From 11218f9067566fa3ace8821b4b890457d6ea17f9 Mon Sep 17 00:00:00 2001
From: sunyuechi <sunyuechi@iscas.ac.cn>
Date: Mon, 25 Dec 2023 00:07:09 +0800
Subject: [PATCH 3/3] lavc/h264dsp: R-V V h264_add_pixels8_clear
C908
h264_add_pixels8_clear_c: 262.0
h264_add_pixels8_clear_rvv_i64: 59.0
---
libavcodec/riscv/h264dsp_init.c | 2 ++
libavcodec/riscv/h264dsp_rvv.S | 22 ++++++++++++++++++++++
2 files changed, 24 insertions(+)
@@ -26,6 +26,7 @@
#include "libavcodec/h264dsp.h"
void ff_h264_add_pixels4_clear_rvv(uint8_t *dst, int16_t *block, int stride);
+void ff_h264_add_pixels8_clear_rvv(uint8_t *dst, int16_t *block, int stride);
av_cold void ff_h264dsp_init_riscv(H264DSPContext *c, const int bit_depth, const int chroma_format_idc)
{
@@ -35,6 +36,7 @@ av_cold void ff_h264dsp_init_riscv(H264DSPContext *c, const int bit_depth, const
if (flags & AV_CPU_FLAG_RVV_I64) {
if (bit_depth == 8) {
c->h264_add_pixels4_clear = ff_h264_add_pixels4_clear_rvv;
+ c->h264_add_pixels8_clear = ff_h264_add_pixels8_clear_rvv;
}
}
#endif
@@ -40,3 +40,25 @@ func ff_h264_add_pixels4_clear_rvv, zve64x
ret
endfunc
+
+func ff_h264_add_pixels8_clear_rvv, zve64x
+ vsetivli zero, 16, e8, m1, ta, ma
+ vle64.v v24, (a1)
+ li t1, 8*8
+ vsetvli zero, t1, e16, m8, ta, ma
+ li t0, 0xff
+ vand.vx v24, v24, t0
+ addi a1, a1, 8*8*2
+ vsetivli zero, 16, e8, m1, ta, ma
+ vse64.v v0, (a1)
+ vsetvli zero, t1, e8, m4, ta, ma
+ vnclipu.wi v24, v24, 0
+ vsetivli zero, 8, e8, mf2, ta, ma
+ vle64.v v8, (a0)
+ vsetvli zero, t1, e8, m4, ta, ma
+ vadd.vv v24, v24, v8
+ vsetivli zero, 8, e8, mf2, ta, ma
+ vse64.v v24, (a0)
+
+ ret
+endfunc
--
2.43.0