Magic-Mirror/llvm-capstone
1
0
Fork 0
mirror of https://github.com/capstone-engine/llvm-capstone.git synced 2024-11-24 22:30:13 +00:00
Code Issues Actions 2 Packages Projects Releases Wiki Activity

[libomptarget][nvptx] Fix build, symbol ordering in target_impl.h

Browse Source
This commit is contained in:
Jon Chesterfield 2019-12-19 01:50:06 +00:00
parent aaa5a5e7ff
commit 89a2bef27a
2 changed files with 11 additions and 11 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
openmp/libomptarget/deviceRTLs/amdgcn/src/target_impl.h
View File

@ -101,8 +101,6 @@ INLINE __kmpc_impl_lanemask_t __kmpc_impl_lanemask_gt() {
return __lanemask_gt();
}
EXTERN bool __kmpc_impl_is_first_active_thread();
INLINE uint32_t __kmpc_impl_smid() {
return __smid();
}
@ -126,6 +124,8 @@ INLINE __kmpc_impl_lanemask_t __kmpc_impl_activemask() {
return __ballot64(1);
}
EXTERN bool __kmpc_impl_is_first_active_thread();
EXTERN int32_t __kmpc_impl_shfl_sync(__kmpc_impl_lanemask_t, int32_t Var,
int32_t SrcLane);

18
openmp/libomptarget/deviceRTLs/nvptx/src/target_impl.h
View File

@ -94,15 +94,6 @@ INLINE __kmpc_impl_lanemask_t __kmpc_impl_lanemask_gt() {
return res;
}
// Return true if this is the first active thread in the warp.
INLINE bool __kmpc_impl_is_first_active_thread() {
unsigned long long Mask = __kmpc_impl_activemask();
unsigned long long ShNum = WARPSIZE - (GetThreadIdInBlock() % WARPSIZE);
unsigned long long Sh = Mask << ShNum;
// Truncate Sh to the 32 lower bits
return (unsigned)Sh == 0;
}
INLINE uint32_t __kmpc_impl_smid() {
uint32_t id;
asm("mov.u32 %0, %%smid;" : "=r"(id));
@ -142,6 +133,15 @@ INLINE __kmpc_impl_lanemask_t __kmpc_impl_activemask() {
#endif
}
// Return true if this is the first active thread in the warp.
INLINE bool __kmpc_impl_is_first_active_thread() {
unsigned long long Mask = __kmpc_impl_activemask();
unsigned long long ShNum = WARPSIZE - (GetThreadIdInBlock() % WARPSIZE);
unsigned long long Sh = Mask << ShNum;
// Truncate Sh to the 32 lower bits
return (unsigned)Sh == 0;
}
// In Cuda 9.0, the *_sync() version takes an extra argument 'mask'.
INLINE int32_t __kmpc_impl_shfl_sync(__kmpc_impl_lanemask_t Mask, int32_t Var,