summaryrefslogtreecommitdiffstats
path: root/src
diff options
context:
space:
mode:
Diffstat (limited to 'src')
-rw-r--r--src/corelib/global/qsimd.h18
-rw-r--r--src/corelib/global/qsimd_p.h29
2 files changed, 15 insertions, 32 deletions
diff --git a/src/corelib/global/qsimd.h b/src/corelib/global/qsimd.h
index 5fd34d0092..1f15c65594 100644
--- a/src/corelib/global/qsimd.h
+++ b/src/corelib/global/qsimd.h
@@ -1,7 +1,7 @@
/****************************************************************************
**
** Copyright (C) 2020 The Qt Company Ltd.
-** Copyright (C) 2018 Intel Corporation.
+** Copyright (C) 2022 Intel Corporation.
** Contact: https://www.qt.io/licensing/
**
** This file is part of the QtCore module of the Qt Toolkit.
@@ -89,20 +89,20 @@
# define QT_COMPILER_USES_mips_dspr2 -1
#endif
-#if defined(Q_PROCESSOR_X86)
-#if defined(Q_CC_MSVC)
+#if defined(Q_PROCESSOR_X86) && defined(Q_CC_MSVC)
// MSVC doesn't define __SSE2__, so do it ourselves
# if (defined(_M_X64) || _M_IX86_FP >= 2)
# define __SSE__ 1
# define __SSE2__ 1
# endif
-#if (defined(_M_AVX) || defined(__AVX__))
+# if (defined(_M_AVX) || defined(__AVX__))
// Visual Studio defines __AVX__ when /arch:AVX is passed, but not the earlier macros
// See: https://msdn.microsoft.com/en-us/library/b0084kay.aspx
# define __SSE3__ 1
# define __SSSE3__ 1
# define __SSE4_1__ 1
# define __SSE4_2__ 1
+# define __POPCNT__ 1
# ifndef __AVX__
# define __AVX__ 1
# endif
@@ -110,7 +110,15 @@
# ifdef __SSE2__
# define QT_VECTORCALL __vectorcall
# endif
-#endif
+# ifdef __AVX2__
+// MSVC defines __AVX2__ with /arch:AVX2
+# define __F16C__ 1
+# define __FMA__ 1
+# define __BMI__ 1
+# define __BMI2__ 1
+# define __LZCNT__ 1
+# endif
+// Starting with /arch:AVX512, MSVC defines all the macros
#endif
#if defined(Q_PROCESSOR_X86) && defined(__SSE2__)
diff --git a/src/corelib/global/qsimd_p.h b/src/corelib/global/qsimd_p.h
index 118a62e6ac..fdf5529135 100644
--- a/src/corelib/global/qsimd_p.h
+++ b/src/corelib/global/qsimd_p.h
@@ -198,33 +198,8 @@
# include <immintrin.h>
# endif
-# if defined(__SSE4_2__) && defined(QT_COMPILER_SUPPORTS_SIMD_ALWAYS) && (defined(Q_CC_INTEL) || defined(Q_CC_MSVC))
-// POPCNT instructions:
-// All processors that support SSE4.2 support POPCNT
-// (but neither MSVC nor the Intel compiler define this macro)
-# define __POPCNT__ 1
-# endif
-
-// AVX intrinsics
-# if defined(__AVX__) && defined(QT_COMPILER_SUPPORTS_SIMD_ALWAYS) && (defined(Q_CC_INTEL) || defined(Q_CC_MSVC))
-// PCLMULQDQ instructions:
-// All processors that support AVX support PCLMULQDQ
-// (but neither MSVC nor the Intel compiler define this macro)
-# define __PCLMUL__ 1
-# endif
-
-# if defined(__AVX2__) && defined(QT_COMPILER_SUPPORTS_SIMD_ALWAYS) && (defined(Q_CC_INTEL) || defined(Q_CC_MSVC))
-// F16C instructions:
-// All processors that support AVX2 support F16C:
-// (but neither MSVC nor the Intel compiler define this macro)
-# define __F16C__ 1
-# endif
-
-# if defined(__BMI__) && !defined(__BMI2__) && defined(Q_CC_INTEL)
-// BMI2 instructions:
-// All processors that support BMI support BMI2 (and AVX2)
-// (but neither MSVC nor the Intel compiler define this macro)
-# define __BMI2__ 1
+# if defined(Q_CC_GNU) && defined(__AVX2__) && (!defined(__BMI__) || !defined(__FMA__))
+# error "Please enable the BMI and FMA extensions; you probably want to use -march=haswell or -march=x86-64-v3 instead of -mavx2"
# endif
# include "qsimd_x86_p.h"