[FFmpeg-devel] [PATCH 07/10] riscv: float vector sum-and-difference with RVV

[remi at remlab.net]

From: Rémi Denis-Courmont <remi at remlab.net>

---
 libavutil/riscv/float_dsp_init.c |  2 ++
 libavutil/riscv/float_dsp_rvv.S  | 18 ++++++++++++++++++
 2 files changed, 20 insertions(+)

diff --git a/libavutil/riscv/float_dsp_init.c b/libavutil/riscv/float_dsp_init.c
index 8539fe9ac5..2165394585 100644
--- a/libavutil/riscv/float_dsp_init.c
+++ b/libavutil/riscv/float_dsp_init.c
@@ -30,6 +30,7 @@ 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_butterflies_float_rvv(float *v1, float *v2, int len);
 
 void ff_vector_dmul_rvv(double *dst, const double *src0, const double *src1,
                          int len);
@@ -47,6 +48,7 @@ av_cold void ff_float_dsp_init_riscv(AVFloatDSPContext *fdsp)
         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;
+        fdsp->butterflies_float = ff_butterflies_float_rvv;
 
         if (flags & AV_CPU_FLAG_ZVE64D) {
             fdsp->vector_dmul = ff_vector_dmul_rvv;
diff --git a/libavutil/riscv/float_dsp_rvv.S b/libavutil/riscv/float_dsp_rvv.S
index 27190c21ff..61beb868b0 100644
--- a/libavutil/riscv/float_dsp_rvv.S
+++ b/libavutil/riscv/float_dsp_rvv.S
@@ -98,6 +98,24 @@ func ff_vector_fmul_add_rvv
         ret
 endfunc
 
+// (a0) = (a0) + (a1), (a1) = (a0) - (a1) [0..a2-1]
+func ff_butterflies_float_rvv
+1:      vsetvli  t0, a2, e32, m8, ta, ma
+        slli     t1, t0, 2
+        vle32.v  v16, (a0)
+        vle32.v  v24, (a1)
+        vfadd.vv v0, v16, v24
+        vfsub.vv v8, v16, v24
+        sub      a2, a2, t0
+        vse32.v  v0, (a0)
+        add      a0, a0, t1
+        vse32.v  v8, (a1)
+        add      a1, a1, t1
+        bnez     a2, 1b
+
+        ret
+endfunc
+
 // (a0) = (a1) * (a2) [0..a3-1]
 func ff_vector_dmul_rvv
 1:      vsetvli  t0, a3, e64, m8, ta, ma
-- 
2.37.2