From 6daaabd19c26b01b948804e4ec84a6f606746ec7 Mon Sep 17 00:00:00 2001
From: sunyuechi <sunyuechi@iscas.ac.cn>
Date: Mon, 19 Feb 2024 18:03:58 +0800
Subject: [PATCH] lavc/vp8dsp: R-V V put_vp8_pixels
C908:
vp8_put_pixels4_c: 87.5
vp8_put_pixels4_rvv_i32: 42.7
vp8_put_pixels8_c: 284.5
vp8_put_pixels8_rvv_i32: 77.7
vp8_put_pixels16_c: 1087.7
vp8_put_pixels16_rvv_i32: 108.0
---
libavcodec/riscv/vp8dsp.h | 75 ++++++++++++++++++++++++++++++++++
libavcodec/riscv/vp8dsp_init.c | 22 ++++++++++
libavcodec/riscv/vp8dsp_rvv.S | 27 ++++++++++++
libavcodec/vp8dsp.c | 2 +
libavcodec/vp8dsp.h | 1 +
5 files changed, 127 insertions(+)
create mode 100644 libavcodec/riscv/vp8dsp.h
new file mode 100644
@@ -0,0 +1,75 @@
+/*
+ * 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
+ */
+
+#ifndef AVCODEC_RISCV_VP8DSP_H
+#define AVCODEC_RISCV_VP8DSP_H
+
+#include "libavcodec/vp8dsp.h"
+
+#define VP8_LF_Y(hv, inner, opt) \
+ void ff_vp8_##hv##_loop_filter16##inner##_##opt(uint8_t *dst, \
+ ptrdiff_t stride, \
+ int flim_E, int flim_I, \
+ int hev_thresh)
+
+#define VP8_LF_UV(hv, inner, opt) \
+ void ff_vp8_##hv##_loop_filter8uv##inner##_##opt(uint8_t *dstU, \
+ uint8_t *dstV, \
+ ptrdiff_t stride, \
+ int flim_E, int flim_I, \
+ int hev_thresh)
+
+#define VP8_LF_SIMPLE(hv, opt) \
+ void ff_vp8_##hv##_loop_filter16_simple_##opt(uint8_t *dst, \
+ ptrdiff_t stride, \
+ int flim)
+
+#define VP8_LF_HV(inner, opt) \
+ VP8_LF_Y(h, inner, opt); \
+ VP8_LF_Y(v, inner, opt); \
+ VP8_LF_UV(h, inner, opt); \
+ VP8_LF_UV(v, inner, opt)
+
+#define VP8_LF(opt) \
+ VP8_LF_HV(, opt); \
+ VP8_LF_HV(_inner, opt); \
+ VP8_LF_SIMPLE(h, opt); \
+ VP8_LF_SIMPLE(v, opt)
+
+#define VP8_MC(n, opt) \
+ void ff_put_vp8_##n##_##opt(uint8_t *dst, ptrdiff_t dststride, \
+ const uint8_t *src, ptrdiff_t srcstride,\
+ int h, int x, int y)
+
+#define VP8_EPEL(w, opt) \
+ VP8_MC(pixels ## w, opt); \
+ VP8_MC(epel ## w ## _h4, opt); \
+ VP8_MC(epel ## w ## _h6, opt); \
+ VP8_MC(epel ## w ## _v4, opt); \
+ VP8_MC(epel ## w ## _h4v4, opt); \
+ VP8_MC(epel ## w ## _h6v4, opt); \
+ VP8_MC(epel ## w ## _v6, opt); \
+ VP8_MC(epel ## w ## _h4v6, opt); \
+ VP8_MC(epel ## w ## _h6v6, opt)
+
+#define VP8_BILIN(w, opt) \
+ VP8_MC(bilin ## w ## _h, opt); \
+ VP8_MC(bilin ## w ## _v, opt); \
