@@ -30,6 +30,8 @@ void ff_vector_fmul_rvv(float *dst, const float *src0, const float *src1,
void ff_vector_fmul_scalar_rvv(float *dst, const float *src, float mul,
int len);
+void ff_vector_dmul_rvv(double *dst, const double *src0, const double *src1,
+ int len);
void ff_vector_dmul_scalar_rvv(double *dst, const double *src, double mul,
int len);
@@ -43,7 +45,9 @@ av_cold void ff_float_dsp_init_riscv(AVFloatDSPContext *fdsp)
fdsp->vector_fmul_scalar = ff_vector_fmul_scalar_rvv;
}
- if (flags & AV_CPU_FLAG_RVV_F64)
+ if (flags & AV_CPU_FLAG_RVV_F64) {
+ fdsp->vector_dmul = ff_vector_dmul_rvv;
fdsp->vector_dmul_scalar = ff_vector_dmul_scalar_rvv;
+ }
#endif
}
@@ -55,6 +55,23 @@ NOHWF mv a2, a3
ret
endfunc
+// (a0) = (a1) * (a2) [0..a3-1]
+func ff_vector_dmul_rvv, zve64d
+1:
+ vsetvli t0, a3, e64, m1, ta, ma
+ vle64.v v16, (a1)
+ sub a3, a3, t0
+ vle64.v v24, (a2)
+ sh3add a1, t0, a1
+ vfmul.vv v16, v16, v24
+ sh3add a2, t0, a2
+ vse64.v v16, (a0)
+ sh3add a0, t0, a0
+ bnez a3, 1b
+
+ ret
+endfunc
+
// (a0) = (a1) * fa0 [0..a2-1]
func ff_vector_dmul_scalar_rvv, zve64d
NOHWD fmv.d.x fa0, a2
From: Rémi Denis-Courmont <remi@remlab.net> --- libavutil/riscv/float_dsp_init.c | 6 +++++- libavutil/riscv/float_dsp_rvv.S | 17 +++++++++++++++++ 2 files changed, 22 insertions(+), 1 deletion(-)