Linander (4):
|
||
|
||
- nir: Add extr_agx opcode
|
||
- agx: Add extr instruction to AGX backend
|
||
- agx: Use AGX extr for tex lowering
|
||
- agx: Optimize lower_resinfo for cube maps
|
||
|
||
Hans-Kristian Arntzen (13):
|
||
|
||
- radv: Fix invalid 64-bit shift.
|
||
- radv: Fix missing VK_ACCESS_2_SHADER_SAMPLED_READ_BIT.
|
||
- radv: Implement VK_ACCESS_2_DESCRIPTOR_BUFFER_READ_BIT_EXT.
|
||
- wsi/common: Add common implementation of vkReleaseSwapchainImagesEXT.
|
||
- wsi/x11: Implement EXT_swapchain_maintenance1.
|
||
- wsi/common: Implement swapchain present fence.
|
||
- wsi/common: Add comment about DEFERRED_ALLOCATION_BIT_EXT.
|
||
- wsi/common: Add function to modify present mode.
|
||
- wsi/wayland: Implement EXT_swapchain_maintenance1.
|
||
- wsi/display: Implement EXT_swapchain_maintenance1.
|
||
- wsi/win32: Implement VK_EXT_swapchain_maintenance1.
|
||
- radv: Expose VK_EXT_swapchain_maintenance1.
|
||
- wsi/x11: Fix present ID signal when IDLE comes before COMPLETE.
|
||
|
||
Harri Nieminen (6):
|
||
|
||
- docs/specs: Fix typos
|
||
- docs/gallium: Fix typos
|
||
- docs/freedreno: Fix typos
|
||
- docs/panfrost: Fix typo
|
||
- docs/svga3d: Fix typo
|
||
- bin: Fix typos
|
||
|
||
Helen Koike (12):
|
||
|
||
- ci/debian-android: move pkgconfig paths to the cross file
|
||
- ci: move patches to patches directory
|
||
- android: allow system = 'android' on cross file
|
||
- ci/android: move sdk version and ndk to a job variable
|
||
- ci: compile deqp for android
|
||
- ci: compile deqp-runner for android
|
||
- ci: debian-android compile virgl
|
||
- ci: export artifacts from debian-android
|
||
- ci/android: add android to the ci
|
||
- android/ci: fix removal of inexistent file
|
||
- android/ci: Fix call to adb
|
||
- android/ci: raise error on script when not related to the tests
|
||
|
||
Hyunjun Ko (1):
|
||
|
||
- vulkan/runtime: match the spec when taking pipeline subsets.
|
||
|
||
Iago Toral Quiroga (20):
|
||
|
||
- broadcom/compiler: produce better code for f2f16 with RTZ rounding
|
||
- v3dv: add paths to handle partial copies of linear images
|
||
- v3dv: drop unused field from v3dv_cmd_buffer
|
||
- v3dv: increase BO allocation size when growing CLs
|
||
- v3dv: ensure we allocate at least the requested space for a CL
|
||
- v3dv: add a cl_advance_and_end helper
|
||
- v3dv: ensure at least V3D_CL_MAX_INSTR_SIZE bytes in last CL instruction
|
||
- v3dv: ensure we apply binning syncs to secondary command buffers
|
||
- v3dv: fix stencil view aspect selection of depth/stencil image
|
||
- v3d: support r{g,gba}16f formats for vertex buffers
|
||
- broadcom/compiler: track pending ldtmu count with each TMU lookup
|
||
- v3dv: pause occlusion queries during vkCmdClearAttachments
|
||
- v3dv: fix format swizzle for buffer views
|
||
- v3dv: drop unused parameter
|
||
- v3dv: always acquire display device before checking if we can present
|
||
- vulkan/wsi/display: set pDisplay to NULL on error
|
||
- v3d,v3dv: stop trying to force 16-bit TMU output for shadow comparisons
|
||
- broadcom/compiler: fix v3d_qpu_uses_sfu
|
||
- broadcom/compiler: add a v3d_qpu_instr_is_legacy_sfu helper
|
||
- broadcom/compiler: fix incorrect check for SFU op
|
||
|
||
Ian Romanick (60):
|
||
|
||
- ntt: Add support for fcsel_gt and fcsel_ge opcodes
|
||
- nir/lower_int_to_float: Add support for i32csel opcodes
|
||
- r300: Enable generation of fcsel_gt and fcsel_ge opcodes
|
||
- i915: Enable generation of fcsel_gt and fcsel_ge opcodes
|
||
- gallium/draw: Enable aapoint NIR helpers to generate bool1, bool32, or float32 Booleans
|
||
- gallium/draw: Enable polygon stipple NIR helpers to generate bool1 or bool32 Booleans
|
||
- nir/builder: Eliminate nir_f2b helper (and use of nir_f2b32 helper)
|
||
- nir/builder: Handle f2b conversions specially in nir_type_convert
|
||
- nir: Eliminate nir_op_f2b
|
||
- lavapipe: Fix bad array index scale factor in lvp_inline_uniforms pass
|
||
- lavapipe: Only check NULL pointers in one place in src_only_uses_uniforms
|
||
- nir/inline_uniforms: Change num_offsets type to uint8_t
|
||
- nir/inline_uniforms: Pass max_num_bo and max_offset around as parameters
|
||
- nir/inline_uniforms: Allow possibility of more than one UBO
|
||
- nir/inline_uniforms: Allow possibility of uni_offsets and num_offsets being NULL
|
||
- nir/inline_uniforms: Make src_only_uses_uniforms public, change name
|
||
- nir/inline_uniforms: Make add_inlinable_uniforms public
|
||
- nir/inline_uniforms: Add inot condition support
|
||
- nir/tests: Don't unconditionally log shaders from this one CF test
|
||
- nir/tests: Refactor creation of loops for loop_analyze test cases
|
||
- nir/tests: Add tests for "inverted" loops
|
||
- nir/tests: Add tests for nir_loop_info::induction_vars tracking
|
||
- nir/loop_analyze: Track induction variables with uniform increments
|
||
- nir/loop_analyze: Use nir_loop_variable::update_src instead of nir_basic_induction_var::alu
|
||
- nir/loop_analyze: Use nir_loop_variable::init_src instead of nir_basic_induction_var::def_outside_loop
|
||
- nir/loop_analyze: Eliminate nir_basic_induction_var
|
||
- nir/loop_analyze: Track induction variables with uniform initializer
|
||
- nir/loop_analyze: Simplify some logic in compute_induction_information
|
||
- nir: ifind_msb_rev can only have int32 sources
|
||
- intel/compiler: Lower find_lsb in NIR
|
||
- nir: intel/compiler: Move ifind_msb lowering to NIR
|
||
- intel/compiler: Tighter src and dest size bounds checking for some opcodes
|
||
- nir/algebraic: Only lower ufind_msb with 32-bit sources
|
||
- nir: intel/compiler: Move ufind_msb lowering to NIR
|
||
- nir/builder: Do not generate 8- or 16-bit find_msb
|
||
- nir/algebraic: Do not generate 8- or 16-bit find_msb
|
||
- nir: Restrict ufind_msb and ufind_msb_rev to 32- or 64-bit sources
|
||
- nir/algebraic: Optimize some ifind_msb to ufind_msb
|
||
- nir/lower_int64: Optionally lower ufind_msb using uadd_sat
|
||
- intel/fs: Don't copy propagate from saturate to sel
|
||
- nir/algebraic: Undistribute fsat from fmax
|
||
- intel/fs: Output opt_combine_constants debug to stderr
|
||
- intel/fs: Refactor part of opt_combine_constants to a separate function
|
||
- intel/fs: Rework the loop of opt_combine_constants that collects constants
|
||
- intel/compiler: Remove one overload of backend_instruction::insert_before
|
||
- intel/compiler: Use NIR_PASS instead of NIR_PASS_V
|
||
- intel/compiler: Micro optimize inst_is_in_block
|
||
- intel/fs: Use specialized version of regions_overlap in opt_copy_propagation
|
||
- intel/compiler: Micro optimize regions_overlap
|
||
- intel/fs: Linked list micro optimizations in brw_nir_move_interpolation_to_top
|
||
- intel/fs: Preserve meta data more often in brw_nir_move_interpolation_to_top
|
||
- intel/fs: White space fixes
|
||
- nir/tests: Add many loop analysis tests for induction vars updated by shifts
|
||
- nir/tests: Add more loop analysis tests for induction vars updated by shifts
|
||
- nir/tests: Add many loop analysis tests for induction variables modified by imul
|
||
- nir/loop_analyze: Add a function to evaluate an ALU as constant
|
||
- nir/loop_analyze: Track induction variable basis information
|
||
- nir/loop_analyze: Change invert_cond instead of changing the condition
|
||
- nir/loop_analyze: Use try_eval_const_alu and induction variable basis info
|
||
- nir/tests: Port almost all loop_analyze tests to new macro-based infastructure
|
||
|
||
Ikshwaku Chauhan (1):
|
||
|
||
- radeonsi: Fix distortion for yuv422 format for GFX10.
|
||
|
||
Illia Abernikhin (1):
|
||
|
||
- util: Extend vk_enum_to_str with bitmasks vk_enum_to_str only generates literals for enums with type: @type="enum", but many enums have type: @type="bitmask" and were not taken into account here.
|
||
|
||
Illia Polishchuk (4):
|
||
|
||
- ANV: Add extra memory types for ANV driver instead of a single one
|
||
- hasvk: Add extra memory types for hasvk driver instead of a single one
|
||
- nir: Add sha1 hash for nir shaders converted from spir-v
|
||
- glx: fix indirect initialization crash
|
||
|
||
Ilya K (1):
|
||
|
||
- intel/vk/grl: don't install libgrl.a
|
||
|
||
Isaac Bosompem (1):
|
||
|
||
- tool/pps: Fix 32-bit build issue with format string
|
||
|
||
Isabella Basso (5):
|
||
|
||
- nir/algebraic: insert patterns inside optimizations list
|
||
- nir/algebraic: extend mediump patterns
|
||
- nir/algebraic: extend lowering patterns for conversions on smaller bit sizes
|
||
- nir/algebraic: make patterns for float conversion lowerings imprecise
|
||
- nir/algebraic: remove duplicate bool conversion lowerings
|
||
|
||
Italo Nicola (10):
|
||
|
||
- panfrost: fix off-by-one when exporting format modifiers
|
||
- panfrost: fix tiny sample_positions BO memory leak
|
||
- hud: use defines for default scale/rotation/visibility values
|
||
- hud: add GALLIUM_HUD_OPACITY envvar
|
||
- panfrost: fix strict-aliasing violations when packing fb ptrs
|
||
- etnaviv: abort() instead of assert(0) on compiler error
|
||
- etnaviv: use stderr for compiler error logging
|
||
- etnaviv: add default clear_buffer and clear_texture APIS
|
||
- etnaviv: lower (un)pack_{2x16,2x32}_split and extract_{byte,word}
|
||
- etnaviv: implement nir_op_uclz and lower find_{msb,lsb} to uclz
|
||
|
||
Iván Briano (7):
|
||
|
||
- anv: uncompressed views of compressed 3d images are now valid
|
||
- vulkan: track the right value on CmdSetColorWriteMasks
|
||
- anv: fix testing for dynamic color blend bits
|
||
- anv: stop tracking color blend state in the pipeline
|
||
- anv: use the parameter passed to the macro
|
||
- intel/fs: handle interpolation modes for at_sample and at_offset too
|
||
- vulkan/wsi/display: do not dereference a NULL pointer
|
||
|
||
Jakub Kulík (1):
|
||
|
||
- mesa: Fix format transform on big endian platforms
|
||
|
||
Jan Beich (1):
|
||
|
||
- util/u_process: implement util_get_command_line for BSDs
|
||
|
||
Janne Grunau (1):
|
||
|
||
- asahi: Fix typo in debug/error message helper macro
|
||
|
||
Jarred Davies (16):
|
||
|
||
- pvr: Use common queue submit implementation
|
||
- pvr: Add support for VK_KHR_timeline_semaphore
|
||
- pvr: Enable threaded submit when supported
|
||
- pvr: Clear wait syncs after job submission
|
||
- pvr: Don't update fragment signal sync when fragment stage is disabled
|
||
- pvr: Fix segfaults when pDepthStencilAttachment is NULL
|
||
- pvr: Generate EOT program at runtime
|
||
- pvr: Generate dummy emit for renders without any emits
|
||
- pvr: Add support for multiple emits from EOT program
|
||
- pvr: Select a single aspect format for the texture state of DS image views
|
||
- pvr: Add initial support for VK_FORMAT_S8_UINT
|
||
- pvr: Don't allocate/upload 0 size coeff programs
|
||
- pvr: Always mark robustBufferAccess as supported
|
||
- pvr: Rename pvr_xgl_pds.c to pvr_pipeline_pds.c
|
||
- pvr: Add robustness buffer support
|
||
- pvr: Mark all normalized formats as supporting with_packed_usc_channel
|
||
|
||
Jesse Natalie (224):
|
||
|
||
- ci/windows: Download updated WARP 1.0.4 package
|
||
- dzn/ci: Remove flakes/fails that don't hit anymore
|
||
- dzn/ci: Add image test group, which is all passing now
|
||
- dzn: Fix clear bind flag logic
|
||
- microsoft/compiler: Lower pack_[u/s]norm_2x16
|
||
- microsoft/compiler: Implement texture sample count query
|
||
- microsoft/compiler: Remove arrays when testing for structs in I/O
|
||
- microsoft/compiler: Always emit float types in the I/O signature for structs
|
||
- microsoft/compiler: Re-work the logic for adding SV_SampleIndex to force sample-rate
|
||
- microsoft/compiler: Use nir info.fs.uses_sample_shading to force sample-rate
|
||
- microsoft/compiler: Set num_components to 4 when updating pos write instructions
|
||
- spirv2dxil: For removing unused vars, consider the whole I/O var size
|
||
- spirv2dxil: When removing unused inputs, make sure they're actually inputs
|
||
- spirv2dxil: Allow killing position as an undef varying
|
||
- spirv2dxil: Replace not-provided inputs with zero instead of undef
|
||
- dzn: Get options13
|
||
- dzn: Support alpha blend factor
|
||
- dzn: When changing root signature, dirty descriptors too
|
||
- dzn: Use R24G8_TYPELESS for 24/8 depth resources
|
||
- dzn: Support int border colors
|
||
- dzn: Storage buffer sizes need to be 4-byte-aligned
|
||
- dzn: Set MultisampleEnable to enable MSAA lines
|
||
- dzn: Use typeless format for creation of depth-only or stencil-only D24S8
|
||
- dzn: Define a symbol that was present in older D3D headers
|
||
- dzn: Support root signature 1.2
|
||
- dzn: Support unnormalized coordinate samplers
|
||
- dzn: Always align cached pipeline header size to input element align
|
||
- dzn: Add a zeroed zsa state when depth or raster is disabled
|
||
- dzn: Disable depth when the rasterizer is disabled due to no position output
|
||
- dzn: Fix format support checks for storage/uniform texel buffers
|
||
- dzn: Remove cmdbuf query 'wait' list
|
||
- microsoft/compiler: Delete incorrect implementation for load_layer_id
|
||
- microsoft/compiler: Subpass textures are supposed to be arrays
|
||
- microsoft/compiler: Delete stale TODO comment
|
||
- microsoft/compiler: Support view instancing
|
||
- spirv2dxil: Pass runtime conf struct to lower_shader_system_values
|
||
- spirv2dxil: Implement lowering for multiview
|
||
- spirv2dxil: Claim multiview support
|
||
- dzn: Put nir compilation options in a struct
|
||
- dzn: Handle multiview pipeline creation
|
||
- dzn: Handle draws and clears for multiview rendering
|
||
- dzn: Implement multiview queries
|
||
- dzn: Enable multiview
|
||
- dzn: Enable independent blending
|
||
- dzn: Delete an unnecessary assert
|
||
- dzn: Rework meta blit VS
|
||
- microsoft/compiler: Add an overload param to unary function helpers
|
||
- microsoft/compiler: Implement a few basic wave/subgroup intrinsics
|
||
- microsoft/compiler: Add lowering passes for basic subgroup vars
|
||
- spirv2dxil: Use 32-bit shared offsets
|
||
- spirv2dxil: Support basic subgroups
|
||
- dzn: Support basic subgroups
|
||
- microsoft/compiler: Fix atomic image umax
|
||
- microsoft/compiler: Lower device index to zero
|
||
- spirv2dxil: Support dispatches with base group indices
|
||
- dzn: Support vkCmdDispatchBase
|
||
- dzn: Use common physical device list/enumeration helpers
|
||
- dzn: Respect suspending/resuming flags to omit clears/resolves
|
||
- dzn: Set dynamic rendering caps
|
||
- dzn: When rendering to 3D, don't treat layers as subresources for barriers
|
||
- dzn: Move patched vertex buffer capability check up a level
|
||
- dzn: Use SHADER_LOAD to indicate SAMPLED_IMAGE support
|
||
- dzn: Usage image view usage instead of image usage
|
||
- dzn: Support EXTENDED_USAGE bit
|
||
- dzn: Usage MULTISAMPLE_LOAD support instead of RT/DS support for MSAA
|
||
- dzn: Descriptor limits are based on binding tier, not heap tier
|
||
- dzn: A single sampler descriptor set needs to support 1024 samplers
|
||
- dzn: Don't expose variable pointers
|
||
- dzn: Fix independent blend check
|
||
- dzn: Enable Vulkan 1.1
|
||
- microsoft/compiler: Don't emit threadgroup barriers for graphics shaders
|
||
- microsoft/compiler: Handle i2i1 and u2u1
|
||
- microsoft/compiler: Handle i1 overloads
|
||
- microsoft/compiler: Implement more wave/quad ops
|
||
- microsoft/compiler: Support emitting the SM6.6 wave size tag
|
||
- spirv2dxil: Lower some wave op properties
|
||
- spirv2dxil: Support subgroup SPIR-V caps
|
||
- dzn: Support more subgroup/quad ops
|
||
- dzn: Implement subgroup size control extension
|
||
- dzn: Use core feature matching logic instead of rolling our own
|
||
- microsoft/compiler: Support float controls
|
||
- dzn: Fix dynamic rendering clear load op for non-multiview
|
||
- dzn: Handle separate stencil usage
|
||
- dzn: Cache GPUVA for buffers
|
||
- dzn: Support float control
|
||
- dzn: Always do clears with copies on non-graphics queues
|
||
- dzn: Enhanced barriers fixes/workarounds
|
||
- dzn: Ensure we don't mix DSV+simultaneous-access
|
||
- dzn: Support Vulkan 1.2
|
||
- dzn: Fix Windows WSI
|
||
- dzn: Don't recursively lock the physical device enum mutex
|
||
- dzn: Report as a software device for non-Windows
|
||
- CI/windows: Don't limit deqp-runner to 4 jobs
|
||
- CI/windows: Apply CI_FDO_CONCURRENT to piglit too
|
||
- dzn: Consider linked shaders when computing DXIL hash
|
||
- wsi/win32: Always use non-SRGB formats for DXGI
|
||
- wsi/win32: Use app-provided timeout instead of arbitrary hardcoded value
|
||
- CI: Lima farm is offline
|
||
- dzn, driconf: Add a driconf entry for NMS to claim wide line support
|
||
- vulkan/wsi: Add a wsi_device param to get_present_modes
|
||
- vulkan/wsi/win32: Support tearing (immediate) and VSync (FIFO) present modes
|
||
- wsi/win32: Don't require buffer blits for software drivers
|
||
- wsi/win32: We don't need a window DC for DXGI
|
||
- clc: Include opencl-c-base.h with LLVM 15 (using builtins)
|
||
- microsoft/clc: Set features that are used by CL tests
|
||
- ci/windows: Update LLVM to 15
|
||
- nir: Add alignment to load_push_constant
|
||
- nir_lower_fp16_casts: Allow opting out of lowering certain rounding modes
|
||
- microsoft/compiler: Handle struct consts in DXIL module dumper
|
||
- microsoft/compiler: Handle frcp for float16/float64
|
||
- microsoft/compiler: Ensure native_low_precision is set for 16-bit bitcasts/stores
|
||
- microsoft/compiler: Handle undef-rounding f2f16 as rtz
|
||
- microsoft/compiler: Move unaligned load/store pass from CL
|
||
- microsoft/compiler: Pass deref modes to unaligned pass and handle push const
|
||
- microsoft/compiler: Simplify bitpacking for load/store lowering with nir_extract_bits
|
||
- microsoft/compiler: Pass an alignment to constant buffer load lowering
|
||
- microsoft/compiler: Handle 48-bit stores to SSBO/shared
|
||
- microsoft/compiler: Support raw buffer load/store intrinsics with 16bit alignment
|
||
- microsoft/compiler: Support lowering SSBO accesses to 16bit vectors
|
||
- spirv2dxil: Set min UBO/SSBO alignments
|
||
- spirv2dxil: Lower unaligned loads and stores
|
||
- spirv2dxil: Move shader model into runtime conf struct
|
||
- spirv2dxil: Support 16bit types
|
||
- dzn: Enable get_surface_capabilities2
|
||
- dzn: Delete unused extensions table
|
||
- dzn: Get options4
|
||
- dzn: Enable 16bit types when supported
|
||
- dzn: Enable KHR_storage_buffer_storage_class
|
||
- vulkan/wsi: Fix Windows build
|
||
- radv: Fix returning an expression from a void function
|
||
- Revert "CI: Disable Windows runners"
|
||
- nir: Propagate alignment when rematerializing cast derefs
|
||
- microsoft/compiler: Implement wave reduce/exclusive scan ops that are supported
|
||
- microsoft/compiler: Add a lowering pass for scan ops that aren't supported
|
||
- spirv2dxil: Handle arithmetic subgroup ops
|
||
- dzn: Claim the arithmetic subgroup bit
|
||
- ci/windows: Update warp to 1.0.5
|
||
- microsoft/compiler: Handle writable buffer UAV size queries
|
||
- d3d12: Report correct texel buffer max size
|
||
- d3d12: Fix buffer SRV/UAV creation
|
||
- d3d12: Remove now-unused UAV format from shader info
|
||
- microsoft/compiler: Fix setting bit 31 in feature flags
|
||
- microsoft/compiler: Only set typed UAV load feature bit for multi-comp loads
|
||
- microsoft/compiler: Refactor type -> resource kind helper
|
||
- microsoft/compiler: Add helpers for getting res_props structs
|
||
- microsoft/compiler: Split handle annotation into two parts
|
||
- microsoft/compiler: Handle "bindless" image/tex sources as heap indices
|
||
- microsoft/compiler: Support descriptor heap indexing for UBO/SSBO
|
||
- microsoft/compiler: Use store_dest instead of store_dest_value more
|
||
- microsoft/compiler: Update header docs for binding modes supported by compiler
|
||
- spirv2dxil: Add a pass to lower deref tex/image and vulkan ubo/ssbo to bindless
|
||
- spirv2dxil: Only lower readonly images to SRVs when the option is set
|
||
- spirv2dxil: Support descriptor indexing capabilities
|
||
- dzn: Remove device pointers from descriptor heaps
|
||
- dzn: Remove descriptor heap type from descriptor heap wrapper
|
||
- dzn: Fix a leak in descriptor set layout creation
|
||
- dzn: Add some docs around descriptor sets and remove redundant/unused data
|
||
- dzn: Put UAVs first for storage images/buffers in descriptor tables
|
||
- dzn: Consistently order depth formats before stencil
|
||
- dzn: Don't use plane slice 1 for depth+stencil SRVs
|
||
- dzn: Set up SRV descs for 3D textures correctly
|
||
- dzn: Skip setting up UAVs for depth resources
|
||
- dzn: Add initial bindless infrastructure
|
||
- dzn: When bindless, only allocate one descriptor per layout entry
|
||
- dzn: Remove defragmenting of descriptor pools
|
||
- dzn: Delete unused function
|
||
- dzn: Allocate descriptor sets in buffers for bindless mode
|
||
- dzn: Don't dirty bindings if root signature doesn't change
|
||
- dzn: Use separate dirty bits for descriptor sets/dynamic buffers
|
||
- dzn: Bind buffers for bindless descriptor sets
|
||
- dzn: Add a binding classification in the pipeline layout remapping
|
||
- dzn: When binding a bindless root signature, bind descriptor heaps first
|
||
- dzn: Ensure root signatures are re-bound after a meta op
|
||
- dzn: Only bind descriptor sets up to the used amount of the current layout
|
||
- dzn: Apply bindless lowering when compiling pipelines
|
||
- dzn: Add a debug option for enabling bindless mode
|
||
- dzn: Support descriptor indexing via bindless
|
||
- dzn: Enable variable size bindings
|
||
- dzn: Use mesa_loge for DXIL validation errors
|
||
- microsoft/clc: Add shader model / validator to compiler API
|
||
- d3d12: Move forward-front-face pass to common DXIL code
|
||
- spirv2dxil: Expose yz flip pass to external callers
|
||
- dzn: Add a helper to generate triangle->point GS
|
||
- dzn: Handle polygon point mode
|
||
- dzn: Claim fillModeNonSolid
|
||
- CI/windows: Update headers and Agility redist to 1.710.0-preview
|
||
- CI/windows: Increase timeout for build container job
|
||
- microsoft/compiler: Fix 8-bit loads and stores when supporting 16-bit DXIL
|
||
- microsoft/compiler: Fix barrier for wave ID computation
|
||
- microsoft/compiler: Assign 1D wave IDs based on local thread ID
|
||
- microsoft/compiler: Fix large shifts
|
||
- spirv2dxil: Add some more supported caps
|
||
- dzn: Add a driconf entry for enabling 8bit loads and stores
|
||
- dzn: Add a driconf option for enabling subgroup ops in VS/GS
|
||
- dzn: Fix SRV barrier state on compute command lists
|
||
- dzn: Raise max number of descriptor sets to 8
|
||
- dzn: Report some more caps correctly that are supported
|
||
- dzn: Align descriptor sets in the bindless buffer
|
||
- dzn: Ensure pipeline variants are used for dynamic stencil masks
|
||
- dzn: Don't use write-combine memory for cache-coherent UMA
|
||
- dzn: Ensure buffer offsets are aligned
|
||
- dzn: Attempt to force depth write states for depth access in LAYOUT_GENERIC
|
||
- dzn: Don't do initial-layout barriers for simultaneous-access resources
|
||
- dzn: Batch command lists together
|
||
- dzn: Fix bindless descriptor sets with multiple dynamic buffers that need custom descriptors
|
||
- dzn: Early-out on no-op barriers
|
||
- dzn: Clean up ABI helpers now that we require DirectX-Headers 606
|
||
- dzn: Use GetResourceAllocationInfo3 for castable formats
|
||
- dzn: Don't leave deleted physical devices in the instance pdev list
|
||
- dzn: Remove skips now that WARP is faster
|
||
- dzn: Support >2K samplers with bindless
|
||
- dzn: Remove xfail for test that passes (if run)
|
||
- microsoft/compiler: Don't split loads/stores that will be split by lower_explicit_io
|
||
- dzn: Changes to descriptor set dirty flag handling
|
||
- dzn: Use a linear allocator for upload data on command buffers
|
||
- dzn: Ignore unnormalized sampling flag if driver doesn't support it
|
||
- dzn: Never set STATE_RENDER_TARGET on a compute command list
|
||
- dzn: Don't enable bindless by default
|
||
- d3d12: Support creating PSOs with no attachments with MSAA without TIR
|
||
- d3d12: Fix buffer reference leak for SO count staging buffer
|
||
- dzn: Handle mismatches in bound descriptor set vs pipeline layout
|
||
- d3d12: Respect buffer offsets for sampler views
|
||
- dzn: Hook up subgroup size to compute shader compilation
|
||
- dzn: Delete queue-level event waits
|
||
|
||
Jonathan Gray (3):
|
||
|
||
- egl/dri2: avoid undefined unlocks
|
||
- intel/dev: remove invalid EHL pci id
|
||
- intel/dev: Add another EHL pci id
|
||
|
||
Jonathan Marek (1):
|
||
|
||
- turnip: fix use of align() instead of util_align_npot() with tile_align_w
|
||
|
||
Jordan Justen (8):
|
||
|
||
- intel/vk/grl: Don't include anv_private.h in genX_grl.h
|
||
- intel/vk/grl: Allow genX_grl.h to be included by C++ files
|
||
- intel/vk/grl: Allow grl/grl_cl_kernel.h to be included by C++ files
|
||
- intel/vk/grl: genX-ify genX_grl_uuid.cpp
|
||
- intel/vk/grl: genX-ify grl_cl_kernel_name()
|
||
- intel/dev: Enable MTL PCI ids
|
||
- intel/compiler: Support fmul_fsign opt for fp64 when int64 isn't supported
|
||
- intel/compiler/gfx12.5+: Lower 64-bit cluster_broadcast with 32-bit ops
|
||
|
||
Joshua Peisach (1):
|
||
|
||
- gallum/asahi: fix memory leak in agx_resource_from_handle
|
||
|
||
José Fonseca (4):
|
||
|
||
- llvmpipe: Ensure floating point SSE state is reset regardless of the write mask.
