[FFmpeg-cvslog] lavc/vorbisdsp: unroll RISC-V V inverse_coupling

[Rémi Denis-Courmont]

ffmpeg | branch: master | Rémi Denis-Courmont <remi at remlab.net> | Thu Jul 13 23:16:03 2023 +0300| [a28aa0475db99cc66c37c98fd5330e313817709c] | committer: Rémi Denis-Courmont

lavc/vorbisdsp: unroll RISC-V V inverse_coupling

This increases the group multiplier as per T-Head C910 benchmarks:

inverse_coupling_c: 4597.0
inverse_coupling_rvv_i32: 1312.7 (m1)
inverse_coupling_rvv_i32: 1116.7 (m2)
inverse_coupling_rvv_i32: 732.2  (m4)
inverse_coupling_rvv_i32: 898.0  (m8)

> http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=a28aa0475db99cc66c37c98fd5330e313817709c
---

 libavcodec/riscv/vorbisdsp_rvv.S | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/libavcodec/riscv/vorbisdsp_rvv.S b/libavcodec/riscv/vorbisdsp_rvv.S
index f45e7dc2f1..81a6c62a65 100644
--- a/libavcodec/riscv/vorbisdsp_rvv.S
+++ b/libavcodec/riscv/vorbisdsp_rvv.S
@@ -23,7 +23,7 @@
 func ff_vorbis_inverse_coupling_rvv, zve32f
         fmv.w.x   ft0, zero
 1:
-        vsetvli   t0, a2, e32, m1, ta, ma
+        vsetvli   t0, a2, e32, m4, ta, ma
         vle32.v   v16, (a1)
         sub       a2, a2, t0
         vle32.v   v24, (a0)