diff options
| author | Oswald Buddenhagen <oswald.buddenhagen@qt.io> | 2018-02-21 13:11:46 +0100 |
|---|---|---|
| committer | Thiago Macieira <thiago.macieira@intel.com> | 2018-03-18 06:11:20 +0000 |
| commit | e28192812168b676b57dc505b31eed3bfcba0e67 (patch) | |
| tree | b596603c8382efe09233287fab242afe73fb5121 | |
| parent | f2fb6659a777cf968957bf8ce8499400d6167a16 (diff) | |
configure: Auto-detect that AVX2 is enabled
If the user has an mkspec that turns AVX2 on (i.e., -march=haswell or
similar option added to QMAKE_CXXFLAGS), the compiler will be using
AVX2 for normal code generation. In this case, we should enable our
AVX2-optimized code as well.
Started-by: Thiago Macieira <thiago.macieira@intel.com>
Change-Id: I940917d6763842499b18fffd15134b004028c6ae
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
| -rw-r--r-- | src/core/configure.json | 2 |
1 files changed, 1 insertions, 1 deletions
diff --git a/src/core/configure.json b/src/core/configure.json index fc938444a..e72c5ed2b 100644 --- a/src/core/configure.json +++ b/src/core/configure.json @@ -92,7 +92,7 @@ "qt3d-simd-avx2": { "label": "Use AVX2 instructions", "purpose": "Use AVX2 SIMD instructions to accelerate matrix operations", - "autoDetect": false, + "autoDetect": "subarch.avx2", "enable": "input.qt3d-simd == 'avx2'", "disable": "input.qt3d-simd == 'sse2' || input.qt3d-simd == 'no' || arch.i386", "output": [ "privateFeature" ] |