diff options
| author | Jussi Kivilinna <jussi.kivilinna@iki.fi> | 2023-02-18 11:14:02 +0200 |
|---|---|---|
| committer | Jussi Kivilinna <jussi.kivilinna@iki.fi> | 2023-02-22 20:27:56 +0200 |
| commit | f359a3ec7e845aa446836bd47994fe18d6d41e08 (patch) | |
| tree | ef2cea6fa0d247dc7276afec4be2150b0ce2b3ed | |
| parent | 855f1551fd921ced652dc0c3c03601dfcd063f1c (diff) | |
| download | libgcrypt-f359a3ec7e845aa446836bd47994fe18d6d41e08.tar.gz | |
aria-avx512: small optimization for aria_diff_m
* cipher/aria-gfni-avx512-amd64.S (aria_diff_m): Use 'vpternlogq' for
3-way XOR operation.
---
Using vpternlogq gives small performance improvement on AMD Zen4. With
Intel tiger-lake speed is the same as before.
Benchmark on AMD Ryzen 9 7900X (zen4, turbo-freq off):
Before:
ARIA128 | nanosecs/byte mebibytes/sec cycles/byte auto Mhz
ECB enc | 0.203 ns/B 4703 MiB/s 0.953 c/B 4700
ECB dec | 0.204 ns/B 4675 MiB/s 0.959 c/B 4700
CTR enc | 0.207 ns/B 4609 MiB/s 0.973 c/B 4700
CTR dec | 0.207 ns/B 4608 MiB/s 0.973 c/B 4700
After (~3% faster):
ARIA128 | nanosecs/byte mebibytes/sec cycles/byte auto Mhz
ECB enc | 0.197 ns/B 4847 MiB/s 0.925 c/B 4700
ECB dec | 0.197 ns/B 4852 MiB/s 0.924 c/B 4700
CTR enc | 0.200 ns/B 4759 MiB/s 0.942 c/B 4700
CTR dec | 0.200 ns/B 4772 MiB/s 0.939 c/B 4700
Cc: Taehee Yoo <ap420073@gmail.com>
Signed-off-by: Jussi Kivilinna <jussi.kivilinna@iki.fi>
| -rw-r--r-- | cipher/aria-gfni-avx512-amd64.S | 16 |
1 files changed, 6 insertions, 10 deletions
diff --git a/cipher/aria-gfni-avx512-amd64.S b/cipher/aria-gfni-avx512-amd64.S index 1076cf8b..0eaa2de8 100644 --- a/cipher/aria-gfni-avx512-amd64.S +++ b/cipher/aria-gfni-avx512-amd64.S @@ -406,21 +406,17 @@ vgf2p8affineinvqb $0, t2, y3, y3; \ vgf2p8affineinvqb $0, t2, y7, y7; - #define aria_diff_m(x0, x1, x2, x3, \ t0, t1, t2, t3) \ /* T = rotr32(X, 8); */ \ /* X ^= T */ \ - vpxorq x0, x3, t0; \ - vpxorq x1, x0, t1; \ - vpxorq x2, x1, t2; \ - vpxorq x3, x2, t3; \ /* X = T ^ rotr(X, 16); */ \ - vpxorq t2, x0, x0; \ - vpxorq x1, t3, t3; \ - vpxorq t0, x2, x2; \ - vpxorq t1, x3, x1; \ - vmovdqu64 t3, x3; + vmovdqa64 x0, t0; \ + vmovdqa64 x3, t3; \ + vpternlogq $0x96, x2, x1, x0; \ + vpternlogq $0x96, x2, x1, x3; \ + vpternlogq $0x96, t0, t3, x2; \ + vpternlogq $0x96, t0, t3, x1; #define aria_diff_word(x0, x1, x2, x3, \ x4, x5, x6, x7, \ |