* [FFmpeg-devel] [PATCH] Revert "avutil/tx_template: extend to 2M" (PR #21184)
@ 2025-12-13 7:31 Zhao Zhili via ffmpeg-devel
0 siblings, 0 replies; only message in thread
From: Zhao Zhili via ffmpeg-devel @ 2025-12-13 7:31 UTC (permalink / raw)
To: ffmpeg-devel; +Cc: Zhao Zhili
PR #21184 opened by Zhao Zhili (quink)
URL: https://code.ffmpeg.org/FFmpeg/FFmpeg/pulls/21184
Patch URL: https://code.ffmpeg.org/FFmpeg/FFmpeg/pulls/21184.patch
This reverts commit 8f48a62, 9af8782, and bd3e71b.
Commit 8f48a62 extends tx to 2M, resulting in the tx_float bss
section reaching a size of 4M.
This isn't a issue on devices with normal memory sizes and OS
supporting virtual memory. But it's a real issue for embedded devices
with realtime OS, which may not support virtual memory, e.g., Nuttx.
This 4M of bss section map to physical memory directly, which is a
scarce resource on embedded devices.
>From 07b26f98e9eb3a2429a439a72e90ac29f28c1658 Mon Sep 17 00:00:00 2001
From: Zhao Zhili <zhilizhao@tencent.com>
Date: Sat, 13 Dec 2025 12:53:58 +0800
Subject: [PATCH] Revert "avutil/tx_template: extend to 2M"
This reverts commit 8f48a62, 9af8782, and bd3e71b.
Commit 8f48a62 extends tx to 2M, resulting in the tx_float bss
section reaching a size of 4M.
This isn't a issue on devices with normal memory sizes and OS
supporting virtual memory. But it's a real issue for embedded devices
with realtime OS, which may not support virtual memory, e.g., Nuttx.
This 4M of bss section map to physical memory directly, which is a
scarce resource on embedded devices.
---
libavutil/tx_template.c | 12 ------------
libavutil/x86/tx_float.asm | 8 ++------
libavutil/x86/tx_float_init.c | 18 +++++++++---------
3 files changed, 11 insertions(+), 27 deletions(-)
diff --git a/libavutil/tx_template.c b/libavutil/tx_template.c
index ec630954b8..ecee572e40 100644
--- a/libavutil/tx_template.c
+++ b/libavutil/tx_template.c
@@ -45,10 +45,6 @@
SR_TABLE(32768) \
SR_TABLE(65536) \
SR_TABLE(131072) \
- SR_TABLE(262144) \
- SR_TABLE(524288) \
- SR_TABLE(1048576) \
- SR_TABLE(2097152) \
#define SR_TABLE(len) \
TABLE_DEF(len, len/4 + 1);
@@ -724,10 +720,6 @@ DECL_SR_CODELET(16384,8192,4096)
DECL_SR_CODELET(32768,16384,8192)
DECL_SR_CODELET(65536,32768,16384)
DECL_SR_CODELET(131072,65536,32768)
-DECL_SR_CODELET(262144,131072,65536)
-DECL_SR_CODELET(524288,262144,131072)
-DECL_SR_CODELET(1048576,524288,262144)
-DECL_SR_CODELET(2097152,1048576,524288)
static av_cold int TX_NAME(ff_tx_fft_init)(AVTXContext *s,
const FFTXCodelet *cd,
@@ -2160,10 +2152,6 @@ const FFTXCodelet * const TX_NAME(ff_tx_codelet_list)[] = {
&TX_NAME(ff_tx_fft32768_ns_def),
&TX_NAME(ff_tx_fft65536_ns_def),
&TX_NAME(ff_tx_fft131072_ns_def),
- &TX_NAME(ff_tx_fft262144_ns_def),
- &TX_NAME(ff_tx_fft524288_ns_def),
- &TX_NAME(ff_tx_fft1048576_ns_def),
- &TX_NAME(ff_tx_fft2097152_ns_def),
/* Prime factor codelets */
&TX_NAME(ff_tx_fft3_ns_def),
diff --git a/libavutil/x86/tx_float.asm b/libavutil/x86/tx_float.asm
index c030147ce8..87be21c2d6 100644
--- a/libavutil/x86/tx_float.asm
+++ b/libavutil/x86/tx_float.asm
@@ -46,7 +46,7 @@
%endif
%assign i 16
-%rep 18
+%rep 14
cextern tab_ %+ i %+ _float ; ff_tab_i_float...