|
||
- llvmpipe: Honor zero sample_mask when multisample is disabled.
|
||
- trace: Don't use italic escape code.
|
||
- wgl: Fix unintentional assignment on assert.
|
||
|
||
José Roberto de Souza (107):
|
||
|
||
- anv: Start to move i915 specific code from anv_device to i915/anv_device
|
||
- anv: Export anv_exec_batch_debug() and chain_command_buffers()
|
||
- anv: Split i915 code from anv_batch_chain.c
|
||
- anv: Move anv_device_check_status() code to i915/anv_device.c
|
||
- intel/dev: Export functions that will be used by different kernel drivers
|
||
- intel/dev: Move i915 code to i915/intel_device_info.c
|
||
- intel/dev: Split hwconfig i915 specific code
|
||
- intel/dev: Detect what is the kernel mode driver loaded
|
||
- intel: Add intel_kmd_type parameter to intel_engine_get_info()
|
||
- intel: Add kmd_type parameter to necessary intel_gem.h functions
|
||
- anv: Nuke anv_queue:index_in_family
|
||
- hasvk: Nuke anv_queue:index_in_family
|
||
- intel/ds: Nuke intel_ds_queue::queue_id
|
||
- intel/ds: Fix crash when allocating more intel_ds_queues than u_vector was initialized
|
||
- intel/genxml/gen125: Add walker configuration fields to 3DSTATE_WM
|
||
- intel/genxml/gen125: Tune 3DSTATE_WM Walker direction
|
||
- intel: Add intel_memory_class_instance
|
||
- anv: Convert drm_i915_gem_memory_class_instance to intel_memory_class_instance
|
||
- anv: Use DRM_IOCTL_I915_GEM_CREATE_EXT in all supported kernels
|
||
- anv: Add basic KMD backend infrastructure
|
||
- anv: Start to move anv_gem_stubs.c to kmd backend
|
||
- anv: Remove remaining bits of anv_i915_query()
|
||
- hasvk: Remove remaining bits of anv_i915_query()
|
||
- anv: Add gem_close to kmd backend
|
||
- anv: Add gem_mmap to kmd backend
|
||
- anv: Move execute_simple_batch() and queue_exec_locked() to kmd backend
|
||
- intel/common: Move i915 files to i915 folder
|
||
- iris: Export batch debug functions
|
||
- iris: Export update_batch_syncobjs()
|
||
- iris: Export num_fences()
|
||
- intel: Make gen12 URB space reservation dependent on compute engine presence
|
||
- intel/blorp: Allocate only necessary amount of VERTEX_BUFFER_STATE
|
||
- intel: Pull in xe_drm.h
|
||
- intel: Add Meson parameter to enable Xe KMD support
|
||
- intel/dev: Add INTEL_KMD_TYPE_XE
|
||
- intel/dev: Implement Xe functions to fill intel_device_info
|
||
- intel/dev: Implement Xe functions to handle hwconfig
|
||
- intel/dev: Query and compute hardware topology for Xe
|
||
- iris: Convert drm_i915_gem_memory_class_instance to intel_memory_class_instance
|
||
- iris/bufmgr: Add i915_gem_set_domain()
|
||
- iris: Use DRM_IOCTL_I915_GEM_CREATE_EXT in all supported kernels
|
||
- iris: Add initial skeleton of kmd backend
|
||
- iris: Move iris_bo_madvise() to i915/iris_bufmgr.c
|
||
- iris: Add iris_bo_set_caching()
|
||
- intel/common: Implement the Xe functions for intel_engine
|
||
- intel/common: Implement the Xe functions for intel_gem
|
||
- iris: Move bo_madvise to kmd backend
|
||
- iris: Move bo_set_caching to kmd backend
|
||
- iris: Move iris_bo_busy_gem() to i915/iris_bufmgr.c
|
||
- iris: Move iris_bo_wait_gem() to i915/iris_bufmgr.c
|
||
- iris: Don't mark protected bo as reusable
|
||
- intel/perf: Disable it for Xe KMD
|
||
- build: Block build of HASVK, Crocus and i915 in non-x86 architectures
|
||
- iris: Add gem_mmap() to kmd backend
|
||
- iris: Add batch_check_for_reset() to kmd backend
|
||
- iris: Move i915 submit_batch() to i915 backend
|
||
- anv: Implement gem_create for Xe backend
|
||
- anv: Implement Xe functions to create and destroy VM
|
||
- anv: Implement gem close and mmap for Xe backend
|
||
- anv: Add gem VM bind and unbind to backend
|
||
- anv: Integrate gem vm bind and unbind kmd backend functions
|
||
- iris: Drop I915_EXEC_FENCE types
|
||
- iris: Drop usage of i915 EXEC_OBJECT_WRITE
|
||
- iris: Move iris_bufmgr_init_global_vm() to i915/iris_bufmgr.c and prepare for Xe KMD
|
||
- anv: Implement Xe version of anv_physical_device_get_parameters()
|
||
- anv: Properly alloc buffers that will be promoted to framebuffer in Xe KMD
|
||
- anv: Handle external objects allocation in Xe
|
||
- iris: Only mark buffer as exported if drmPrimeHandleToFD() succeed
|
||
- iris: Implement the Xe version of iris_bufmgr_init_global_vm()
|
||
- iris: Implement the function to destroy VM in Xe
|
||
- iris: Implement gem_create() in Xe kmd backend
|
||
- iris: Implement gem_mmap() in Xe kmd backend
|
||
- iris: Store iris_context's priority
|
||
- iris: Move to i915/iris_batch.c code to create and replace i915 context
|
||
- iris: Move to iris_i915_batch.c code to destroy i915 context
|
||
- intel: Move memory aligment information to intel_device_info
|
||
- anv: Use intel_device_info memory alignment
|
||
- intel: Set mem_alignment in Xe kmd
|
||
- anv: Apply memory alignment requirements in Xe kmd
|
||
- intel: Add TODO about removal of 2Mb alignment in i915
|
||
- anv: Replace I915_ENGINE_CLASS_VIDEO by INTEL_ENGINE_CLASS_VIDEO
|
||
- anv: Create Xe engines
|
||
- anv: Implement Xe version of check_status()
|
||
- anv: Handle Xe queue/engine priority
|
||
- anv: Implement Xe version of execute_simple_batch()
|
||
- iris: Prepare iris_bufmgr functions for vm bind error paths
|
||
- iris: Add vm bind and unbind to kmd backend
|
||
- iris: Implement gem_vm_bind() and gem_vm_unbind() in Xe kmd backend
|
||
- iris: Ajust gem buffer allocation size in Xe kmd
|
||
- intel: Sync xe_drm.h
|
||
- anv: Partialy import drm-uapi/gpu_scheduler.h and use it
|
||
- anv: Fetch max_context_priority from drm_xe_query_config
|
||
- intel: Allocate mesh shader URB space before task shader
|
||
- anv: Move to a function code to clflush batch buffers
|
||
- anv: Implement Xe version of anv_queue_exec_locked() and queue_exec_trace()
|
||
- anv: Disable anv_bo_sync_type for Xe kmd
|
||
- anv: Add assert in functions not supported by Xe kmd
|
||
- iris: Add BO_ALLOC_SHARED
|
||
- iris: Handle allocation of exported buffers in Xe kmd
|
||
- iris: Handle allocation of scanout buffers in Xe
|
||
- iris: Implement Xe version of bo_madvise() and bo_set_caching()
|
||
- anv: Fix vm bind of imported buffers
|
||
- iris: Add function to close gem bos
|
||
- iris: Handle Xe syncronization with syncobjs
|
||
- loader: Add Xe KMD support
|
||
- iris: Fix close of exported bos
|
||
- iris: Allow shared scanout buffer to be placed in smem as well
|
||
|
||
Juan A. Suarez Romero (8):
|
||
|
||
- v3d/v3dv: define performance counters in common
|
||
- v3d: cache pipe query results
|
||
- v3d: include offset as part of streamout target
|
||
- v3d: implement NV_conditional_render extension
|
||
- v3d: fix condition for EZ disabling when stencil on
|
||
- v3d: set depth compare function correctly
|
||
- v3d: use primitive type to get stream output offset
|
||
- v3d: apply 1D texture miplevel alignment in arrays
|
||
|
||
Julia Tatz (3):
|
||
|
||
- zink: zink_heap isn't 1-to-1 with memoryTypeIndex
|
||
- zink: trival renames heap_idx -> memoryTypeIndex
|
||
- zink: correct sparse bo mem_type_idx placement
|
||
|
||
Juston Li (29):
|
||
|
||
- venus: refactor out vn_feedback_event_cmd_record2
|
||
- venus: refactor VK_KHR_synchronization2 ext sync fd requirements
|
||
- venus: require importable external semaphores for WSI
|
||
- venus: require exportable bit for ext fence sync fd
|
||
- venus: require exportable/importable bit for ext semaphores sync fd
|
||
- venus: remove filtering external semaphores for QueueSubmit
|
||
- venus: drop VkQueueBindSparse
|
||
- venus: append fence feedback batch
|
||
- venus: refactor QueueSubmit/QueueSubmit2
|
||
- venus: vn_queue: align vulkan object variable naming
|
||
- docs/envvars: add missing mesa disk cache envvars
|
||
- util/fossilize_db: don't destroy foz on RO load fail
|
||
- utils/fossilize_db: refactor out loading RO foz dbs
|
||
- util/fossilize_db: add runtime RO foz db loading via FOZ_DBS_DYNAMIC_LIST
|
||
- util/fossilize_db: fix macOS inotify build error
|
||
- util/fossilize_db: add ifdef for inotify header
|
||
- util/tests/cache_test: Skip Cache.List if not supported
|
||
- anv: check initial cmd_buffer is chainable
|
||
- venus: refactor semaphore helper functions
|
||
- venus: refactor batch submission fixup
|
||
- venus: add NO_TIMELINE_SEM_FEEDBACK perf option
|
||
- venus: add timeline semaphore feedback cmds
|
||
- venus: enable timeline semaphore feedback
|
||
- venus: add SHADER_DEVICE_ADDRESS_BIT to buffer cache
|
||
- venus: switch to lazy VkBuffer cache
|
||
- venus: add VN_DEBUG_CACHE flag
|
||
- venus: Add VkBuffer cache statistics for debug
|
||
- venus: shader cache fossilize replay fix
|
||
- util/disk_cache: use posix_fallocate() for index files
|
||
|
||
Kai Wasserbäch (5):
|
||
|
||
- fix: gallivm: limit usage of LLVMContextSetOpaquePointers() to LLVM 15
|
||
- fix(FTBFS): gallivm: fix LLVM #include of Triple.h, moved to TargetParser
|
||
- fix(FTBFS): clover: fix LLVM #include of Triple.h, moved to TargetParser
|
||
- fix: clover/llvm: replace llvm::None with std::nullopt for LLVM 17+
|
||
- fix: gallivm: fix LLVM #include of Host.h, moved to TargetParser
|
||
|
||
Kai-Heng Feng (2):
|
||
|
||
- iris: Retry DRM_IOCTL_I915_GEM_EXECBUFFER2 on ENOMEM
|
||
- Revert "iris: Avoid abort() if kernel can't allocate memory"
|
||
|
||
Karmjit Mahil (28):
|
||
|
||
- pvr: Process set and reset event sub commands.
|
||
- pvr: Process wait event sub command.
|
||
- pvr: Add SPM scratch buffer infrastructure.
|
||
- pvr: Acquire scratch buffer on framebuffer creation.
|
||
- pvr: Update comment about ZS and MSAA buffers for pvrsrvkm submission.
|
||
- pvr: Set SPMSCRATCHBUFFER flag.
|
||
- pvr: Add SPM load usc empty programs
|
||
- pvr: Upload spm load programs to device.
|
||
- pvr: Add support for VK_ATTACHMENT_LOAD_OP_LOAD.
|
||
- pvr: Move descriptor write into pvr_write_descriptor_set()
|
||
- pvr: Add support to copy descriptors on vkUpdateDescriptorSets()
|
||
- pvr: Handle VK_QUERY_RESULT_WAIT_BIT.
|
||
- pvr: Store enum pvr_stage_allocation instead of VkShaderStageFlags
|
||
- pvr: Put old descriptor set approach behind a hardcoding check
|
||
- pvr: Change last_DMA to last_dma
|
||
- pvr: Write descriptor set addrs table dev addr into shareds
|
||
- pvr: Add PVR_SELECT() helper macro
|
||
- pvr: Add push consts support to descriptor program.
|
||
- pvr: Add support for dynamic buffers descriptors
|
||
- pvr: Add support for blend constants.
|
||
- pvr: Move PBE START_POS into csb enum helpers header
|
||
- pvr: Setup SPM EOT state
|
||
- pvr: Remove unused msaa_mode field
|
||
- pvr: Remove component_alignment
|
||
- pvr: Setup SPM background object
|
||
- pvr: Don't advertise currently unsupported features
|
||
- pvr: Advertise STORAGE_IMAGE_BIT for B10G11R11_UFLOAT_PACK32
|
||
- pvr: Don't advertise S8_UINT support
|
||
|
||
Karol Herbst (44):
|
||
|
||
- rusticl: fix build error with valgrind being enabled
|
||
- rusticl/util: extract offset_of macro
|
||
- rusticl/icd: Make it work in case Rustc shuffles struct around
|
||
- rusticl/kernel: fix clGetKernelInfo CL_KERNEL_ATTRIBUTES for non source programs
|
||
- rusticl/program: enable spirv
|
||
- llvmpipe/ci: increase deqp-runner timeout
|
||
- rusticl/device: fix some device limits
|
||
- rusticl/device: limit CL_DEVICE_MAX_CONSTANT_ARGS
|
||
- rusticl: no compute only
|
||
- rusticl: allocate printf buffer as staging
|
||
- nir: Skip samplers and textures in lower_explicit_io
|
||
- nir/deref: don't replace casts with deref_struct if we'd lose the stride
|
||
- ci/zink: move threading tests to flakes
|
||
- rusticl/kernel: Images arg sizes also have to match the host pointer size
|
||
- gallivm: fix lp_vec_add_offset_ptr for 32 bit builds
|
||
- nvc0: enable fp helper invocation memory loads on Turing+
|
||
- nir: track existence of variable shared memory
|
||
- rusticl/kernel: set has_variable_shared_mem on the nir
|
||
- gallium: add get_compute_state_info
|
||
- lp: implement get_compute_state_info
|
||
- iris: implement get_compute_state_info
|
||
- nv50: implement get_compute_state_info
|
||
- nvc0: implement get_compute_state_info
|
||
- panfrost: move max_thread_count and take reg_count into account
|
||
- panfrost: implement get_compute_state_info
|
||
- rusticl/kernel: make use of cso info
|
||
- radeonsi: implement get_compute_state_info
|
||
- radeonsi: use default float mode for CL
|
||
- rusticl: enable radeonsi
|
||
- rusticl: split platform into core and api parts
|
||
- rusticl/platform: rename _cl_platform_id to Platform
|
||
- rusticl/platform: move getter into the type
|
||
- rusticl/platform: move device initialization to the platform
|
||
- rusticl/program: allow dumping compilation logs through RUSTICL_DEBUG
|
||
- rusticl/program: make IL programs look closer to CLC ones
|
||
- clc: add clc_validate_spirv
|
||
- rusticl/program: validate the SPIR-V when created from IL
|
||
- rusticl/program: extract common code of compile and build
|
||
- rusticl/program: rework source code tracking
|
||
- rusticl/event: drop work item before updating status
|
||
- radeonsi: lower mul_high
|
||
- ac/llvm: support shifts on 16 bit vec2
|
||
- rusticl: don't set size_t-is-usize for >=bindgen-0.65
|
||
- nvc0: do not randomly emit fences.
|
||
|
||
Kenneth Graunke (37):
|
||
|
||
- intel/blorp: Lower base_workgroup_id to zero
|
||
- intel/compiler: Move atomic op translation into emit_*_atomic()
|
||
- intel/compiler: Use LSC opcode enum rather than legacy BRW_AOPs
|
||
- intel/compiler: Add an lsc_op_num_data_values() helper
|
||
- intel/compiler: Eliminate SHADER_OPCODE_UNTYPED_ATOMIC_FLOAT
|
||
- intel/compiler: Drop redundant 32-bit expansion for shared float atomics
|
||
- intel/compiler: Delete fs_visitor::nir_emit_{ssbo,shared}_atomic_float()
|
||
- intel/compiler: Combine nir_emit_{ssbo,shared}_atomic into one helper
|
||
- intel/compiler: Delete all the A64 atomic variants for type sizes
|
||
- intel/compiler: Drop dest checking in atomic code
|
||
- intel/compiler: Use more symbolic source names in components_read()
|
||
- anv: Add missing untyped data port flush on PIPELINE_SELECT
|
||
- iris: Add missing untyped data port flush on PIPELINE_SELECT
|
||
- loader: Add infrastructure for tracking active CRTC resources
|
||
- egl: Rewrite eglGetMscRateANGLE to avoid probes and handle multi-monitor
|
||
- iris: Perform load_constant address math in 32-bit rather than 64-bit
|
||
- anv: Perform load_constant address math in 32-bit rather than 64-bit
|
||
- anv: Make a batch decoder for each queue family
|
||
- nir: Print divergence information for registers as well as SSA defs
|
||
- nir: Fix merge_set_dump() to compile again
|
||
- nir: Fix typos in the from-SSA pass comments
|
||
- intel: Use common helpers for TCS passthrough shaders
|
||
- intel/fs: Fix inferred_sync_pipe for F16TO32 opcodes
|
||
- intel/fs: Add builder helpers for F32TO16/F16TO32 that work on Gfx7.x
|
||
- intel/fs: Delete a TODO about using brw_F32TO16.
