Support vpinsrq in delocater #1543
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
nit: |
dkostic
approved these changes
Apr 22, 2024
skmcgrail
approved these changes
Apr 22, 2024
skmcgrail
pushed a commit
to skmcgrail/aws-lc
that referenced
this pull request
May 1, 2024
Instead of generating portable code, gcc can be configured with e.g. march=cpu-type that allows it to generate code using instructions from instruction sets supported on up-to cpu-type. This fails because vpinsrq is not supported in the delocater. This PR adds support for vpinsrq an instruction from the set AVX512DQ. I don't think there are other 4-argument instructions, for where a GOT reloc can be emitted, that we need to support right now. This also makes the implementation a tad easier, because we do not need to cater for the relocation being either the first or second argument - it can only be the second. Otherwise, the implementation follows the one for the type instrThreeArg. (cherry picked from commit d940162)
skmcgrail
added a commit
that referenced
this pull request
May 2, 2024
d52018b Minor functions to build with Ruby's cipher module (#1564) 364d28b Changed SSL_client_hello_get0_ciphers to align with OpenSSL behavior (#1542) e8eb7de ppc64le: EVP_has_aes_hardware is false w/ no-asm (#1566) d726d06 OpenBSD 7.4 and 7.5 Support (#1437) a66c66e Remove comments about overread for entropy generation (#1551) f8a575f Migrate from __FreeBSD__ to __FreeBSD_version (#1562) c31d1ce Centralize handling of s2n-bignum alt/non-alt function selection (#1547) 00f3c45 CI for other MacOS versions (#1558) 0541314 Cleanup remaing duplicate symbol definitions and turn Wredundant-decls on (#1561) 4d280eb Fix ec2 CI testing framework (#1541) 9a4b43e Update x25519_test.cc array initialization to avoid a bug with a GCC 13 warning (#1555) 388cbe7 Remove duplicate X509_OBJECT_new and X509_OBJECT_free declarations (#1560) 2ea6706 Avoid 'z' format with MSVCRT (#1559) c25dc2a Add dependency to python3-six in github action grpc (#1554) 2bdcba3 Link porting guide table to header documentation (#1540) 311ca38 Basic GH CI build/test with full range of gcc/clang (#1546) 1f19717 Add SHA3-256 KAT to FIPS self-test (#1549) 0f3548a Add EC point add/dbl to speed.cc (#1545) d7ddfc4 Fix the NTP integration test (NTP website changed) (#1548) 8ccd85b Fix skipped tests in Mariadb integration CI (#1533) d940162 Support vpinsrq in delocater (#1543) 4cd6d21 Remove redundant test exec libraries (#1544) 56f3569 [ML-KEM] Add experimental support for ML-KEM-512-IPD (#1516) c295aef Upstream merge 2024 04 16 (#1535) 2e51629 Re-add function 0aebf17 Define OPENSSL_NO_TLS_PHA, typedef PSK callback signatures (#1526) 46056cf Pull the string-based extensions APIs into their own section 960ea42 Unexport X509_VERIFY_PARAM_lookup 3c597b1 Remove X509_VERIFY_PARAM_get0_peername 9c399e5 Document some key usage accessors 2fe70b5 Simplify and document X509_supported_extension 2e04897 Const-correct X509_LOOKUP_METHOD 9826568 Replace X509_LOOKUP_ctrl with real functions e47c056 Tidy up x509_lu.c functions a little 62e019f Clean up the by_file_ctrl x509 code to be slightly less obtuse 45c46c2 Use relative links in markdown files By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license and the ISC license.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Issues:
P125414272
Description of changes:
Instead of generating portable code, gcc can be configured with e.g.
march=cpu-typethat allows it to generate code using instructions from instruction sets supported on up-tocpu-type.In one of these cases, we for example saw:
In this case, the compiler tries to be smart and get the necessary memory addresses for
EVP_PKEY_CTX_freeandEVP_PKEY_CTX_dupto populateEVP_MD_pctx_opse.g.:This fails because
vpinsrqis not supported in the delocater.This PR adds support for
vpinsrqan instruction from the setAVX512DQ. I don't think there are other 4-argument instructions, for where a GOT reloc can be emitted, that we need to support right now. This also makes the implementation a tad easier, because we do not need to cater for the relocation being either the first or second argument - it can only be the second.Otherwise, the implementation follows the one for the type
instrThreeArg.Testing:
Tested this in an environment with conditions that triggers the missing support for
vpinsrq. Before it failed, after this PR is applied, build and tests succeeded.By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license and the ISC license.