T-Head C908: h263dsp.dct_unquantize_inter_c: 3.7 h263dsp.dct_unquantize_inter_rvv_i32: 1.7 h263dsp.dct_unquantize_intra_c: 4.0 h263dsp.dct_unquantize_intra_rvv_i32: 1.5
SpacemiT X60: h263dsp.dct_unquantize_inter_c: 3.5 h263dsp.dct_unquantize_inter_rvv_i32: 1.2 h263dsp.dct_unquantize_intra_c: 3.5 h263dsp.dct_unquantize_intra_rvv_i32: 1.2 --- libavcodec/riscv/h263dsp_init.c | 15 ++++++++++++--- libavcodec/riscv/h263dsp_rvv.S | 27 +++++++++++++++++++++++++++ 2 files changed, 39 insertions(+), 3 deletions(-) diff --git a/libavcodec/riscv/h263dsp_init.c b/libavcodec/riscv/h263dsp_init.c index 21b536366c..78d537d1e1 100644 --- a/libavcodec/riscv/h263dsp_init.c +++ b/libavcodec/riscv/h263dsp_init.c @@ -25,6 +25,8 @@ #include "libavutil/riscv/cpu.h" #include "libavcodec/h263dsp.h" +void ff_h263_dct_unquantize_intra_rvv(int16_t *, ptrdiff_t len, int, int); +void ff_h263_dct_unquantize_inter_rvv(int16_t *, ptrdiff_t len, int, int); void ff_h263_h_loop_filter_rvv(uint8_t *src, int stride, int q); void ff_h263_v_loop_filter_rvv(uint8_t *src, int stride, int q); @@ -33,9 +35,16 @@ av_cold void ff_h263dsp_init_riscv(H263DSPContext *c) #if HAVE_RVV int flags = av_get_cpu_flags(); - if ((flags & AV_CPU_FLAG_RVV_I32) && ff_rv_vlen_least(128)) { - c->h263_h_loop_filter = ff_h263_h_loop_filter_rvv; - c->h263_v_loop_filter = ff_h263_v_loop_filter_rvv; + if (flags & AV_CPU_FLAG_RVV_I32) { + if (ff_rv_vlen_least(128) || (flags & AV_CPU_FLAG_RVB_ADDR)) { + c->h263_dct_unquantize_intra = ff_h263_dct_unquantize_intra_rvv; + c->h263_dct_unquantize_inter = ff_h263_dct_unquantize_inter_rvv; + } + + if (ff_rv_vlen_least(128)) { + c->h263_h_loop_filter = ff_h263_h_loop_filter_rvv; + c->h263_v_loop_filter = ff_h263_v_loop_filter_rvv; + } } #endif } diff --git a/libavcodec/riscv/h263dsp_rvv.S b/libavcodec/riscv/h263dsp_rvv.S index 97503d527c..f2b7eb9a87 100644 --- a/libavcodec/riscv/h263dsp_rvv.S +++ b/libavcodec/riscv/h263dsp_rvv.S @@ -20,6 +20,33 @@ #include "libavutil/riscv/asm.S" +func ff_h263_dct_unquantize_intra_rvv, zve32x + addi a0, a0, 2 + j 1f +endfunc + +func ff_h263_dct_unquantize_inter_rvv, zve32x + addi a1, a1, 1 +1: + vsetvli t0, a1, e16, m8, ta, mu + vle16.v v8, (a0) +#if defined(__riscv_v_min_vlen) && __riscv_v_min_vlen < 128 + sub a1, a1, t0 +#endif + vmv.v.x v16, a3 + vmslt.vi v0, v8, 0 + vneg.v v16, v16, v0.t + vmsne.vi v0, v8, 0 + vmadd.vx v8, a2, v16, v0.t + vse16.v v8, (a0) +#if defined(__riscv_v_min_vlen) && __riscv_v_min_vlen < 128 + sh1add a0, t0, a0 + bnez a1, 1b +#endif + + ret +endfunc + .option push .option norelax func ff_h263_h_loop_filter_rvv, zve32x -- 2.45.2 _______________________________________________ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-devel To unsubscribe, visit link above, or email ffmpeg-devel-requ...@ffmpeg.org with subject "unsubscribe".