diff options
| author | Thiago Macieira <thiago.macieira@intel.com> | 2017-04-25 12:02:09 -0300 |
|---|---|---|
| committer | Thiago Macieira <thiago.macieira@intel.com> | 2017-11-22 17:58:21 +0000 |
| commit | d813c66bfcfac1837814ec4d174d0389172f0d4c (patch) | |
| tree | 0fb335520ed92297e7db89dbe8b4b95826a1ac90 | |
| parent | b71b7461b0b95bbf02c215019381b19e4070e07c (diff) | |
Fix the build when AVX2 is enabled but __F16C__ isn't defined
If -mavx2 is used, __AVX2__ is defined, which enables the F16C code
after commit 280e321e52fd4e86545f3f0d4bd4e047786a897e, but that was
wrong since we aren't allowed to use the F16C intrinsics with either
Clang or GCC (we can only do that with GCC 4.9 and Clang 4.8, and only
with an __attribute__ decoration).
With ICC and MSVC, we are allowed to use the intrinsics, but the
#include was missing.
[ChangeLog][QtCore] Fixed a compilation issue with qfloat16 if AVX2
support is enabled in the compiler. Since all processors that support
AVX2 also support F16C, for GCC and Clang it is recommended to either
add -mf16c to your build or to use the corresponding -march= switch.
Task-number: QTBUG-64529
Change-Id: I84e363d735b443cb9beefffd14b8ac1fd4baa978
Reviewed-by: Allan Sandfeld Jensen <allan.jensen@qt.io>
| -rw-r--r-- | src/corelib/global/qfloat16.h | 15 |
1 files changed, 12 insertions, 3 deletions
diff --git a/src/corelib/global/qfloat16.h b/src/corelib/global/qfloat16.h index 89a62a93db..a0aa9496b4 100644 --- a/src/corelib/global/qfloat16.h +++ b/src/corelib/global/qfloat16.h @@ -44,7 +44,16 @@ #include <QtCore/qmetatype.h> #include <string.h> -#if defined __F16C__ +#if defined(QT_COMPILER_SUPPORTS_F16C) && defined(__AVX2__) && !defined(__F16C__) +// All processors that support AVX2 do support F16C too. That doesn't mean +// we're allowed to use the intrinsics directly, so we'll do it only for +// the Intel and Microsoft's compilers. +# if defined(Q_CC_INTEL) || defined(Q_CC_MSVC) +# define __F16C__ 1 +# endif +#endif + +#if defined(QT_COMPILER_SUPPORTS_F16C) && defined(__F16C__) #include <immintrin.h> #endif @@ -116,7 +125,7 @@ QT_WARNING_DISABLE_CLANG("-Wc99-extensions") QT_WARNING_DISABLE_GCC("-Wold-style-cast") inline qfloat16::qfloat16(float f) Q_DECL_NOTHROW { -#if defined(QT_COMPILER_SUPPORTS_F16C) && (defined(__F16C__) || defined(__AVX2__)) +#if defined(QT_COMPILER_SUPPORTS_F16C) && defined(__F16C__) __m128 packsingle = _mm_set_ss(f); __m128i packhalf = _mm_cvtps_ph(packsingle, 0); b16 = _mm_extract_epi16(packhalf, 0); @@ -134,7 +143,7 @@ QT_WARNING_POP inline qfloat16::operator float() const Q_DECL_NOTHROW { -#if defined(QT_COMPILER_SUPPORTS_F16C) && (defined(__F16C__) || defined(__AVX2__)) +#if defined(QT_COMPILER_SUPPORTS_F16C) && defined(__F16C__) __m128i packhalf = _mm_cvtsi32_si128(b16); __m128 packsingle = _mm_cvtph_ps(packhalf); return _mm_cvtss_f32(packsingle); |