[FFmpeg-cvslog] aarch64: Use .data.rel.ro for const data with relocations
Martin Storsjö
git at videolan.org
Tue Dec 9 12:22:23 CET 2014
ffmpeg | branch: master | Martin Storsjö <martin at martin.st> | Sun Nov 16 12:52:21 2014 +0200| [780cd20b00a69e26bbfffbb8eec16fbe999ea793] | committer: Martin Storsjö
aarch64: Use .data.rel.ro for const data with relocations
This reverts commit c00365b46d464ce47716315c1801818d811bdb9a
in addition to using a different section.
Signed-off-by: Martin Storsjö <martin at martin.st>
> http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=780cd20b00a69e26bbfffbb8eec16fbe999ea793
---
libavcodec/aarch64/fft_neon.S | 35 +++++++++++++++++-----------------
libavcodec/aarch64/opus_imdct_neon.S | 20 +++++++++----------
libavutil/aarch64/asm.S | 10 ++++++++--
3 files changed, 35 insertions(+), 30 deletions(-)
diff --git a/libavcodec/aarch64/fft_neon.S b/libavcodec/aarch64/fft_neon.S
index ae9e1c6..e205e23 100644
--- a/libavcodec/aarch64/fft_neon.S
+++ b/libavcodec/aarch64/fft_neon.S
@@ -376,8 +376,7 @@ function ff_fft_calc_neon, export=1
ld1 {v30.16b}, [x10]
mov x7, #-8
movrel x12, pmmp
- ldr x4, [x3, x2, lsl #3]
- add x3, x3, x4
+ ldr x3, [x3, x2, lsl #3]
movrel x13, mppm
movrel x14, X(ff_cos_16)
ld1 {v31.16b}, [x11]
@@ -416,22 +415,22 @@ function ff_fft_permute_neon, export=1
ret
endfunc
-const fft_tab_neon
- .quad fft4_neon - fft_tab_neon
- .quad fft8_neon - fft_tab_neon
- .quad fft16_neon - fft_tab_neon
- .quad fft32_neon - fft_tab_neon
- .quad fft64_neon - fft_tab_neon
- .quad fft128_neon - fft_tab_neon
- .quad fft256_neon - fft_tab_neon
- .quad fft512_neon - fft_tab_neon
- .quad fft1024_neon - fft_tab_neon
- .quad fft2048_neon - fft_tab_neon
- .quad fft4096_neon - fft_tab_neon
- .quad fft8192_neon - fft_tab_neon
- .quad fft16384_neon - fft_tab_neon
- .quad fft32768_neon - fft_tab_neon
- .quad fft65536_neon - fft_tab_neon
+const fft_tab_neon, relocate=1
+ .quad fft4_neon
+ .quad fft8_neon
+ .quad fft16_neon
+ .quad fft32_neon
+ .quad fft64_neon
+ .quad fft128_neon
+ .quad fft256_neon
+ .quad fft512_neon
+ .quad fft1024_neon
+ .quad fft2048_neon
+ .quad fft4096_neon
+ .quad fft8192_neon
+ .quad fft16384_neon
+ .quad fft32768_neon
+ .quad fft65536_neon
endconst
const pmmp, align=4
diff --git a/libavcodec/aarch64/opus_imdct_neon.S b/libavcodec/aarch64/opus_imdct_neon.S
index c98931c..d99edf4 100644
--- a/libavcodec/aarch64/opus_imdct_neon.S
+++ b/libavcodec/aarch64/opus_imdct_neon.S
@@ -438,8 +438,8 @@ function fft_b15_calc_neon
uzp1 v12.4s, v4.4s, v5.4s // exp[11 - 14].re
uzp2 v13.4s, v4.4s, v5.4s // exp[11 - 14].im
zip1 v14.4s, v6.4s, v7.4s // exp[5,10].re/exp[5,10].im
- ldr x6, [x5, x3, lsl #3]
- add x5, x5, x6
+ add x5, x5, x3, lsl #3
+ ldr x5, [x5]
mov x10, x0
blr x5
ldp x20, x30, [sp]
@@ -451,14 +451,14 @@ function fft_b15_calc_neon
ret
endfunc
-const fft_tab_neon
- .quad fft15_neon - fft_tab_neon
- .quad fft30_neon - fft_tab_neon
- .quad fft60_neon - fft_tab_neon
- .quad fft120_neon - fft_tab_neon
- .quad fft240_neon - fft_tab_neon
- .quad fft480_neon - fft_tab_neon
- .quad fft960_neon - fft_tab_neon
+const fft_tab_neon, relocate=1
+ .quad fft15_neon
+ .quad fft30_neon
+ .quad fft60_neon
+ .quad fft120_neon
+ .quad fft240_neon
+ .quad fft480_neon
+ .quad fft960_neon
endconst
function ff_celt_imdct_half_neon, export=1
diff --git a/libavutil/aarch64/asm.S b/libavutil/aarch64/asm.S
index b766e67..6a7f506 100644
--- a/libavutil/aarch64/asm.S
+++ b/libavutil/aarch64/asm.S
@@ -52,12 +52,18 @@ FUNC .func \name
.endif
.endm
-.macro const name, align=2
+.macro const name, align=2, relocate=0
.macro endconst
ELF .size \name, . - \name
.purgem endconst
.endm
-#ifndef __MACH__
+#if HAVE_SECTION_DATA_REL_RO
+.if \relocate
+ .section .data.rel.ro
+.else
+ .section .rodata
+.endif
+#elif !defined(__MACH__)
.section .rodata
#else
.const_data
More information about the ffmpeg-cvslog
mailing list