lavu/floatdsp: RISC-V V vector_fmul_add

libavutil/riscv/float_dsp_init.c

@@ -31,6 +31,8 @@ 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);
+void ff_vector_fmul_add_rvv(float *dst, const float *src0, const float *src1,
+                             const float *src2, int len);
 
 void ff_vector_dmul_rvv(double *dst, const double *src0, const double *src1,
                          int len);
@@ -48,6 +50,7 @@ av_cold void ff_float_dsp_init_riscv(AVFloatDSPContext *fdsp)
         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;
+        fdsp->vector_fmul_add = ff_vector_fmul_add_rvv;
     }
 
     if (flags & AV_CPU_FLAG_RVV_F64) {

libavutil/riscv/float_dsp_rvv.S

@@ -74,6 +74,25 @@ NOHWF   mv       a2, a3
         ret
 endfunc
 
+// (a0) = (a1) * (a2) + (a3) [0..a4-1]
+func ff_vector_fmul_add_rvv, zve32f
+1:
+        vsetvli   t0, a4, e32, m1, ta, ma
+        vle32.v   v8, (a1)
+        sub       a4, a4, t0
+        vle32.v   v16, (a2)
+        sh2add    a1, t0, a1
+        vle32.v   v24, (a3)
+        sh2add    a2, t0, a2
+        vfmadd.vv v8, v16, v24
+        sh2add    a3, t0, a3
+        vse32.v   v8, (a0)
+        sh2add    a0, t0, a0
+        bnez      a4, 1b
+
+        ret
+endfunc
+
 // (a0) = (a1) * (a2) [0..a3-1]
 func ff_vector_dmul_rvv, zve64d
 1: