History

lavc/alacdsp: unroll RISC-V V loops

This increases the group multiplier as per T-Head C910 benchmarks: alac_append_extra_bits_mono_c: 803.0 alac_append_extra_bits_stereo_c: 1604.2 alac_decorrelate_stereo_c: 1077.5 LMUL=1 alac_append_extra_bits_mono_rvv_i32: 418.2 alac_append_extra_bits_stereo_rvv_i32: 693.2 alac_decorrelate_stereo_rvv_i32: 673.5 LMUL=2 alac_append_extra_bits_mono_rvv_i32: 382.2 alac_append_extra_bits_stereo_rvv_i32: 648.2 alac_decorrelate_stereo_rvv_i32: 542.7 LMUL=4 alac_append_extra_bits_mono_rvv_i32: 241.5 alac_append_extra_bits_stereo_rvv_i32: 512.7 alac_decorrelate_stereo_rvv_i32: 364.2 LMUL=8 alac_append_extra_bits_mono_rvv_i32: 239.7 alac_append_extra_bits_stereo_rvv_i32: 497.2 alac_decorrelate_stereo_rvv_i32: 426.7

Rémi Denis-Courmont committed 3 years ago

Commit c541ecf1 parent a28aa041 file · +3 −3Browse files

Unified · Split

libavcodec/riscv/alacdsp_rvv.S+3−3
@@ -25,7 +25,7 @@ func ff_alac_decorrelate_stereo_rvv, zve32x · 24 unmodified lines
2525 ld a4, 8(a0)
2626 ld a0, 0(a0)
2727 1:
28− vsetvli t0, a1, e32, m1, ta, ma
28+ vsetvli t0, a1, e32, m4, ta, ma
2929 vle32.v v24, (a4)
3030 sub a1, a1, t0
3131 vle32.v v16, (a0)
@@ -47,7 +47,7 @@ func ff_alac_append_extra_bits_mono_rvv, zve32x · 15 unmodified lines
4747 ld a0, (a0)
4848 ld a1, (a1)
4949 1:
50− vsetvli t0, a4, e32, m1, ta, ma
50+ vsetvli t0, a4, e32, m8, ta, ma
5151 vle32.v v16, (a0)
5252 sub a4, a4, t0
5353 vle32.v v24, (a1)
@@ -67,7 +67,7 @@ func ff_alac_append_extra_bits_stereo_rvv, zve32x · 13 unmodified lines
6767 ld a7, 8(a1)
6868 ld a1, (a1)
6969 1:
70− vsetvli t0, a4, e32, m1, ta, ma
70+ vsetvli t0, a4, e32, m8, ta, ma
7171 vle32.v v16, (a0)
7272 sub a4, a4, t0
7373 vle32.v v0, (a6)

Keyboard shortcuts

General

Show keyboard shortcuts
?
Focus the current search
/

Issues and pull requests

Open filters
f

Issues, pull requests and Inbox

Select next item
j↓
Select previous item
k↑
Open selected item
Enter
Dismiss filters or clear selection
Esc

Repository code

Go to file
t