From: toqsxw@outlook.com
To: ffmpeg-devel@ffmpeg.org
Cc: Wu Jianhua <toqsxw@outlook.com>
Subject: [FFmpeg-devel] [PATCH v3 5/8] avcodec/vvcdec: reuse h26x/2656_inter.asm to enable x86 optimizations
Date: Tue, 23 Jan 2024 01:46:25 +0800
Message-ID: <OSZP286MB2173741FCD2CC63F31AD2E9ACA752@OSZP286MB2173.JPNP286.PROD.OUTLOOK.COM> (raw)
In-Reply-To: <20240122174628.1206503-1-toqsxw@outlook.com>
From: Wu Jianhua <toqsxw@outlook.com>
Signed-off-by: Wu Jianhua <toqsxw@outlook.com>
---
libavcodec/Makefile | 1 +
libavcodec/vvc/vvcdsp.c | 4 +
libavcodec/vvc/vvcdsp.h | 2 +
libavcodec/x86/vvc/Makefile | 6 +
libavcodec/x86/vvc/vvcdsp_init.c | 202 +++++++++++++++++++++++++++++++
5 files changed, 215 insertions(+)
create mode 100644 libavcodec/x86/vvc/Makefile
create mode 100644 libavcodec/x86/vvc/vvcdsp_init.c
diff --git a/libavcodec/Makefile b/libavcodec/Makefile
index bb42095165..ce33631b60 100644
--- a/libavcodec/Makefile
+++ b/libavcodec/Makefile
@@ -65,6 +65,7 @@ OBJS = ac3_parser.o \
# subsystems
include $(SRC_PATH)/libavcodec/vvc/Makefile
+include $(SRC_PATH)/libavcodec/x86/vvc/Makefile
OBJS-$(CONFIG_AANDCTTABLES) += aandcttab.o
OBJS-$(CONFIG_AC3DSP) += ac3dsp.o ac3.o ac3tab.o
OBJS-$(CONFIG_ADTS_HEADER) += adts_header.o mpeg4audio_sample_rates.o
diff --git a/libavcodec/vvc/vvcdsp.c b/libavcodec/vvc/vvcdsp.c
index c82ea7be30..c542be5258 100644
--- a/libavcodec/vvc/vvcdsp.c
+++ b/libavcodec/vvc/vvcdsp.c
@@ -138,4 +138,8 @@ void ff_vvc_dsp_init(VVCDSPContext *vvcdsp, int bit_depth)
VVC_DSP(8);
break;
}
+
+#if ARCH_X86
+ ff_vvc_dsp_init_x86(vvcdsp, bit_depth);
+#endif
}
diff --git a/libavcodec/vvc/vvcdsp.h b/libavcodec/vvc/vvcdsp.h
index b5a63c5833..6f59e73654 100644
--- a/libavcodec/vvc/vvcdsp.h
+++ b/libavcodec/vvc/vvcdsp.h
@@ -167,4 +167,6 @@ typedef struct VVCDSPContext {
void ff_vvc_dsp_init(VVCDSPContext *hpc, int bit_depth);
+void ff_vvc_dsp_init_x86(VVCDSPContext *hpc, const int bit_depth);
+
#endif /* AVCODEC_VVC_VVCDSP_H */
diff --git a/libavcodec/x86/vvc/Makefile b/libavcodec/x86/vvc/Makefile
new file mode 100644
index 0000000000..b4acc22501
--- /dev/null
+++ b/libavcodec/x86/vvc/Makefile
@@ -0,0 +1,6 @@
+clean::
+ $(RM) $(CLEANSUFFIXES:%=libavcodec/x86/vvc/%)
+
+OBJS-$(CONFIG_VVC_DECODER) += x86/vvc/vvcdsp_init.o
+X86ASM-OBJS-$(CONFIG_VVC_DECODER) += x86/h26x/h2656dsp.o \
+ x86/h26x/h2656_inter.o
diff --git a/libavcodec/x86/vvc/vvcdsp_init.c b/libavcodec/x86/vvc/vvcdsp_init.c
new file mode 100644
index 0000000000..c197cdb4cc
--- /dev/null
+++ b/libavcodec/x86/vvc/vvcdsp_init.c
@@ -0,0 +1,202 @@
+/*
+ * VVC DSP init for x86
+ *
+ * Copyright (C) 2022-2024 Nuo Mi
