[FFmpeg-cvslog] lavu/floatdsp: RISC-V V vector_fmac_scalar

Rémi Denis-Courmont git at videolan.org
Tue Sep 27 14:59:04 EEST 2022


ffmpeg | branch: master | Rémi Denis-Courmont <remi at remlab.net> | Mon Sep 26 17:52:32 2022 +0300| [c3db27ba9552fc3cb764bf9eb717840d981c5779] | committer: Lynne

lavu/floatdsp: RISC-V V vector_fmac_scalar

> http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=c3db27ba9552fc3cb764bf9eb717840d981c5779
---

 libavutil/riscv/float_dsp_init.c |  3 +++
 libavutil/riscv/float_dsp_rvv.S  | 19 +++++++++++++++++++
 2 files changed, 22 insertions(+)

diff --git a/libavutil/riscv/float_dsp_init.c b/libavutil/riscv/float_dsp_init.c
index 29114dfb82..9e19413d5d 100644
--- a/libavutil/riscv/float_dsp_init.c
+++ b/libavutil/riscv/float_dsp_init.c
@@ -27,6 +27,8 @@
 
 void ff_vector_fmul_rvv(float *dst, const float *src0, const float *src1,
                          int len);
+void ff_vector_fmac_scalar_rvv(float *dst, const float *src, float mul,
+                                int len);
 void ff_vector_fmul_scalar_rvv(float *dst, const float *src, float mul,
                                 int len);
 
@@ -42,6 +44,7 @@ av_cold void ff_float_dsp_init_riscv(AVFloatDSPContext *fdsp)
 
     if (flags & AV_CPU_FLAG_RVV_F32) {
         fdsp->vector_fmul = ff_vector_fmul_rvv;
+        fdsp->vector_fmac_scalar = ff_vector_fmac_scalar_rvv;
         fdsp->vector_fmul_scalar = ff_vector_fmul_scalar_rvv;
     }
 
diff --git a/libavutil/riscv/float_dsp_rvv.S b/libavutil/riscv/float_dsp_rvv.S
index 710e122444..4c325db9fd 100644
--- a/libavutil/riscv/float_dsp_rvv.S
+++ b/libavutil/riscv/float_dsp_rvv.S
@@ -38,6 +38,25 @@ func ff_vector_fmul_rvv, zve32f
         ret
 endfunc
 
+// (a0) += (a1) * fa0 [0..a2-1]
+func ff_vector_fmac_scalar_rvv, zve32f
+NOHWF   fmv.w.x   fa0, a2
+NOHWF   mv        a2, a3
+1:
+        vsetvli   t0, a2, e32, m1, ta, ma
+        slli      t1, t0, 2
+        vle32.v   v24, (a1)
+        sub       a2, a2, t0
+        vle32.v   v16, (a0)
+        sh2add    a1, t0, a1
+        vfmacc.vf v16, fa0, v24
+        vse32.v   v16, (a0)
+        sh2add    a0, t0, a0
+        bnez      a2, 1b
+
+        ret
+endfunc
+
 // (a0) = (a1) * fa0 [0..a2-1]
 func ff_vector_fmul_scalar_rvv, zve32f
 NOHWF   fmv.w.x  fa0, a2



More information about the ffmpeg-cvslog mailing list