mirror of
https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/
synced 2026-11-10 20:48:12 +00:00
crypto: x86/aes-gcm - remove VAES+AVX10/256 optimized code
Remove the VAES+AVX10/256 optimized implementation of AES-GCM.
It's no longer expected to be useful for future CPUs, since Intel
changed the AVX10 specification to require 512-bit vectors.
In addition, it's no longer very useful to serve as the 256-bit fallback
for older Intel CPUs (Ice Lake and Tiger Lake) that downclock too
eagerly when 512-bit vectors are used. This is because I ended up
writing another 256-bit implementation anyway, using VAES+AVX2. The
VAES+AVX2 implementation is almost as fast as the VAES+AVX10/256 one, as
shown by the following tables. So, let's just use it instead.
Table 1: AES-256-GCM encryption throughput change,
CPU vs. message length in bytes:
| 16384 | 4096 | 4095 | 1420 | 512 | 500 |
----------------------+-------+-------+-------+-------+-------+-------+
Intel Ice Lake Server | -2% | -1% | 0% | -2% | -2% | 3% |
| 300 | 200 | 64 | 63 | 16 |
----------------------+-------+-------+-------+-------+-------+
Intel Ice Lake Server | 1% | 0% | 4% | 2% | -6% |
Table 2: AES-256-GCM decryption throughput change,
CPU vs. message length in bytes:
| 16384 | 4096 | 4095 | 1420 | 512 | 500 |
----------------------+-------+-------+-------+-------+-------+-------+
Intel Ice Lake Server | -1% | -1% | 1% | -2% | 0% | 2% |
| 300 | 200 | 64 | 63 | 16 |
----------------------+-------+-------+-------+-------+-------+
Intel Ice Lake Server | -1% | 4% | 1% | 0% | -5% |
Acked-by: Ard Biesheuvel <ardb@kernel.org>
Tested-by: Ard Biesheuvel <ardb@kernel.org>
Link: https://lore.kernel.org/r/20251002023117.37504-3-ebiggers@kernel.org
Signed-off-by: Eric Biggers <ebiggers@kernel.org>
This commit is contained in:
parent
fae3b96ba6
commit
f65e908606
2 changed files with 6 additions and 59 deletions
Loading…
Add table
Add a link
Reference in a new issue