diff options
| author | Jussi Kivilinna <jussi.kivilinna@iki.fi> | 2022-03-10 21:36:34 +0200 |
|---|---|---|
| committer | Jussi Kivilinna <jussi.kivilinna@iki.fi> | 2022-03-10 22:26:16 +0200 |
| commit | 089223aa3b554e5c9a07d9176470308dab10ac8a (patch) | |
| tree | 76989cf83a83168f31901222b1291a34f38bd857 /cipher/Makefile.am | |
| parent | d820d27a3bce6365523fbcb6ec607b23dd4ca4e2 (diff) | |
| download | libgcrypt-089223aa3b554e5c9a07d9176470308dab10ac8a.tar.gz | |
SHA512: Add AVX512 implementation
* LICENSES: Add 'cipher/sha512-avx512-amd64.S'.
* cipher/Makefile.am: Add 'sha512-avx512-amd64.S'.
* cipher/sha512-avx512-amd64.S: New.
* cipher/sha512.c (USE_AVX512): New.
(do_sha512_transform_amd64_ssse3, do_sha512_transform_amd64_avx)
(do_sha512_transform_amd64_avx2): Add ASM_EXTRA_STACK to return value
only if assembly routine returned non-zero value.
[USE_AVX512] (_gcry_sha512_transform_amd64_avx512)
(do_sha512_transform_amd64_avx512): New.
(sha512_init_common) [USE_AVX512]: Use AVX512 implementation if HW
feature supported.
---
Benchmark on Intel Core i3-1115G4 (tigerlake):
Before:
| nanosecs/byte mebibytes/sec cycles/byte auto Mhz
SHA512 | 1.51 ns/B 631.6 MiB/s 6.17 c/B 4089
After (~29% faster):
| nanosecs/byte mebibytes/sec cycles/byte auto Mhz
SHA512 | 1.16 ns/B 819.0 MiB/s 4.76 c/B 4090
GnuPG-bug-id: T4460
Signed-off-by: Jussi Kivilinna <jussi.kivilinna@iki.fi>
Diffstat (limited to 'cipher/Makefile.am')
| -rw-r--r-- | cipher/Makefile.am | 2 |
1 files changed, 1 insertions, 1 deletions
diff --git a/cipher/Makefile.am b/cipher/Makefile.am index 3339c463..6eec9bd8 100644 --- a/cipher/Makefile.am +++ b/cipher/Makefile.am @@ -127,7 +127,7 @@ EXTRA_libcipher_la_SOURCES = \ sha256-armv8-aarch32-ce.S sha256-armv8-aarch64-ce.S \ sha256-intel-shaext.c sha256-ppc.c \ sha512.c sha512-ssse3-amd64.S sha512-avx-amd64.S \ - sha512-avx2-bmi2-amd64.S \ + sha512-avx2-bmi2-amd64.S sha512-avx512-amd64.S \ sha512-armv7-neon.S sha512-arm.S \ sha512-ppc.c sha512-ssse3-i386.c \ sm3.c sm3-avx-bmi2-amd64.S sm3-aarch64.S \ |