129 lines
2.5 KiB
Text
129 lines
2.5 KiB
Text
|
- place a table in rodata
|
||
|
|
- use _CET_ENDBR in functions header
|
||
|
|
|
||
|
|
Index: libavcodec/x86/fft.asm
|
||
|
|
--- libavcodec/x86/fft.asm.orig
|
||
|
|
+++ libavcodec/x86/fft.asm
|
||
|
|
@@ -325,6 +325,7 @@ INIT_YMM avx
|
||
|
|
%if HAVE_AVX_EXTERNAL
|
||
|
|
align 16
|
||
|
|
fft8_avx:
|
||
|
|
+ _CET_ENDBR
|
||
|
|
mova m0, Z(0)
|
||
|
|
mova m1, Z(1)
|
||
|
|
T8_AVX m0, m1, m2, m3, m4
|
||
|
|
@@ -335,6 +336,7 @@ fft8_avx:
|
||
|
|
|
||
|
|
align 16
|
||
|
|
fft16_avx:
|
||
|
|
+ _CET_ENDBR
|
||
|
|
mova m2, Z(2)
|
||
|
|
mova m3, Z(3)
|
||
|
|
T4_SSE m2, m3, m7
|
||
|
|
@@ -372,6 +374,7 @@ fft16_avx:
|
||
|
|
|
||
|
|
align 16
|
||
|
|
fft32_avx:
|
||
|
|
+ _CET_ENDBR
|
||
|
|
call fft16_avx
|
||
|
|
|
||
|
|
mova m0, Z(4)
|
||
|
|
@@ -396,6 +399,7 @@ fft32_avx:
|
||
|
|
ret
|
||
|
|
|
||
|
|
fft32_interleave_avx:
|
||
|
|
+ _CET_ENDBR
|
||
|
|
call fft32_avx
|
||
|
|
mov r2d, 32
|
||
|
|
.deint_loop:
|
||
|
|
@@ -419,6 +423,7 @@ INIT_XMM sse
|
||
|
|
align 16
|
||
|
|
fft4_avx:
|
||
|
|
fft4_sse:
|
||
|
|
+ _CET_ENDBR
|
||
|
|
mova m0, Z(0)
|
||
|
|
mova m1, Z(1)
|
||
|
|
T4_SSE m0, m1, m2
|
||
|
|
@@ -428,6 +433,7 @@ fft4_sse:
|
||
|
|
|
||
|
|
align 16
|
||
|
|
fft8_sse:
|
||
|
|
+ _CET_ENDBR
|
||
|
|
mova m0, Z(0)
|
||
|
|
mova m1, Z(1)
|
||
|
|
T4_SSE m0, m1, m2
|
||
|
|
@@ -442,6 +448,7 @@ fft8_sse:
|
||
|
|
|
||
|
|
align 16
|
||
|
|
fft16_sse:
|
||
|
|
+ _CET_ENDBR
|
||
|
|
mova m0, Z(0)
|
||
|
|
mova m1, Z(1)
|
||
|
|
T4_SSE m0, m1, m2
|
||
|
|
@@ -465,6 +472,7 @@ fft16_sse:
|
||
|
|
%macro FFT48_3DNOW 0
|
||
|
|
align 16
|
||
|
|
fft4 %+ SUFFIX:
|
||
|
|
+ _CET_ENDBR
|
||
|
|
T2_3DNOW m0, m1, Z(0), Z(1)
|
||
|
|
mova m2, Z(2)
|
||
|
|
mova m3, Z(3)
|
||
|
|
@@ -479,6 +487,7 @@ fft4 %+ SUFFIX:
|
||
|
|
|
||
|
|
align 16
|
||
|
|
fft8 %+ SUFFIX:
|
||
|
|
+ _CET_ENDBR
|
||
|
|
T2_3DNOW m0, m1, Z(0), Z(1)
|
||
|
|
mova m2, Z(2)
|
||
|
|
mova m3, Z(3)
|
||
|
|
@@ -532,6 +541,7 @@ FFT48_3DNOW
|
||
|
|
%macro DECL_PASS 2+ ; name, payload
|
||
|
|
align 16
|
||
|
|
%1:
|
||
|
|
+ _CET_ENDBR
|
||
|
|
DEFINE_ARGS zc, w, n, o1, o3
|
||
|
|
lea o3q, [nq*3]
|
||
|
|
lea o1q, [nq*8]
|
||
|
|
@@ -548,10 +558,6 @@ DEFINE_ARGS zc, w, n, o1, o3
|
||
|
|
%macro FFT_DISPATCH 2; clobbers 5 GPRs, 8 XMMs
|
||
|
|
lea r2, [dispatch_tab%1]
|
||
|
|
mov r2, [r2 + (%2q-2)*gprsize]
|
||
|
|
-%ifdef PIC
|
||
|
|
- lea r3, [$$]
|
||
|
|
- add r2, r3
|
||
|
|
-%endif
|
||
|
|
call r2
|
||
|
|
%endmacro ; FFT_DISPATCH
|
||
|
|
|
||
|
|
@@ -731,11 +737,7 @@ DECL_PASS pass_interleave_3dnow, PASS_BIG 0
|
||
|
|
%define pass_interleave_3dnowext pass_interleave_3dnow
|
||
|
|
%endif
|
||
|
|
|
||
|
|
-%ifdef PIC
|
||
|
|
-%define SECTION_REL - $$
|
||
|
|
-%else
|
||
|
|
%define SECTION_REL
|
||
|
|
-%endif
|
||
|
|
|
||
|
|
%macro DECL_FFT 1-2 ; nbits, suffix
|
||
|
|
%ifidn %0, 1
|
||
|
|
@@ -759,6 +761,7 @@ DECL_PASS pass_interleave_3dnow, PASS_BIG 0
|
||
|
|
|
||
|
|
align 16
|
||
|
|
fft %+ n %+ fullsuffix:
|
||
|
|
+ _CET_ENDBR
|
||
|
|
call fft %+ n2 %+ SUFFIX
|
||
|
|
add r0, n*4 - (n&(-2<<%1))
|
||
|
|
call fft %+ n4 %+ SUFFIX
|
||
|
|
@@ -773,8 +776,10 @@ fft %+ n %+ fullsuffix:
|
||
|
|
%endrep
|
||
|
|
%undef n
|
||
|
|
|
||
|
|
+[SECTION .data.rel.ro write]
|
||
|
|
align 8
|
||
|
|
dispatch_tab %+ fullsuffix: pointer list_of_fft
|
||
|
|
+__?SECT?__
|
||
|
|
%endmacro ; DECL_FFT
|
||
|
|
|
||
|
|
%if HAVE_AVX_EXTERNAL
|