Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel
os
linux
lib/crypto: x86/blake2s: Use vpternlogd for 3-input XORs
AVX-512 supports 3-input XORs via the vpternlogd (or vpternlogq)
instruction with immediate 0x96. This approach, vs. the alternative of
two vpxor instructions, is already used in the CRC, AES-GCM, and AES-XTS
code, since it reduces the instruction count and is faster on some CPUs.
Make blake2s_compress_avx512() take advantage of it too.
Reviewed-by: Ard Biesheuvel <ardb@kernel.org>
Link: https://lore.kernel.org/r/20251102234209.62133-7-ebiggers@kernel.org
Signed-off-by: Eric Biggers <ebiggers@kernel.org>
+2
-4
+2
-4
lib/crypto/x86/blake2s-core.S
+2
-4
lib/crypto/x86/blake2s-core.S
···
278
278
jne .Lavx512_roundloop
279
279
280
280
// Compute the new h: h[0..7] ^= v[0..7] ^ v[8..15]
281
-
vpxor %xmm10,%xmm0,%xmm0
282
-
vpxor %xmm11,%xmm1,%xmm1
283
-
vpxor %xmm2,%xmm0,%xmm0
284
-
vpxor %xmm3,%xmm1,%xmm1
281
+
vpternlogd $0x96,%xmm10,%xmm2,%xmm0
282
+
vpternlogd $0x96,%xmm11,%xmm3,%xmm1
285
283
decq NBLOCKS
286
284
jne .Lavx512_mainloop
287
285