%assign i (i << 1)
%endrep
@@ -1385,11 +1385,7 @@ FFT_SPLIT_RADIX_DEF 8192, .16384pt
FFT_SPLIT_RADIX_DEF 16384, .32768pt
FFT_SPLIT_RADIX_DEF 32768, .65536pt
FFT_SPLIT_RADIX_DEF 65536, .131072pt
-FFT_SPLIT_RADIX_DEF 131072, .262144pt
-FFT_SPLIT_RADIX_DEF 262144, .524288pt
-FFT_SPLIT_RADIX_DEF 524288, .1048576pt
-FFT_SPLIT_RADIX_DEF 1048576, .2097152pt
-FFT_SPLIT_RADIX_DEF 2097152
+FFT_SPLIT_RADIX_DEF 131072
;===============================================================================
; Final synthesis + deinterleaving code
diff --git a/libavutil/x86/tx_float_init.c b/libavutil/x86/tx_float_init.c
index 3e99c21eac..f69a5a6d77 100644
--- a/libavutil/x86/tx_float_init.c
+++ b/libavutil/x86/tx_float_init.c
@@ -271,15 +271,15 @@ const FFTXCodelet * const ff_tx_codelet_list_float_x86[] = {
AV_TX_INPLACE | FF_TX_PRESHUFFLE | FF_TX_ASM_CALL, AV_CPU_FLAG_AVXSLOW),
TX_DEF(fft32_ns, FFT, 32, 32, 2, 0, 352, b8_i2, fma3, FMA3, AV_TX_INPLACE | FF_TX_PRESHUFFLE,
AV_CPU_FLAG_AVXSLOW),
- TX_DEF(fft_sr, FFT, 64, 2097152, 2, 0, 256, b8_i2, avx, AVX, 0, AV_CPU_FLAG_AVXSLOW),
- TX_DEF(fft_sr_asm, FFT, 64, 2097152, 2, 0, 320, b8_i2, avx, AVX,
+ TX_DEF(fft_sr, FFT, 64, 131072, 2, 0, 256, b8_i2, avx, AVX, 0, AV_CPU_FLAG_AVXSLOW),
+ TX_DEF(fft_sr_asm, FFT, 64, 131072, 2, 0, 320, b8_i2, avx, AVX,
AV_TX_INPLACE | FF_TX_PRESHUFFLE | FF_TX_ASM_CALL, AV_CPU_FLAG_AVXSLOW),
- TX_DEF(fft_sr_ns, FFT, 64, 2097152, 2, 0, 320, b8_i2, avx, AVX, AV_TX_INPLACE | FF_TX_PRESHUFFLE,
+ TX_DEF(fft_sr_ns, FFT, 64, 131072, 2, 0, 320, b8_i2, avx, AVX, AV_TX_INPLACE | FF_TX_PRESHUFFLE,
AV_CPU_FLAG_AVXSLOW),
- TX_DEF(fft_sr, FFT, 64, 2097152, 2, 0, 288, b8_i2, fma3, FMA3, 0, AV_CPU_FLAG_AVXSLOW),
- TX_DEF(fft_sr_asm, FFT, 64, 2097152, 2, 0, 352, b8_i2, fma3, FMA3,
+ TX_DEF(fft_sr, FFT, 64, 131072, 2, 0, 288, b8_i2, fma3, FMA3, 0, AV_CPU_FLAG_AVXSLOW),
+ TX_DEF(fft_sr_asm, FFT, 64, 131072, 2, 0, 352, b8_i2, fma3, FMA3,
AV_TX_INPLACE | FF_TX_PRESHUFFLE | FF_TX_ASM_CALL, AV_CPU_FLAG_AVXSLOW),
- TX_DEF(fft_sr_ns, FFT, 64, 2097152, 2, 0, 352, b8_i2, fma3, FMA3, AV_TX_INPLACE | FF_TX_PRESHUFFLE,
+ TX_DEF(fft_sr_ns, FFT, 64, 131072, 2, 0, 352, b8_i2, fma3, FMA3, AV_TX_INPLACE | FF_TX_PRESHUFFLE,