|
||
- intel/fs: Use new F16TO32 helpers for unpack_half_split_* opcodes
|
||
- Revert "intel/fs: Fix inferred_sync_pipe for F16TO32 opcodes"
|
||
- intel/fs: Use F32TO16/F16TO32 helpers in fquantize16 handling
|
||
- intel/fs: Move packHalf2x16 handling to lower_pack()
|
||
- intel/eu: Simplify brw_F32TO16 and brw_F16TO32
|
||
- intel/vec4: Retype texture/sampler indexes to UD
|
||
- intel/fs: Make bld.F16TO32 actually emit F16TO32 not F32TO16
|
||
- i965/vec4: Implement uclz in the vec4 backend
|
||
- st/mesa, iris: Add optional CPU-based ASTC void extent denorm flushing
|
||
- intel/compiler: Use nir_dest_bit_size() for ballot bit size check
|
||
- iris: Extend resource creation helpers to allow for explicit strides
|
||
- iris: Hack around gbm_gralloc stride restrictions
|
||
|
||
Konrad Dybcio (3):
|
||
|
||
- freedreno/registers: Add RBBM_GPR0_CNTL for non-GMU operation
|
||
- freedreno: Add A2xx perf counter reg values
|
||
- freedreno: Add A2xx REG_A2XX_RBBM_PM_OVERRIDE2 bitfields
|
||
|
||
Konstantin Seurer (78):
|
||
|
||
- vulkan: Track the nullDescriptor feature
|
||
- radv: Add a helper for finding memory indices
|
||
- radv: Create a null TLAS as meta state
|
||
- radv: Use the null accel struct instead of emitting 0
|
||
- radv/rt: Get rid of accel struct null checks
|
||
- radv: Advertise rt pipelines for Control (DX12)
|
||
- radv/bvh/meson: Add the option to set defines
|
||
- radv/bvh: Add a define for extended SAH
|
||
- radv: Add a shader variant for PLOC with extended SAH
|
||
- radv: Wrap internal build type inside a build_config struct
|
||
- radv: Enable extended SAH for shallow BVHs
|
||
- radv: Merge the leaf and internal converter
|
||
- radv: Improve the BVH size estimation
|
||
- radv: Fix creating accel structs with unbound buffers
|
||
- radv: Work around shader_call_data variables in raygen shaders
|
||
- radv/rq: Use 16 stack entries if there is only one ray query
|
||
- radv/llvm: Use the shader names as module name
|
||
- ac/llvm: Fix validation error with global io
|
||
- radv: Scalarize global IO with LLVM enabled
|
||
- radv: Make radv_compute_dispatch non-static
|
||
- radv: Implement ordered compute dispatches
|
||
- radv: Use an ordered dispatch for BVH encoding
|
||
- radv: Remove radv_indirect_unaligned_dispatch
|
||
- vulkan,nir: Refactor ycbcr conversion state into a struct
|
||
- radv: Use common ycbcr conversion lowering
|
||
- radv/rra: Find copy memory index when initializing the trace state
|
||
- radv/rra: Hide deferred accel struct data destruction behind an env var
|
||
- radv: Hash VK_PIPELINE_CREATE_RAY_TRACING_NO_NULL_* flags
|
||
- radv: Clean up dynamic RT stack allocation
|
||
- ac/llvm: Implement bvh64_intersect_ray_amd
|
||
- radv: Make accel struct meta state initialization thread safe
|
||
- radv: Force ACO for BVH build shaders
|
||
- radv: Pre-compile BVH build shaders if there is a cache
|
||
- radv: Advertise ray query support with LLVM
|
||
- radv/rt: Skip instances after loading the entire node
|
||
- vulkan: Add vk_acceleration_structure
|
||
- radv: Use vk_acceleration_structure
|
||
- anv: Use vk_acceleration_structure
|
||
- radv/bvh/encode: Use the node type for identifying internal nodes
|
||
- radv/bvh: Replace is_final_tree with bvh_offset
|
||
- radv/bvh/encode: Move bvh_offset NULL check to the top of the loop
|
||
- radv/bvh/encode: Introduce is_root_node
|
||
- radv/bvh/encoder: Move dst_node initialization into the loop
|
||
- radv: Add a build config for compact builds
|
||
- radv/bvh: Implement compact encoding
|
||
- radv: Use compact encoding
|
||
- radv: Move the geometry infos before the BVH
|
||
- radv/bvh: Move the size header field up
|
||
- radv/bvh: Add a shader for filling the header
|
||
- radv: Use indirect header filling for compact builds
|
||
- nir: Add cull_mask_and_flags_amd intrinsic
|
||
- radv/rt: Merge cull_mask and flags
|
||
- radv/rt: Pre shift cull_mask
|
||
- radv: Move header and geometry info init into separate functions
|
||
- radv: Only init geometry infos if RRA is enabled
|
||
- radv/rt: Use ushr for extracting the cull mask
|
||
- radv/rt: Fix updating stack_size if the shader uses scratch
|
||
- radv/rt: Use vk_pipeline_hash_shader_stage for RT stages
|
||
- vulkan: Add vk_shader_module_init
|
||
- radv/rt: Properly handle pNext of pipeline library stages
|
||
- radv/sqtt: Skip dumping pipeline libraries
|
||
- radv: Fix loading stack_size from the cache
|
||
- radv: Fix inserting stack_size into the cache
|
||
- radv/rt: Handle load_constant instructions when inlining shaders
|
||
- nir/lower_shader_calls: Remat derefs before lowering resumes
|
||
- radv/rt: Refactor rq_load lowering
|
||
- radv/rq: Rematerialize inv_dir before proceed
|
||
- radv: Set user SGPR locations when declaring args
|
||
- radv: Stop counting user SGPRS separately
|
||
- radv/ci: Update ray tracing pipeline fail/skip lists
|
||
- radv: Add radv_shader_type to fix gs_copy and trap handler handling
|
||
- radv: Remove some dead radv_shader_args setup
|
||
- aco: Remove is_gs_copy_shader
|
||
- radv: Remove has_previous_stage
|
||
- radv: Pack and encode geometry id and flags on the CPU
|
||
- radv/bvh: Remove calculate_node_bounds
|
||
- radv: Remove radv_bvh_aabb_node::aabb
|
||
- nir/lower_fp16_casts: Fix SSA dominance
|
||
|
||
Kurt Kartaltepe (1):
|
||
|
||
- drirc: Set limit_trig_input_range option for Nier games
|
||
|
||
Lang Yu (1):
|
||
|
||
- amd/common: fix a typo
|
||
|
||
Lepton Wu (1):
|
||
|
||
- egl/dri2: Use primary device in EGL device platform for kms_swrast
|
||
|
||
Lina Versace (8):
|
||
|
||
- util/glsl2spirv: Fix build with Python 3.6
|
||
- docs: Lower Python requirement to 3.6
|
||
- mailmap: Lina is Chad's new name
|
||
- mailmap: Add Lina's new google.com address
|
||
- venus: Update protocol for VK_EXT_memory_budget
|
||
- venus: Delete vn_renderer_info::has_cache_management
|
||
- venus: Refactor vn_physical_device_init_memory_properties
|
||
- venus: Enable VK_EXT_memory_budget
|
||
|
||
Lionel Landwerlin (151):
|
||
|
||
- pps: print out message when we get the first counters
|
||
- anv: record secondaries' traces into primaries
|
||
- intel/ds: track secondary cmdbuffers in perfetto
|
||
- intel/ds: move event_id access to perfetto lambda
|
||
- util/u_trace: add support for variable length trace points
|
||
- vulkan/debug_utils: copy debug util labels
|
||
- anv: add support for command buffer tagging in traces
|
||
- intel/ds: add INTEL_GPU_TRACEPOINT envvar to toggle tracepoints
|
||
- intel/ds: remove unused trace point
|
||
- intel/utrace: document tracepoints
|
||
- Revert "ci: build hasvk if we're building anv"
|
||
- intel: use a shared UUID with other drivers
|
||
- nir/divergence: add missing RT intrinsinc handling
|
||
- anv: fix generated indirect draw shader stats checks
|
||
- nir/lower_io: fix bounds checking for 64bit_bounded_global
|
||
- anv: fix preemption enable emission in gpu_memcpy
|
||
- intel/fs: avoid cmod optimization on instruction with different write_mask
|
||
- intel/decoder: print out compute push constants
|
||
- intel/common: add a INTEL_DECODE variable to parameter decoder at runtime
|
||
- vulkan/wsi/wayland: improve same gpu detection
|
||
- intel/fs: drop FS_OPCODE_UNIFORM_PULL_CONSTANT_LOAD_GFX7
|
||
- anv: fix null descriptors
|
||
- docs: list anv in EXT_extended_dynamic_state3 support
|
||
- intel: add missing PS restriction on BDW+
|
||
- anv: expose EXT_load_store_op_none
|
||
- intel/fs: make alpha_to_coverage a tristate
|
||
- anv: rename RT pipeline function helper
|
||
- anv: rename a few internal functions to highlight gfx use
|
||
- intel/ds: track end of pipe bits
|
||
- anv: use malloc for host only descriptor sets
|
||
- anv: put the video extension behind a variable
|
||
- intel/fs: fix mesh indirect movs
|
||
- intel/dev: add a default urb value for intel_stub_gpu on dg2
|
||
- anv/hasvk: handle a SAMPLED_READ/STORAGE_READ access flags
|
||
- anv: remove copied information from runtime graphics state
|
||
- anv: move 3DSTATE_VERTEX_ELEMENT emission to dynamic path
|
||
- anv: implement VK_EXT_vertex_input_dynamic_state
|
||
- intel/fs: bound subgroup invocation read to dispatch size
|
||
- hasvk: fix KHR_shader_float_controls reporting
|
||
- intel/perf: also add the oa timestamp shift on MTL
|
||
- anv: move debug submit to helper and call it on execbuf failure
|
||
- anv: track vram only BOs to print things out on ENOMEM execbuf
|
||
- anv: fix vma heap memory leak
|
||
- anv: fix invalid masking of 48bit address
|
||
- anv: remove assert typed write support when using NULL surface
|
||
- vulkan/runtime: store parameters of VK_EXT_sliced_view_of_3d
|
||
- anv: fixup Wa_16011107343 for Gfx12 only
|
||
- iris: fix Wa_16011107343 for Gfx12
|
||
- anv: remove more Gfx7 code
|
||
- genxml: Fix STATE_BASE_ADDRESS::BindlessSurfaceStateSize field size
|
||
- genxml: fix border color offset field on Gfx12+
|
||
- anv/hasvk: speed up null image/view descriptor writes
|
||
- anv: fix scratch buffer reloc in 3DSTATE_HS
|
||
- anv: fixup condition for Wa_14016118574
|
||
- anv: pull Wa_14016118574 out of some loop not changing state
|
||
- util/glsl2spirv: add support for include directive
|
||
- anv: fix incorrect parameter
|
||
- anv: correctly reset generation address on command buffer reset
|
||
- anv: fix generated forward jump with more than 67M draws
|
||
- anv: remove copied code from generation shader
|
||
- anv: remove BTI related flush in generation shaders
|
||
- anv: correctly program 3DSTATE_SF in generation shaders
|
||
- anv: limit push constant dirtyness with generation shaders
|
||
- anv: remove pre hasvk split assert
|
||
- anv: remove commented code
|
||
- anv: fix 3DSTATE_PS emission in generation shaders
|
||
- anv: fix indirect draws VF cache tracking of index buffer
|
||
- anv: make sure mi_memcpy lands before push constant loads
|
||
- anv: remove MI_NOOPs at the end of the generation batch
|
||
- anv: use a single generation shader for indirect draws
|
||
- anv: rename generated draws for Gfx11
|
||
- anv: use 64bit int support in generation shaders
|
||
- anv: pack more data into generated draws input
|
||
- anv: move common shader code into header
|
||
- anv: use a list of generated shaders
|
||
- anv: remove unused item_count parameter
|
||
- anv: add gfx9 generated draw support
|
||
- blorp: add dependency on idep_intel_dev
|
||
- vulkan/runtime: only consider slice info with 3D image views
|
||
- anv: VK_EXT_image_sliced_view_of_3d
|
||
- nir: fix nir_ishl_imm
|
||
- anv: enable VK_EXT_pipeline_library_group_handles
|
||
- anv/iris: report counter symbols with debug option
|
||
- intel/fs: report max register pressure in shader stats
|
||
- anv: report max register pressure in pipeline properties
|
||
- anv: force MEDIA_INTERFACE_DESCRIPTOR_LOAD reemit after 3D->GPGPU switch
|
||
- radv: use 1ull for alignment computations
|
||
- util: allow align64() to do alignments >= 4Gb
|
||
- docs: fix invalid link
|
||
- iris: trace frames with u_trace
|
||
- anv: export EXT_pipeline_library_group_handles only with RT
|
||
- docs: update Anv features support
|
||
- anv: more formats for acceleration structure vertices
|
||
- intel/fs: don't SEND messages as partial writes
|
||
- intel/fs: fix nir_opt_peephole_ffma max vec assumption
|
||
- intel/fs: fixup sources number from opt_algebraic
|
||
- intel/fs: add MOV source count validation
|
||
- intel/fs: prevent large vector ops generated by peephole_ffma
|
||
- intel/fs: fix subgroup invocation read bounds checking
|
||
- vulkan/wsi: add a headless swapchain implementation/option
|
||
- intel/compiler: report max dispatch width statistic
|
||
- anv: report shader max dispatch width in pipeline props
|
||
- intel/devinfo: add an option to pick platform to print
|
||
- intel/devinfo: printout URB entries
|
||
- intel/dev: use generated WA helpers for Wa_22012575642
|
||
- intel/devinfo: dedicated entries for XeHP
|
||
- intel/devinfo: initialize pci_device_id with from_pci_id()
|
||
- intel/dev: fold Gfx12 URB entries in Gfx12 HW info
|
||
- util/u_trace: move needs_cs_param option to tracepoints
|
||
- vulkan/runtime: also copy strings on queue debug utils
|
||
- intel/ds: rename frame timeline row to queue
|
||
- anv: fix incorrect utrace bo release
|
||
- anv: fixup locking for utrace submission increments
|
||
- anv: rename anv_utrace_flush_copy in anv_utrace_submit
|
||
- anv: add utrace support for queue debug utils
|
||
- anv: implement recommended flush/wait of AUX-TT invalidation
|
||
- iris: implement recommended flush/wait of AUX-TT invalidation
|
||
- anv: hash immutable sampler conversion data not pointers
|
||
- anv: compute the largest GRL kernel scratch size
|
||
- anv: move queue check helpers to anv_private
|
||
- anv: take care of maxStorageBufferRange being uint32_t
|
||
- isl: update max buffer size for SKL+
|
||
- intel/dev: set a default valid kmd_type
|
||
- intel/perf: fix OA format selection on MTL
|
||
- intel/fs: run VGRF compaction just before max live register accounting
|
||
- intel/fs: don't consider fixup_nomask_control_flow SENDs predicate
|
||
- intel/fs: UNDEF fixup_nomask_control_flow temp register
|
||
- intel/fs: copy instruction sources in logical send lowering
|
||
- intel/fs: factor out lsc surface descriptor settings
|
||
- nir: reuse nir_component_mask() where it makes sense
|
||
- nir: add 2 new intel intrinsics for uniform ssbo/shared loads
|
||
- intel/fs: optimize uniform SSBO & shared loads
|
||
- intel/fs: also allow vec8+ vectorization of load_global_const_block_intel
|
||
- anv: pass steam output as argument for anv_dump_pipe_bits
|
||
- anv: replace query flush before gpu copy by semaphore wait
|
||
- anv: fixup streamout write barriers
|
||
- intel/fs: use nomask for setting cr0 for float controls
|
||
- anv: exclude performance queries from blorp clears
|
||
- intel/ds: add a new timeline row for frames
|
||
- anv: add utrace tracking of frame boundaries
|
||
- vulkan/runtime: discard unused graphics stages in libraries
|
||
- intel/vec4: force exec_all on float control instruction
|
||
- vulkan/overlay: deal with unknown pNext structures
|
||
- isl: don't set inconsistent fields for depth when using stencil only
|
||
- isl: fix a number of errors on storage format support on Gfx9/12.5
|
||
- anv: rework Wa_14017076903 to only apply with occlusion queries
|
||
- nir/divergence: add missing load_global_constant_* intrinsics
|
||
- anv: fix anv_nir_lower_ubo_loads pass
|
||
- intel/fs: fix per vertex input clamping
|
||
- intel/compiler: make uses_pos_offset a tri-state
|
||
- intel/fs: fix scheduling of HALT instructions
|
||
|
||
Liviu Prodea (1):
|
||
|
||
- meson: Ignore unused variables when assertions are disabled
|
||
|
||
Lone_Wolf (3):
|
||
|
||
- compiler/clc: Fix embedded clang headers (microsoft-clc) for LLVM 16+
|
||
- clc: Add clangASTMatchers to fix static llvm build of microsoft-clc with LLVM 16+
|
||
- clc: Add clang frontendhlsl module to fix build of microsoft-clc with llvm 16+
|
||
|
||
Luc Ma (2):
|
||
|
||
- xlib: fix glXDestroyContext in Gallium frontends
|
||
- meson: keep Mako version checking in accord with build msg
|
||
|
||
Lucas Fryzek (11):
|
||
|
||
- crocus: Add support for \`get_screen_fd`
|
||
- tegra: Add support for \`get_screen_fd`
|
||
- nouveau: Add support for \`get_screen_fd`
|
||
- zink: Add support for \`get_screen_fd`
|
||
- iris: Add support for \`get_screen_fd`
|
||
- i915: Add support for \`get_screen_fd`
|
||
- svga: Add support for \`get_screen_fd`
|
||
- virgl: Add support for \`get_screen_fd`
|
||
- r300/r600/radeon_si: Add support for \`get_screen_fd`
|
||
- d3d12/llvmpipe/softpipe: Add support for \`get_screen_fd`
|
||
- gallium: Modify default path for DMABUF to use DRM
|
||
|
||
Lucas Stach (7):
|
||
|
||
- etnaviv: don't drop TS capability on GPUs with MMUv2
|
||
- etnaviv: drm: fix BO array leaks
|
||
- etnaviv: free pm queries dynarray on screen destroy
|
||
- etnaviv: drm: fix check if BO is on a deferred destroy list
|
||
- etnaviv: fix double scanout import of multiplanar resources
|
||
- etnaviv: flush VS texture cache when texture data is changed
|
||
- etnaviv: fix texture barrier implementation
|
||
|
||
Luigi Santivetti (4):
|
||
|
||
- pvr: fix uses_tile_buffers in clear color attachment
|
||
- pvr: add support for tile buffer output clear
|
||
- pvr: add padding bytes when allocating buffer memory
|
||
- pvr: fix segfault in dEQP-VK.ycbcr.query.*
|
||
|
||
Luna Nova (3):
|
||
|
||
- device_select_layer: fix inverted strcmp in device_select_find_dri_prime_tag_default (v1)
|
||
- device_select_layer: apply DRI_PRIME even if default device is > 1 to match opengl behavior
|
||
- device_select_layer: pick a default device before applying DRI_PRIME
|
||
|
||
Lynne (1):
|
||
|
||
- aco_validate: allow for wave32 in p_dual_src_export_gfx11
|
||
|
||
M Henning (1):
|
||
|
||
- nouveau/codegen: Check nir_dest_num_components
|
||
|
||
Maarten Lankhorst (1):
|
||
|
||
- iris: Place scanout buffers only into lmem for discrete GPUs
|
||
|
||
Marcin Ślusarz (23):
|
||
|
||
- intel/compiler: fix generation of vec8/vec16 alu instruction
|
||
- intel/compiler/mesh: handle const data in task & mesh programs
|
||
- intel/compiler: fine-grained control of dispatch widths
|
||
- nir: add nir_mod_analysis & its tests
|
||
- intel/compiler/mesh: optimize indirect writes
|
||
- intel/compiler/mesh: support longer write messages
|
||
- intel/compiler/mesh: remove dead code path supporting >4 dword writes
|
||
- intel/compiler/mesh: use U888X packed index format
|
||
- anv: bump ANV_MAX_QUEUE_FAMILIES
|
||
- intel/compiler: replace gl_Layer & gl_ViewportIndex by 0 in fs if ms doesn't write it
|
||
- anv: fix how unset gl_Viewport & gl_Layer are handled in mesh case
|
||
- intel/compiler/mesh: use slice id of task urb handles in mesh shaders
|
||
- anv: enable task redistribution
|
||
- intel/compiler/mesh: apply URB payload mask once per program
|
||
- intel/compiler/mesh: follow the type of offset variable
|
||
- intel/compiler: remove unused field from fs_thread_payload
|
||
- anv: halve the push constants space in mesh pipelines
|
||
- crocus/meson: add back dependency on libintel_dev
|
||
- anv,hasvk: remove stale comments
|
||
- anv: call nir_shader_gather_info early
|
||
- anv: work around for per-prim attributes corruption
|
||
- intel/compiler: compactify locations of mesh outputs
|
||
- anv: ignore structure types handled in vk_device_memory_create
|
||
|
||
Marek Olšák (212):
|
||
|
||
- glthread: fix an upload buffer leak
|
||
- util: fix util_is_vbo_upload_ratio_too_large
|
||
- mesa: allow GL_UNSIGNED_INT64_ARB as vertex format for ARB_bindless_texture
|
||
- glapi: autogenerate function parameters with no space between * and variable
|
||
- glthread: handle GL_*_ARRAY in glEnable/Disable
|
||
- glthread: set GL_OUT_OF_MEMORY if we fail to upload indices
|
||
- glthread: set GL_OUT_OF_MEMORY if we fail to upload vertices
|
||
- glthread: execute glMultiDrawArrays(draw_count < 0) asynchronously
|
||
- glthread: change multi_draw_elements_async() to never fail due to large size
|
||
- glthread: do vertex uploads if an index buffer is present for glDrawElements
|
||
- mesa: move gl_vertex_format_user definition into glthread.h
|
||
- glthread: pack and name the type of glthread_vao::Attrib
|
||
- glthread: make marshal functions for glBegin/End attribs non-static
|
||
- glthread: remove the vbo_upload_ratio_too_large fallback for glMultiDrawElements
|
||
- glthread: do vertex uploads if an index buffer is present for MultiDrawElements
|
||
- glthread: disallow glthread if buffer uploads are unsupported
|
||
- ac/llvm: run the LLVM sinking pass because LLVM will stop running it
|
||
- ac/llvm: run the IPSCCP pass
|
||
- ac/llvm: remove llvm:: now that we use "using namespace llvm"
|
||
- amd: update amdgpu_drm.h
|
||
- ac/gpu_info: add PCIe info
|
||
- radeonsi/ci: update gfx10.3 results
|
||
- radeonsi/ci: add gfx1100 results
|
||
- radeonsi: fix RB+ blending with sRGB formats
|
||
- radeonsi/gfx11: unset SAMPLE_MASK_TRACKER_WATERMARK to fix hangs
|
||
- amd: split GFX1103 into GFX1103_R1 and GFX1103_R2
|
||
- amd: fix tile_swizzle on gfx11 - should be shifted by 10 bits, not 8
|
||
- amd: update SX_BLEND_OPT_EPSILON.MRT0_EPSILON enum definitions
|
||
- amd: update shadowed register tables for gfx11
|
||
- amd: improve RB+ blending precision
|
||
- radeonsi: implement RB+ depth-only rendering for better perf
|
||
- radeonsi/gfx11: remove the INST_PREF_SIZE workaround
|
||
- radeonsi/gfx11: add a comment why we use PRIM_GRP_SIZE <= 252
|
||
- radeonsi/gfx11: adjust ACCUM_* fields for tessellation
|
||
- radeonsi/gfx11: fix blend->cb_target_mask dependency for shader keys
|
||
- radeonsi/gfx11: move the PIXEL_PIPE_STAT_CONTROL event into the GFX preambles
|
||
- radeonsi/gfx11: use new packet EVENT_WRITE_ZPASS
|
||
- radeonsi: deduplicate VS/TES/GS update code
|
||
- radeonsi/gfx11: always set MSAA_NUM_SAMPLES=0 for DCC_DECOMPRESS
|
||
- radeonsi: merge si_ps_key_update_framebuffer_blend & .._update_blend_rasterizer
|
||
- radeonsi: determine alpha_to_coverage robustly in si_update_framebuffer_blend_rasterizer
|
||
- radeonsi: never set INTERPOLATE_COMP_Z
|
||
- amd: unify and tune the attribute ring size for gfx11
|
||
- amd: change pbb_max_alloc_count for gfx11
|
||
- amd: update the cache size for gfx1103_r1
|
||
- amd: update late_alloc_wave64 for gfx11
|
||
- amd: sort and re-indent packet definitions
|
||
- amd: fix typo in shadowed uconfig registers on gfx11
|
||
- amd: document OOB behavior on gfx11
|
||
- amd/registers: remove confusing definitions from gfx10-rsrc.json
|
||
- radeonsi: set NEVER as the depth compare func if depth compare is disabled
|
||
- amd/llvm: fix LLVM 15 & 16 crashes in SelectionDAG.cpp
|
||
- radeonsi: call ac_init_llvm_once before any util_queue initialization
|
||
- radeonsi: set sampler COMPAT_MODE in the corresponding branch
|
||
- amd/ci: update sanctuary trace sha1
|
||
- radeonsi/gfx11: don't add mrt0 export for alpha-to-coverage if mrtz is present
|
||
- radeonsi/gfx11: don't add alpha to mrt0 format for A2C if exporting via mrtz
|
||
- amd: define new SET_*_REG_PAIRS packets
|
||
- radeonsi: clean up si_set_mutable_tex_desc_fields
|
||
- amd/surface: clean up is_dcc_supported_by_L2
|
||
- amd,util: fix how lod bias is converted to fixed-point
|
||
- amd: don't hardcode real VGPR allocation granularity on gfx10.3 and gfx11
|
||
- glthread: track the current element array buffer in the Core profile too
|
||
- mesa: ignore indices[i] if count[i] == 0 for MultiDrawElements
|
||
- glthread: initialize indices[i] for no-op MultiDrawElements
|
||
- glthread: upload non-BO indices in the core profile to fix GStreamer
|
||
- glthread: add a heuristic to stop locking global mutexes with multiple contexts
|
||
- glthread: ignore non-VBO vertex arrays with NULL data pointers
|
||
- Revert "ci/zink: Disable Amnesia trace until the linked issue gets fixed."
|
||
- glthread: rewrite glMultiDrawArrays to never fail to upload vertices
|
||
- glthread: change glMultiDrawElements to execute draw_count < 0 asynchronously
|
||
- glthread: don't execute glDraw code if we're inside glBegin/End
|
||
- glthread: don't pass index bounds to the driver for async calls
|
||
- glthread: move some draw call parameters closer to their use
|
||
- glthread: don't bind/unbind uploaded indexbuf, pass it to glDraw directly
|
||
- glthread: don't bind/unbind uploaded indexbuf, pass it to glMultiDraw directly
|
||
- glthread: track vertex formats for all attributes
|
||
- glthread: add a vertex upload path that unrolls indices for glDrawElements
|
||
- glthread: reorder draw code a little
|
||
- glthread: add ctx->GLThread.draw_always_async to simplify draw checking
|
||
- glthread: remove goto statements and add unlikely() into draw functions
|
||
- glthread: inline draw functions that have only one use
|
||
- glthread: don't execute Draw and BufferSubData calls if the context is lost
|
||
- glthread: handle non-VBO uploads for glMultiModeDraw{Arrays,Elements}IBM
|
||
- glthread: add API to allow passing DrawID from glthread to mesa
|
||
- glthread: convert (Multi)DrawIndirect into direct if user buffers are present
|
||
- glthread: remove unnecessary debug code
|
||
- glthread: don't free glthread for GL_DEBUG_OUTPUT_SYNCHRONOUS, only disable it
|
||
- glthread: don't restore non-VBO vertex arrays after all draws
|
||
- Revert "radeonsi/ci: Update stoney test expectations"
|
||
- radeonsi: fix COMPAT_MODE on gfx8-9
|
||
- amd: fix LOD_BIAS on gfx6-9 and adjust the lod bias CAP
|
||
- amd: add missing gfx11 register definitions
|
||
- amd: bump AMD_MAX_SE and change the CU mask type to 16 bits
|
||
- radeonsi/gfx11: fix the CU_EN clear mask for RSRC4_GS
|
||
- radeonsi/gfx11: don't set non-existent VGT_STRMOUT_BUFFER_CONFIG
|
||
- radeonsi/gfx11: set CB_COLORi_INFO.MAX_COMP_FRAG on GFX1103_R2
|
||
- radeonsi: move a few DB_SHADER_CONTROL states into si_shader_ps
|
||
- radeonsi: change si_shader::ctx_reg to a nameless union for better readability
|
||
- radeonsi: remove no-op setting of THDS_PER_SUBGRP
|
||
- radeonsi: use SPI_SHADER_USER_DATA_HS_0 definition instead of LS_0
|
||
- radeonsi: set PA_SU_VTX_CNTL consecutively with PA_CL_GB_VERT_CLIP_ADJ
|
||
- radeonsi/gfx11: ignore alpha_is_on_msb because the hw ignores it
|
||
- radeonsi: replace si_screen::has_out_of_order_rast with the radeon_info field
|
||
- radeonsi: disable Smart Access Memory because CPU access has large overhead
|
||
- amd,radeonsi: remove unused LLVM functions
|
||
- amd/registers: unify VRS combiner definition names between gfx103 and gfx11
|
||
- amd: replace SI_BIG_ENDIAN with UTIL_ARCH_BIG_ENDIAN
|
||
- radeonsi: remove returns from si_emit_global_shader_pointers
|
||
- radeonsi: reformat emit_cb_render_state, create_blend_state, create_rs_state
|
||
- radeonsi: remove a gfx11 check in si_shader_gs (legacy GS)
|
||
- radeonsi: remove unused VS_STATE_LS_OUT_PATCH_SIZE
|
||
- radeonsi: always add 1 to lshs_vertex_stride now that LS_OUT_PATCH_SIZE is gone
|
||
- radeonsi: correct and clean up obsolete vs_state_bits comments
|
||
- radeonsi: rename esgs_itemsize -> esgs_vertex_stride
|
||
- amd: query the per-SIMD VGPR counts from the kernel, don't hardcode them
|
||
- radeonsi: don't clamp z_samples to fix Unreal Tournament 99
|
||
- amd/registers: only define SPI and COMPUTE registers in the 0xB000 range
|
||
- radeonsi: reorganize emit_db_render_state and simplify VRS code
|
||
- radeonsi: reorganize si_initialize_color_surface for better readability
|
||
- radeonsi: reorganize si_init_depth_surface for better readability
|
||
- radeonsi: don't set PACKET_TO_ONE_PA for line stippling
|
||
- radeonsi/gfx11: change the default of COMPUTE_DISPATCH_INTERLEAVE to 256
|
||
- amd: implement conformant TRUNC_COORD behavior for gfx11
|
||
- amd/gpu_info: add a workaround for SI_FORCE_FAMILY=gfx1100
|
||
- nir,amd: add and use nir_intrinsic_load_esgs_vertex_stride_amd
|
||
- nir: lower to fragment_mask_fetch/load_amd with EQAA correctly
|
||
- glthread: fix a perf regression due to draw_always_async flag, fix DrawIndirect
|
||
- mesa: fix glPopClientAttrib with fixed-func VP and zero-stride varyings
|
||
- mesa: remove a redundant call to _mesa_update_edgeflag_state_vao
|
||
- mesa: initialize VertexProgram._VaryingInputs before the first use
|
||
- amd: update amdgpu_drm.h
|
||
- amd,radeonsi: change enabled_rb_mask to 64 bits
|
||
- amd: query cache sizes from the kernel
|
||
- ac/nir: don't use load_esgs_vertex_stride_amd on gfx6-8
|
||
- amd: massively simplify how info->spi_cu_en is applied
|
||
- amd/rtld: allow 64K LDS for all shader stages except for gfx6
|
||
- radeonsi/ci: update flakes and gfx8-polaris11 results
|
||
- radeonsi: remove Smart Access Memory because CPU access has large overhead
|
||
- radeonsi: reorganize si_emit_framebuffer_state for better readability
|
||
- radeonsi: don't merge SET_* packets that have a different index in si_pm4_state
|
||
- radeonsi: reindent code in si_state_binning.c
|
||
- radeonsi: add si_pm4_set_reg_va to simplify setting reg_va_low_idx for RGP
|
||
- radeonsi: check the pm4.reg_va_low_idx assertion unconditionally
|
||
- radeonsi: simplify encoding VGPRS and SGPRS
|
||
- radeonsi: assume shader is never NULL in si_emit_shader_*
|
||
- nir: return progress from nir_lower_io_to_scalar
|
||
- nir: skip nir_op_unpack_32_4x8 in nir_lower_alu_width
|
||
- ac/nir: add ac_nir_lower_subdword_loads to lower 8/16-bit loads to 32 bits
|
||
- aco: implement nir_op_unpack_32_4x8
|
||
- ac/llvm: implement nir_op_unpack_32_4x8
|
||
- amd: lower subdword UBO loads in NIR
|
||
- amd: lower multi-component subdword SSBO loads in NIR
|
||
- lavapipe/ci: add a new flake
|
||
- amd: add nir_intrinsic_xfb_counter_sub_amd and fix overflowed streamout offsets
|
||
- amd/llvm,radeonsi/gfx11: switch to using GDS_STRMOUT registers
|
||
- radeonsi/gfx11: only allocate GDS OA for streamout, GDS memory is not needed
|
||
- radeonsi: emulate VGT_ESGS_RING_ITEMSIZE in the shader on gfx9-11
|
||
- radeonsi: merge si_emit_initial_compute_regs with si_init_cs_preamble_state
|
||
- radeonsi: separate nir_texop_descriptor_amd lowering
|
||
- radeonsi: lower nir_texop_sampler_descriptor_amd
|
||
- radeonsi: set pm4.atom.emit in si_get_shader_pm4_state
|
||
- radeonsi: reindent si_shader_ls, si_shader_es, si_shader_gs, si_shader_vs
|
||
- radeonsi: reorganize si_shader_hs
|
||
- radeonsi: reorganize si_shader_ngg
|
||
- radeonsi: reorganize si_shader_ps
|
||
- radeonsi: other cosmetic changes in si_state_shaders.cpp
|
||
- radeonsi: allow using 64K LDS for NGG to allow larger workgroups
|
||
- radeonsi: increase NGG workgroup size to 256 for VS/TES with streamout and GS
|
||
- glapi: move files specific to shared-glapi into the shared-glapi subdirectory
|
||
- glapi: inline the meson list files_mapi_util
|
||
- mesa: move ctx->Table -> ctx->Dispatch.Table except Client & MarshalExec
|
||
- mesa: rename CurrentClientDispatch to GLApi
|
||
- mesa: put dispatch table initialization into one place
|
||
- glthread: qualify the \*cmd unmarshal parameter with restrict
|
||
- vbo: fix current attribs not updating gallium vertex elements
|
||
- radeonsi: remove unused TCS/TES SGPR fields
|
||
- radeonsi: dump shader stats only if dumping asm shaders
|
||
- radeonsi: replace nonir,noir,noasm,preoptir options with new reworked options
|
||
- radeonsi: remove duplicated gfx11 check in si_msaa_resolve_blit_via_CB
|
||
- radeonsi: rework MSAA resolve averaging to exploit instruction-level parallelism
|
||
- radeonsi: add AMD_DEBUG=nowcstream to enable caching for stream_uploader
|
||
- radeonsi: don't print the base non-view texture format for AMD_TEST=computeblit
|
||
- radeonsi: fix AMD_TEST=computeblit being rejected on gfx < 11
|
||
- radeonsi: don't convert to fp16 in the compute blit if not testing
|
||
- radeonsi: don't use fp16_rtz for FP formats in the compute blit
|
||
- radeonsi: correct an assertion if we get a display list with no vertex buffers
|
||
- ac/nir: don't emit duplicated parameter exports
|
||
- ac/nir: use plural correctly in the ac_nir_export_parameters name
|
||
- radeonsi: remove unused vs_output_param_mask
|
||
- egl: reorder code in _eglQueryDevicesEXT, add \*swrast variable
|
||
- egl: don't expose swrast device if swrast is not built
|
||
- amd/llvm: fix handling of unsupported vec3 loads on gfx6
|
||
- amd/llvm: remove no-op code for vec3 loads in ac_build_tbuffer_load
|
||
- amd: update addrlib
|
||
- amd: rename GFX1036 -> RAPHAEL_MENDOCINO
|
||
- amd: set the correct LLVM processor name for gfx1036
|
||
- radeonsi/gfx11: reduce MSAA samples to 8 for no-attachment framebuffer
|
||
- radeonsi: simplify binning settings to work around GPU hangs
|
||
- amd: add gfx940 register definitions
|
||
- amd: add initial code for gfx940
|
||
- radeonsi: use COMPUTE_DISPATCH_SCRATCH_BASE on gfx940
|
||
- radeonsi: always use ffma32 on gfx940
|
||
- ac/surface: force linear image layout for chips not supporting image opcodes
|
||
- radeonsi: add an emulated image descriptor for gfx940
|
||
- ac/nir: implement image opcode emulation for CDNA, enable it in radeonsi
|
||
- radeonsi: don't set registers that don't exist on gfx940
|
||
- amd/registers: simplify integer division by 0x1000 in the parser
|
||
- amd/registers: fix the parser to include CP_COHER registers for gfx940
|
||
- amd/registers: update gfx940.json
|
||
- amd/registers: use gfx9 packet definitions for gfx940
|
||
- nir: fix 2 bugs in nir_create_passthrough_tcs
|
||
|
||
Mario Kleiner (1):
|
||
|
||
- v3dv: Enable (leased) direct display extensions.
