* [FFmpeg-devel] [PATCH 2/2] lavu/float_dsp: optimise R-V V fmul_reverse & fmul_window
@ 2023-11-19 11:39 Rémi Denis-Courmont
0 siblings, 0 replies; only message in thread
From: Rémi Denis-Courmont @ 2023-11-19 11:39 UTC (permalink / raw)
To: ffmpeg-devel
Roll the loop to avoid slow gathers.
Before:
vector_fmul_reverse_c: 1561.7
vector_fmul_reverse_rvv_f32: 2410.2
vector_fmul_window_c: 2068.2
vector_fmul_window_rvv_f32: 1879.5
After:
vector_fmul_reverse_c: 1561.7
vector_fmul_reverse_rvv_f32: 916.2
vector_fmul_window_c: 2068.2
vector_fmul_window_rvv_f32: 1202.5
---
libavutil/riscv/float_dsp_rvv.S | 8 +++++---
1 file changed, 5 insertions(+), 3 deletions(-)
diff --git a/libavutil/riscv/float_dsp_rvv.S b/libavutil/riscv/float_dsp_rvv.S
index a2f9488249..ce5b6823d4 100644
--- a/libavutil/riscv/float_dsp_rvv.S
+++ b/libavutil/riscv/float_dsp_rvv.S
@@ -75,18 +75,19 @@ endfunc
func ff_vector_fmul_window_rvv, zve32f
// a0: dst, a1: src0, a2: src1, a3: window, a4: length
- vsetvli t0, zero, e16, m2, ta, ma
+ // e16/m2 and e32/m4 are possible but slower due to gather.
+ vsetvli t0, zero, e16, m1, ta, ma
sh2add a2, a4, a2
vid.v v0
sh3add t3, a4, a3
vadd.vi v0, v0, 1
sh3add t0, a4, a0
1:
- vsetvli t2, a4, e16, m2, ta, ma
+ vsetvli t2, a4, e16, m1, ta, ma
slli t4, t2, 2
vrsub.vx v2, v0, t2
sub t3, t3, t4
- vsetvli zero, zero, e32, m4, ta, ma
+ vsetvli zero, zero, e32, m2, ta, ma
sub a2, a2, t4
vle32.v v8, (t3)
sub t0, t0, t4
@@ -133,6 +134,7 @@ endfunc
// TODO factor vrsub, separate last iteration?
// (a0) = (a1) * reverse(a2) [0..a3-1]
func ff_vector_fmul_reverse_rvv, zve32f
+ // e16/m4 and e32/m8 are possible but slower due to gather.
vsetvli t0, zero, e16, m4, ta, ma
sh2add a2, a3, a2
vid.v v0
--
2.42.0
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
To unsubscribe, visit link above, or email
ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".
^ permalink raw reply [flat|nested] only message in thread
only message in thread, other threads:[~2023-11-19 11:39 UTC | newest]
Thread overview: (only message) (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2023-11-19 11:39 [FFmpeg-devel] [PATCH 2/2] lavu/float_dsp: optimise R-V V fmul_reverse & fmul_window Rémi Denis-Courmont
Git Inbox Mirror of the ffmpeg-devel mailing list - see https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
This inbox may be cloned and mirrored by anyone:
git clone --mirror https://master.gitmailbox.com/ffmpegdev/0 ffmpegdev/git/0.git
# If you have public-inbox 1.1+ installed, you may
# initialize and index your mirror using the following commands:
public-inbox-init -V2 ffmpegdev ffmpegdev/ https://master.gitmailbox.com/ffmpegdev \
ffmpegdev@gitmailbox.com
public-inbox-index ffmpegdev
Example config snippet for mirrors.
AGPL code for this site: git clone https://public-inbox.org/public-inbox.git