AV_CPU_FLAG_AVXSLOW),
TX_DEF(fft15, FFT, 15, 15, 15, 0, 320, factor_init, avx2, AVX2,
@@ -287,11 +287,11 @@ const FFTXCodelet * const ff_tx_codelet_list_float_x86[] = {
TX_DEF(fft15_ns, FFT, 15, 15, 15, 0, 384, factor_init, avx2, AVX2,
AV_TX_INPLACE | FF_TX_PRESHUFFLE, AV_CPU_FLAG_AVXSLOW),
- TX_DEF(fft_sr, FFT, 64, 2097152, 2, 0, 320, b8_i2, avx2, AVX2, 0,
+ TX_DEF(fft_sr, FFT, 64, 131072, 2, 0, 320, b8_i2, avx2, AVX2, 0,
AV_CPU_FLAG_AVXSLOW | AV_CPU_FLAG_SLOW_GATHER),
- TX_DEF(fft_sr_asm, FFT, 64, 2097152, 2, 0, 384, b8_i2, avx2, AVX2,
+ TX_DEF(fft_sr_asm, FFT, 64, 131072, 2, 0, 384, b8_i2, avx2, AVX2,
AV_TX_INPLACE | FF_TX_PRESHUFFLE | FF_TX_ASM_CALL, AV_CPU_FLAG_AVXSLOW | AV_CPU_FLAG_SLOW_GATHER),
- TX_DEF(fft_sr_ns, FFT, 64, 2097152, 2, 0, 384, b8_i2, avx2, AVX2, AV_TX_INPLACE | FF_TX_PRESHUFFLE,
+ TX_DEF(fft_sr_ns, FFT, 64, 131072, 2, 0, 384, b8_i2, avx2, AVX2, AV_TX_INPLACE | FF_TX_PRESHUFFLE,
AV_CPU_FLAG_AVXSLOW | AV_CPU_FLAG_SLOW_GATHER),
TX_DEF(fft_pfa_15xM, FFT, 60, TX_LEN_UNLIMITED, 15, 2, 320, fft_pfa_init, avx2, AVX2,
--
2.49.1
_______________________________________________
ffmpeg-devel mailing list -- ffmpeg-devel@ffmpeg.org
To unsubscribe send an email to ffmpeg-devel-leave@ffmpeg.org
^ permalink raw reply [flat|nested] only message in thread
only message in thread, other threads:[~2025-12-13 7:32 UTC | newest]
Thread overview: (only message) (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2025-12-13 7:31 [FFmpeg-devel] [PATCH] Revert "avutil/tx_template: extend to 2M" (PR #21184) Zhao Zhili via ffmpeg-devel
Git Inbox Mirror of the ffmpeg-devel mailing list - see https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
This inbox may be cloned and mirrored by anyone:
git clone --mirror https://master.gitmailbox.com/ffmpegdev/0 ffmpegdev/git/0.git
# If you have public-inbox 1.1+ installed, you may
# initialize and index your mirror using the following commands:
public-inbox-init -V2 ffmpegdev ffmpegdev/ https://master.gitmailbox.com/ffmpegdev \
ffmpegdev@gitmailbox.com
public-inbox-index ffmpegdev
Example config snippet for mirrors.
AGPL code for this site: git clone https://public-inbox.org/public-inbox.git