|
||
|
||
Mark Collins (4):
|
||
|
||
- meson: update flex/bison requirement to cover all usages
|
||
- meson: forcefully disable libdrm when host doesn't have it
|
||
- tu: KGSL backend rewrite
|
||
- tu: fix tu_GetInstanceProcAddr not handling null instance
|
||
|
||
Mark Janes (13):
|
||
|
||
- intel: Implement Wa_16011448509
|
||
- util: add macro to support gcc/clang poison
|
||
- intel/dev: generate helpers to identify platform workarounds
|
||
- intel/dev: Print required workarounds with intel_dev_info
|
||
- intel/fs: use generated workaround helpers for Wa_14010017096
|
||
- intel/fs: use generated helpers for Wa_1209978020 / Wa_18012201914
|
||
- intel/fs: use generated workaround helpers for Wa_14017989577
|
||
- intel: use generated workaround helpers for Wa_1409600907
|
||
- intel: use generated helpers for Wa_1409433168/Wa_16011107343
|
||
- intel/fs: use generated helpers for Wa_14013363432 / Wa_14012688258
|
||
- intel/dev: fix macro string concatenation for INTEL_WA_{id}_GFX_VER
|
||
- intel/dev: fix macro naming convention in gen_wa_helpers.py
|
||
- intel/dev: use GFX_VERx10 to detect genX compilation
|
||
|
||
Martin Roukala (né Peres) (22):
|
||
|
||
- ci/deqp-runner: compress results.csv before uploading it to GitLab
|
||
- ci/piglit: compress results.csv before uploading it to GitLab
|
||
- zink/ci/radv: remove a test from the fails list
|
||
- zink/ci: add a fail to the VG flake list
|
||
- zink/ci: relocate radv testing from radv's gitlab-ci.yml
|
||
- zink/ci: add spec\@!opengl 1.1\@line-smooth-stipple to the fails list
|
||
- ci/b2c: uprev to b2c v0.9.9
|
||
- ci/debian/x86_test-vk: drop an outdated dependency
|
||
- ci/core-manual-rules: enclose the whole condition in quotes
|
||
- zink/ci: allow running manual jobs again on RADV
|
||
- ci/init-stage2: allow sourcing the job env vars from the CWD
|
||
- ci/init-stage2: always set XDG_RUNTIME_DIR
|
||
- ci/b2c: move away from the hand-rolled initscript
|
||
- ci: bring back the valve farm online
|
||
- ci/valve-farm-rules: allow running jobs from outside the mesa namespace
|
||
- radv/ci: reduce the parallelism for vkcts-vangogh
|
||
- zink/ci: increase the parallelism of zink-radv-vangogh-valve
|
||
- zink/ci: update the radv expectations
|
||
- radv/ci: update VanGogh's expectations
|
||
- ci/b2c: increase the console timeout to 4 minutes
|
||
- radv/ci: update the navi10 expectations
|
||
- zink/ci: add a test to the fails list
|
||
|
||
Matt Coster (6):
|
||
|
||
- pvr: Extract setup of winsys job submit flags into separate functions
|
||
- pvr: Add support for geometry-only render jobs
|
||
- pvr: Add pvr_csb_bake()
|
||
- pvr: Rename global_queue_job_count to global_cmd_buffer_submit_count
|
||
- pvr: Split render job submission for multi-layer framebuffers
|
||
- pvr: Add firmware stream support for transfer submit
|
||
|
||
Matthieu Bouron (1):
|
||
|
||
- lavapipe: honor dst base array layer when resolving color attachments
|
||
|
||
Mauro Rossi (1):
|
||
|
||
- hasvk: include "vk_android.h" header in anv_android.c
|
||
|
||
Maíra Canal (1):
|
||
|
||
- v3dv: remove unused clamp_to_transparent_black_border property
|
||
|
||
Michel Dänzer (29):
|
||
|
||
- mesa/st: Fix GL_EXT_texture_type_2_10_10_10_REV name in comment
|
||
- mesa/st: Handle all 10 bpc types in st_choose_format
|
||
- glsl/standalone: Fix up _mesa_reference_shader_program_data signature
|
||
- glsl/standalone: Do not pass memory allocated with ralloc_size to free
|
||
- anv/grl: Use union for reinterpreting integer as float
|
||
- clover: Reserve vector memory in make_text_section
|
||
- ci: Update Fedora image to 36
|
||
- ci: Re-enable intel-clc in fedora-release job
|
||
- ci: Enable i915 Gallium driver in fedora-release job
|
||
- ci: Enable the hasvk Vulkan driver in the fedora-release job
|
||
- frontend/dri: Initialize callbacks in dri_swrast_kms_init_screen
|
||
- nouveau: Make getSize return unsigned int
|
||
- r600: Use container_of instead of direct pointer cast
|
||
- crocus: Use ralloc_free for memory allocated with rzalloc
|
||
- iris: Use ralloc_free for memory allocated with rzalloc
|
||
- ci: Remove some -Werror workarounds for debian-android job
|
||
- ci: Split up -Werror workarounds for debian-mingw32-x86_64 job
|
||
- intel/vk/grl: Do not use no_override_init_args for C++
|
||
- ci: Pass -Werror to compiler linking stage for LTO
|
||
- ci: Allow passing c{,pp}_link_args to meson
|
||
- ci: Make ccache optional
|
||
- ci: Drop ccache from Fedora image
|
||
- ci: Install procps-ng in Fedora image
|
||
- ci: Enable LTO for fedora-release job
|
||
- vulkan: Fix GetPhysicalDeviceSparseImageFormatProperties definitions
|
||
- svga: Make vmw_svga_winsys_buffer_map definition match declaration
|
||
- svga: Make declaration of emit_input_declaration match definition
|
||
- clover/llvm: Use llvm::DataLayout::getABITypeAlign with LLVM >= 16
|
||
- clover/llvm: Use std::nullopt already with LLVM 16
|
||
|
||
Michel Zou (4):
|
||
|
||
- ci/mingw: drop useless -Wno-error flags
|
||
- vulkan/wsi: fix -Wnarrowing warning
|
||
- vk/entry_points:: fix mingw build
|
||
- mesa/draw: fix -Wformat warning
|
||
|
||
Mike Blumenkrantz (536):
|
||
|
||
- zink: simplify get_slot_components() for xfb emission
|
||
- zink: add renderdoc handling
|
||
- zink: prune old swapchains on present
|
||
- zink: break out implicit feedback loop detection into separate function
|
||
- zink: set textures_used in analyze_io
|
||
- zink: outdent code in add_implicit_color_feedback_loop()
|
||
- zink: make implicit feedback loop application stricter
|
||
- zink: skip implicit feedback loop layout changes if feedback loop not present
|
||
- zink: store drm format as internal_format for imported resources
|
||
- zink: handle modifier nplanes queries correctly for planar formats
|
||
- zink: NV_compute_shader_derivatives
|
||
- zink: preserve present resources during async presentation
|
||
- zink: add a util function for creating semaphores
|
||
- zink: add a binary semaphore cache
|
||
- zink: move semaphore caching to zink_reset_batch_state()
|
||
- zink: consolidate semaphore creation where possible
|
||
- zink: simplify some dynarray concat descriptor code
|
||
- zink: delete need_blend_constants
|
||
- zink: don't use ds3 blend states without color attachments
|
||
- radv: repack radv_graphics_pipeline struct
|
||
- radv: reorder dynamic state checks during bind
|
||
- radv: simplify depth aspect check in radv_handle_image_transition()
|
||
- radv: add some graphics pipeline hints to optimize pipeline bind
|
||
- radv: remove redundant type sizing
|
||
- radv: add an early out in radv_cmd_buffer_flush_dynamic_state()
|
||
- zink: use actual swapchain object for surface comparison
|
||
- radv: stop using radv_pipeline_has_stage() in BindPipeline
|
||
- zink: flag old-style shadow tex mask for fragment shaders
|
||
- zink: break out tex dest rewriting into separate function
|
||
- zink: add an extra_data param to zink_shader_compile
|
||
- zink: track depth swizzle on samplerviews
|
||
- zink: add a fs shader key member to indicate depth texturing mode
|
||
- zink: rework depth sampler splatting in shaders
|
||
- zink: block pipeline fast-pathing for any programs using depth texture modes
|
||
- zink: plug in the program/module parts of shadow texture mode emulation
|
||
- zink: create another samplerview for shadow textures
|
||
- zink: remove old depth swizzle workaround
|
||
- zink: pass depth swizzle data block to shader compile
|
||
- mesa: remove dead parameter doc for _mesa_new_texture_object()
|
||
- mesa: populate gl_program::ShadowSamplers mask from shader data
|
||
- mesa: (more) correctly handle incomplete depth textures
|
||
- zink: fix implicit feedback loop detection
|
||
- radv: Move constant flushing check out to callers.
|
||
- zink: fix VK_DYNAMIC_STATE_LINE_WIDTH usage
|
||
- zink: move barrier jit to zink_context.c
|
||
- zink: don't skip repeated handling feedback loops
|
||
- zink: return false for implicit feedback loop check with image binds
|
||
- zink: update sampler layout when detecting feedback loop for first time
|
||
- zink: force GENERAL layout for all fb attachments with image binds
|
||
- zink: validation ci updates
|
||
- zink: reorder commands more aggressively
|
||
- Revert "zink: allow direct memory mapping for any COHERENT+CACHED buffer"
|
||
- zink: fix heap/memory type selection
|
||
- zink: add VK_BUFFER_USAGE_SHADER_DEVICE_ADDRESS_BIT for buffers if ext is enabled
|
||
- zink: set vkusage/vkflags for buffer resource objects
|
||
- zink: assert that buffer descriptor usage is populated before bind
|
||
- zink: always set RESOURCE usage for descriptor buffers
|
||
- zink: disable bindless texture ext with descriptor buffer
|
||
- zink: set VK_PIPELINE_CREATE_DESCRIPTOR_BUFFER_BIT_EXT when using DB
|
||
- zink: don't add dmabuf export type if dmabuf isn't supported
|
||
- zink: stop leaking push descriptor templates
|
||
- zink: always unref old images when adding new binds
|
||
- zink: hook up VK_EXT_multisampled_render_to_single_sampled
|
||
- zink: shrink zink_render_pass_state::msaa_expand_mask
|
||
- zink: use VK_EXT_multisampled_render_to_single_sampled for EXT_multisample_render_to_texture
|
||
- lavapipe: move noop fs creation to device
|
||
- lavapipe: add refcounting for shader nir
|
||
- lavapipe: refcount nir shaders instead of cloning
|
||
- lavapipe: break out (and slightly refactor) gallium shader cso creation
|
||
- lavapipe: create gfx gallium csos at pipeline bind
|
||
- lavapipe: delete unused pipelines immediately
|
||
- lavapipe: delete lvp_pipeline::mem_ctx
|
||
- lavapipe: try harder to reuse pipeline layouts during merge
|
||
- zink: only set VkPipelineColorBlendStateCreateInfo::attachmentCount without full ds3
|
||
- zink: fix zink_mem_type_idx_from_bits()
|
||
- zink: rework descriptor buffer templating to use offsets
|
||
- Revert "zink: fix zink_mem_type_idx_from_bits()"
|
||
- zink: enable PIPE_CAP_ALLOW_GLTHREAD_BUFFER_SUBDATA_OPT
|
||
- zink: make bindless buffer_infos a union
|
||
- zink: fix bindless struct member comments
|
||
- zink: skip updating descriptor buffer sets that aren't active
|
||
- zink: set VK_PIPELINE_CREATE_DESCRIPTOR_BUFFER_BIT_EXT on compute pipelines
|
||
- zink: break out descriptor binding into separate function
|
||
- zink: add a flag to indicate whether a descriptor buffer is bound
|
||
- zink: implement descriptor buffer handling of bindless texture
|
||
- zink: enable bindless texture with ZINK_DESCRIPTORS=db
|
||
- zink: free descriptor buffer maps on batch state destroy
|
||
- zink: fix more cases of heap/memtype suballocator mismatch
|
||
- zink: cache and reuse dummy inputattachment for fbfetch
|
||
- zink: handle missing line rasterization modes with ds3
|
||
- zink: add back VK_DESCRIPTOR_BINDING_PARTIALLY_BOUND_BIT for bindless
|
||
- gallium: add PIPE_CAP_NULL_TEXTURES
|
||
- radeonsi: set PIPE_CAP_NULL_TEXTURES
|
||
- zink: conditionally enable PIPE_CAP_NULL_TEXTURES
|
||
- zink: fix max acquired image count
|
||
- lavapipe: disable VK_FORMAT_FEATURE_2_COLOR_ATTACHMENT_BLEND_BIT for int formats
|
||
- zink: set PIPE_CAP_VALIDATE_ALL_DIRTY_STATES
|
||
- zink: move bindless_layout to screen and init on creation
|
||
- zink: take screen param in init_db_template_entry()
|
||
- zink: const-ify a bunch of shader key inlines
|
||
- zink: move gpl usability checks to static inline for reuse
|
||
- zink: remove duplicated gpl output blend initializations
|
||
- zink: store last pipeline directly for zink_gfx_program::last_pipeline
|
||
- zink: pass screen to descriptor_util_pool_key_get()
|
||
- zink: delete zink_screen::framebuffer_cache
|
||
- zink: make last_vertex_stage the first bit in zink_vs_key_base
|
||
- zink: ralloc zink_shader structs
|
||
- zink: add a define for the "default" optimal key
|
||
- zink: add a define for testing that an optimal key is the default
|
||
- zink: add VK_PIPELINE_CREATE_DESCRIPTOR_BUFFER_BIT_EXT for gpl libs
|
||
- zink: don't set blend_id with full_ds3
|
||
- zink: set gfx feedback loop bit in pipeline state for driver workaround
|
||
- zink: set zs feedback loop bit from driver workaround on ctx create
|
||
- zink: fix gpl lib hashing
|
||
- zink: use screen indexing for bindless descriptor set in db bind
|
||
- zink: use screen indexing for bindless descriptor set in template bind
|
||
- util/vbuf: fix multidraw unrolling
|
||
- zink: flag bindless_init before calling zink_batch_bind_db() in init
|
||
- zink: avoid the descriptor set multiplier for bindless buffers
|
||
- zink: split out VkShaderModule creation
|
||
- zink: add flags param to zink_pipeline_layout_create()
|
||
- zink: split out gfx pipeline library creation
|
||
- zink: add gpl flags for libraries based on shaders passed
|
||
- zink: allow multiple gpl libraries in zink_create_gfx_pipeline_combined()
|
||
- zink: move gpl input/output funcs to zink_pipeline.c
|
||
- zink: enable combining intermediate gpl libs from combine function
|
||
- zink: use GPL to handle (simple) separate shader objects
|
||
- zink: set PIPE_CAP_SURFACE_REINTERPRET_BLOCKS
|
||
- zink: store gfx_hash on zink_gfx_program
|
||
- zink: break out zink_gfx_program::libs into refcounted object
|
||
- zink: implement cross-program pipeline library sharing
|
||
- zink: add newlines to some debug printfs
|
||
- zink: rename some variables in zink_set_shader_images()
|
||
- zink: unset gfx shader read when unbinding shader images
|
||
- zink: remove stale comment
|
||
- zink: unref image buffer descriptors on unbind
|
||
- zink: rework set_shader_images() hook
|
||
- zink: pull out image descriptor updating in set_shader_images
|
||
- zink: add a local is_compute var for set_shader_images
|
||
- zink: allocate all batch command buffers in one call
|
||
- zink: sync LTO compiles for GPL pipelines on shader free
|
||
- zink: fix descriptor pool free iterating
|
||
- zink: don't fetch/update pipeline cache for separate shader programs
|
||
- zink: assert that the found program matches the expected one in shader_free
|
||
- zink: flag gfx programs as removed-from-cache by default
|
||
- zink: rework separate shader descriptor iterating
|
||
- zink: use a single descriptor buffer for all non-bindless types
|
||
- zink: add an io assignment pass for separate shaders
|
||
- zink: rename a struct member for clarity
|
||
- zink: move zink_batch_state::db_bound reset to zink_batch_descriptor_reset()
|
||
- zink: move db_bound to batch descriptor data
|
||
- zink: ensure db is bound before separate shader update
|
||
- zink: store base descriptor size on the screen
|
||
- zink: implement a scaling descriptor buffer size
|
||
- zink: calloc separable program zink_gfx_library_key struct
|
||
- zink: also replace hash_entry::key when replacing separable program
|
||
- zink: always use NEAREST for zs blits
|
||
- zink: fix indentation of rebind_image()
|
||
- zink: only try for a fb rebind if fb binds exist in rebind_image()
|
||
- zink: account for null surface when trying to retain clears on fb bind
|
||
- zink: break out pipe_surface init for new surface creation
|
||
- zink: const-ify a surface param
|
||
- zink: don't handle mutable init on surface creation with tc enabled
|
||
- zink: verify compressed format layer count when creating surfaces
|
||
- zink: set ZINK_DESCRIPTORS=db for radv jobs
|
||
- zink: enable renderpass optimizations by default for selected drivers
|
||
- Revert "zink: always use NEAREST for zs blits"
|
||
- zink: block LINEAR filtered blits for zs formats
|
||
- lavapipe: enable linear filtering for depth formats
|
||
- gallium: plumb resolve attachments through from frontends -> pipe_framebuffer_state
|
||
- dri3: avoid deadlocking when polling deleted windows for events
|
||
- vulkan/wsi: avoid deadlocking dri3 when polling deleted windows for events
|
||
- vl/dri3: avoid deadlocking when polling deleted windows for events
|
||
- zink: delete some now-broken ntv dref sampling code
|
||
- zink: more accurately handle i/o for separate shaders
|
||
- zink: handle semi-matching i/o for separate shaders
|
||
- zink: enable renderpass optimizing on lavapipe
|
||
- kopper: fix loop iterating for msaa texture creation
|
||
- zink: move db input attachment size check to screen init
|
||
- zink: make ZINK_DESCRIPTOR_MODE=db the default
|
||
- driconf: add zink glthread disable for a game
|
||
- aux/tc: add a 'has_resolve' member to tc_renderpass_info
|
||
- zink: actually hook up ZINK_DEBUG=norp
|
||
- zink: add ZINK_DEBUG=map
|
||
- zink: add debug marker tracing for qbo updates
|
||
- util/box: add intersection test functions for 1d/3d
|
||
- zink: add some tracking for copy box regions
|
||
- zink: add a util function for optimizing TRANSFER_DST image barriers
|
||
- zink: utilize copy box tracking to avoid barrier emission for buf2img copies
|
||
- zink: fix slab allocator sizing
|
||
- zink: delete dead uniform variables
|
||
- zink: fix shader read access removal for barrier generation
|
||
- zink: rework descriptor unbind params to use is_compute directly
|
||
- zink: fix bindless texture barrier generation
|
||
- zink: delete unused emit_image param in ntv
|
||
- zink: simplify/rework image typing in ntv
|
||
- zink: avoid adding ubo/ssbo bindings multiple times for different bitsizes
|
||
- zink: add locking for zink_screen::copy_context and defer creation
|
||
- zink: allow direct memory mapping for any COHERENT+CACHED buffer
|
||
- lavapipe: EXT_image_sliced_view_of_3d
|
||
- vulkan/wsi: switch to using an options struct for last param
|
||
- vulkan/wsi/x11: make 4 image minimum for xwayland driver-specific
|
||
- docs: add pipeline library support for tu
|
||
- aux/tc: track whether queries have been terminated in a renderpass
|
||
- aux/tc: only call tc_parse_draw() when parsing renderpass info
|
||
- zink: move zink_batch_no_rp call for query reset
|
||
- zink: remove suspended queries from list before resuming
|
||
- zink: reset queries on the promoted cmdbuf when possible
|
||
- zink: rewrite zink_query_start struct
|
||
- zink: fix zink_query_start initialization
|
||
- zink: fix possible query destroy leak
|
||
- zink: make zink_vk_query unref consistent
|
||
- zink: delete zink_query::last_start_idx
|
||
- zink: handle multiple query starts in qbo update
|
||
- zink: don't auto-sync qbos on query end
|
||
- zink: add zink_query::suspended to indicate suspended state of queries
|
||
- zink: un-suspend queries if they end while suspended
|
||
- zink: don't double suspend queries
|
||
- zink: refuse to start cs invocation queries in renderpass
|
||
- zink: only try doing qbo updates on query suspend if !in_rp
|
||
- zink: handle null query results
|
||
- zink: handle null query results for conditional render
|
||
- zink: only update qbo for TIME_ELAPSED on start if !in_rp
|
||
- zink: try updating qbos on query resume if !in_rp
|
||
- zink: reorder some query code
|
||
- zink: rework find_or_allocate_qp()
|
||
- zink: create/use query pools dynamically
|
||
- zink: rework query pool overflow
|
||
- zink: track whether queries were started in a renderpass
|
||
- zink: break out query suspend functionality for reuse
|
||
- zink: use more consistent check for deleting zink_query::stats_list links
|
||
- zink: pull 'was_line_loop' into ctx for query updating
|
||
- zink: always start/stop/resume queries inside renderpasses
|
||
- zink: use tc renderpass optimizing to more optimally start queries
|
||
- zink: skip buffer barriers for ACCESS_NONE -> ACCESS_READ / ACCESS_WRITE
|
||
- zink: disable queries for clear_texture()
|
||
- zink: resume queries after conditional render and clears are processed
|
||
- zink: only resume queries inside renderpasses from set_active_query_state
|
||
- zink: track whether a primgen query is suspended and needing color write hacks
|
||
- zink: rework xfb queries for drivers with poor primgen support
|
||
- zink: merge qbo update copies when possible
|
||
- zink: set predicate_dirty on query creation
|
||
- zink: eliminate internal qbo copy barrier
|
||
- util/box: fix off-by-one calc error in intersection funcs
|
||
- zink: avoid ballooning of copy box tracking
|
||
- zink: add perfetto tracing for barriers
|
||
- zink: avoid unnecessary read-only layout changes for zs attachments
|
||
- zink: skip attachment barrier for redundant layout-setting if !valid
|
||
- zink: add functions for faster batch-usage completion checks
|
||
- zink: fix zink_resource_access_is_write()
|
||
- zink: use split image barriers if the image can be easily proved idle
|
||
- zink: skip buffer barriers if the buffer can be easily proved idle
|
||
- zink: determine whether debug markers are used on screen create
|
||
- zink: be more descriptive with perfetto buffer barriers
|
||
- zink: reorder some blit debug markers
|
||
- zink: pass cmdbuf to debug marker begin
|
||
- zink: remove debug markers for u_blitter ops
|
||
- zink: when skipping a TRANSFER_DST image barrier, set access tracking
|
||
- zink: add frame trace markers on the queue
|
||
- zink: unbind fb on context destroy
|
||
- zink: only add deferred barrier on fb unbind when layout needs to change
|
||
- zink: fix descriptor update flagging on null ssbo set
|
||
- zink: propagate valid_buffer_range when replacing buffer storage
|
||
- zink: check for layout updates when unbinding samplerviews
|
||
- zink: eliminate pre barrier for adding resource binds
|
||
- zink: don't unset existing access when adding resource binds
|
||
- zink: explicitly flush src clears when u_blittering
|
||
- zink: always set color writes on the unordered cmdbuf
|
||
- zink: bind descriptor buffers to unordered cmdbuf
|
||
- zink: set dynamic pcp for unordered cmdbuf
|
||
- zink: always set batch usage for descriptors after barrier
|
||
- llvmpipe: fix LP_PERF=no_depth to ignore depth format
|
||
- zink: track the last write access for resources
|
||
- zink: add a mechanism for managing TRANSFER_DST buffer barriers
|
||
- zink: add a mechanism to trigger copy box resets from batch state reset
|
||
- zink: add a driver workaround to disable copy box optimizations
|
||
- zink: hook up buffer TRANSFER_DST barrier optimizing
|
||
- aux/tc: fix rp info resizing clobbering current info
|
||
- vulkan/wsi: fix crash in failed swapchain creation for wayland
|
||
- lavapipe: split out spirv compile of shaders
|
||
- lavapipe: split out shader struct members into their own struct
|
||
- lavapipe: pass shader struct and layout to scan_pipeline_info()
|
||
- lavapipe: more small shader struct usage tweaks
|
||
- lavapipe: move xfb init to shader struct
|
||
- lavapipe: rename inline uniform function params
|
||
- lavapipe: move uniform inline functions to shader struct
|
||
- lavapipe: break out main shader lowering into separate function
|
||
- Revert "Revert "ci: disable mesa-swrast runner jobs""
|
||
- zink: ignore renderdoc if ZINK_RENDERDOC isn't in use
|
||
- radv: delete radv_graphics_pipeline_compile() asserts
|
||
- radv: avoid a huge memset in radv_graphics_pipeline_compile()
|
||
- aux/tc: use renderpass tracking to optimize texture_subdata calls
|
||
- lavapipe: fix dynamic depth clamping
|
||
- lavapipe: set render_condition_enabled=false for vkCmdClearDepthStencilImage
|
||
- lavapipe: add command debugging
|
||
- lavapipe: beef up LVP_POISON_MEMORY
|
||
- ci: fix LVP_POISON_MEMORY usage
|
||
- zink: rework zink_resource::valid_buffer_range
|
||
- zink: return the unordered state from zink_resource_buffer_transfer_dst_barrier()
|
||
- zink: unify image TRANSFER_DST barrier checks
|
||
- zink: rename zink_check_transfer_dst_barrier()
|
||
- zink: super reorder buffer copies
|
||
- tu: don't set startup debug on debug builds
|
||
- zink: fix copy box iteration
|
||
- glthread: align small buffer uploads to 4 bytes
|
||
- zink: fix copy box iteration when adding
|
||
- zink: fix copy box merging adjacency
|
||
- aux/trace: delete GALLIUM_TRACE_NIR log message
|
||
