devtools/eigen
1
0
Fork 0
mirror of https://gitlab.com/libeigen/eigen.git synced 2026-04-10 11:34:33 +08:00
Code Issues Packages Projects Releases Wiki Activity

Altivec fixes for Darwin: do not use unsupported VSX insns

Browse Source
This commit is contained in:
Sergey Fedorov
2023-01-12 16:33:33 +00:00
committed by Antonio Sánchez
parent 6156797016
commit 4d05765345
6 changed files with 43 additions and 38 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
Eigen/src/Core/util/ConfigureVectorization.h
View File

@@ -352,10 +352,10 @@
#endif
} // end extern "C"
#elif defined __VSX__
#elif defined(__VSX__) && !defined(__APPLE__)
#define EIGEN_VECTORIZE
#define EIGEN_VECTORIZE_VSX
#define EIGEN_VECTORIZE_VSX 1
#include <altivec.h>
// We need to #undef all these ugly tokens defined in <altivec.h>
// => use __vector instead of vector

15
Eigen/src/Core/util/Macros.h
View File

@@ -371,7 +371,7 @@
#endif
/// \internal EIGEN_ARCH_PPC set to 1 if the architecture is PowerPC
#if defined(__powerpc__) || defined(__ppc__) || defined(_M_PPC)
#if defined(__powerpc__) || defined(__ppc__) || defined(_M_PPC) || defined(__POWERPC__)
#define EIGEN_ARCH_PPC 1
#else
#define EIGEN_ARCH_PPC 0
@@ -976,11 +976,16 @@ namespace Eigen {
// directly for std::complex<T>, Eigen::half, Eigen::bfloat16. For these,
// you will need to apply to the underlying POD type.
#if EIGEN_ARCH_PPC && EIGEN_COMP_GNUC_STRICT
// This seems to be broken on clang. Packet4f is loaded into a single
// register rather than a vector, zeroing out some entries. Integer
// This seems to be broken on clang. Packet4f is loaded into a single
// register rather than a vector, zeroing out some entries. Integer
// types also generate a compile error.
// General, Altivec, VSX.
#define EIGEN_OPTIMIZATION_BARRIER(X) __asm__ ("" : "+r,v,wa" (X));
#if EIGEN_OS_MAC
// General, Altivec for Apple (VSX were added in ISA v2.06):
#define EIGEN_OPTIMIZATION_BARRIER(X) __asm__ ("" : "+r,v" (X));
#else
// General, Altivec, VSX otherwise:
#define EIGEN_OPTIMIZATION_BARRIER(X) __asm__ ("" : "+r,v,wa" (X));
#endif
#elif EIGEN_ARCH_ARM_OR_ARM64
#ifdef __ARM_FP
// General, VFP or NEON.