diff options
| author | Lion Yang <lion@aosc.xyz> | 2017-01-05 06:25:59 +0800 |
|---|---|---|
| committer | Ian Lance Taylor <iant@golang.org> | 2017-01-05 17:50:56 +0000 |
| commit | b820ef5c33c1b508c7b0df8827b6f13c9480acea (patch) | |
| tree | bc26cf9150ec949afe490a11b5095cc11f3117f7 | |
| parent | a2b615d5270f0bc2ee1dfcdd7849bdd05ee76a14 (diff) | |
| download | go-b820ef5c33c1b508c7b0df8827b6f13c9480acea.tar.gz go-b820ef5c33c1b508c7b0df8827b6f13c9480acea.zip | |
x/crypto/chacha20poly1305: fix detection of BMI on AMD64
This change uses runtime.support_bmi2 as an additional condition
to examine the usability of AVX2 version algorithm, fixes
the crash on the platfrom which supports AVX2 but not support BMI2.
Fixes #18512
Change-Id: I408c0844ae2eb242dacf70cb9e8cec1b8f3bd941
Reviewed-on: https://go-review.googlesource.com/34851
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
| -rw-r--r-- | src/vendor/golang_org/x/crypto/chacha20poly1305/chacha20poly1305_amd64.s | 14 |
1 files changed, 9 insertions, 5 deletions
diff --git a/src/vendor/golang_org/x/crypto/chacha20poly1305/chacha20poly1305_amd64.s b/src/vendor/golang_org/x/crypto/chacha20poly1305/chacha20poly1305_amd64.s index 2fa9b555a8..7c6a1a5e68 100644 --- a/src/vendor/golang_org/x/crypto/chacha20poly1305/chacha20poly1305_amd64.s +++ b/src/vendor/golang_org/x/crypto/chacha20poly1305/chacha20poly1305_amd64.s @@ -278,9 +278,11 @@ TEXT ·chacha20Poly1305Open(SB), 0, $288-97 MOVQ ad+72(FP), adp // Check for AVX2 support - CMPB runtime·support_avx2(SB), $1 - JE chacha20Poly1305Open_AVX2 - + CMPB runtime·support_avx2(SB), $0 + JE noavx2bmi2Open + CMPB runtime·support_bmi2(SB), $1 // for MULXQ + JE chacha20Poly1305Open_AVX2 +noavx2bmi2Open: // Special optimization, for very short buffers CMPQ inl, $128 JBE openSSE128 // About 16% faster @@ -1485,9 +1487,11 @@ TEXT ·chacha20Poly1305Seal(SB), 0, $288-96 MOVQ ad+72(FP), adp // Check for AVX2 support - CMPB runtime·support_avx2(SB), $1 + CMPB runtime·support_avx2(SB), $0 + JE noavx2bmi2Seal + CMPB runtime·support_bmi2(SB), $1 // for MULXQ JE chacha20Poly1305Seal_AVX2 - +noavx2bmi2Seal: // Special optimization, for very short buffers CMPQ inl, $128 JBE sealSSE128 // About 15% faster |