- zink: fix layer check for compressed format surface creation
|
||
- zink: track current queue for resources
|
||
- zink: remove redundant dmabuf_acquire setting
|
||
- zink: use res->queue to auto-handle queue transitions back to gfx queue
|
||
- zink: track tc fences better
|
||
- zink: add an assert to catch renderpass optimizing bugs
|
||
- zink: don't use/update tc rp info while blitting
|
||
- zink: reset tc fb info upon splitting a renderpass
|
||
- zink: add and use a function for "safely" ending renderpasses
|
||
- zink: disable queries when flushing clears from set_fb
|
||
- zink: disable tc flush notify with rp optimizing
|
||
- zink: trigger oom flushes more aggressively from copy ops
|
||
- aux/tc: flag late zs clears as partial clears
|
||
- aux/tc: use a local 'deferred' variable in tc_flush()
|
||
- aux/tc: fix renderpass splitting on flush
|
||
- aux/tc: track the number of active queries
|
||
- aux/tc: don't sync for get_sample_position
|
||
- aux/tc: add a function to reset rp info
|
||
- aux/tc: fix initial rp info allocation
|
||
- aux/tc: make some of the rp tracking api private
|
||
- aux/tc: rework inter-batch renderpass info handling
|
||
- zink: only flag rp info for updating on flush, don't actually update
|
||
- zink: add tracing for copy ops
|
||
- zink: expand ZINK_DEBUG=sync to cover copy ops
|
||
- zink: add some asserts for zs layout in dynamic render
|
||
- zink: double check layouts for possible feedback loop images
|
||
- zink: end rp earlier in set_framebuffer_state
|
||
- zink: add a function for applying u_blitter barriers
|
||
- zink: add a dynamic render version of clear_texture hook
|
||
- zink: reset fb clears using the clears_enabled mask
|
||
- zink: manually apply barriers whenever zink_context::blitting is set
|
||
- zink: split out pipeline rp info update function
|
||
- zink: add zink_context::unordered_blitting to preserve unordered flags
|
||
- zink: implement unordered u_blitter calls
|
||
- zink: round geometry for u_blitter debug markers
|
||
- zink: catch zs u_blitter ops for draw markers
|
||
- zink: add debug markers for draws
|
||
- zink: track zsbuf info even when rp optimizing is disabled
|
||
- zink: fix dsa state parsing for tc info
|
||
- zink: track whether the fb zsbuf is readonly
|
||
- zink: add a fixup case for readonly zsbuf clears
|
||
- zink: improve no-oping of write -> readonly zsbuf layouts
|
||
- zink: don't flag rp layout check on next draw when rp optimizing
|
||
- zink: allow zink_is_zsbuf_used() without tc / rp optimizing
|
||
- zink: rename add_implicit_color_feedback_loop()
|
||
- zink: explicitly eliminate feedback loops for unused zsbufs
|
||
- zink: further eliminate zs implicit feedback loops for read-only access
|
||
- zink: split out luminance/alpha clear conversion code for reuse
|
||
- zink: convert luminance/alpha clear colors in dynamic texture clear
|
||
- zink: handle swapchain creation failure less lazily
|
||
- zink: simplify resource_check_defer_buffer_barrier()
|
||
- zink: delete unused barrier api
|
||
- zink: never split a renderpass for a loadop change
|
||
- zink: flag some rp ends as unsafe
|
||
- zink: add batch refs for framebuffer surfaces on bind and ref update
|
||
- zink: fix unordered access for image descriptors
|
||
- zink: force unordered_write=false when binding image descriptors
|
||
- zink: add an assert to ensure zsbuf invalidation doesn't break rendering
|
||
- zink: only run post-fb-unbind layout stuff if the resource isn't being destroyed
|
||
- zink: always set sampler layouts when unbinding fb images while rp optimizing
|
||
- zink: rework handling of unordered->ordered write buffer barriers
|
||
- zink: don't update fbfetch in db mode if inputAttachmentDescriptorSize==0
|
||
- zink: add ZINK_DEBUG=flushsync
|
||
- zink: track whether zsbuf is unused
|
||
- zink: flag rp layout change if zsbuf usedness changes on dsa/fs state bind
|
||
- driconf: make glthread=true default for source games
|
||
- aux/trace: dump blend states with enums
|
||
- aux/trace: fix GALLIUM_TRACE_NIR handling
|
||
- zink: fix some type mismatches for c++ compilation
|
||
- zink: break out a src region barrier check for reuse
|
||
- zink: move all barrier-related functions to c++
|
||
- zink: use c++ template to deduplicate all the buffer barrier code
|
||
- zink: minor tweaks for image barriers
|
||
- zink: use c++ template to deduplicate image barrier functions
|
||
- zink: stop leaking separate shader nir
|
||
- lavapipe: always copy streamout info when creating shaders
|
||
- lavapipe: don't memcpy tess_ccw when copying pipeline library shaders
|
||
- lavapipe: refactor shader compile functions to not take pipeline params
|
||
- lavapipe: track bound shader stages on rendering_state
|
||
- lavapipe: add a device member to rendering_state
|
||
- lavapipe: stop using rendering_state::pipeline
|
||
- lavapipe: refactor compute shader binding
|
||
- lavapipe: merge some loops in handle_graphics_pipeline()
|
||
- lavapipe: PIPE_SHADER\_ -> MESA_SHADER\_
|
||
- lavapipe: don't access pipeline shader structs as much during bind
|
||
- lavapipe: pull out dynamic tess origin check in gfx pipeline bind
|
||
- lavapipe: break out all the important parts of gfx pipeline setting for reuse
|
||
- lavapipe: delete unused struct member
|
||
- lavapipe: refactor pipeline destroy a bit
|
||
- lavapipe: add a ref for the tess_ccw nir on creation
|
||
- lavapipe: unify lvp_pipeline_nir creation
|
||
- lavapipe: dynamically bind noop fs at draw time when needed
|
||
- lavapipe: don't double unbind gfx stages on pipeline bind
|
||
- lavapipe: split out gfx stage unbinding
|
||
- lavapipe: only update shader access for bind/unbind stages
|
||
- lavapipe: only unset tess_states pointers on tes bind
|
||
- lavapipe: avoid uniformly unsetting gs_output_lines
|
||
- lavapipe: move default rasterizer state values to rendering_state init
|
||
- mapi: add InternalInvalidateFramebufferAncillaryMESA
|
||
- glthread: add _mesa_glthread_invalidate_zsbuf()
|
||
- kopper: apply ancillary invalidation through glthread on swapbuffers
|
||
- llvmpipe: fix linear fs analysis with nonzero fs outputs
|
||
- llvmpipe: fix handling of unused color attachments
|
||
- zink: add spirv builder function for terminate
|
||
- zink: set src access when rebinding buffers, unset unordered_*
|
||
- zink: fix quads emulation gs with array variables
|
||
- zink: block resolves where src extents > dst extents
|
||
- zink: omit VkPipelineVertexInputStateCreateInfo with dynamic vinput
|
||
- zink: flag vertex buffers for rebind after vstate draws
|
||
- zink: use search_or_add for masking vstate
|
||
- zink: bind vertex state directly from draw hook
|
||
- zink: add another vstate draw template for popcnt presence
|
||
- zink: explicitly pass null velems when creating pipelines with dynamic vinput
|
||
- zink: don't swizzle velems state for vstate draws
|
||
- zink: use fast popcnt for vstate draws
|
||
- zink: stop caching vertex states
|
||
- lavapipe: break out pipeline layout creation for reuse
|
||
- lavapipe: implement EXT_shader_object
|
||
- lavapipe: advertise EXT_shader_object
|
||
- zink: delete shader reordering in assign_io()
|
||
- zink: add and populate a shader_info struct to zink_shader
|
||
- zink: pass nir_shader to update_so_info()
|
||
- zink: generate flat_flags during shader creation
|
||
- zink: use zink_shader::info instead of zink_shader::nir::info
|
||
- zink: simplify fbfetch output detection from fs
|
||
- zink: pass nir directly to zink_shader_tcs_create()
|
||
- zink: swap nir pointers when compiling compute shaders
|
||
- zink: directly return nir from zink_shader_tcs_create
|
||
- zink: streamline nir cloning for assign_io
|
||
- zink: store nir as serialized on zink_shader structs
|
||
- zink: simplify assign_io() further
|
||
- zink: break out nir blob deserializing
|
||
- zink: move nir cloning out to callers of zink_shader_compile
|
||
- zink: store num_inlinable_uniforms separately for cs programs
|
||
- zink: always store nir serialized
|
||
- zink: be explicit about separate shader dsl indexing during creation
|
||
- zink: rework choose_pdev (again)
|
||
- glthread: use id 0 for internal buffer objects
|
||
- radv: fix leak of nir from retained shaders
|
||
- zink: don't try copying multiple results for conditional render copy
|
||
- zink: more explicitly track/check rp optimizing per-context
|
||
- zink: don't access non_fs part of zink_shader from fs
|
||
- zink: reuse d3d12 variable copying to make passthrough gs more robust
|
||
- zink: reuse copy_vars for generated tcs
|
||
- zink: don't trigger shader variants on pcp change if driver supports dynamic pcp
|
||
- Revert "zink: don't trigger shader variants on pcp change if driver supports dynamic pcp"
|
||
- zink: try to prune resources from barrier jit on fb unbind
|
||
- lavapipe: copy fragment shader when merging GPL pipelines
|
||
- lavapipe: refactor/consolidate GPL shader copying
|
||
- lavapipe: don't double-inline ubo0
|
||
- lavapipe: implement inline variant caching
|
||
- zink: block oom flushes during unordered blits
|
||
- zink: unroll array loop when copying vars for passthrough shaders
|
||
- zink: free GPL input/output libs on context destroy to avoid leaking
|
||
- zink: fix GPL lib leaking
|
||
- zink: remove redundant 'blitting' check in zink_prep_fb_attachment()
|
||
- zink: break out feedback loop pipeline state flagging for reuse
|
||
- zink: pre-convert attachment id to attachment idx
|
||
- zink: eliminate implicit feedback loops on rp begin
|
||
- zink: track per-image swapchain layouts
|
||
- zink: handle swapchain handoffs around makecurrent
|
||
- zink: remove a fixed validation error for ci
|
||
- mesa/st/program: don't init xfb info if there are no outputs
|
||
- zink: remove atomics from zink_query
|
||
- zink: pass ctx through query destroy paths
|
||
- zink: always defer query pool deletion
|
||
- zink: move memoryTypeIndex selection down in general bo allocation
|
||
- zink: slightly rework memoryTypeIndex selection to pre-determine heap
|
||
- zink: restore BAR allocation failure demotion
|
||
- zink: make general bo allocation more robust by iterating
|
||
- zink: avoid zero-sized memcmp for descriptor layouts
|
||
- iris: use util_framebuffer_get_num_samples when setting ps dispatch samples
|
||
- zink: manually re-set framebuffer after msrtss replicate blit
|
||
- zink: handle 'blitting' flag better in msrtss replication
|
||
- zink: skip msrtss replicate if the attachment will be full-cleared
|
||
- zink: avoid recursion during msrtss blits from flushing clears
|
||
- nir/lower_alpha_test: rzalloc state slots
|
||
- zink: fix non-db bindless texture buffers
|
||
- zink: emit demote cap when using demote
|
||
- zink: only print copy box warning once per resource
|
||
- util/debug: move null checks out of debug message macro
|
||
- zink: don't bitcast bool deref loads/stores
|
||
- drisw: don't leak the winsys
|
||
- zink: check for extendedDynamicState3DepthClipNegativeOneToOne for ds3 support
|
||
- draw: fix viewmask iterating
|
||
- zink: don't pin flush queue threads if no threads exist
|
||
- zink: add z32s8 as mandatory GL3.0 profile attachment format
|
||
- nir/gs: fix array type copying for passthrough gs
|
||
- zink: fix array copying in pv lowering
|
||
- gallivm: break out native vector width calc for reuse
|
||
- llvmpipe: do late init for llvm builder
|
||
- zink: break out VkImageViewUsageCreateInfo applying for reuse
|
||
- zink: reapply VkImageViewUsageCreateInfo when rebinding a surface
|
||
- draw: fix robust ubo size calc
|
||
- llvmpipe: fix native vector width init
|
||
- zink: add extendedDynamicState3DepthClipNegativeOneToOne to profile
|
||
- zink: only unset a generated tcs if the bound tcs is the generated one
|
||
- zink: set depth dynamic state values unconditionally
|
||
- zink: null some descriptor buffer pointers during destruction
|
||
- zink: sync queries at the end of cmdbufs
|
||
- cso: unbind fb state when unbinding the context
|
||
- i915: use util_copy_framebuffer_state to set fb state
|
||
- i915: use util_unreference_framebuffer_state to unref fb state
|
||
- iris: use util_unreference_framebuffer_state to unref fb state
|
||
- softpipe: use util_unreference_framebuffer_state to unref fb state
|
||
- v3d: use util_unreference_framebuffer_state to unref fb state
|
||
- vc4: use util_unreference_framebuffer_state to unref fb state
|
||
- llvmpipe: use util_unreference_framebuffer_state to unref fb state
|
||
- svga: use util_unreference_framebuffer_state to unref fb state
|
||
- zink: don't init mutable resource bit for swapchain images
|
||
- zink: don't init mutable for swapchain src during blit
|
||
- zink: allow vk 1.2 timelineSemaphore feature if extension isn't supported
|
||
- zink: stringify unsupported prim restart log error
|
||
- zink: delete persistent map tracking
|
||
- zink: add PERSISTENT for db buffer maps
|
||
- zink: delete unnecessary pipeline stage flags from inference
|
||
- zink: use an intermediate variable for binding ssbo slots
|
||
- zink: unbind the ssbo slot being iterated, not the index of the buffer
|
||
- zink: flush INDIRECT_BUFFER mem barrier for compute
|
||
- zink: disable batched unordered barries with ZINK_DEBUG=noreorder
|
||
- zink: block batching of unordered barriers if previous usage was write
|
||
- zink: fix uncached memory readback
|
||
- glsl/lower_samplers_as_deref: apply bindings for unused samplers
|
||
- zink: bind bindless db set when updating separate shader db sets
|
||
- zink: compare desc set to detect bindless vars in separate shaders
|
||
- zink: adjust bindless texel buffer handle before indexing
|
||
- zink: block more flushes during unordered blits
|
||
- zink: also cache swapchain semaphores
|
||
|
||
Mohamed Ahmed (3):
|
||
|
||
- vulkan/runtime: move common buffer related entrypoints to vk_buffer.c
|
||
- vulkan/runtime: implement vkGetBufferMemoryRequirements2()
|
||
- anv: remove GetBufferMemoryRequirements2()
|
||
|
||
Nanley Chery (16):
|
||
|
||
- docs: Document the implicit barriers around blits
|
||
- glsl: Add compute shaders to encode DXT5/BC3
|
||
- glsl: Modify the #includes in the DXT5 shaders
|
||
- mesa: Create _mesa_CreateShaderProgramv_impl
|
||
- mesa/st: Add get_compute_program
|
||
- mesa/st: Add and use create_bc1_endpoint_ssbo
|
||
- mesa/st: Add st_compute_transcode_astc_to_dxt5
|
||
- mesa/st: Add st_texture_image_resource_level
|
||
- mesa/st: Enable compute-based transcoding to DXT5
|
||
- mesa/st: Measure compressed fallback unmap paths
|
||
- iris: Update comment in iris_cache_flush_for_render
|
||
- iris: Flush caches for aux-mode changes more often
|
||
- iris: Drop iris_cache_flush_for_render
|
||
- iris: Allocate ZEROED BOs for shared resources
|
||
- iris/bufmgr: Add and use zero_bo
|
||
- iris/bufmgr: Handle flat_ccs for BO_ALLOC_ZEROED
|
||
|
||
Nataraj Deshpande (1):
|
||
|
||
- anv: Bump VkDeviceMemory objects limit to 4GB
|
||
|
||
Neha Bhende (1):
|
||
|
||
- docs: Add GL 4.3 support info in mesa docs
|
||
|
||
Nicolas Dufresne (1):
|
||
|
||
- util/format: Fix wrong colors when importing YUYV and UYVY
|
||
|
||
Nicolas F (1):
|
||
|
||
- driconf: remove the adaptive sync special case for mpv
|
||
|
||
Oleksii Bozhenko (5):
|
||
|
||
- glsl: fix gl_CullDistance lowering from float[8] to vec4[2]
|
||
- ci: Uprev Piglit
|
||
- Move combining clip and cull optimization before linking
|
||
- wsi: add rgb_component_bits_are_equal
|
||
- wsi: remove get_sorted_vk_formats duplication
|
||
|
||
Patrick Lerda (25):
|
||
|
||
- lima: fix memory leak related to u_transfer_helper_create()
|
||
- mesa/program: fix memory leak triggered by parser errors
|
||
- mesa/st: fix possible crash related to arb invalid memory access
|
||
- r600: fix shader blob memory leak
|
||
- vbo/save: fix possible crash related to fixup_vertex()
|
||
- mesa/shaderapi: fix path memory leak
|
||
- mesa/framebuffer: fix gl_framebuffer.resolve refcnt imbalance
|
||
- mesa/program: fix memory leak triggered by invalid extended swizzle selector
|
||
- mesa/program: fix memory leak triggered by multiple targets used on one texture image unit
|
||
- mesa/program: fix memory leak triggered by arb alias
|
||
- radeonsi: fix memory leak related to ureg_get_tokens()
|
||
- glx: fix memory leak related to __glXCloseDisplay()
|
||
- r600: fix refcnt imbalance related to shader
|
||
- intel: fix memory leak related to brw_nir_create_passthrough_tcs()
|
||
- r600: fix typo that could lead to a possible crash
|
||
- egl: fix memory leak related to _eglRefreshDeviceList()
|
||
- r600: fix refcnt imbalance related to r600_set_vertex_buffers()
|
||
- r600: fix refcnt imbalance related to evergreen_set_shader_images()
|
||
- lima: fix refcnt imbalance related to framebuffer
|
||
- r600/sfn: fix memory leak related to sh_info->arrays
|
||
- aux/draw: fix memory leak related to ureg_get_tokens()
|
||
- crocus: fix refcnt imbalance related to framebuffer
|
||
- crocus: fix refcnt imbalance related to crocus_create_surface()
|
||
- r600: fix refcnt imbalance related to atomic_buffer_state
|
||
- radeonsi: set proper drm_amdgpu_cs_chunk_fence alignment
|
||
|
||
Paul Gofman (1):
|
||
|
||
- driconf: add a workaround for Kaiju-A-Gogo
|
||
|
||
Paulo Zanoni (8):
|
||
|
||
- anv: don't leave undefined values in exec->syncobj_values
|
||
- anv: check the return value of anv_execbuf_add_bo_bitset()
|
||
- anv: run buf_finish() if add_bo() fails during execute_simple_batch()
|
||
- anv: rename anv_execbuf->array_length to bo_array_length
|
||
- anv: use vk_realloc for the anv_execbuf arrays
|
||
- hasvk: don't leave undefined values in exec->syncobj_values
|
||
- hasvk: check the return value of anv_execbuf_add_bo_bitset()
|
||
- anv: there's no need to set exec_obj offsets twice
|
||
|
||
Pavel Ondračka (16):
|
||
|
||
- nir/lower_bool: ntt: Generate a good opcode for bcsel
|
||
- r300: update rv515 ci failures list
|
||
- r300: skip sin/cos input range transformation for nine and ntt
|
||
- r300: remove backend input range transformation for sin and cos
|
||
- ntt: pass ubo_vec4_max nir_opt_offsets flag through ntt options
|
||
- r300: set ubo_vec4_max ntt option properly
|
||
- r300: remove backend negative addressing emulation
|
||
- nir: nir opt_shrink_vectors whitespace fix
|
||
- nir: mark progress when removing trailing unused alu channels
|
||
- nir: mark progress when removing trailing unused load_const channels
|
||
- r300: set register file to none if swizzles are constant only
|
||
- nir: shrink phi nodes in nir_opt_shrink_vectors
|
||
- r300: drop VDPAU support
|
||
- r300: simplify KILL transformation
|
||
- nine: use separate register for aL emulation
|
||
- r300: fix unconditional KIL on R300/R400
|
||
|
||
Pedro J. Estébanez (4):
|
||
|
||
- spirv_to_dxil: Unify spirv_to_nir_options
|
||
- spirv2dxil: Split read-only image as SRV logic into declared and inferred
|
||
- spirv: Assume input attachments are read-only
|
||
- Revert "microsoft/compiler: Use SRVs for read-only images"
|
||
|
||
Philip Langdale (1):
|
||
|
||
- radeonsi: correctly declare YUV420_10 RT Format support for AV1
|
||
|
||
Philipp Zabel (2):
|
||
|
||
- vulkan/wsi/wayland: fix acquire_next_image to report timeouts properly
|
||
- zink: fix build with -Dvulkan-beta=true
|
||
|
||
Pierre-Eric Pelloux-Prayer (37):
|
||
|
||
- radeonsi: simplify dpbb settings
|
||
- ac/info: move pci bus info in a struct
|
||
- ac: add ac_query_pci_bus_info helper
|
||
- ac: don't call ac_query_pci_bus_info from ac_query_gpu_info
|
||
- radeonsi/sqtt: don't read results for disabled SEs
|
||
- radeonsi/sqtt: disable SE1+ on GFX11
|
||
- radeonsi/sqtt: update registers for gfx11
|
||
- radeonsi/sqtt: implement offset workaround for gfx11
|
||
- vbo: remove bogus assert
|
||
- vbo: lower VBO_SAVE_BUFFER_SIZE to avoid large VRAM usage
|
||
- glthread: fix glArrayElement handling
|
||
- drm-uapi/dma-buf.h: use __u32/__u64 types
|
||
- winsys/amdgpu: use DMA_BUF_SET_NAME_B if available
|
||
- radeonsi/gfx11: clamp PRIM_GRP_SIZE
|
||
- radeonsi/gfx11: fix ge_cntl programming
|
||
- amd/surface: fix base_mip_width of subsampled formats
|
||
- winsys/amdgpu: use amdgpu_device_get_fd
|
||
- radeonsi/video: use specific PIPE_BIND\_ value for video buffers
|
||
- radeonsi: fix incorrect vgpr indices in the ps_prolog
|
||
- radeonsi/test: use gbm-skips.txt
|
||
- radeonsi/test: update test results
|
||
- radeonsi: don't use PKT3_SET_SH_REG_INDEX on gfx9 and older
|
||
- radeonsi: fix fast depth_clear_value/stencil_clear_value
|
||
- yegl/wayland: fix glthread deadlocks
|
||
- Revert "driconf: add a workaround for plasmashell freezing"
|
||
- ac/llvm: fix build with LLVM 17
|
||
- mesa: fix CopyImageSubDataOES with GL_TEXTURE_EXTERNAL_OES
|
||
- amd/surface: rename metadata functions
|
||
- ac/surface: introduce umd metadata v2
|
||
- radeonsi: add AMD_DEBUG=extra_md
|
||
- radeonsi: don't use si_decompress_dcc if the blitter is running
|
||
- radv: add RADV_DEBUG=extra_md
|
||
- radeonsi: don't use alignment_log2 of imported buffers
|
||
- mesa: fix invalid index_bo refcounting
|
||
- util/vbuf: clarify indirect draws handling
|
||
- util/vbuf: fix index_bo leak
|
||
- radeonsi: update test results
|
||
|
||
Pino Toscano (1):
|
||
|
||
- symbols-check: support OSes based on GNU toolchain
|
||
|
||
Qiang Yu (78):
|
||
|
||
- radeonsi: implement nir_load_ring_gsvs_amd
|
||
- radeonsi: implement nir_load_ring_gs2vs_offset_amd
|
||
- radeonsi: lower nir streamout intrinsics in abi
|
||
- radeonsi: use nir_print_xfb_info to replace si_dump_streamout
|
||
- radeonsi: use ac_nir_lower_legacy_vs to replace si_llvm_vs_build_end
|
||
- radeonsi: add nir implementation of gs copy shader generation
|
||
- radeonsi: build legacy gs output info when shader compile
|
||
- radeonsi: replace llvm gs copy shader generation with nir
|
||
- radeonsi: remove llvm gs copy shader generate
|
||
- radeonsi: replace llvm legacy gs code with nir lowering
|
||
- radeonsi: move gfx10_ngg_export_vertex to si_shader_llvm.c
|
||
- gallium/aux: remove nir_helpers
|
||
- nir/xfb_info: nir_gather_xfb_info_from_intrinsics update nir xfb_info
|
||
- radeonsi: update nir xfb info after medium io lowering
|
||
- nir: add nir_export_amd intrinsic
|
||
- ac/llvm: implement nir_export_amd
|
||
- aco: implement nir_export_amd
|
||
- ac/nir: gs and nogs use ac_nir_export_primitive
|
||
- ac/nir: add ac_nir_export_position
|
||
- ac/nir: add ac_nir_export_parameter
|
||
- ac/nir: add force_vrs to ac_nir_export_position
|
||
- amd,radeonsi: implement nir_load_force_vrs_rates_amd in driver abi
|
||
- radeonsi: clamp vertex color in legacy gs instead of gs copy shader
|
||
- radeonsi: update outputs written nir info
|
||
- radeonsi: remove the extra handling for VS/TES primitive id
|
||
- radeonsi: set nr_pos_exports outside of llvm translation
|
||
- ac/nir,radv,radeonsi: legacy vs use ac_nir_export_(position|parameter)
|
||
- ac/nir,radv,radeonsi: gs copy shader use ac_nir_export_(position|parameter)
|
||
- ac/nir/ngg: fix clip dist culling mask uninitialized
|
||
- ac/nir/ngg: change clipdist_neg_mask_var type to uint32
|
||
- ac/nir/ngg,radv,radeonsi: nogs use ac_nir_export_(position|parameter)
|
||
- ac/nir/ngg: prepare gather_vs_outputs to be used by gs
|
||
- ac/nir/ngg: gs use ac_nir_export_(position|parameter)
|
||
- ac/nir/ngg,radv: ms use ac_nir_export_(primitive|position|parameter)
|
||
- nir,ac/llvm,aco: remove nir_export_primitive_amd
|
||
- nir,ac/llvm,aco,radv,radeonsi: remove nir_export_vertex_amd
|
||
- aco: remove early_rast wait insert
|
||
- radv: move radv_consider_force_vrs above radv_fill_shader_info
|
||
- radv: use amd common force_vrs option
|
||
- ac/llvm,radeonsi: lower nir_load_barycentric_at_sample in abi
|
||
- radeonsi: add num_component param to load_internal_binding
|
||
- ac/llvm,radeonsi: lower fbfetch in abi
|
||
- radeonsi: only init llvm output when needed.