+ * Copyright (c) 2023-2024 Wu Jianhua
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include "config.h"
+
+#include "libavutil/cpu.h"
+#include "libavutil/x86/asm.h"
+#include "libavutil/x86/cpu.h"
+#include "libavcodec/vvc/vvcdec.h"
+#include "libavcodec/vvc/vvc_ctu.h"
+#include "libavcodec/vvc/vvcdsp.h"
+#include "libavcodec/x86/h26x/h2656dsp.h"
+
+#define FW_PUT(name, depth, opt) \
+static void ff_vvc_put_ ## name ## _ ## depth ## _##opt(int16_t *dst, const uint8_t *src, ptrdiff_t srcstride, \
+ int height, const int8_t *hf, const int8_t *vf, int width) \
+{ \
+ ff_h2656_put_## name ## _ ## depth ## _##opt(dst, 2 * MAX_PB_SIZE, src, srcstride, height, hf, vf, width); \
+}
+
+#define FW_PUT_TAP(fname, bitd, opt ) \
+ FW_PUT(fname##4, bitd, opt ); \
+ FW_PUT(fname##8, bitd, opt ); \
+ FW_PUT(fname##16, bitd, opt ); \
+ FW_PUT(fname##32, bitd, opt ); \
+ FW_PUT(fname##64, bitd, opt ); \
+ FW_PUT(fname##128, bitd, opt ); \
+
+#define FW_PUT_4TAP(fname, bitd, opt) \
+ FW_PUT(fname ## 2, bitd, opt) \
+ FW_PUT_TAP(fname, bitd, opt)
+
+#define FW_PUT_4TAP_SSE4(bitd) \
+ FW_PUT_4TAP(pixels, bitd, sse4) \
+ FW_PUT_4TAP(4tap_h, bitd, sse4) \
+ FW_PUT_4TAP(4tap_v, bitd, sse4) \
+ FW_PUT_4TAP(4tap_hv, bitd, sse4)
+
+#define FW_PUT_8TAP_SSE4(bitd) \
+ FW_PUT_TAP(8tap_h, bitd, sse4) \
+ FW_PUT_TAP(8tap_v, bitd, sse4) \
+ FW_PUT_TAP(8tap_hv, bitd, sse4)
+
+#define FW_PUT_SSE4(bitd) \
+ FW_PUT_4TAP_SSE4(bitd) \
+ FW_PUT_8TAP_SSE4(bitd)
+
+FW_PUT_SSE4( 8);
+FW_PUT_SSE4(10);
+FW_PUT_SSE4(12);
+
+#define FW_PUT_TAP_AVX2(n, bitd) \
+ FW_PUT(n ## tap_h32, bitd, avx2) \
+ FW_PUT(n ## tap_h64, bitd, avx2) \
+ FW_PUT(n ## tap_h128, bitd, avx2) \
+ FW_PUT(n ## tap_v32, bitd, avx2) \
+ FW_PUT(n ## tap_v64, bitd, avx2) \
+ FW_PUT(n ## tap_v128, bitd, avx2)
+
+#define FW_PUT_AVX2(bitd) \
+ FW_PUT(pixels32, bitd, avx2) \
+ FW_PUT(pixels64, bitd, avx2) \
+ FW_PUT(pixels128, bitd, avx2) \
+ FW_PUT_TAP_AVX2(4, bitd) \
+ FW_PUT_TAP_AVX2(8, bitd) \
+
+FW_PUT_AVX2( 8)
+FW_PUT_AVX2(10)
+FW_PUT_AVX2(12)
+
+#define FW_PUT_TAP_16BPC_AVX2(n, bitd) \
+ FW_PUT(n ## tap_h16, bitd, avx2) \
+ FW_PUT(n ## tap_v16, bitd, avx2) \
+ FW_PUT(n ## tap_hv16, bitd, avx2) \
+ FW_PUT(n ## tap_hv32, bitd, avx2) \
+ FW_PUT(n ## tap_hv64, bitd, avx2) \
+ FW_PUT(n ## tap_hv128, bitd, avx2)
+
+#define FW_PUT_16BPC_AVX2(bitd) \
+ FW_PUT(pixels16, bitd, avx2) \
+ FW_PUT_TAP_16BPC_AVX2(4, bitd) \
+ FW_PUT_TAP_16BPC_AVX2(8, bitd);
+
+FW_PUT_16BPC_AVX2(10);
+FW_PUT_16BPC_AVX2(12);
+
+#define PEL_LINK(dst, C, W, idx1, idx2, name, D, opt) \
+ dst[C][W][idx1][idx2] = ff_vvc_put_## name ## _ ## D ## _##opt; \
