aboutsummaryrefslogtreecommitdiff
diff options
context:
space:
mode:
authorMichał Górny <mgorny@gentoo.org>2017-10-24 08:54:19 +0200
committerBenjamin Peterson <benjamin@python.org>2017-10-23 23:54:19 -0700
commit1aa00ff383c43335e4a5044274617dbf59bc839e (patch)
tree7a9435e859119e00b0159268a8456a2d7a7db6ea /Modules/_blake2
parentbpo-31847: Fix commented out tests in test_syntax. (#4084) (diff)
downloadcpython-1aa00ff383c43335e4a5044274617dbf59bc839e.tar.gz
cpython-1aa00ff383c43335e4a5044274617dbf59bc839e.tar.bz2
cpython-1aa00ff383c43335e4a5044274617dbf59bc839e.zip
fixes bpo-31834: Use optimized code for BLAKE2 only with SSSE3+ (#4066)
Rework the code choosing BLAKE2 code paths from using the optimized variant on all x86_64 machines to using it when SSSE3 or better supported instructions sets are available. Firstly, this solves the problem of using pure SSE2 code path on x86_64 machines. As reported in the bug, this code is slower than the reference code on all tested x86_64 machines. Furthermore, on Athlon64 that lacks SSSE3, it is even 2.5 times slower than the reference code! Checking for SSSE3 therefore ensures that the optimized implementation will only be used when it has a chance of performing better. Secondly, this makes it possible to use SSSE3+ optimizations on 32-bit x86 systems. This allows for even 2 times speed gain on modern 32-bit x86 systems (tested in a 32-bit chroot).
Diffstat (limited to 'Modules/_blake2')
-rw-r--r--Modules/_blake2/blake2b_impl.c4
-rw-r--r--Modules/_blake2/blake2s_impl.c4
2 files changed, 6 insertions, 2 deletions
diff --git a/Modules/_blake2/blake2b_impl.c b/Modules/_blake2/blake2b_impl.c
index b1ae3e9b628..3c2a035f3d7 100644
--- a/Modules/_blake2/blake2b_impl.c
+++ b/Modules/_blake2/blake2b_impl.c
@@ -26,7 +26,9 @@
#include "impl/blake2.h"
#include "impl/blake2-impl.h" /* for secure_zero_memory() and store48() */
-#ifdef BLAKE2_USE_SSE
+/* pure SSE2 implementation is very slow, so only use the more optimized SSSE3+
+ * https://bugs.python.org/issue31834 */
+#if defined(__SSSE3__) || defined(__SSE4_1__) || defined(__AVX__) || defined(__XOP__)
#include "impl/blake2b.c"
#else
#include "impl/blake2b-ref.c"
diff --git a/Modules/_blake2/blake2s_impl.c b/Modules/_blake2/blake2s_impl.c
index 3615a383db3..2c5697299fc 100644
--- a/Modules/_blake2/blake2s_impl.c
+++ b/Modules/_blake2/blake2s_impl.c
@@ -26,7 +26,9 @@
#include "impl/blake2.h"
#include "impl/blake2-impl.h" /* for secure_zero_memory() and store48() */
-#ifdef BLAKE2_USE_SSE
+/* pure SSE2 implementation is very slow, so only use the more optimized SSSE3+
+ * https://bugs.python.org/issue31834 */
+#if defined(__SSSE3__) || defined(__SSE4_1__) || defined(__AVX__) || defined(__XOP__)
#include "impl/blake2s.c"
#else
#include "impl/blake2s-ref.c"