|
||
- ac/llvm: only init outputs when fragment shader for radv
|
||
- aco: only ls and ps use store output now
|
||
- aco, radv: Add load_grid_size_from_user_sgpr to aco options.
|
||
- aco, radv: Move is_trap_handler_shader to aco info.
|
||
- ac/nir: move store_var_components to common place
|
||
- ac/nir: tcs write tess factor support pass by reg
|
||
- ac/nir: init tess factor location with IO remap
|
||
- ac/nir: handle tess factor output missing case
|
||
- ac/llvm,radeonsi: lower nir_load_ring_tess_factors_amd
|
||
- radeonsi: lower nir_load_ring_tess_factors_offset_amd
|
||
- radeonsi: monolithic TCS emit tessfactor in nir directly
|
||
- ac/llvm: respect channel_type when ac_build_buffer_load
|
||
- ac/llvm: add missing type convert for nir_load_buffer_amd
|
||
- nir: pack_(s|u)norm_2x16 support float16 as input
|
||
- ac/llvm: implement float16 nir_op_pack_(s|u)norm_2x16
|
||
- aco: implement float16 nir_op_pack_(s|u)norm_2x16
|
||
- nir,radeonsi: add and implement nir_load_alpha_reference_amd
|
||
- nir: add nir_fisnan helper function
|
||
- ac/nir: add ac_nir_lower_ps
|
||
- radeonsi: monolithic PS emit epilog in nir directly
|
||
- radeonsi: expose si_nir_load_internal_binding
|
||
- ac/nir: add ac_nir_load_arg_at_offset
|
||
- radeonsi: add si_nir_lower_vs_inputs
|
||
- ac/llvm: vs_rel_patch_id can also be fixed up
|
||
- ac/llvm: move ac_fixup_ls_hs_input_vgprs to amd common
|
||
- radeonsi: monolithic VS emit prolog in nir directly
|
||
- ac/llvm,radeonsi: remove abi->load_inputs implementation
|
||
- ac/llvm: remove ac_build_opencoded_load_format
|
||
- radeonsi: fix max scrach lds size calculation when ngg
|
||
- ac/nir/ngg: fix gs culling vertex liveness check for odd vertices
|
||
- ac/nir/ngg: fix store shared alignment
|
||
- ac/llvm: remove some unused code replaced by nir
|
||
- ac,aco: move gfx10 ngg prim count zero workaround to nir
|
||
- aco: fix nir_f2u64 translation
|
||
- ac/nir/cull: fix line position w culling
|
||
|
||
Raun (2):
|
||
|
||
- dzn: Enable VK_KHR_bind_memory2
|
||
- dzn: Enable VK_KHR_get_memory_requirements2
|
||
|
||
Rhys Perry (48):
|
||
|
||
- radv: implement GS load_ring_gsvs_amd/load_ring_gs2vs_offset_amd
|
||
- radv,aco: use ac_nir_lower_legacy_gs
|
||
- aco: restore semantic_can_reorder for GS output stores
|
||
- ac/nir: use store_buffer_amd's base index
|
||
- ac/llvm: add support for fp32 addition atomics
|
||
- aco: add support for fp32 addition atomics
|
||
- radv: load ssbo_atomic_fadd descriptor
|
||
- radv/gfx11: expose shaderBufferFloat32AtomicAdd
|
||
- aco/tests: fix assembler.gfx11.vop12c_v128 with LLVM 15
|
||
- aco/tests: update assembler tests for latest LLVM 16
|
||
- radv: skip creation of null TLAS for null winsys
|
||
- aco: set has_color_exports with GPL
|
||
- aco: end reduce tmp after control flow, when used within control flow
|
||
- aco/tests: add setup_reduce_temp.divergent_if_phi
|
||
- aco/spill: always end spill vgpr after control flow
|
||
- aco: limit VALUPartialForwardingHazard search
|
||
- radv: set state.vbo_misaligned_mask_invalid in radv_bind_vs_input_state
|
||
- ac: move ring_offsets to ac_shader_args
|
||
- ac/llvm: let ring_offsets be accessed like a normal arg
|
||
- radv/llvm: use the ring_offsets shader arg
|
||
- aco: fix out-of-bounds access when moving s_mem(real)time across SMEM
|
||
- aco: don't modify exec in p_interp_gfx11
|
||
- aco: don't apply modifiers through DPP to unsupported instructions
|
||
- aco: fix pathological case in LdsDirectVALUHazard
|
||
- aco: always update orig_names in get_reg_phi()
|
||
- radv: remove is_internal pipeline creation parameter
|
||
- aco/tests: add tests for v_fma_f32 with 2 fp16 literals
|
||
- aco: make IDSet sparse
|
||
- nir/range_analysis: fix vectorized phis and intrinsics
|
||
- nir: use xyzw order for precise fdot
|
||
- nir: make fdph lowering match fdot
|
||
- nir: add nir_lower_alu_width_test.fdot_order
|
||
- aco/gfx11: fix RT prolog scratch initialization
|
||
- aco: set needs_flat_scr=true for RT
|
||
- util/dynarray: allow an initial stack allocation to be used
|
||
- nir/range_analysis: add missing masking of shift amounts
|
||
- nir/range_analysis: add helpers for limiting stack usage
|
||
- nir/range_analysis: use perform_analysis() in nir_unsigned_upper_bound()
|
||
- nir/range_analysis: use perform_analysis() in nir_analyze_range()
|
||
- radv: fix setting radv_shader_info::user_data_0 with rt
|
||
- aco: don't optimize s_or_b64(v_cmp_u_f32(a, b), cmp(a, a))
|
||
- aco: fix nir_var_shader_out barriers for task shaders
|
||
- radv/gfx11: improve RT scratch allocation
|
||
- nir: make nir_fisnan helper exact
|
||
- aco: remove SMEM_instruction::prevent_overflow
|
||
- ac/nir/ps: fix null export write mask miss set to 0xf
|
||
- aco: don't move exec reads around exec writes
|
||
- aco: don't move exec writes around exec writes
|
||
|
||
Rob Clark (180):
|
||
|
||
- freedreno/ci: Switch a630 jobs over to manual
|
||
- freedreno/ci: Cleanup a618 yaml
|
||
- freedreno/ci: Add a618 egl/skqp/piglit jobs
|
||
- Revert "freedreno/ci: Switch also performance a630 job to manual"
|
||
- Revert "freedreno/ci: Switch a630 jobs over to manual"
|
||
- freedreno/ci: Add an a618 flake
|
||
- freedreno/drm: Remove assert
|
||
- freedreno: Fix tracking of enabled SSBOs
|
||
- freedreno/a6xx: Workaround for no pos/psize
|
||
- freedreno: Don't re-install a flushed batch
|
||
- freedreno/a6xx: Rework barrier handling
|
||
- freedreno/ir3: Stop copying options
|
||
- freedreno/ir3: Let driver specify fb-read descriptor
|
||
- freedreno: Track image/SSBO usage for all stages
|
||
- freedreno/ir3: Add descriptor set lowering
|
||
- freedreno/a6xx: Pre-bake IBO descriptor sets
|
||
- freedreno/a6xx: Add bindless state
|
||
- freedreno/a6xx: Switch over to bindless IBO
|
||
- freedreno/a6xx: Remove bindfull IBO state
|
||
- freedreno/a6xx: Removing munging of tex state for IBO
|
||
- freedreno/a6xx: Remove tex fb_read state
|
||
- freedreno/a6xx: Move compute to tex state group
|
||
- freedreno/a6xx: Move tex state building
|
||
- freedreno/a6xx: Expose SSBO/image for all shader stages
|
||
- freedreno: Restore GL_VENDOR string
|
||
- gallium/util: Add util_writes_depth() helper
|
||
- freedreno/a6xx: Add LRZ perf warn for ztest direction changes
|
||
- freedreno/a6xx: Invalidate LRZ on blend+depthwrite
|
||
- turnip: Rename lrz force_disable_mask
|
||
- turnip: Invalidate LRZ on blend+depthwrite
|
||
- util/xmlconfig: Use os_get_option()
|
||
- freedreno: Add driconf to disable conservative LRZ
|
||
- freedreno/a6xx: Add a few kernel regs/etc
|
||
- freedreno/drm: Add some ref/unref debugging
|
||
- freedreno/drm: Detect zombie BOs
|
||
- freedreno/drm: Remove bo_del_or_recycle()
|
||
- freedreno/drm: Split out bo->finalize()
|
||
- freedreno/drm: Synchronize handle close and lookup
|
||
- freedreno/drm/virtio: Flush before CREATE_BLOB
|
||
- freedreno/drm: Restart import on zombie race
|
||
- freedreno/gmem: Fix for partial z/s fast-clear
|
||
- freedreno/decode: Increase size of offsets table
|
||
- freedreno/a6xx: LRZ for MSAA
|
||
- freedreno/ir3: Scalarize load_ssbo
|
||
- freedreno/a6xx: Add missing CS_BINDLESS mapping
|
||
- freedreno/a6xx: Add CS instrlen workaround
|
||
- freedreno: nondraw-batch
|
||
- freedreno: Skip flush_resource with explicit sync
|
||
- freedreno/a6xx: Don't double-write SP_CS_OBJ_START
|
||
- freedreno: Don't open-code setting dirty CS state
|
||
- freedreno/a6xx: Make shader state independent of grid info
|
||
- freedreno/a6xx: Also FLUSH_CACHE on image barrier
|
||
- freedreno/a6xx: Remove excess CS flushing
|
||
- freedreno+ir3: Move storage_16bit to compiler options
|
||
- freedreno/a6xx: Move CS state to PROG state group
|
||
- freedreno/drm: Move sa_cpu_prep() to core
|
||
- freedreno/drm/virtio: Limit guest handles passed to virtgpu
|
||
- ir3: Quiet unused variable warning
|
||
- freedreno: Quiet unused variable warnings
|
||
- freedreno/a2xx: Move pack_rgba()
|
||
- freedreno: Indent fixes
|
||
- freedreno/a6xx: Move num_driver_params to program state
|
||
- freedreno: Move num_vertices calc to backend
|
||
- freedreno: Remove impossible NULL check
|
||
- freedreno: Add FD_DIRTY_QUERY
|
||
- freedreno: Avoid screen lock when no rsc tracking needed
|
||
- freedreno: Account for multi-draw in num_draws
|
||
- freedreno: Push num_draws down to backend
|
||
- freedreno/a6xx: Drop unused return
|
||
- freedreno/a6xx: Split out flush_streamout() helper
|
||
- freedreno/a6xx: Multi-draw support
|
||
- freedreno/a6xx: Do tex-state invalidates in same ctx
|
||
- freedreno/drm: Make rb refcnt non-atomic
|
||
- freedreno/a6xx: Remove tex-state refcnting
|
||
- freedreno: Move blend out of dirty-rsc tracking
|
||
- freedreno: Move FD_MESA_DEBUG cases out of draw_vbo
|
||
- freedreno/a6xx: Pre-compute PROG related LRZ state
|
||
- freedreno: Avoid taking screen lock
|
||
- freedreno/batch: Stop tracking cross-context deps
|
||
- freedreno: Drop batch lock
|
||
- freedreno: Add seqno helper
|
||
- freedreno/drm: Optimize stateobj re-emit
|
||
- freedreno/a6xx: Move rsc seqno out of tex cache key
|
||
- freedreno/a6xx: Fix set_sampler_views(start != 0)
|
||
- freedreno/a6xx: Drop unneed fd6_texture_state() arg
|
||
- freedreno/a6xx: Fix sampler view rsc_seqno for X32_S8X24
|
||
- freedreno/a6xx: Add a way to assert valid format
|
||
- freedreno/a6xx: Remove needs_invalidate flag
|
||
- freedreno/a6xx: Small cleanup
|
||
- freedreno/a6xx: Static-ify sampler_view_update()
|
||
- freedreno/a6xx: Fix view_seqno in tex cache key
|
||
- freedreno/cffdec: Fix unitialized count for pkt2
|
||
- freedreno/cffdec: Add helper to find next pkt
|
||
- freedreno/cffdec: Add helper to parse CP_INDIRECT_BUFFER
|
||
- freedreno/cffdec: Fix hang location detection
|
||
- freedreno/crashdec: Refactor crashdec tests
|
||
- freedreno/crashdec: Add another prefetch test
|
||
- freedreno/crashdec: Handle multi-IB prefetching
|
||
- freedreno/crashdec: Disable GALLIUM_DUMP_CPU
|
||
- vk/runtime: Allow enumerate and try_create_for_drm to coexist
|
||
- turnip: Move things to prep for multi-kernel support
|
||
- turnip: drm code-motion
|
||
- turnip: Split out vfuncs for kernel interface
|
||
- turnip: Allow knl backend specific entrypoints
|
||
- turnip: Move QueueWaitIdle entrypoint to kgsl
|
||
- turnip: Handle kgsl vs drm specifics at runtime
|
||
- turnip: Refactor device loading
|
||
- tu+meson: Re-work KMD selection
|
||
- tu/kgsl: Propagate tu_physical_device_init() errors
|
||
- Revert "freedreno: Account for multi-draw in num_draws"
|
||
- freedreno/a6xx: Namespace reg/pkt packer vars
|
||
- freedreno/a6xx: Convert blitter to OUT_REG()
|
||
- freedreno/a6xx: Fix mirror x/y blits
|
||
- util: Add a simple no-op libdrm shim
|
||
- turnip: Use libdrm shim
|
||
- loader: Use libdrm shim
|
||
- vk/runtime: Use libdrm shim
|
||
- freedreno/common: Replace or_mask() with BitsetEnum<T>
|
||
- freedreno: Promote non-drawing batches to sysmem
|
||
- freedreno: Nerf strict-aliasing warning for all of gcc
|
||
- freedreno/registers: Schema validation for gen_header.py
|
||
- freedreno/registers: Add regs for a690
|
||
- freedreno: Quiet c++ warning about designated initializers
|
||
- freedreno/ir3: Un-inline enums
|
||
- freedreno/ir3: Don't use negative opc for meta instructions
|
||
- freedreno/ir3: c++-proof the headers
|
||
- freedreno/ir3+tu: Calculate subgroup size in ir3
|
||
- freedreno/ir3: Add missing driver params
|
||
- freedreno: Un-inline buffer-mask enum
|
||
- freedreno: c++-proofing
|
||
- freedreno/a6xx: Rework texture_clear fallback
|
||
- freedreno/a6xx: Add missing "inline"
|
||
- freedreno/a6xx: Fix designator initializer order
|
||
- freedreno/a6xx: Convert to c++
|
||
- freedreno/registers: Fix designator order
|
||
- freedreno/registers: Add prefix="variant"
|
||
- freedreno/registers: Merge a6xx and a7xx regs
|
||
- freedreno/registers: Start adding a7xx pipe/control regs
|
||
- freedreno/decode: Start adding a7xx support
|
||
- freedreno/registers: Start adding stuff for a7xx
|
||
- freedreno/registers: Track varset
|
||
- freedreno/registers: Split out regpair builder helper
|
||
- freedreno/registers: Add c++ magic for register variants
|
||
- freedreno/registers: Fix nameless fields
|
||
- freedreno/registers: Define rest of CP_REG_WRITE
|
||
- freedreno/a6xx: Simplify iova emit
|
||
- mesa: Rework discard_framebuffer()
|
||
- driconf: Add ignore_discard_framebuffer option
|
||
- driconf: Work around incorrect GI discard/invalidate
|
||
- freedreno: Specify GMEM tile alignment per GPU
|
||
- freedreno+tu: Big GMEM support
|
||
- freedreno+tu: Add a690 support
|
||
- freedreno/a6xx: Restore mode
|
||
- freedreno/rnn: Fix reg names for regs with variants
|
||
- freedreno/afuc: Add raw mode for disasm
|
||
- freedreno/registers: Add control reg for zap fw base
|
||
- Revert "CI: Disable freedreno"
|
||
- dri2/android: Bypass throttling
|
||
- freedreno/drm: Fast path for idle check
|
||
- freedreno/drm: Stop cleanup at first active BO
|
||
- mesa: Add a few more function traces
|
||
- freedreno/drm: Make threaded-submit optional
|
||
- freedreno/drm: Disable threaded-submit for msm
|
||
- freedreno: Optimize repeated finishes
|
||
- freedreno: Stop being too clever by half
|
||
- freedreno: Hoist dirty vars
|
||
- freedreno: Extract out a helper
|
||
- freedreno: Inline single-use helpers
|
||
- freedreno: Re-work dirty-resource tracking
|
||
- freedreno: Avoid looping shader stages if nothing dirty
|
||
- freedreno: Move driconf settings into sub-struct
|
||
- freedreno: Support the disable_throttling=true driconf option
|
||
- util/disk_cache: Split out queue initialization
|
||
- util/disk_cache: Add NONE type
|
||
- util/disk_cache: Use queue state to skip put
|
||
- util/disk_cache: Move blob_put_cb to the async queue
|
||
- freedreno/a6xx: Allow z24s8 format casts
|
||
- freedreno/a6xx: Fix valid_format_cast logic for newer a6xx
|
||
- freedreno: Fix resource tracking vs rebind/invalidate
|
||
- dri/android: Fix MSAA resolve
|
||
|
||
Rohan Garg (9):
|
||
|
||
- iris: Don't flush the render cache for a compute batch
|
||
- anv: drop unused headers
|
||
- anv: reuse the VK_IMAGE_ASPECT_PLANES_BITS_ANV macro
|
||
- isl: fix some documentation
|
||
- anv/blorp: use existing function to convert the op to a string
|
||
- anv: break out of the loop when the first color attachment is found
|
||
- anv,hasvk: cleanup unused enum
|
||
- intel/genxml: Add the preferred slm size enum for gen125
|
||
- anv,blorp,iris: Set PreferredSLMAllocationSize on gfx125+
|
||
|
||
Roland Scheidegger (2):
|
||
|
||
- llvmpipe: only use accurate_a0 hack if there are no textures bound
|
||
- lavapipe, nir: Fix wrong array index scaling in nir_collect_src_uniforms
|
||
|
||
Rose Hudson (4):
|
||
|
||
- radeonsi: report 0 block size for Polaris HEVC encoding
|
||
- asahi: wire up shader disk cache support
|
||
- agx: isolate compiler debug flags
|
||
- asahi: disable disk cache in debug runs
|
||
|
||
Ruijing Dong (5):
|
||
|
||
- frontends/va: revert commit 0b02db30
|
||
- raseonsi/vcn: fix a h264 decoding issue
|
||
- frontends/va: disable skip_frame_enable in vaapi interface.
|
||
- radeonsi/vcn: correct cropping for hevc case
|
||
- radeonsi/vcn: fix decoding bs buffer alignement issue.
|
||
|
||
Ryan Neph (17):
|
||
|
||
- ci: fix directory existence racing in parallel test execution
|
||
- util/u_process: add MESA_PROCESS_NAME override to util_get_process_name()
|
||
- util/u_process: remove util_get_process_name_may_override()
|
||
- util/xmlconfig: add MESA_DRICONF_EXECUTABLE_OVERRIDE
|
||
- venus: update venus-protocol headers to partially fix WA1
|
||
- venus: temporarily redirect VkDrmFormatModifierPropertiesListEXT to "2" variant
|
||
- ci: uprev virglrenderer
|
||
- venus: update venus-protocol headers to fix WA1
|
||
- Revert "venus: temporarily redirect VkDrmFormatModifierPropertiesListEXT to "2" variant"
|
||
- venus: add vn_relax_init/_fini()
|
||
- venus: set/check ring status bits independently
|
||
- venus: init exp features before ring init again
|
||
- venus: update to latest protocol for ringMonitoring
|
||
- venus: check and configure new ringMonitoring feature
|
||
- venus: re-use VN_DEBUG_NO_ABORT to disable ring monitoring abort()
|
||
- virgl: hook new get_fd proc for drm winsys
|
||
- i915: hook new get_fd proc for drm winsys
|
||
|
||
Sagar Ghuge (10):
|
||
|
||
- iris: Stop marking context unconditionally as guilty
|
||
- intel/fs: Always stall between the fences on Gen11+
|
||
- nir: Handle other variants of image_samples properly while lowering
|
||
- intel/compiler: Add swsb_stall debug option
|
||
- anv: Implement Wa_14015297576
|
||
- iris: Implement Wa_14015297576
|
||
- intel/compiler: Add Wa_14014063774 for slm_fence
|
||
- intel/decoder: Bump the binding table guess value to 32
|
||
- anv: Drop unused param from add_surface_reloc
|
||
- anv: Drop dead code that sets the L3BypassDisable field
|
||
|
||
Sai Teja Pottumuttu (2):
|
||
|
||
- iris: Fix to release BO immediately if not busy
|
||
- anv: Fix stride mismatch in mesa and minigbm
|
||
|
||
Sajeesh Sidharthan (3):
|
||
|
||
- radeonsi/vcn: disable fence for JPEG decoding
|
||
- radeonsi/vcn: set bitstream buffer size to encoded bitstream size
|
||
- radeonsi/vcn: optimize bitstream buffer resize logic
|
||
|
||
Sam Edwards (1):
|
||
|
||
- nouveau: Fix null dereference in nouveau_pushbuf_destroy
|
||
|
||
Samuel Iglesias Gonsálvez (1):
|
||
|
||
- docs/developers: Add Igalia as Mesa consultancy
|
||
|
||
Samuel Pitoiset (279):
|
||
|
||
- radv: fix missing implementation of creating images from swapchains
|
||
- radv: fix hashing pipeline keys if RADV_PERFTEST=ngg_streamout is used
|
||
- radv: fix re-emitting RB+ when the non-compacted color format changes
|
||
- ac/nir: clear unused components before storing XFB outputs to LDS
|
||
- ac: add TC_OP_ATOMIC_SUB_32
|
||
- radv: fix setting MAX_MIP for BC views
|
||
- radv: fix buffer to image copies with BC views on the graphics queue
|
||
- radv: fix creating BC image views when the base layer is > 0
|
||
- radv: rename ac_surf_nbc_view::max_mip to num_levels
|
||
- radv: move some color blend helpers to radv_private.h
|
||
- radv: add a new helper for normalizing blend factors
|
||
- radv: add support for dynamic blend equation
|
||
- radv: enable compiling PS epilogs on-demand for dynamic color blend equations
|
||
- radv: fix detecting that blend is enabled when all CB states are dynamic
|
||
- radv: advertise extendedDynamicState3ColorBlendEquation
|
||
- radv: remove an old FIXME about a possible bug with TC-compat HTILE
|
||
- radv/winsys: fix incorrect PCIID for GFX11 in the null winsys
|
||
- radv: print depth image size with RADV_DEBUG=img
|
||
- radv: fix RADV_DEBUG=hang with multiple cmdbuffer per submission
|
||
- radv/winsys: prefix all error messages with RADV
|
||
- radv: fix creating libraries with PS epilog and all CB states as dynamic
|
||
- radv: fix ignoring graphics shader stages that don't need to be imported
|
||
- radv: add a layer for fixing rendering issues with RAGE2
|
||
- radv: simplify VK_PIPELINE_CREATE_FAIL_ON_PIPELINE_COMPILE_REQUIRED
|
||
- radv: pass the number of stages to radv_hash_shaders()
|
||
- radv: split radv_create_shaders() between graphics and compute shaders
|
||
- radv: rename radv_create_shaders() to radv_graphics_pipeline_compile()
|
||
- radv: ignore all CB dynamic states when there is no color attachments
|
||
- radv: regroup dynamic states initialization
|
||
- radv: only initialize non-zero values for the default dynamic state
|
||
- radv: stop setting INTERPOLATE_COMP_Z
|
||
- radv: fix RB+ for SRGB formats
|
||
- radv: adjust ACCUM tessellation fields on GFX11+
|
||
- radv: fix GPL fast-linking with libs that have retained NIR shaders
|
||
- radv: skip shaders cache for fast-linked pipelines with GPL
|
||
- radv: remove useless check about CS in radv_lower_io()
|
||
- radv: simplify pipeline_has_ngg during graphics shaders compilation
|
||
- radv: add helpers for capturing shaders and statistics
|
||
- radv: pass radv_graphics_pipeline to radv_graphics_pipeline_compile()
|
||
- radv: move retained shaders info to radv_graphics_pipeline
|
||
- radv: pass radv_compute_pipeline to radv_compute_pipeline_compile()
|
||
- radv: pass pCreateInfo to radv_graphics_pipeline_compile()
|
||
- radv: optimize radv_pipeline_layout_add_set() slightly
|
||
- radv: remove redundant zero initialization of pipeline layout
|
||
- radv: remove radv_pipeline_stage::spirv::sha1
|
||
- radv: allow to create a noop FS in a library with GPL
|
||
- radv: remove one unused variable in radv_graphics_lib_pipeline_init()
|
||
- radv: pass the lib flags for generating the pipeline key
|
||
- radv: return a boolean value in radv_pipeline_needs_dynamic_ps_epilog()
|
||
- radv: stop using the graphics pipeline key after compilation
|
||
- radv: determine the last VGT API stage earlier
|
||
- radv: skip compilation when possible with GPL fast-linking
|
||
- radv: simplify an assertion after considering RADV_FORCE_VRS
|
||
- radv: do not insert fast-linked libraries to the shaders cache
|
||
- radv: fix skipping graphics pipeline compilation when the FS is NULL
|
||
- radv: cleanup graphics pipeline library flags uses
|
||
- radv: simplify determining when the fragment shader needs an epilog
|
||
- radv: regroup PS epilog info when generating the graphics pipeline key
|
||
- radv: fix disabling MRT compaction for on-demand PS epilogs
|
||
- radv: make sure to disable MRT compaction when compiling a PS epilog with GPL
|
||
- radv: simplify creating a FS epilog from a library
|
||
- radv: stop skipping the cache for compute/raytracing pipelines with GPL
|
||
- radv: stop skipping the cache for monolithic graphics pipelines with GPL
|
||
- docs: add missing RADV_PERFTEST=video_decode
|
||
- docs: stop reporting RADV_PERFTEST=gpl as experimental/suboptimal
|
||
- radv/ci: set RADV_PERFTEST=GPL for all VKCTS jobs
|
||
- radv/ci: bump the number of runners to 3 for vkcts-navi21-valve
|
||
- radv: restore uploading shaders individually instead of consecutively
|
||
- radv: implement graphics shaders relocation for a RGP workaround
|
||
- radv: fix importing retained NIR shaders when a lib uses the RETAIN bit
|
||
- radv: use last_vgt_api_stage for determining the last stage with XFB
|
||
- radv: only initialize shader arguments for the active stages
|
||
- radv simplify compiling graphics shaders with a mask of active NIR stages
|
||
- radv: disable DCC for mipmaps on GFX11
|
||
- radv: ignore registering pipeline libaries with SQTT
|
||
- radv/ci: add missing expected failures with RADV_PERFTEST=gpl on GFX1100
|
||
- radv: reduce maximum line width to 8.0
|
||
- radv: add support for rectangularLines
|
||
- Revert "radv: acquire pstate on-demand when capturing with RGP"
|
||
- radv/amdgpu: only set a new pstate if the current one is different
|
||
- radv: only skip emitting the pipeline blend state if the FS uses an epilog
|
||
- radv: stop using a PS epilog when the FS doesn't write any color outputs
|
||
- ci: uprev vkd3d-proton
|
||
- zink/ci: skip KHR-GL46.texture_swizzle.functional with RADV
|
||
- zink/ci: set RADV_PERFTEST=gpl for RADV jobs
|
||
- radv/ci: disable vkcts-kabini-valve
|
||
- radv/ci: move CI lists for external GPUs in separate folder
