diff options
author | Paul Davis <paul@linuxaudiosystems.com> | 2015-05-12 21:07:09 -0400 |
---|---|---|
committer | Paul Davis <paul@linuxaudiosystems.com> | 2015-06-29 14:18:13 -0400 |
commit | 6410aa896f974002f8539ee3ca2f70bf66c0a0af (patch) | |
tree | 7ef2a62226a6ac5927c563efe21ea6d1ecc0da72 /wscript | |
parent | e2a76746e65c42fd10a892ffd82300f1cf776ac6 (diff) |
Added optimized AVX function for sample processing
Added AVX versions of existing 5 SSE functions. Added 6th AVX function to copy vectors which is 1.5 times faster then memcpy.
Data consistency and validness is fully tested after processing with new AVX functions on aligned and non aligned buffers.
Diffstat (limited to 'wscript')
-rw-r--r-- | wscript | 4 |
1 files changed, 2 insertions, 2 deletions
@@ -417,12 +417,12 @@ int main() { return 0; }''', if (re.search ("(x86_64|AMD64)", cpu) != None): # on Windows sse is supported by 64 bit platforms only build_host_supports_sse = True - + # mingw GCC compiler to uses at&t (Unix specific) assembler dialect by default # compiler_flags.append (["--mmnemonic=att", "msyntax=att") compiler_flags.extend ([ flags_dict['sse'], flags_dict['fpmath-sse'], flags_dict['xmmintrinsics'], flags_dict['attasm'] ]) - + # end of processor-specific section # optimization section |