[PATCH] D120076: [OpenMP] Add RTL function to externalization RAII

[Joseph Huber via Phabricator via llvm-commits]

jhuber6 created this revision.
jhuber6 added reviewers: jdoerfert, tianshilei1992.
Herald added subscribers: ormris, guansong, hiraditya, yaxunl.
jhuber6 requested review of this revision.
Herald added subscribers: llvm-commits, sstefan1.
Herald added a project: LLVM.

This patch adds the '_kmpc_get_hardware_num_threads_in_block'
OpenMP RTL function to the externalization RAII struct. This was getting
optimized out and then being replaced with an undefined value once added
back in, causing bugs for complex reductions.

Fixes #53909.


Repository:
  rG LLVM Github Monorepo

https://reviews.llvm.org/D120076

Files:
  llvm/lib/Transforms/IPO/OpenMPOpt.cpp
  llvm/test/Transforms/OpenMP/get_hardware_num_threads_in_block_fold.ll


Index: llvm/test/Transforms/OpenMP/get_hardware_num_threads_in_block_fold.ll
===================================================================
--- llvm/test/Transforms/OpenMP/get_hardware_num_threads_in_block_fold.ll
+++ llvm/test/Transforms/OpenMP/get_hardware_num_threads_in_block_fold.ll
@@ -178,7 +178,16 @@
   ret void
 }
 
-declare i32 @__kmpc_get_hardware_num_threads_in_block()
+define internal i32 @__kmpc_get_hardware_num_threads_in_block() {
+; CHECK-LABEL: define {{[^@]+}}@__kmpc_get_hardware_num_threads_in_block
+; CHECK-SAME: () #[[ATTR1]] {
+; CHECK-NEXT:    [[RET:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block_dummy()
+; CHECK-NEXT:    ret i32 [[RET]]
+;
+  %ret = call i32 @__kmpc_get_hardware_num_threads_in_block_dummy()
+  ret i32 %ret
+}
+declare i32 @__kmpc_get_hardware_num_threads_in_block_dummy()
 declare i32 @__kmpc_target_init(%struct.ident_t*, i8, i1 zeroext, i1 zeroext) #1
 declare void @__kmpc_target_deinit(%struct.ident_t* nocapture readnone, i8, i1 zeroext) #1
 declare void @__kmpc_parallel_51(%struct.ident_t*, i32, i32, i32, i32, i8*, i8*, i8**, i64)
Index: llvm/lib/Transforms/IPO/OpenMPOpt.cpp
===================================================================
--- llvm/lib/Transforms/IPO/OpenMPOpt.cpp
+++ llvm/lib/Transforms/IPO/OpenMPOpt.cpp
@@ -2119,6 +2119,8 @@
                                        OMPRTL___kmpc_barrier_simple_generic);
     ExternalizationRAII ThreadId(OMPInfoCache,
                                  OMPRTL___kmpc_get_hardware_thread_id_in_block);
+    ExternalizationRAII NumThreads(
+        OMPInfoCache, OMPRTL___kmpc_get_hardware_num_threads_in_block);
     ExternalizationRAII WarpSize(OMPInfoCache, OMPRTL___kmpc_get_warp_size);
 
     registerAAs(IsModulePass);


-------------- next part --------------
A non-text attachment was scrubbed...
Name: D120076.409713.patch
Type: text/x-patch
Size: 1756 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20220217/02280942/attachment.bin>