diff mbox series

[FFmpeg-devel,2/2] lavc/h264dsp: R-V V high-depth add_pixels8

Message ID 20240713130154.77672-2-remi@remlab.net
State New
Headers show
Series [FFmpeg-devel,1/2] lavc/h264dsp: R-V V add_pixels4 and 8-bit add_pixels8 | expand

Checks

Context Check Description
andriy/configure_x86 warning Failed to apply patch

Commit Message

Rémi Denis-Courmont July 13, 2024, 1:01 p.m. UTC 
---
 libavcodec/riscv/h264addpx_rvv.S | 22 ++++++++++++++++++++++
 libavcodec/riscv/h264dsp_init.c  |  2 ++
 2 files changed, 24 insertions(+)

Comments

Rémi Denis-Courmont July 13, 2024, 5:50 p.m. UTC | #1 
T-Head C908 (cycles);
h264_add_pixels8_9bpp_c:        270.5
h264_add_pixels8_9bpp_rvv_i32:  164.2
h264_add_pixels8_10bpp_c:       270.5
h264_add_pixels8_10bpp_rvv_i32: 164.2
h264_add_pixels8_12bpp_c:       270.5
h264_add_pixels8_12bpp_rvv_i32: 164.2
h264_add_pixels8_14bpp_c:       270.5
h264_add_pixels8_14bpp_rvv_i32: 164.2
diff mbox series

Patch

diff --git a/libavcodec/riscv/h264addpx_rvv.S b/libavcodec/riscv/h264addpx_rvv.S
index fd36bd4896..3d307b2cb8 100644
--- a/libavcodec/riscv/h264addpx_rvv.S
+++ b/libavcodec/riscv/h264addpx_rvv.S
@@ -87,3 +87,25 @@  func ff_h264_add_pixels8_8_rvv, zve64x
         vsse64.v        v8, (a0), a2
         ret
 endfunc
+
+func ff_h264_add_pixels8_16_rvv, zve32x
+        li      t0, 8
+        vsetivli    zero, 8, e16, m1, ta, ma
+1:
+        vle32.v     v16, (a1)
+        addi    t0, t0, -1
+        vle16.v     v8, (a0)
+        addi    a1, a1, 8 * 4
+        vncvt.x.x.w v24, v16
+        .equ    offset, 0
+        .rept   256 / __riscv_xlen
+        sx      zero, offset(a1)
+        .equ    offset, offset + (__riscv_xlen / 8)
+        .endr
+        vadd.vv     v8, v8, v24
+        vse16.v     v8, (a0)
+        add     a0, a0, a2
+        bnez    t0, 1b
+
+        ret
+endfunc
diff --git a/libavcodec/riscv/h264dsp_init.c b/libavcodec/riscv/h264dsp_init.c
index a6d06b3ac4..ea89314952 100644
--- a/libavcodec/riscv/h264dsp_init.c
+++ b/libavcodec/riscv/h264dsp_init.c
@@ -63,6 +63,7 @@  void ff_h264_idct8_add_14_rvv(uint8_t *dst, int16_t *block, int stride);
 
 void ff_h264_add_pixels8_8_rvv(uint8_t *dst, int16_t *block, int stride);
 void ff_h264_add_pixels4_8_rvv(uint8_t *dst, int16_t *block, int stride);
+void ff_h264_add_pixels8_16_rvv(uint8_t *dst, int16_t *block, int stride);
 void ff_h264_add_pixels4_16_rvv(uint8_t *dst, int16_t *block, int stride);
 
 extern int ff_startcode_find_candidate_rvb(const uint8_t *, int);
@@ -126,6 +127,7 @@  av_cold void ff_h264dsp_init_riscv(H264DSPContext *dsp, const int bit_depth,
             dsp->h264_idct8_add = ff_h264_idct8_add_14_rvv;
         }
         if (bit_depth > 8 && zvl128b) {
+            dsp->h264_add_pixels8_clear = ff_h264_add_pixels8_16_rvv;
             if (flags & AV_CPU_FLAG_RVV_I64)
                 dsp->h264_add_pixels4_clear = ff_h264_add_pixels4_16_rvv;
         }