| commit | 0fd879120d6f25b65084b288c61e0412cf265ba4 | [log] [tgz] |
|---|---|---|
| author | Eric Biggers <ebiggers@google.com> | Tue Dec 17 03:54:56 2024 +0000 |
| committer | Boringssl LUCI CQ <boringssl-scoped@luci-project-accounts.iam.gserviceaccount.com> | Mon Jan 06 14:17:42 2025 -0800 |
| tree | a072def6560978025bd91549827c630d4fadd85e | |
| parent | 3b6e1be4391d96e81cee022f77f7bab85d51cf4e [diff] |
Disable the "AVX10/256" AES-GCM functions for now
Since there is now a VAES+AVX2 implementation of AES-GCM, and the future
of AVX10/256 is uncertain, disable the AES-GCM functions that use
AVX10/256 (equivalently AVX512 with a maximum vector length of 256
bits). This leaves VAES+AVX2 as the sole 256-bit support for now.
For now this just affects Intel Ice Lake and Tiger Lake (which actually
support AVX512, but where downclocking issues make 256-bit arguably
preferable to 512-bit), where a slight performance loss is seen on long
messages. The following tables compare AES-256-GCM throughput in MB/s
on Ice Lake server for various message lengths:
Encryption:
| 16384 | 4096 | 4095 | 1420 | 512 | 500 |
--------+-------+-------+-------+-------+-------+-------+
Before | 7533 | 6990 | 6220 | 5096 | 4200 | 2702 |
After | 7403 | 6879 | 6236 | 4980 | 4040 | 2868 |
| 300 | 200 | 64 | 63 | 16 |
--------+-------+-------+-------+-------+-------+
Before | 2086 | 1555 | 1031 | 657 | 433 |
After | 2069 | 1635 | 1045 | 667 | 430 |
Decryption:
| 16384 | 4096 | 4095 | 1420 | 512 | 500 |
--------+-------+-------+-------+-------+-------+-------+
Before | 7703 | 7140 | 6524 | 5283 | 4244 | 2990 |
After | 7572 | 7056 | 6494 | 5155 | 4224 | 3073 |
| 300 | 200 | 64 | 63 | 16 |
--------+-------+-------+-------+-------+-------+
Before | 2276 | 1733 | 1070 | 680 | 447 |
After | 2249 | 1743 | 1100 | 692 | 447 |
This change should be reconsidered if AVX10/256 sees widespread support,
as we shouldn't carry forward a restriction to AVX2 unnecessarily.
This change also replaces gcm_init_vpclmulqdq_avx10 with
gcm_init_vpclmulqdq_avx10_512, now instantiated using 512-bit vectors.
Otherwise it would be the only avx10 function left using 256-bit.
Change-Id: I7fd21568482118a2ce7a382e9042b187cd2739f7
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/74369
Reviewed-by: David Benjamin <davidben@google.com>
Commit-Queue: David Benjamin <davidben@google.com>
BoringSSL is a fork of OpenSSL that is designed to meet Google's needs.
Although BoringSSL is an open source project, it is not intended for general use, as OpenSSL is. We don't recommend that third parties depend upon it. Doing so is likely to be frustrating because there are no guarantees of API or ABI stability.
Programs ship their own copies of BoringSSL when they use it and we update everything as needed when deciding to make API changes. This allows us to mostly avoid compromises in the name of compatibility. It works for us, but it may not work for you.
BoringSSL arose because Google used OpenSSL for many years in various ways and, over time, built up a large number of patches that were maintained while tracking upstream OpenSSL. As Google's product portfolio became more complex, more copies of OpenSSL sprung up and the effort involved in maintaining all these patches in multiple places was growing steadily.
Currently BoringSSL is the SSL library in Chrome/Chromium, Android (but it's not part of the NDK) and a number of other apps/programs.
Project links:
To file a security issue, use the Chromium process and mention in the report this is for BoringSSL. You can ignore the parts of the process that are specific to Chromium/Chrome.
There are other files in this directory which might be helpful: