This is an archive of the discontinued LLVM Phabricator instance.

Differential D40198

[CUDA] Tweak CUDA wrappers to make cuda-9 work with libc++
ClosedPublic

Authored by tra on Nov 17 2017, 2:31 PM.

Details

Reviewers
jlebar
EricWF
Commits
rG05914bf482ad: [CUDA] Tweak CUDA wrappers to make cuda-9 work with libc++
rC319485: [CUDA] Tweak CUDA wrappers to make cuda-9 work with libc++
rL319485: [CUDA] Tweak CUDA wrappers to make cuda-9 work with libc++
Summary

CUDA-9 headers check for specific libc++ version and ifdef out
some of the definitions we need if LIBCPP_VERSION >= 3800.

Diff Detail

Repository
rL LLVM

Event Timeline

tra created this revision.Nov 17 2017, 2:31 PM
Herald added a reviewer: EricWF. · View Herald TranscriptNov 17 2017, 2:31 PM
Herald added a subscriber: sanjoy. · View Herald Transcript
tra added a comment.Nov 29 2017, 3:17 PM

ping.

jlebar accepted this revision.Nov 30 2017, 11:49 AM
This revision is now accepted and ready to land.Nov 30 2017, 11:49 AM
Closed by commit rL319485: [CUDA] Tweak CUDA wrappers to make cuda-9 work with libc++ (authored by tra). · Explain WhyNov 30 2017, 2:22 PM
This revision was automatically updated to reflect the committed changes.

Revision Contents

PathSize
cfe/
trunk/
lib/
Headers/
6 lines

Diff 125021

cfe/trunk/lib/Headers/__clang_cuda_runtime_wrapper.h

Show First 20 Lines • Show All 264 Lines • ▼ Show 20 Lines
#pragma push_macro("__GNUC__") #pragma push_macro("__GNUC__")
#undef __GNUC__ #undef __GNUC__
#define signbit __ignored_cuda_signbit #define signbit __ignored_cuda_signbit
// CUDA-9 omits device-side definitions of some math functions if it sees // CUDA-9 omits device-side definitions of some math functions if it sees
// include guard from math.h wrapper from libstdc++. We have to undo the header // include guard from math.h wrapper from libstdc++. We have to undo the header
// guard temporarily to get the definitions we need. // guard temporarily to get the definitions we need.
#pragma push_macro("_GLIBCXX_MATH_H") #pragma push_macro("_GLIBCXX_MATH_H")
#pragma push_macro("_LIBCPP_VERSION")
#if CUDA_VERSION >= 9000 #if CUDA_VERSION >= 9000
#undef _GLIBCXX_MATH_H #undef _GLIBCXX_MATH_H
// We also need to undo another guard that checks for libc++ 3.8+
#ifdef _LIBCPP_VERSION
#define _LIBCPP_VERSION 3700
#endif
#endif #endif
#include "math_functions.hpp" #include "math_functions.hpp"
#pragma pop_macro("_GLIBCXX_MATH_H") #pragma pop_macro("_GLIBCXX_MATH_H")
#pragma pop_macro("_LIBCPP_VERSION")
#pragma pop_macro("__GNUC__") #pragma pop_macro("__GNUC__")
#pragma pop_macro("signbit") #pragma pop_macro("signbit")
#pragma pop_macro("__host__") #pragma pop_macro("__host__")
#include "texture_indirect_functions.h" #include "texture_indirect_functions.h"
// Restore state of __CUDA_ARCH__ and __THROW we had on entry. // Restore state of __CUDA_ARCH__ and __THROW we had on entry.
▲ Show 20 Lines • Show All 89 Lines • Show Last 20 Lines