Commit Graph

7 Commits

Author SHA1 Message Date
Andrey Semashev
3e6fee4751 Counter compiler optimizations when checking for SSSE3 and AVX2.
The compiler is able to optimize away most of the instructions that we test,
which results in checking only the compiler support for intrinsics, and not
the assembler capability to translate SSSE3 and AVX2 instructions from assembler
into the actual binary code. This commit forces the compiler to actually emit
instructions in the assembler, which then tests binutils capability to translate
those into binary code.
2019-01-08 00:56:24 +03:00
Andrey Semashev
e5d5c66683 Changed the AVX2 flag for Intel compilers to -xCORE-AVX2. This flag is said to be supported since Intel Compiler 13 and is supposed to be better than -march=core-avx2 since it does not set the target architecture, which may be set by the user. 2016-04-14 19:32:03 +03:00
Andrey Semashev
18df30ed33 Followup for the previous commit. 2015-03-25 00:48:52 +03:00
Andrey Semashev
5bb604ee39 Updated copyright dates. 2015-01-25 17:32:33 +03:00
Andrey Semashev
da1f448fd5 Corrected compiler switches to enable AVX2 and SSSE3 for Intel Compiler. Disabled some bogus warnings as well. 2014-06-21 22:12:17 +04:00
Andrey Semashev
a21d096e4a Updated copyright notes to 2014. 2014-01-19 18:46:09 +04:00
Andrey Semashev
34e7b55f48 To resolve problems with some testers that have non-numeric toolset versions compiler checks for SSSE3/AVX2 support have been reworked. The checks now use configure framework instead of relying on compiler version checks.
[SVN r84291]
2013-05-15 19:09:30 +00:00