|
||
- radv: configure SQ_THREAD_TRACE_CTRL.REG_AT_HWM on GFX11
|
||
- radv: only enable SQTT for SE0 on GFX11
|
||
- radv: make sure to wait for the trace buffer also on GFX11
|
||
- radv: implement a workaround for SQTT on GFX11
|
||
- radv: disable SPM counters with RGP on GFX11
|
||
- radv: enable SQTT tracing on GFX11
|
||
- radv: set VS_OUT_MISC_SIDE_BUS_ENA for clip distances on GFX10.3+
|
||
- radv/ci: cleanup CI lists for dEQP-VK.memory.* tests that timeout
|
||
- ac/nir: add resinfo lowering for sliced storage 3D views
|
||
- radv: implement VK_EXT_image_sliced_view_of_3d on GFX10+
|
||
- radv: advertise VK_EXT_image_sliced_view_of_3d on GFX10+
|
||
- radv: cleanup radv_emit_{conservative,msaa}_state() functions
|
||
- radv: stop setting ENABLE_POSTZ_OVERRASTERIZATION to 1
|
||
- radv: set MSAA_NUM_SAMPLES to 0 for underestimate rasterization
|
||
- radv: enable primitiveUnderestimation on GFX9+
|
||
- zink/ci: skip one more test that timeout with RADV
|
||
- radv: fix flushing non-coherent images inside secondaries on GFX9+
|
||
- radv: fix flushing non-coherent images in EndCommandBuffer()
|
||
- radv: fix draw calls with 0-sized index buffers and robustness on NAVI10
|
||
- radv: only expose EXT_pipeline_library_group_handles if RT is enabled
|
||
- amd,ac/rgp: fix SQTT memory types
|
||
- radv: ignore alpha_is_on_msb on GFX11 because the hw ignores it
|
||
- radv: use new EVENT_WRITE_ZPASS packet3 on GFX11
|
||
- radv: fix DCC decompress on GFX11
|
||
- radv: stop allocationg the attr ring BO for compute queues on GFX11
|
||
- ci: uprev CTS to 1.3.5.0
|
||
- radv/ci: adjust timeouts for Vega10 and Renoir
|
||
- radv/ci: stop skipping some graphics pipeline library tests
|
||
- radv/ci: update CI lists for CTS 1.3.5.0 on GFX110/POLARIS10/PITCAIRN
|
||
- wsi: move an assertion in wsi_xxx_surface_get_capabilities2()
|
||
- radv: do not add descriptor BOs on update when the global BO list is used
|
||
- radv: fix incorrect stride for primitives generated query with GDS
|
||
- radv: fix border color swizzle for stencil-only format on GFX9+
|
||
- radv: fix defining RADV_USE_WSI_PLATFORM
|
||
- radv: move disabling DCC for VRS rate images in radv_get_surface_flags()
|
||
- ac/surface: add RADEON_SURF_VRS_RATE for selecting swizzle mode on GFX11
|
||
- radv: add support for VRS attachment on GFX11
|
||
- radv: do not emit PA_SC_VRS_OVERRIDE_CNTL from the pipeline on GFX11
|
||
- radv: advertise attachmentFragmentShadingRate on GFX11
|
||
- radv: enable VK_KHR_fragment_shading_rate on GFX11
|
||
- radv: disable DCC with signedness reinterpretation on GFX11
|
||
- radv: move instance related code to radv_instance.c
|
||
- radv: move physical device related code to radv_physical_device.c
|
||
- radv: move queue related code to radv_queue.c
|
||
- radv: move sampler related code to radv_sampler.c
|
||
- radv: move event related code to radv_event.c
|
||
- radv: move buffer related code to radv_buffer.c
|
||
- radv: move device memory related code to radv_device_memory.c
|
||
- radv: zero-initialize radv_shader_args right before declaring them
|
||
- radv: zero-initialize radv_shader_info earlier for graphics pipeline
|
||
- radv: fix the error code when the driver fails to create a PS epilog
|
||
- radv: determine if a graphics pipeline needs a noop FS earlier
|
||
- radv: keep track of the retained NIR shaders sha1 for LTO pipelines
|
||
- radv: allow to cache optimized (LTO) pipelines with GPL
|
||
- radv: rename RADV_PIPELINE_LIBRARY to RADV_PIPELINE_RAY_TRACING_LIB
|
||
- radv: add helpers for destroying various pipeline types
|
||
- radv: fix NGG streamout with VS and GPL on GFX11
|
||
- spirv: add SpvCapabilityFragmentFullyCoveredEXT
|
||
- spirv,nir: add support for SpvBuiltInFullyCoveredEXT
|
||
- radv: lower nir_intrinsic_load_fully_covered
|
||
- radv: enable SAMPLE_COVERAGE_ENA if the fully covered built-in is used
|
||
- radv: implement fullyCoveredFragmentShaderInputVariable
|
||
- radv: enable fullyCoveredFragmentShaderInputVariable on GFX9+
|
||
- radv: remove set but never used num_preserved_sgprs
|
||
- radv: stop storing the binary as part of radv_shader_part
|
||
- radv: store spi_shader_col_format to radv_shader_part_binary
|
||
- radv: store the total radv_shader_part_binary size
|
||
- radv: upload prologs/epilogs as part of radv_shader_part_create()
|
||
- radv: allow to return the PS epilog binary to the pipeline
|
||
- radv: make radv_shader_part_create() non-static
|
||
- radv: add support for caching PS epilogs
|
||
- radv: stop using radv_get_shader_shader() for task shaders
|
||
- radv: replace radv_lookup_user_sgpr() by radv_get_user_sgpr()
|
||
- radv: pass shader/base_reg to radv_emit_descriptor_pointers()
|
||
- radv: pass shader/base_reg to radv_emit_inline_push_consts()
|
||
- radv: pass shader/base_reg to radv_emit_userdata_address()
|
||
- radv: use a separate compute path in radv_flush_constants()
|
||
- radv: pass radv_shader to radv_dump_shader_stats()
|
||
- radv: rework dumping shaders when a GPU hang is reported
|
||
- radv: pass radv_ray_tracing_pipeline to radv_rt_pipeline_compile()
|
||
- radv: pass radv_shader to radv_shader_need_indirect_descriptor_sets()
|
||
- radv: implement VK_KHR_map_memory2
|
||
- radv: use common GetBufferMemoryRequirements2()
|
||
- radv: move cs_regalloc_hang_bug to radv_shader_info
|
||
- radv: pass a radv_shader to radv_emit_dispatch_packets()
|
||
- radv: remove radv_pipeline::device completely
|
||
- radv: rework binding shaders to cmdbuf by introducing new helpers
|
||
- radv: move radv_meta_* to a new folder
|
||
- radv: copy the multisample state to radv_cmd_state
|
||
- radv: move uses_user_sample_locations to radv_multisample_state
|
||
- radv: separate the sample shading state between FS and graphics pipeline
|
||
- radv: add DI_PT_RECTLIST to si_conv_prim_to_gs_out()
|
||
- radv: stop checking dynamic states when emitting the guardband state
|
||
- radv: rename gfx9_gs_info to radv_legacy_gs_info
|
||
- radv: move {esgs,gsvs}_ring_size to radv_legacy_gs_info
|
||
- radv/rt: bind the pipeline stack when it's not dynamic
|
||
- radv/ci: update CI lists for Polaris10 and Pitcairn
|
||
- radv: stop using get_vs_output_info() when emitting VS/NGG shaders
|
||
- radv: emit the GS copy shader outside of radv_pipeline_emit_hw_gs()
|
||
- radv: add radv_get_last_vgt_shader() helper
|
||
- radv: stop using the pipeline for emitting PS inputs
|
||
- radv: use the shader info stage to simplify emitting NGG shaders
|
||
- radv: use the ES type to apply a workaround for NGG on GFX10
|
||
- radv: pass the ES shader to radv_pipeline_emit_hw_ngg()
|
||
- radv: stop using the pipeline for emitting shaders
|
||
- radv: pass shader/base_reg to radv_emit_view_index_per_stage
|
||
- radv: pass a shaders array to radv_get_shader()
|
||
- radv: add radv_bind_shader() helper
|
||
- radv: add an assertion about shader stage to radv_bind_pre_rast_shader()
|
||
- radv: keep track of active stages as part of the cmdbuf state
|
||
- radv: determine the last VGT shader at pipeline bind time
|
||
- radv: stop using last_vgt_api_stage_{locs} during cmdbuf recording
|
||
- radv: move dirtying flags for mesh shading to radv_bind_pre_rast_shader()
|
||
- radv: copy bound shaders to the cmdbuf state
|
||
- radv: determine and store the next graphics stage to radv_shader_info
|
||
- radv: move user_data_0 to the shader info pass
|
||
- radv: replace pipeline->is_ngg occurrences during cmdbuf recording
|
||
- radv: replace pipeline->force_vrs_per_vertex during cmdbuf recording
|
||
- radv: use serialized NIR for graphics libs with the RETAIN flag
|
||
- radv: remove radv_graphics_pipeline::use_per_attribute_vb_descs
|
||
- radv: remove radv_graphics_pipeline::last_vertex_attrib_bit
|
||
- radv: remove radv_graphics_pipeline::next_vertex_stage
|
||
- radv: remove radv_graphics_pipeline::can_use_simple_input
|
||
- aco: remove unused aco_shader_info::vb_desc_usage_mask
|
||
- radv: adjust vb_desc_usage_mask for dynamic VS inputs in the info pass
|
||
- radv: remove radv_graphics_pipeline::vb_desc_usage_mask
|
||
- radv: remove radv_graphics_pipeline::vb_desc_alloc_size
|
||
- radv: rework emitting inner coverage when a fragment shader is bound
|
||
- radv: copy custom blend mode to the cmdbuf state
|
||
- radv: add a helper that returns the current rasterized primitive
|
||
- radv: copy rast_prim to the cmdbuf state
|
||
- radv: copy uses_{drawid,baseinstance} to the cmdbuf state
|
||
- radv: copy ia_multi_vgt_param to the cmdbuf state
|
||
- radv: add a helper to convert a VkPipelineBindPoint
|
||
- radv: copy need_indirect_descriptor_sets to radv_cmd_state
|
||
- radv: add push constant state to the cmdbuf state
|
||
- radv: fix sample shading when a new fragment shader is bound
|
||
- vulkan: add dynamic support for rectangles enable/mode
|
||
- radv: add dynamic support for rectangles enable/mode
|
||
- vulkan: Update XML and headers to 1.3.246
|
||
- radv: copy db_render_control to the cmdbuf state
|
||
- radv: set PS_ITER_SAMPLE(1) for sample shading during cmdbuf recording
|
||
- radv: configure PA_SC_MODE_CNTL_1 during cmdbuf recording
|
||
- radv: add the raygen shader BO to the cmdbuf list
|
||
- radv: fix binding raytracing/compute pipelines
|
||
- zink/ci: remove primitive-id-no-gs-quads from the NAVI10 fail list
|
||
- radv/ci: add one more flake
|
||
- radv: only copy non-NULL shaders when loaded from the cache
|
||
- radv: rely on non-NULL binaries when inserting shaders to the cache
|
||
- radv: allow to create/insert PS epilogs from/to the cache for libs
|
||
- radv: remove dead code in radv_pipeline_get_nir()
|
||
- radv: add VkGraphicsPipelineLibraryFlag to the graphics pipeline key
|
||
- radv: ensure to retain NIR shaders for GPL libs found in the cache
|
||
- radv: enable shaders cache for libraries with GPL
|
||
- radv: fix VS prologs with GPL and static binding stride
|
||
- radv: emit the PS epilog after the graphics pipeline
|
||
- radv: add a helper for retaining NIR shaders
|
||
- radv: move the serialized NIR to radv_graphics_lib_pipeline
|
||
- radv: simplify a check when retaining NIR shaders
|
||
- radv: do not retain noop FS for libs when a cache hit happened
|
||
- radv: import retained NIR shaders later in the compilation process
|
||
- radv/rt: stop storing unused hashes/identifiers
|
||
- radv: create a helper for copying VkPipelineShaderStageCreateInfo
|
||
- radv: copy stages instead of serializing NIR for GPL with the RETAIN flag
|
||
- radv: enable VK_EXT_graphics_pipeline_library by default
|
||
- radv/ci: update expected failures for PITCAIRN
|
||
- radv/ci: remove no longer existing tests for PITCAIRN
|
||
- radv/ci: update expected failures with BONAIRE
|
||
- docs: add more release notes for RADV
|
||
- radv: fix re-emitting vertex user SGPRs when binding a graphics pipeline
|
||
- radv/ci: remove one RT test from the expected failures on RDNA3
|
||
- radv: split radv_pipeline.c into radv_pipeline_{compute,graphics}.c
|
||
- radv: fix pipeline creation feedback with imported graphics libs
|
||
- radv: cleanup after splitting radv_pipeline.c
|
||
- radv: fix detecting FMASK_DECOMPRESS/DCC_DECOMPRESS meta pipelines
|
||
- vulkan: ignore rasterizationSamples when the state is dynamic
|
||
- radv: try to keep HTILE compressed for READ_ONLY_OPTIMAL layout
|
||
- radv: re-emit the guardband state when related PSO are bound
|
||
- radv: disable fast-clears with CMASK for 128-bit formats
|
||
- radv: do not allow 1D block-compressed images with (extended) storage on GFX6
|
||
- radv: fix usage flag for 3D compressed 128 bpp images on GFX9
|
||
- radv: update binning settings to work around GPU hangs
|
||
- radv/amdgpu: fix adding continue preambles and postambles BOs to the list
|
||
- radv: wait for occlusion queries in the resolve query shader
|
||
- radv: delay enabling/disabling occlusion queries at draw time
|
||
- radv: track DB_COUNT_CONTROL changes to avoid context rolls
|
||
- radv: add the perf counters BO to the preambles BO list
|
||
- radv: only enable extendedDynamicState3ConservativeRasterizationMode on GFX9+
|
||
- ac/nir: fix 8-bit/10-bit PS exports clamping
|
||
- radv: fix dynamic depth clamp enable support
|
||
- radv: fix fast-clearing images with VK_REMAINING_{ARRAY_LAYERS,MIP_LEVELS}
|
||
- radv: disable RB+ blend optimizations on GFX11 when a2c is enabled
|
||
|
||
Sarah Walker (1):
|
||
|
||
- pvr: Update FWIF transfer queue register structures
|
||
|
||
Sathishkumar S (8):
|
||
|
||
- radeonsi/vcn: add register definitions for JPEG 4.0.3
|
||
- radeonsi/vcn: use register versions for jpeg
|
||
- radeonsi/vcn: add support for picture crop on JPEG 4.0.3
|
||
- radeonsi/vcn: support ARGB/RGBA conversion on JPEG 4.0.3
|
||
- radeonsi/vcn: set jpeg reg version for gfx940
|
||
- radeonsi/vcn: reset to default value when ROI/FC is not used
|
||
- frontends/va: support crop region in jpeg decode
|
||
- radeonsi/vcn: enable RGBA/ARGB formats on gfx940 jpeg
|
||
|
||
Sebastian Wick (1):
|
||
|
||
- loader: do not check the mesa DRI_Mesa version if it was not found
|
||
|
||
Sergi Blanch Torne (8):
|
||
|
||
- ci: disable Collabora's LAVA lab for maintance
|
||
- Revert "ci: Collabora's LAVA lab for maintance"
|
||
- ci: Uprev kernel to 6.1.7
|
||
- ci: disable Collabora's LAVA lab for maintance
|
||
- ci: disable Collabora's LAVA lab for maintance
|
||
- ci: include setup test environment script in the output artifacts
|
||
- Revert "ci: disable Collabora's LAVA lab for maintance"
|
||
- ci: disable Collabora's LAVA lab for maintance
|
||
|
||
Sidney Just (4):
|
||
|
||
- zink: Fix non debug builds failing to compile on
|
||
- loader: Add missing brace to fix compile
|
||
- zink: add check for samplerMirrorClampToEdge Vulkan 1.2 feature
|
||
- zink: Add missing features to the profile file
|
||
|
||
Sil Vilerino (15):
|
||
|
||
- d3d12: Honor suggested driver profile/level for H264/HEVC encode
|
||
- d3d12: Video processing - Fix out of bounds array access
|
||
- d3d12: Video Encode - Fix ID3D12CommandAllocator leak
|
||
- d3d12: Fix VP9 Decode - Checking 0xFF instead of 0x7F for invalid frame_ref[i].Index7Bits
|
||
- frontend/va: Add format support checks for VA_RT_FORMAT_* in VaCreateConfig/VaGetConfigAttributes
|
||
- frontend/va: Remove duplicate code in format support checking/reporting.
|
||
- frontend/va: Keep track of some VP9 previous frame data for current frame use_prev_in_find_mvs_refs
|
||
- d3d12: VP9 Decode - Fix use_prev_in_find_mvs_refs calculation
|
||
- d3d12: Fix video decode for interlaced streams with reference only textures required
|
||
- d3d12: H264/HEVC Encode - Set both VBV InitialCapacity/Size in CBR Rate Control to same value when requested
|
||
- d3d12: Encode H264/HEVC - Do not write PPS unless different from active
|
||
- d3d12: Encode - Only upload headers when written headers size is > 0
|
||
- nir: Fix use of alloca() without #include c99_alloca.h
|
||
- Revert "d3d12: Honor suggested driver profile/level for H264/HEVC encode"
|
||
- d3d12: Video processor to only promote resources to permanent residency when there is work to be flushed
|
||
|
||
Simon Fels (2):
|
||
|
||
- venus: allow vtest socket being specified by env variable
|
||
- virgl/vtest: allow socket being specified by env variable
|
||
|
||
Simon Perretta (38):
|
||
|
||
- pvr: Add new Rogue compiler framework
|
||
- pvr: Add support for optional instruction params
|
||
- pvr: Support dual-destination ALU instructions
|
||
- pvr: Commonise some instruction member defs
|
||
- pvr: Drop the ENUM_PACKED macro
|
||
- pvr: Keep NIR SSA defs instead of registers
|
||
- pvr: Adjust instruction repeat offset
|
||
- pvr: Validate instruction repeat and src/dst sizes
|
||
- pvr: Add block printing support during validation
|
||
- pvr: Clarify unreachable text
|
||
- pvr: Add ADD64 support
|
||
- pvr: Add memory load support
|
||
- pvr: Add bitwise instruction support
|
||
- pvr: Additional register subarray support
|
||
- pvr: Support loading immediate values
|
||
- pvr: Load descriptors from memory
|
||
- pvr: Split pvr_private.h
|
||
- pvr: Use descriptor/set/table offsets from driver
|
||
- pvr: Add NIR pass to lower vars to SSA
|
||
- pvr: Amend subarray ownership code
|
||
- pvr: Add support for fitr.pixel
|
||
- pvr: Add support for sample instructions
|
||
- pvr: Add support for validating modifier combos
|
||
- pvr: Add support for emitpix
|
||
- pvr: Add support for WOP
|
||
- pvr: Register allocation improvements
|
||
- pvr: Fix descriptor set address calculation
|
||
- pvr: Add support for generating per-job EOT program
|
||
- pvr: Add support for generating NOP program
|
||
- pvr: Add support for IDF
|
||
- pvr: Add support for ST
|
||
- pvr: Add branch support
|
||
- pvr: Add support for TST
|
||
- pvr: Add basic support for manual instruction grouping
|
||
- pvr: Add support for MOVC
|
||
- pvr: Add late op lowering pass and conditional execution
|
||
- pvr: Amend definitions for ST and IDF
|
||
- pvr: Add encodings for index registers
|
||
|
||
Simon Ser (1):
|
||
|
||
- egl: fix fd_display_gpu on surfaceless and device platforms
|
||
|
||
Sonny Jiang (5):
|
||
|
||
- radeonsi: Add NV12 support for AV1
|
||
- gallium/pipe: change PIPE_DEFAULT_DECODER_FEEDBACK_TIMEOUT_NS to 1 second
|
||
- amd/common: Add gfx940 codec query support
|
||
- radeonsi/vcn: Add video capabilities support for gfx940
|
||
- radeonsi/vcn: Add decode support for gfx940
|
||
|
||
SoroushIMG (31):
|
||
|
||
- zink: add pass checking for lod overflow in txf
|
||
- zink: add zink_cs_key
|
||
- zink: add VK_EXT_image_robustness
|
||
- zink: add robust_access field to shader key
|
||
- zink: lower LOD-invalid txf when imageRobustAccess2 is missing
|
||
- zink: update gl43 profile to allow imageRobustAccess
|
||
- zink: fix sparse residency query and minLOD feature checks
|
||
- zink: fix cap check for arb sparse texture2
|
||
- zink: only save frag const buffers when used by blit
|
||
- zink: fix leak when rebinding same image surface
|
||
- zink: clear null image surfaces to 0
|
||
- zink: fix pointcoord y inversion
|
||
- zink: relax bresenhamLines requirement for non-strictLine drivers
|
||
- zink: fix compute shader leaks
|
||
- zink: allocate program shader caches from the program's mem ctx
|
||
- zink: stop creating pipeline library cache for non-optimal_key drivers
|
||
- zink: free resource objects' views array during destruction
|
||
- zink: fix stale point sprite mode state
|
||
- zink: fix shadow mask change logic when binding sampler views
|
||
- zink: track shadow swizzle for all shader stages
|
||
- zink: minor formatting change
|
||
- zink: add needs_zs_shader_swizzle shader key
|
||
- zink: extend shadow swizzle pass to all zs textures
|
||
- zink: add depth/stencil needs shader swizzle workaround field
|
||
- zink: workaround undefined swizzle 1 for z/s textures
|
||
- zink: rename shadow key to zs swizzle
|
||
- zink: Add driver name and API version to renderer name
|
||
- zink: do not emit line stipple dynamic state when emulating
|
||
- zink: take location_frac into account in lower_line_smooth_gs
|
||
- zink: fix incorrect line mode check for bresenham
|
||
- zink: refcount the correct query pool
|
||
|
||
Sui Jingfeng (1):
|
||
|
||
- meson: add basic support for loongarch
|
||
|
||
SureshGuttula (1):
|
||
|
||
- radeonsi: Add support for DPB resize
|
||
|
||
Sviatoslav Peleshko (9):
|
||
|
||
- anv: Handle VkAccelerationStructureBuildRangeInfoKHR::transformOffset
|
||
- driconf/anv: Apply limit_trig_input_range WA to Rise of the Tomb Raider
|
||
- iris: Avoid creating uncompressed view with unaligned tile offsets on BDW
|
||
- anv: Handle all fields in VkAccelerationStructureBuildRangeInfoKHR
|
||
- anv: Move WA MEDIA_VFE_STATE after stalling PIPE_CONTROL
|
||
- glsl: Fix codegen for constant ir_binop_{l,r}shift with mixed types
|
||
- isl: Check all channels in isl_formats_have_same_bits_per_channel
|
||
- anv: Handle UNDEFINED format in image format list
|
||
- anv: Improve image/view usage bits verification
|
||
|
||
Tapani Pälli (42):
|
||
|
||
- intel/compiler: add cpp_std=c++17 when building tests
|
||
- intel/hasvk: remove some stale comments, wa was removed
|
||
- anv: add restrictions for 3DSTATE_RASTER::AntiAliasingEnable
|
||
- hasvk: add restrictions for 3DSTATE_RASTER::AntiAliasingEnable
|
||
- iris: add restrictions for 3DSTATE_RASTER::AntiAliasingEnable
|
||
- mesa: move component bits queries as GL ES only
|
||
- intel/genxml: set unused 3DSTATE_PS_EXTRA field as mbz
|
||
- intel: enable existing workaround for ICL platform
|
||
- intel/blorp: disable REP16 for gfx12+ with R10G10B10_FLOAT_A2
|
||
- iris: disable preemption for 3DPRIMITIVE during streamout
|
||
- iris: handle error in iris_resource_from_handle
|
||
- spirv: add workaround for Metro Exodus in spirv_to_nir
|
||
- radv: revert Metro Exodus workaround which was moved to common code
|
||
- mesa/st: refactor st_destroy_texcompress_compute condition
|
||
- mesa/st: add astc decoder lookup tables
|
||
- mesa/st: initialize resources for ASTC decoding
|
||
- mesa: add astc decoder shader template (glsl es version)
|
||
- mesa/st: support compute shader decoding of ASTC
|
||
- anv: Wa_14016407139, add required pc when SBA programmed
|
||
- iris: implement emission of 3DSTATE_HS for Wa_1306463417
|
||
- anv: emit 3DSTATE_HS in cmd_buffer_flush_gfx_state
|
||
- anv: limit generated draws to pipelines without HS stage
|
||
- anv: implement emission of 3DSTATE_HS for Wa_1306463417
|
||
- iris: emit 3DSTATE_HS for each primitive on gfx12
|
||
- anv: emit 3DSTATE_HS for each primitive on gfx12
|
||
- intel/compiler: add comment about workaround on simd width
|
||
- anv: fix sends_count_expectation assert on simd32
|
||
- intel/isl: disable TILE64 for YCRCB formats
|
||
- anv: implement occlusion query related Wa_14017076903
|
||
- iris: implement occlusion query related Wa_14017076903
|
||
- intel/fs: restore message layout changes for cube array
|
||
- anv: use primitive ID override when shader does not supply it
|
||
- anv: take primitive ID override to account Wa_14015297576
|
||
- anv: check for MESA_SHADER_TESS_CTRL with get_tcs_prog_data
|
||
- intel/common: limit the amount of SLM with Wa_14017341140
|
||
- intel/fs: use intel_needs_workaround for Wa_22013689345
|
||
- intel/compiler: use intel_needs_workaround for Wa_14012437816
|
||
- isl: disable mcs (and mcs+ccs) for color msaa on gfxver 125
|
||
- iris: implement state cache invalidate for Wa_16013063087
|
||
- anv: cleanup bitmask construction for PIPELINE_SELECT
|
||
- anv: implement state cache invalidate for Wa_16013063087
|
||
- isl: fix layout for comparing surf and view properties
|
||
|
||
Tatsuyuki Ishi (22):
|
||
|
||
- radv: Fix depth-only-with-discard when epilogs are used.
|
||
- radv: Fix emitting tess indirect descriptors twice.
|
||
- radv: Loop over shader stages in flush_indirect_descriptor_sets.
|
||
- radv: Fix noop FS not getting constructed for GPL pipelines.
|
||
- radv: Fix missing rbplus_allowed check for dynamic PS epilogs.
|
||
- radv: Assert the hardware support rbplus when emitting rbplus state.
|
||
- radv: Keep shader code ptr in a separately allocated buffer.
|
||
- radv/sqtt: Use code buffer from radv_shader directly instead of copying.
