Enable AVX code for SHA-*.

SHA-1, SHA-256, and SHA-512 get a 12-26%, 17-23%, and 33-37% improvement,
respectively on x86-64. SHA-1 and SHA-256 get a 8-20% and 14-17% improvement on
x86. (x86 does not have AVX code for SHA-512.) This costs us 12k of binary size
on x86-64 and 8k of binary size on x86.

$ bssl speed SHA- (x86-64, before)
Did 4811000 SHA-1 (16 bytes) operations in 1000013us (4810937.5 ops/sec): 77.0 MB/s
Did 1414000 SHA-1 (256 bytes) operations in 1000253us (1413642.3 ops/sec): 361.9 MB/s
Did 56000 SHA-1 (8192 bytes) operations in 1002640us (55852.5 ops/sec): 457.5 MB/s
Did 2536000 SHA-256 (16 bytes) operations in 1000140us (2535645.0 ops/sec): 40.6 MB/s
Did 603000 SHA-256 (256 bytes) operations in 1001613us (602028.9 ops/sec): 154.1 MB/s
Did 25000 SHA-256 (8192 bytes) operations in 1010132us (24749.2 ops/sec): 202.7 MB/s
Did 1767000 SHA-512 (16 bytes) operations in 1000477us (1766157.5 ops/sec): 28.3 MB/s
Did 638000 SHA-512 (256 bytes) operations in 1000933us (637405.3 ops/sec): 163.2 MB/s
Did 32000 SHA-512 (8192 bytes) operations in 1025646us (31199.8 ops/sec): 255.6 MB/s

$ bssl speed SHA- (x86-64, after)
Did 5438000 SHA-1 (16 bytes) operations in 1000060us (5437673.7 ops/sec): 87.0 MB/s
Did 1590000 SHA-1 (256 bytes) operations in 1000181us (1589712.3 ops/sec): 407.0 MB/s
Did 71000 SHA-1 (8192 bytes) operations in 1007958us (70439.4 ops/sec): 577.0 MB/s
Did 2955000 SHA-256 (16 bytes) operations in 1000251us (2954258.5 ops/sec): 47.3 MB/s
Did 740000 SHA-256 (256 bytes) operations in 1000628us (739535.6 ops/sec): 189.3 MB/s
Did 31000 SHA-256 (8192 bytes) operations in 1019619us (30403.5 ops/sec): 249.1 MB/s
Did 2348000 SHA-512 (16 bytes) operations in 1000285us (2347331.0 ops/sec): 37.6 MB/s
Did 878000 SHA-512 (256 bytes) operations in 1001064us (877066.8 ops/sec): 224.5 MB/s
Did 43000 SHA-512 (8192 bytes) operations in 1002485us (42893.4 ops/sec): 351.4 MB/s

$ bssl speed SHA- (x86, before, SHA-512 redacted because irrelevant)
Did 4319000 SHA-1 (16 bytes) operations in 1000066us (4318715.0 ops/sec): 69.1 MB/s
Did 1306000 SHA-1 (256 bytes) operations in 1000437us (1305429.5 ops/sec): 334.2 MB/s
Did 58000 SHA-1 (8192 bytes) operations in 1014807us (57153.7 ops/sec): 468.2 MB/s
Did 2291000 SHA-256 (16 bytes) operations in 1000343us (2290214.5 ops/sec): 36.6 MB/s
Did 594000 SHA-256 (256 bytes) operations in 1000684us (593594.0 ops/sec): 152.0 MB/s
Did 25000 SHA-256 (8192 bytes) operations in 1030688us (24255.6 ops/sec): 198.7 MB/s

$ bssl speed SHA- (x86, after, SHA-512 redacted because irrelevant)
Did 4673000 SHA-1 (16 bytes) operations in 1000063us (4672705.6 ops/sec): 74.8 MB/s
Did 1484000 SHA-1 (256 bytes) operations in 1000453us (1483328.1 ops/sec): 379.7 MB/s
Did 69000 SHA-1 (8192 bytes) operations in 1008305us (68431.7 ops/sec): 560.6 MB/s
Did 2684000 SHA-256 (16 bytes) operations in 1000196us (2683474.0 ops/sec): 42.9 MB/s
Did 679000 SHA-256 (256 bytes) operations in 1000525us (678643.7 ops/sec): 173.7 MB/s
Did 29000 SHA-256 (8192 bytes) operations in 1033251us (28066.8 ops/sec): 229.9 MB/s

Change-Id: I952a3b4fc4c52ebb50690da3b8c97770e8342e98
Reviewed-on: https://boringssl-review.googlesource.com/6470
Reviewed-by: Adam Langley <agl@google.com>
4 files changed
tree: 8246d7b6d01ecc3e955c799ac047f3ba8f10e18f
  1. .clang-format
  2. .gitignore
  3. BUILDING.md
  4. CMakeLists.txt
  5. FUZZING.md
  6. LICENSE
  7. PORTING.md
  8. README.md
  9. STYLE.md
  10. codereview.settings
  11. crypto/
  12. decrepit/
  13. fuzz/
  14. include/
  15. ssl/
  16. tool/
  17. util/
README.md

BoringSSL

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.

There are other files in this directory which might be helpful:

  • PORTING.md: how to port OpenSSL-using code to BoringSSL.
  • BUILDING.md: how to build BoringSSL
  • STYLE.md: rules and guidelines for coding style.
  • include/openssl: public headers with API documentation in comments. Also available online.
  • FUZZING.md: information about fuzzing BoringSSL.