+ VP8_MC(bilin ## w ## _hv, opt)
+
+#endif /* AVCODEC_RISCV_VP8DSP_H */
@@ -24,11 +24,33 @@
#include "libavutil/cpu.h"
#include "libavutil/riscv/cpu.h"
#include "libavcodec/vp8dsp.h"
+#include "vp8dsp.h"
void ff_vp8_idct_dc_add_rvv(uint8_t *dst, int16_t block[16], ptrdiff_t stride);
void ff_vp8_idct_dc_add4y_rvv(uint8_t *dst, int16_t block[4][16], ptrdiff_t stride);
void ff_vp8_idct_dc_add4uv_rvv(uint8_t *dst, int16_t block[4][16], ptrdiff_t stride);
+VP8_EPEL(16, rvv);
+VP8_EPEL(8, rvv);
+VP8_EPEL(4, rvv);
+
+av_cold void ff_vp78dsp_init_riscv(VP8DSPContext *c)
+{
+#if HAVE_RVV
+ int flags = av_get_cpu_flags();
+
+ if (flags & AV_CPU_FLAG_RVV_I32 && ff_get_rv_vlenb() >= 16) {
+ c->put_vp8_epel_pixels_tab[0][0][0] = ff_put_vp8_pixels16_rvv;
+ c->put_vp8_epel_pixels_tab[1][0][0] = ff_put_vp8_pixels8_rvv;
+ c->put_vp8_epel_pixels_tab[2][0][0] = ff_put_vp8_pixels4_rvv;
+
+ c->put_vp8_bilinear_pixels_tab[0][0][0] = ff_put_vp8_pixels16_rvv;
+ c->put_vp8_bilinear_pixels_tab[1][0][0] = ff_put_vp8_pixels8_rvv;
+ c->put_vp8_bilinear_pixels_tab[2][0][0] = ff_put_vp8_pixels4_rvv;
+ }
+#endif
+}
+
av_cold void ff_vp8dsp_init_riscv(VP8DSPContext *c)
{
#if HAVE_RVV
@@ -71,3 +71,30 @@ func ff_vp8_idct_dc_add4uv_rvv, zve32x
ret
endfunc
+
+.macro put_vp8_pixels
+1:
+ addi a4, a4, -1
+ vle8.v v0, (a2)
+ vse8.v v0, (a0)
+ add a2, a2, a3
+ add a0, a0, a1
+ bnez a4, 1b
+
+ ret
+.endm
+
+func ff_put_vp8_pixels16_rvv, zve32x
+ vsetivli zero, 16, e8, m1, ta, ma
+ put_vp8_pixels
+endfunc
+
+func ff_put_vp8_pixels8_rvv, zve32x
+ vsetivli zero, 8, e8, mf2, ta, ma
+ put_vp8_pixels
+endfunc
+
+func ff_put_vp8_pixels4_rvv, zve32x
+ vsetivli zero, 4, e8, mf4, ta, ma
+ put_vp8_pixels
+endfunc
@@ -681,6 +681,8 @@ av_cold void ff_vp78dsp_init(VP8DSPContext *dsp)
ff_vp78dsp_init_arm(dsp);
#elif ARCH_PPC
ff_vp78dsp_init_ppc(dsp);
+#elif ARCH_RISCV
+ ff_vp78dsp_init_riscv(dsp);
#elif ARCH_X86
ff_vp78dsp_init_x86(dsp);
#endif
@@ -87,6 +87,7 @@ void ff_vp78dsp_init(VP8DSPContext *c);
void ff_vp78dsp_init_aarch64(VP8DSPContext *c);
void ff_vp78dsp_init_arm(VP8DSPContext *c);
void ff_vp78dsp_init_ppc(VP8DSPContext *c);
+void ff_vp78dsp_init_riscv(VP8DSPContext *c);
void ff_vp78dsp_init_x86(VP8DSPContext *c);
void ff_vp8dsp_init(VP8DSPContext *c);
--
2.43.2