|
||
- radv: Replace radv_trap_handler_shader with radv_shader.
|
||
- radeonsi: SDMA v4 size field is size - 1
|
||
- radv: SDMA v4 size field is size - 1
|
||
- radv: Remove SDMA padding from copy helpers.
|
||
- radv: Use common helpers to translate format in SDMA copy.
|
||
- radv/rt: Don't upload the prolog twice.
|
||
- radv: Use radeon_cmdbuf for sdma_copy_image.
|
||
- radv: Introduce sdma_copy_buffer for GFX7+.
|
||
- radv: Upload shaders to invisible VRAM on small BAR systems.
|
||
- radv: Wait for shader uploads asynchronously.
|
||
- radv: Fix missing wait of GS copy shader upload for dmashaders.
|
||
- amd: Add radv_foreach_stage to ForEachMacros.
|
||
- radv: Pre-compute descriptor set layout hash.
|
||
- ci/android: Make armv8's arch aarch64 instead of arm.
|
||
|
||
Teng, Jin Chung (2):
|
||
|
||
- frontend/va: Add large_scale_tile from VADecPictureParameterBufferAV1
|
||
- d3d12: AV1 Dec - Set anchor_frame_idx only when large_scale_tile equals 1
|
||
|
||
Thomas H.P. Andersen (4):
|
||
|
||
- docs/panvk: VK_KHR_descriptor_update_template
|
||
- meson: use summary()
|
||
- meson: use sections in summary()
|
||
- v3dv: use common code for descriptor update template
|
||
|
||
Thong Thai (6):
|
||
|
||
- gallium/auxiliary/vl: clean-up progressive shader
|
||
- radeonsi/vcn: use encoder/decoder caps reported by kernel
|
||
- gallium/auxiliary/vl: add crop to compute shader
|
||
- mesa/main: rework locale setup/teardown
|
||
- util: check and initialize locale before using it
|
||
- tgsi: use locale independent float and double parsing
|
||
|
||
Timothy Arceri (15):
|
||
|
||
- nir/nir_opt_copy_prop_vars: remove extra loop
|
||
- nir/nir_opt_copy_prop_vars: avoid comparison explosion
|
||
- nir/nir_opt_copy_prop_vars: reuse hash tables
|
||
- nir/nir_opt_copy_prop_vars: reuse dynamic arrays
|
||
- nir/nir_opt_copy_prop_vars: reorder clone calls
|
||
- nir/nir_opt_copy_prop_vars: don't call memset when cloning
|
||
- ci: enable dEQP-VK.ubo.random.all_shared_buffer.48
|
||
- glsl: copy prop vars before scalarizing alus
|
||
- glsl: add _token_list_prepend() helper to the parser
|
||
- glsl: isolate object macro replacments
|
||
- glsl: remove do_copy_propagation_elements() optimisation pass
|
||
- glsl: allow 64-bit integer on RHS of shift
|
||
- util/00-mesa-defaults: add Akka Arrh workaround
|
||
- mesa: add _mesa_is_api_gles2() helper
|
||
- glsl: move some GL ES checks to the NIR linker
|
||
|
||
Timur Kristóf (155):
|
||
|
||
- aco/optimizer: Add missing v_lshlrev condition to can_apply_extract.
|
||
- aco/optimizer: Optimize p_extract + v_mul_u32_u24 to v_mad_u32_u16.
|
||
- radv: Make NGG query emission a dirty flag.
|
||
- radv: Get rid of app_shaders_internal.
|
||
- radv, aco: Add uses_full_subgroups to compute shader info.
|
||
- aco: Enable constant exec mask based optimization on compute shaders.
|
||
- radv: Lower dynamic VS inputs in NIR.
|
||
- aco: Remove dynamic VS input loads.
|
||
- nir: Add pack_half_2x16_rtz_split opcode.
|
||
- radv, aco, ac: Implement pack_half_2x16_rtz_split.
|
||
- nir: Lower pack_half_2x16_split to RTZ if available.
|
||
- nir: Add algebraic optimization for VKD3D-Proton fp32->fp16 conversion.
|
||
- ac/gpu_info: Add has_pcie_bandwidth_info.
|
||
- radv: Don't place CS in VRAM when bandwidth is too low.
|
||
- nir/opt_algebraic: Add optimization for ieq/ine and right-shift.
|
||
- radv: Disable NGG culling when conservative overestimation is used.
|
||
- ac/nir/cull: Alway remove zero-area triangles in face culling.
|
||
- ac/nir/ngg: Include culled primitives in query.
|
||
- radv: Don't change LDS_SIZE for NGG culling shaders.
|
||
- radv: Move checking primitive topology to radv_get_ngg_culling_settings.
|
||
- radv: Use shader code to skip NGG culling in small workgroups.
|
||
- radv: Remove NGG culling skip from command buffer.
|
||
- radv: Refactor radv_emit_ngg_culling_state so it's based on dirty flags.
|
||
- nir: Clarify comment above load_buffer_amd.
|
||
- ac: Port ACO's get_fetch_format to ac_get_safe_fetch_size.
|
||
- ac/llvm: Remove "structurized" argument and instead check vindex.
|
||
- ac/llvm: Fix buffer_load_amd with larger than 32-bit channel sizes.
|
||
- ac/llvm: Fix ac_build_buffer_load to work with more than 4 channels.
|
||
- ac/llvm: Change ac_build_tbuffer_load to take format and channel type.
|
||
- radv: Move VS input lowering to new file: radv_nir_lower_vs_inputs.
|
||
- aco: Get rid of redundant load_vmem_mubuf function.
|
||
- aco: Don't set scalar offset on buffer load instructions when it's zero.
|
||
- aco: Remove MTBUF zero operand.
|
||
- radv: Call nir_lower_array_deref_of_vec in radv_lower_io_to_scalar_early.
|
||
- aco/optimizer: Change v_cmp with subgroup invocation to constant.
|
||
- radv: Emulate VGT_ESGS_ITEMSIZE in shaders on GFX9+.
|
||
- util: Add util_format_get_array.
|
||
- ac: Add pending_vmem field to args.
|
||
- radv: Set pending_vmem on dynamic VS input args.
|
||
- aco: Generalize vs_inputs to args_pending_vmem.
|
||
- aco, radv: Rename aco_*_key to aco_*_info.
|
||
- aco, radv: Move PS epilog and VS prolog args to their info structs.
|
||
- aco, radv: Don't use radv_shader_args in aco.
|
||
- aco: Don't include headers from radv.
|
||
- ac/nir: clear nir_var_shader_out from TCS barriers
|
||
- aco: Remove vtx_binding from MUBUF/MTBUF instructions.
|
||
- nir: Add load_typed_buffer_amd intrinsic.
|
||
- aco: Implement load_typed_buffer_amd.
|
||
- ac/llvm: Implement typed buffer load intrinsic.
|
||
- radv: Lower non-dynamic VS inputs in NIR.
|
||
- radv: Apply swizzle and alpha adjust in radv_nir_lower_vs_inputs.
|
||
- aco: Remove VS inputs from visit_load_input.
|
||
- aco: Rename visit_load_input to visit_load_fs_input.
|
||
- radv: Remove VS inputs code from LLVM backend.
|
||
- ac/llvm: Remove unused function ac_build_struct_tbuffer_load.
|
||
- aco, radv: Remove VS IO information from ACO.
|
||
- aco: Don't add soffset to swizzled MUBUF base.
|
||
- aco: Use zero for MUBUF/MTBUF when soffset is undefined.
|
||
- aco: Disable MUBUF/MTBUF offsets when they are zero.
|
||
- aco: Always enable idxen for swizzled buffer access on GFX11.
|
||
- ac/nir/ngg: Remove usused lds_es enum values.
|
||
- ac/nir/ngg: Rename saved_uniform to reusable_nondeferred_variable.
|
||
- ac/nir/ngg: Split some functions out of save_reusable_variables.
|
||
- ac/nir/ngg: Move divergence analysis call to analyze_shader_before_culling.
|
||
- ac/nir/ngg: Rename state variables to "s".
|
||
- ac/nir/ngg: Remove some superfluous variables.
|
||
- ac/nir/ngg: Create separate variable for repacked rel_patch_id.
|
||
- ac/nir/ngg: Rename repacked variables to clarify their name.
|
||
- ac: Add more defines for mesh shading packets.
|
||
- radv: Use new mesh shading packet defines.
|
||
- radv: Add per-prim attributes to ring_attr stride.
|
||
- radv: Use per-prim params in has_param_exports.
|
||
- radv: Add extra offset to per-prim params.
|
||
- radv: Use PRIM_ATTR for PS inputs on GFX11.
|
||
- radv: Include per-prim params in NUM_INTERP on GFX11.
|
||
- radv: Adjust mesh draw packets for GFX11.
|
||
- ac/nir/ngg: Clarify mesh shader scratch ring.
|
||
- ac/nir/ngg: Use attribute ring for mesh shader params.
|
||
- ac/nir/ngg: Split legacy workgroup index function.
|
||
- ac/nir/ngg: Fix mesh shader layer on GFX11.
|
||
- ac/nir/ngg: Store special MS outputs in attribute ring for PS to read.
|
||
- radv: Enable mesh shading on GFX11.
|
||
- radv: Fix swizzled VS input loads when some components are unused.
|
||
- radv: Don't expose NV_mesh_shader and don't use it in CI.
|
||
- radv: Remove NV_mesh_shader API entrypoints.
|
||
- radv: Remove first_task and ib_addr/ib_stride.
|
||
- radv: Clean up emitting zero mesh shader draw id.
|
||
- ac/nir/ngg: Remove NV_mesh_shader support.
|
||
- ac/nir: Remove ac_nir_apply_first_task_to_task_shader.
|
||
- nir: Remove IB address and stride intrinsics.
|
||
- radv: Move radv_nir_* to a new folder.
|
||
- radv: Move radv_nir_lower_primitive_shading_rate to new file.
|
||
- radv: Move radv_nir_lower_fs_intrinsics to new file.
|
||
- radv: Move radv_nir_lower_intrinsics_early to new file.
|
||
- radv: Move radv_nir_lower_view_index to new file.
|
||
- radv: Move radv_nir_lower_viewport_to_zero to new file.
|
||
- radv: Move radv_nir_export_multiview to new file.
|
||
- radv, ac/nir: Move sin/cos lowering to a common pass.
|
||
- radv: Move I/O lowering functions into a new file.
|
||
- radv: Use radv_get_shader to get vertex shader when binding pipeline.
|
||
- ac/nir/ngg: Slightly improve attribute ring offset calculation.
|
||
- ac/nir: Store only lowest 8 bits for task draw ring DWORD3.
|
||
- ac/nir: When task->mesh dispatch Y or Z are 0, also set X to 0.
|
||
- aco: Consider p_cbranch_nz as divergent branch too.
|
||
- aco: Don't remove exec writes that also write other registers.
|
||
- aco: Simplify get_phi_operand using Operand::c32_or_c64.
|
||
- aco: Don't verify branch exec read when eliminating exec writes.
|
||
- aco: Pop branch operands when targets are same in SSA elimination.
|
||
- aco: Call dominator_tree before lower_phis.
|
||
- aco: Better phi lowering for merge block when else-side is const.
|
||
- nir: Gather compile time constant task->mesh dispatch size.
|
||
- radv: Use linear_dispatch info in GFX11 task/mesh draw packet.
|
||
- radv/amdgpu: Extract CS chain and unchain functions.
|
||
- radv/amdgpu: Expose CS chain and unchain on the winsys.
|
||
- radv/amdgpu: Extract radv_amdgpu_add_cs_to_bo_list function.
|
||
- radv/amdgpu: Remember which CS the current one is chained to.
|
||
- radv/amdgpu: Walk chained CS objects for BO list.
|
||
- radv/amdgpu: Unchain CS array in queue code not in winsys.
|
||
- radv: Chain cmd buffers in queue code when possible, not in winsys.
|
||
- radv/amdgpu: Remove can_patch and chained submit code path.
|
||
- ac/llvm: Cover runtime 0 in GFX10 gs_alloc_req workaround.
|
||
- aco: Fix optimization of v_cmp with subgroup invocation.
|
||
- aco: Don't use nir_selection_control in aco_ir.
|
||
- aco: Only include nir.h in instruction selection.
|
||
- radv: Don't include nir.h in radv_shader.h
|
||
- radv: Create continue preamble on GFX6 even when no shader rings are used.
|
||
- ac: Add maximum number of submitted IBs.
|
||
- radv/amdgpu: Fix mismatching return type of radv_amdgpu_cs_submit.
|
||
- radv/amdgpu: Only allow IB BOs on graphics and compute queues.
|
||
- radv/amdgpu: Use correct alignment when creating CS BOs.
|
||
- radv/amdgpu: Extract radv_amdgpu_cs_add_old_ib_buffer.
|
||
- radv/amdgpu: Add a few assertions during submit.
|
||
- radv/amdgpu: Remove hw_can_chain in favour of use_ib.
|
||
- radv/amdgpu: Rewrite fallback code path so it can split submissions.
|
||
- radv/amdgpu: Allow multiple continue preambles.
|
||
- radv/amdgpu: Add continue preambles to fallback submit.
|
||
- radv/amdgpu: Add postambles to fallback submit.
|
||
- radv/amdgpu: Add ability to submit non-chained CS to fallback.
|
||
- radv/amdgpu: Split gang submissions correctly when not chained.
|
||
- radv: Fill continue preambles and postambles properly.
|
||
- radv: Split submission in winsys instead of radv_queue.
|
||
- radv/amdgpu: Use fallback submit for queues that can't use IBs.
|
||
- radv/amdgpu: Clean up submission functions.
|
||
- radv/amdgpu: Respect maximum number of submitted IBs per IP type.
|
||
- radv: Allow task/mesh shaders with RADV_DEBUG=noibs.
|
||
- radv/amdgpu: Add bool is_secondary argument to cs_create function.
|
||
- radv/amdgpu: Extract radv_amdgpu_cs_bo_create function.
|
||
- radv/amdgpu: Place secondary CS without IB2 in non-WC GTT.
|
||
- ac, aco, radv: Clarify LDS size on GFX6, and NGG shaders.
|
||
- radv: Don't hardcode LDS granularity in gfx9_get_gs_info.
|
||
- aco: Remove setup_*_variables and add setup_lds_size instead.
|
||
- aco, radv: Remove "key" from aco_compiler_options.
|
||
- aco, radv: Remove redundant enable_mrt_output_nan_fixup from PS epilog info.
|
||
- ac/nir/ngg: Don't store primitive IDs from culled primitives.
|
||
- aco: Disallow constant propagation on SOPP and fixed operands.
|
||
|
||
Tomeu Vizoso (7):
|
||
|
||
- android: Make libbacktrace optional again
|
||
- android: Cleanup unneeded headers from the sync stub
|
||
- ci: Build for Android with libbacktrace=false
|
||
- ci: Use NDK 25b to build for the Android ABI level 33
|
||
- etnaviv: handle missing alu conversion opcodes
|
||
- etnaviv: print writemask of store operations
|
||
- etnaviv: don't read too much from uniform arrays
|
||
|
||
Turo Lamminen (4):
|
||
|
||
- radv: Change radeon_cmdbuf counters to uint64_t to make alias analysis optimize radeon_emit better
|
||
- radv: Clean up variables in si_get_ia_multi_vgt_param
|
||
- radv: Avoid redundant fetch of radv_device
|
||
- radv: Optimize emitting prefetches
|
||
|
||
Val Packett (1):
|
||
|
||
- mailmap: Remap name and email for Val Packett
|
||
|
||
Vincent Davis Jr (1):
|
||
|
||
- gbm/backend: fix gbm compile without dri
|
||
|
||
Vinson Lee (2):
|
||
|
||
- radv: Fix memory leak.
|
||
- pps: Fix build errors.
|
||
|
||
Vitaliy Triang3l Kuzmin (2):
|
||
|
||
- radv: Set DB_Z_INFO.NUM_SAMPLES to MSAA_EXPOSED_SAMPLES without Z/S
|
||
- r600: Alpha to coverage dithering on Evergreen+
|
||
|
||
Väinö Mäkelä (12):
|
||
|
||
- intel/vec4: Set the rounding mode
|
||
- intel/vec4: Don't optimize multiply by 1.0 away
|
||
- hasvk: Don't claim shaderDenormPreserveFloat32 on gfx7
|
||
- hasvk: Tell spirv_to_nir float controls are always supported
|
||
- hasvk: Enable PixelShaderKillsPixel when omask is used
|
||
- hasvk: Mark VK_IMAGE_LAYOUT_ATTACHMENT_OPTIMAL as stencil write optimal
|
||
- hasvk: Handle subpass self-dependencies for stencil shadow copies
|
||
- ci/intel: Update hasvk HSW xfails
|
||
- hasvk: Disable non-zero fast clears for 8xMSAA images
|
||
- intel/ci: Remove outdated hasvk copy_and_blit xfails
|
||
- intel/ci: Remove hasvk xfails missing from the CTS
|
||
- intel/ci: Remove skipped float_controls tests from hasvk xfails
|
||
|
||
X512 (3):
|
||
|
||
- hgl: remove
|
||
- haiku: fix build
|
||
- EGL: implement Haiku driver over Gallium
|
||
|
||
Yevhenii Kolesnikov (2):
|
||
|
||
- nir/loop_analyze: Track induction variables incremented by more operations
|
||
- nir/loop_analyze: Determine iteration counts for more kinds of loops
|
||
|
||
Yiwei Zhang (46):
|
||
|
||
- venus: log upon device creation
|
||
- venus: lazily query and cache gralloc front rendering usage
|
||
- venus: disable non AHB external memory bits on Android
|
||
- venus: fix formating
|
||
- venus: fix tracing init to include instance creation
|
||
- venus: render server enforces blob_id_0
|
||
- venus: move exp features init back to use ring submit
|
||
- venus: further disallow sparse resource
|
||
- venus: replace binary search with hardcode for max buffer size
|
||
- venus: start requiring all experimental features
|
||
- venus: clean up memoryResourceAllocationSize
|
||
- venus: clean up globalFencing
|
||
- venus: refactor sync fd fence and sempahore features
|
||
- venus: tighten up the sync fd requirements for Android wsi
|
||
- venus: distinguish external memory from mappable memory support
|
||
- venus: fix external buffer creation
|
||
- venus: remove redundant abstractions for wsi struct search
|
||
- venus: refactor image create info pnext tracking
|
||
- venus: simplify ahb image creation
|
||
- venus: simplify support for non-AHB external images
|
||
- venus: fix external image creation
|
||
- venus: fix device memory export alloc info
|
||
- venus: fix VK_EXT_image_view_min_lod feature query
|
||
- venus: ensure invariance of buffer memory requirement size
|
||
- venus: sync to latest protocol for ring status enum
|
||
- venus: abort ring submit when ring is in fatal status
|
||
- venus: propagate vn_ring to vn_relax
|
||
- venus: vn_relax to abort on ring fatal status upon warn order
|
||
- venus: revert back the warn order
|
||
- venus: sync to latest protocol for asyncRoundtrip
|
||
- venus: switch to use 64bit roundtrip seqno
|
||
- venus: make vn_instance_wait_roundtrip asynchronous
|
||
- venus: let vn_instance_submit_command track ring seqno
|
||
- venus: make common wsi bo submission async
|
||
- venus: refactor to add vn_sync_payload_external
|
||
- venus: make external fence and semaphore export async
|
||
- Revert "zink/kopper: Add extra swapchain images for Venus"
|
||
- venus: sync latest protocol for layering extensions
|
||
- venus: add VK_EXT_load_store_op_none support
|
||
- venus: add VK_EXT_rasterization_order_attachment_access support
|
||
- venus/docs: sync to latest venus supported extensions
|
||
- venus: requires asyncRoundtrip
|
||
- venus: requires ringMonitoring
|
||
- venus: move exp feature init back to ring and remove unused function
|
||
- venus: forward ARM driverVersion for ANGLE workarounds
|
||
- radv: respect VK_QUERY_RESULT_WAIT_BIT in GetQueryPoolResults
|
||
|
||
Yogesh Mohan Marimuthu (18):
|
||
|
||
- egl: add render_gpu tag to dri2_dpy->fd and dri2_dpy->dri_screen variable
|
||
- loader,glx: add render_gpu tag psc->driScreen and psc->fd
|
||
- loader,glx,egl,vl,d3d: loader_get_user_preferred_fd() function to return original_fd
|
||
- egl: remove is_different_gpu variable from struct dri2_egl_display
|
||
- glx: remove is_different_gpu variable from struct dri_screen
|
||
- loader,glx,egl: remove is_different_gpu variable from loader
|
||
- ac,radeonsi: move shadow regs create ib preamble function to amd common
|
||
- radv: add shadowregs variable to RADV_DEBUG environment variable
|
||
- radv: add support for register shadowing
|
||
- radv: set preemp flag and pre_ena bit for shadowregs
|
||
- radv: INDEX_TYPE and NUM_INSTANCES PKT3 are not shadowed
|
||
- radv: fence complete struct is 4 qw size
|
||
- radv: allow NULL initial_preamble_cs in radv_amdgpu_winsys_cs_submit_sysmem()
|
||
- radeonsi: remove some shadow reg optimization for bf1 game
|
||
- wsi/display: check alloc failure in wsi_display_alloc_connector()
|
||
- ac/surface: only adjust pitch if surf_pitch was modified
|
||
- amd/surface: add RADEON_SURF_NO_TEXTURE flag
|
||
- radv: set RADEON_SURF_NO_TEXTURE flag in radv_get_surface_flags()
|
||
|
||
Yogesh Mohanmarimuthu (7):
|
||
|
||
- egl: add fd_display_gpu to struct dri2_egl_display
|
||
- egl,egl/x11: keep display fd open for prime
|
||
- egl: create DRI screen for display GPU in case of prime
|
||
- loader,glx,egl/x11: init dri_screen_display_gpu in struct loader_dri3_drawable
|
||
- egl/wayland: keep display fd open for prime
|
||
- loader: make image_format_to_fourcc() non-static
|
||
- egl/wayland: for prime, allocate linear_copy from display GPU VRAM
|
||
|
||
Yonggang Luo (8):
|
||
|
||
- util: Implement util_iround with lrintf unconditionally
|
||
- util: Fixes error: no previous prototype for 'mesa_cache_db_entry_remove' Fixes: c92c99481fd ("util/mesa-db: Support removal of cache entries")
|
||
- vulkan: Use static_assert for check HWVULKAN_DISPATCH_MAGIC == ICD_LOADER_MAGIC
|
||
- meson: Split c_cpp_args from pre_args
|
||
- meson: Combine duplicated c_args and cpp_args
|
||
- meson: When sse2 enabled, both c and cpp using sse2 options
|
||
- meson: Split sse2_arg and sse2_args out of c_cpp_args
|
||
- meson: Use sse2_arg and sse2_args to replace usage of c and c_sse2_args
|
||
|
||
Yusuf Khan (2):
|
||
|
||
- nvc0/nv50: support and enable EXT_memory_object*
|
||
- gallium: create query_memory_info implementation for sw drivers
|
||
|
||
Yuxuan Shui (1):
|
||
|
||
- loader: unregister special event in loader_dri3_drawable_fini
|
||
|
||
antonino (54):
|
||
|
||
- zink: fix line smooth lowering
|
||
- zink: add \`zink_emulate_point_smooth` driconf
|
||
- zink: add \`lower_point_smooth` to \`zink_fs_key`
|
||
- zink/nir_to_spirv: add support for \`nir_intrinsic_load_point_coord`
|
||
- nir: handle output beeing written to deref in \`nir_lower_point_smooth`
|
||
- zink: handle point_smooth emulation
|
||
- drirc: set \`zink_emulate_point_smooth` for Quake II
|
||
- zink: fix stipple pattern in oblique lines
|
||
- zink: fix \`final_hash` update in \`zink_gfx_program_update`
|
||
- mesa: correctly allocate space for converted primtives
|
||
- gallium: decompose quad strips into quads if supported
|
||
- zink: handle switching between primitives
|
||
- nir: handle primitives with adjacency
|
||
- nir: avoid generating conflicting output variables
|
||
- nir: calculate number of vertices in nir_create_passthrough_gs
|
||
- nir: handle edge flags in nir_create_passthrough_gs
|
||
- zink: add \`has_edgeflags` flag to zink_shader and zink_gfx_program
|
||
- zink: handle edgeflags
|
||
- nir: allow to force line strip out in nir_create_passthrough_gs
|
||
- zink: force line strip out when emulating stipple
|
||
- zink: filled quad emulation gs generation function
|
||
- zink: add \`zink_rast_prim` enum
|
||
- zink: handle quads
|
||
- zink: fix flat shading on filled quads
|
||
- zink: add flags to \`zink_gfx_program` and \`zink_context`
|
||
- zink: add \`needs_inlining` to \`zink_shader`
|
||
- zink: implement flat shading using inlined uniforms
|
||
- nir/zink: handle provoking vertex mode in \`nir_create_passthrough_gs`
|
||
- zink: handle provoking vertex mode for filled quads
|
||
- nir: keep xfb properties in nir_create_passthrough_gs
|
||
- zink: keep xfb properties in quad emulation gs
|
||
- zink: advertise support for the quad primitive
|
||
- zink: prevent crash when freeing
|
||
- zink: unified \`zink_set_primitive_emulation_keys` and \`zink_create_primitive_emulation_gs`
|
||
- zink: zink: add \`parent` to \`zink_shader::non_fs`
|
||
- zink: improve generated gs unbinding
|
||
- zink: unbind generated gs in \`bind_last_vertex_stage`
|
||
- zink/ci: remove \`primitive-id-no-gs-quads` from radv-vangogh-fails
|
||
- nir: only handle flat interpolation when needed in \`nir_create_passthrough_gs`
|
||
- zink: simplify logic to call \`zink_set_primitive_emulation_keys`
|
||
- zink: add field to 'zink_gs_key' and enum
|
||
- zink: add provoking vertex mode lowering
|
||
- zink: always advertize provoking vertex mode support
|
||
- zink: update requirements now that pv mode can be emulated
|
||
- zink: add \`descriptor_bindless_id` to \`zink_shader_info`
|
||
- zink: fix sampler array collision in \`nir_to_spirv`
|
||
- zink: don't emulate edgeflags for patches
|
||
- zink: use correct primitives for passthrough gs with tess
|
||
- zink: fix pv mode lowring index calculation
|
||
- zink: use ring buffer to preserve last element
|
||
- zink: fix exit condition on pv emulation loop
|
||
- zink: fix line strip offsets in pv mode emulation
|
||
- zink: fix store subsitution in \`lower_pv_mode_gs_store`
|
||
- zink: take location_frac into account in pv emulation
|
||
|
||
driver1998 (1):
|
||
|
||
- gallium: Use DETECT_OS_WINDOWS instead of 'WIN32'
|
||
|
||
osy (1):
|
||
|
||
- virgl: enable timer queries only if host supports it
|
||
|
||
t0b3 (1):
|
||
|
||
- nir/nir_opt_move: fix ALWAYS_INLINE compiler error
|
||
|
||
volodymyr.o (1):
|
||
|
||
- mesa ctx->API --> _mesa_is_foo(ctx)
|
||
|
||
xurui (2):
|
||
|
||
- panfrost: Check the return value of drmGetVersion
|
||
- zink: bs->dd.push_pool[1].pool should be freed
|