+ dst ## _uni[C][W][idx1][idx2] = ff_h2656_put_uni_ ## name ## _ ## D ## _##opt; \
+
+#define MC_TAP_LINKS(pointer, C, my, mx, fname, bitd, opt ) \
+ PEL_LINK(pointer, C, 1, my , mx , fname##4 , bitd, opt ); \
+ PEL_LINK(pointer, C, 2, my , mx , fname##8 , bitd, opt ); \
+ PEL_LINK(pointer, C, 3, my , mx , fname##16, bitd, opt ); \
+ PEL_LINK(pointer, C, 4, my , mx , fname##32, bitd, opt ); \
+ PEL_LINK(pointer, C, 5, my , mx , fname##64, bitd, opt ); \
+ PEL_LINK(pointer, C, 6, my , mx , fname##128, bitd, opt );
+
+#define MC_8TAP_LINKS(pointer, my, mx, fname, bitd, opt) \
+ MC_TAP_LINKS(pointer, LUMA, my, mx, fname, bitd, opt)
+
+#define MC_8TAP_LINKS_SSE4(bd) \
+ MC_8TAP_LINKS(c->inter.put, 0, 0, pixels, bd, sse4); \
+ MC_8TAP_LINKS(c->inter.put, 0, 1, 8tap_h, bd, sse4); \
+ MC_8TAP_LINKS(c->inter.put, 1, 0, 8tap_v, bd, sse4); \
+ MC_8TAP_LINKS(c->inter.put, 1, 1, 8tap_hv, bd, sse4)
+
+#define MC_4TAP_LINKS(pointer, my, mx, fname, bitd, opt) \
+ PEL_LINK(pointer, CHROMA, 0, my , mx , fname##2 , bitd, opt ); \
+ MC_TAP_LINKS(pointer, CHROMA, my, mx, fname, bitd, opt) \
+
+#define MC_4TAP_LINKS_SSE4(bd) \
+ MC_4TAP_LINKS(c->inter.put, 0, 0, pixels, bd, sse4); \
+ MC_4TAP_LINKS(c->inter.put, 0, 1, 4tap_h, bd, sse4); \
+ MC_4TAP_LINKS(c->inter.put, 1, 0, 4tap_v, bd, sse4); \
+ MC_4TAP_LINKS(c->inter.put, 1, 1, 4tap_hv, bd, sse4)
+
+#define MC_LINK_SSE4(bd) \
+ MC_4TAP_LINKS_SSE4(bd) \
+ MC_8TAP_LINKS_SSE4(bd)
+
+#define MC_TAP_LINKS_AVX2(C,tap,bd) do { \
+ PEL_LINK(c->inter.put, C, 4, 0, 0, pixels32, bd, avx2) \
+ PEL_LINK(c->inter.put, C, 5, 0, 0, pixels64, bd, avx2) \
+ PEL_LINK(c->inter.put, C, 6, 0, 0, pixels128, bd, avx2) \
+ PEL_LINK(c->inter.put, C, 4, 0, 1, tap##tap_h32, bd, avx2) \
+ PEL_LINK(c->inter.put, C, 5, 0, 1, tap##tap_h64, bd, avx2) \
+ PEL_LINK(c->inter.put, C, 6, 0, 1, tap##tap_h128, bd, avx2) \
+ PEL_LINK(c->inter.put, C, 4, 1, 0, tap##tap_v32, bd, avx2) \
+ PEL_LINK(c->inter.put, C, 5, 1, 0, tap##tap_v64, bd, avx2) \
+ PEL_LINK(c->inter.put, C, 6, 1, 0, tap##tap_v128, bd, avx2) \
+ } while (0)
+
+#define MC_LINKS_AVX2(bd) \
+ MC_TAP_LINKS_AVX2(LUMA, 8, bd); \
+ MC_TAP_LINKS_AVX2(CHROMA, 4, bd);
+
+#define MC_TAP_LINKS_16BPC_AVX2(C, tap, bd) do { \
+ PEL_LINK(c->inter.put, C, 3, 0, 0, pixels16, bd, avx2) \
+ PEL_LINK(c->inter.put, C, 3, 0, 1, tap##tap_h16, bd, avx2) \
+ PEL_LINK(c->inter.put, C, 3, 1, 0, tap##tap_v16, bd, avx2) \
+ PEL_LINK(c->inter.put, C, 3, 1, 1, tap##tap_hv16, bd, avx2) \
+ PEL_LINK(c->inter.put, C, 4, 1, 1, tap##tap_hv32, bd, avx2) \
+ PEL_LINK(c->inter.put, C, 5, 1, 1, tap##tap_hv64, bd, avx2) \
