History

lavc/g722dsp: optimise R-V V apply_qmf

This stores the constant coefficients deinterleaved, so that they can be loaded directly with NF=0. Unfortunately, we cannot optimise loading the input, due to insufficient memory alignment (not 32-bit). Before: g722_apply_qmf_c: 82.5 g722_apply_qmf_rvv_i32: 78.2 After: g722_apply_qmf_c: 82.5 g722_apply_qmf_rvv_i32: 65.2

Rémi Denis-Courmont committed 2 years ago

Commit b88d4051 parent e33ce0d1 file · +13 −11Browse files

Showing the changes to libavcodec (1 of 1 file). Show all

Unified · Split

libavcodec/riscv/g722dsp_rvv.S+13−11
@@ -24,7 +24,9 @@ func ff_g722_apply_qmf_rvv, zve32x · 23 unmodified lines
2424 lla t0, qmf_coeffs
2525 vsetivli zero, 12, e16, m2, ta, ma
2626 vlseg2e16.v v28, (a0)
27− vlseg2e16.v v24, (t0)
27+ addi t1, t0, 12 * 2
28+ vle16.v v24, (t0)
29+ vle16.v v26, (t1)
2830 vwmul.vv v16, v28, v24
2931 vwmul.vv v20, v30, v26
3032 vsetivli zero, 12, e32, m4, ta, ma
@@ -41,26 +43,26 @@ endfunc · 10 unmodified lines
4143 const qmf_coeffs, align=2
4244 .short 3
4345 .short -11
44− .short -11
45− .short 53
4646 .short 12
47− .short -156
4847 .short 32
49− .short 362
5048 .short -210
51− .short -805
5249 .short 951
5350 .short 3876
54− .short 3876
55− .short 951
5651 .short -805
57− .short -210
5852 .short 362
59− .short 32
6053 .short -156
61− .short 12
6254 .short 53
6355 .short -11
6456 .short -11
57+ .short 53
58+ .short -156
59+ .short 362
60+ .short -805
61+ .short 3876
62+ .short 951
63+ .short -210
64+ .short 32
65+ .short 12
66+ .short -11
6567 .short 3
6668 endconst

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