[Openmp-commits] [openmp] r373793 - Use named constant to indicate all lanes, to handle 32 and 64 wide architectures

[Jon Chesterfield via Openmp-commits]

Author: jonchesterfield
Date: Fri Oct  4 14:39:22 2019
New Revision: 373793

URL: http://llvm.org/viewvc/llvm-project?rev=373793&view=rev
Log:
Use named constant to indicate all lanes, to handle 32 and 64 wide architectures

Summary: Use named constant to indicate all lanes, to handle 32 and 64 wide architectures

Reviewers: ABataev, jdoerfert, grokos, ronlieb

Reviewed By: grokos

Subscribers: ronlieb, openmp-commits

Tags: #openmp

Differential Revision: https://reviews.llvm.org/D68369

Modified:
    openmp/trunk/libomptarget/deviceRTLs/nvptx/src/parallel.cu
    openmp/trunk/libomptarget/deviceRTLs/nvptx/src/reduction.cu
    openmp/trunk/libomptarget/deviceRTLs/nvptx/src/target_impl.h

Modified: openmp/trunk/libomptarget/deviceRTLs/nvptx/src/parallel.cu
URL: http://llvm.org/viewvc/llvm-project/openmp/trunk/libomptarget/deviceRTLs/nvptx/src/parallel.cu?rev=373793&r1=373792&r2=373793&view=diff
==============================================================================
--- openmp/trunk/libomptarget/deviceRTLs/nvptx/src/parallel.cu (original)
+++ openmp/trunk/libomptarget/deviceRTLs/nvptx/src/parallel.cu Fri Oct  4 14:39:22 2019
@@ -320,7 +320,7 @@ EXTERN bool __kmpc_kernel_parallel(void
     // can be changed incorrectly because of threads divergence.
     bool IsActiveParallelRegion = threadsInTeam != 1;
     IncParallelLevel(IsActiveParallelRegion,
-                     IsActiveParallelRegion ? 0xFFFFFFFF : 1u);
+                     IsActiveParallelRegion ? __kmpc_impl_all_lanes : 1u);
   }
 
   return isActive;
@@ -347,7 +347,7 @@ EXTERN void __kmpc_kernel_end_parallel()
   // be changed incorrectly because of threads divergence.
     bool IsActiveParallelRegion = threadsInTeam != 1;
     DecParallelLevel(IsActiveParallelRegion,
-                     IsActiveParallelRegion ? 0xFFFFFFFF : 1u);
+                     IsActiveParallelRegion ? __kmpc_impl_all_lanes : 1u);
 }
 
 ////////////////////////////////////////////////////////////////////////////////

Modified: openmp/trunk/libomptarget/deviceRTLs/nvptx/src/reduction.cu
URL: http://llvm.org/viewvc/llvm-project/openmp/trunk/libomptarget/deviceRTLs/nvptx/src/reduction.cu?rev=373793&r1=373792&r2=373793&view=diff
==============================================================================
--- openmp/trunk/libomptarget/deviceRTLs/nvptx/src/reduction.cu (original)
+++ openmp/trunk/libomptarget/deviceRTLs/nvptx/src/reduction.cu Fri Oct  4 14:39:22 2019
@@ -24,14 +24,14 @@ EXTERN
 void __kmpc_nvptx_end_reduce_nowait(int32_t global_tid) {}
 
 EXTERN int32_t __kmpc_shuffle_int32(int32_t val, int16_t delta, int16_t size) {
-  return __kmpc_impl_shfl_down_sync(0xFFFFFFFF, val, delta, size);
+  return __kmpc_impl_shfl_down_sync(__kmpc_impl_all_lanes, val, delta, size);
 }
 
 EXTERN int64_t __kmpc_shuffle_int64(int64_t val, int16_t delta, int16_t size) {
    uint32_t lo, hi;
    __kmpc_impl_unpack(val, lo, hi);
-   hi = __kmpc_impl_shfl_down_sync(0xFFFFFFFF, hi, delta, size);
-   lo = __kmpc_impl_shfl_down_sync(0xFFFFFFFF, lo, delta, size);
+   hi = __kmpc_impl_shfl_down_sync(__kmpc_impl_all_lanes, hi, delta, size);
+   lo = __kmpc_impl_shfl_down_sync(__kmpc_impl_all_lanes, lo, delta, size);
    return __kmpc_impl_pack(lo, hi);
 }
 
@@ -82,7 +82,7 @@ int32_t __kmpc_nvptx_simd_reduce_nowait(
                                         kmp_ShuffleReductFctPtr shflFct,
                                         kmp_InterWarpCopyFctPtr cpyFct) {
   __kmpc_impl_lanemask_t Liveness = __kmpc_impl_activemask();
-  if (Liveness == 0xffffffff) {
+  if (Liveness == __kmpc_impl_all_lanes) {
     gpu_regular_warp_reduce(reduce_data, shflFct);
     return GetThreadIdInBlock() % WARPSIZE ==
            0; // Result on lane 0 of the simd warp.
@@ -143,7 +143,7 @@ static int32_t nvptx_parallel_reduce_now
   return BlockThreadId == 0;
 #else
   __kmpc_impl_lanemask_t Liveness = __kmpc_impl_activemask();
-  if (Liveness == 0xffffffff) // Full warp
+  if (Liveness == __kmpc_impl_all_lanes) // Full warp
     gpu_regular_warp_reduce(reduce_data, shflFct);
   else if (!(Liveness & (Liveness + 1))) // Partial warp but contiguous lanes
     gpu_irregular_warp_reduce(reduce_data, shflFct,
@@ -318,7 +318,7 @@ static int32_t nvptx_teams_reduce_nowait
 
   // Reduce across warps to the warp master.
   __kmpc_impl_lanemask_t Liveness = __kmpc_impl_activemask();
-  if (Liveness == 0xffffffff) // Full warp
+  if (Liveness == __kmpc_impl_all_lanes) // Full warp
     gpu_regular_warp_reduce(reduce_data, shflFct);
   else // Partial warp but contiguous lanes
     gpu_irregular_warp_reduce(reduce_data, shflFct,

Modified: openmp/trunk/libomptarget/deviceRTLs/nvptx/src/target_impl.h
URL: http://llvm.org/viewvc/llvm-project/openmp/trunk/libomptarget/deviceRTLs/nvptx/src/target_impl.h?rev=373793&r1=373792&r2=373793&view=diff
==============================================================================
--- openmp/trunk/libomptarget/deviceRTLs/nvptx/src/target_impl.h (original)
+++ openmp/trunk/libomptarget/deviceRTLs/nvptx/src/target_impl.h Fri Oct  4 14:39:22 2019
@@ -27,6 +27,8 @@ INLINE uint64_t __kmpc_impl_pack(uint32_
 }
 
 typedef uint32_t __kmpc_impl_lanemask_t;
+static const __kmpc_impl_lanemask_t __kmpc_impl_all_lanes =
+    UINT32_C(0xffffffff);
 
 INLINE __kmpc_impl_lanemask_t __kmpc_impl_lanemask_lt() {
   __kmpc_impl_lanemask_t res;