+ PEL_LINK(c->inter.put, C, 6, 1, 1, tap##tap_hv128, bd, avx2) \
+ } while (0)
+
+#define MC_LINKS_16BPC_AVX2(bd) \
+ MC_TAP_LINKS_16BPC_AVX2(LUMA, 8, bd); \
+ MC_TAP_LINKS_16BPC_AVX2(CHROMA, 4, bd);
+
+void ff_vvc_dsp_init_x86(VVCDSPContext *const c, const int bd)
+{
+ const int cpu_flags = av_get_cpu_flags();
+
+ if (ARCH_X86_64) {
+ if (bd == 8) {
+ if (EXTERNAL_SSE4(cpu_flags)) {
+ MC_LINK_SSE4(8);
+ }
+ if (EXTERNAL_AVX2_FAST(cpu_flags)) {
+ MC_LINKS_AVX2(8);
+ }
+ } else if (bd == 10) {
+ if (EXTERNAL_SSE4(cpu_flags)) {
+ MC_LINK_SSE4(10);
+ }
+ if (EXTERNAL_AVX2_FAST(cpu_flags)) {
+ MC_LINKS_AVX2(10);
+ MC_LINKS_16BPC_AVX2(10);
+ }
+ } else if (bd == 12) {
+ if (EXTERNAL_SSE4(cpu_flags)) {
+ MC_LINK_SSE4(12);
+ }
+ if (EXTERNAL_AVX2_FAST(cpu_flags)) {
+ MC_LINKS_AVX2(12);
+ MC_LINKS_16BPC_AVX2(12);
+ }
+ }
+ }
+}
--
2.34.1
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
To unsubscribe, visit link above, or email
ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".
next prev parent reply other threads:[~2024-01-22 17:47 UTC|newest]
Thread overview: 12+ messages / expand[flat|nested] mbox.gz Atom feed top
[not found] <20240122174628.1206503-1-toqsxw@outlook.com>
2024-01-22 17:46 ` [FFmpeg-devel] [PATCH v3 2/8] avcodec/hevcdsp_template: reuse put/put_luma/put_chroma from h2656_inter_template toqsxw
2024-01-22 17:46 ` [FFmpeg-devel] [PATCH v3 3/8] avcodec/x86/hevc_mc: move put/put_uni to h26x/h2656_inter.asm toqsxw
2024-01-22 17:46 ` [FFmpeg-devel] [PATCH v3 4/8] avcodec/x86/h26x/h2656_inter: add dststride to put toqsxw
2024-01-22 17:46 ` toqsxw [this message]
2024-01-23 13:01 ` [FFmpeg-devel] [PATCH v3 5/8] avcodec/vvcdec: reuse h26x/2656_inter.asm to enable x86 optimizations Nuo Mi
2024-01-25 12:43 ` Nuo Mi
2024-01-22 17:46 ` [FFmpeg-devel] [PATCH v3 6/8] tests/checkasm: add checkasm_check_vvc_mc toqsxw
2024-01-22 17:46 ` [FFmpeg-devel] [PATCH v3 7/8] avcodec/x86/vvc: add avg and avg_w AVX2 optimizations toqsxw
2024-01-22 22:46 ` Michael Niedermayer
2024-01-23 18:23 ` [FFmpeg-devel] 回复: " Wu Jianhua
2024-01-22 17:46 ` [FFmpeg-devel] [PATCH v3 8/8] tests/checkasm/vvc_mc: add check_avg toqsxw
[not found] <20240122152527.601122-1-toqsxw@outlook.com>
2024-01-22 15:25 ` [FFmpeg-devel] [PATCH v3 5/8] avcodec/vvcdec: reuse h26x/2656_inter.asm to enable x86 optimizations toqsxw
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=OSZP286MB2173741FCD2CC63F31AD2E9ACA752@OSZP286MB2173.JPNP286.PROD.OUTLOOK.COM \
--to=toqsxw@outlook.com \
--cc=ffmpeg-devel@ffmpeg.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
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