Codegen: Selectively copy in array addresses for OpenMP code

The detection of values that need to be copied in to the generated OpenMP
subfunction also detects the array base addresses needed in the SCoP. Hence, it
is not necessary to unconditionally copy all the base addresses to the generated
function.

Test cases are modified to reflect this change. Arrays which are global
variables do not occur in the struct passed to the subfunction anymore. A test
case for base address copy-in is added in copy_in_array.{c,ll}.

Committed with slight modifications

Contributed by:  Armin Groesslinger <armin.groesslinger@uni-passau.de>

llvm-svn: 167215
This commit is contained in:
Tobias Grosser 2012-11-01 05:34:55 +00:00
parent 177982c478
commit ebe8c8cea2
5 changed files with 7 additions and 19 deletions

View File

@ -504,16 +504,6 @@ SetVector<Value*> ClastStmtCodeGen::getOMPValues(const clast_stmt *Body) {
I != E; I++)
Values.insert(I->second);
// The memory reference base addresses
for (Scop::iterator SI = S->begin(), SE = S->end(); SI != SE; ++SI) {
ScopStmt *Stmt = *SI;
for (SmallVector<MemoryAccess*, 8>::iterator I = Stmt->memacc_begin(),
E = Stmt->memacc_end(); I != E; ++I) {
Value *BaseAddr = const_cast<Value*>((*I)->getBaseAddr());
Values.insert((BaseAddr));
}
}
// Find the temporaries that are referenced in the clast statements'
// basic blocks but are not defined by these blocks (e.g., references
// to function arguments or temporaries defined before the start of

View File

@ -31,4 +31,4 @@ for.end: ; preds = %for.body
ret void
}
; CHECK: %omp.userContext = alloca { [100 x float]*, float }
; CHECK: %omp.userContext = alloca { float }

View File

@ -32,4 +32,4 @@ for.end: ; preds = %for.body
ret void
}
; CHECK: %omp.userContext = alloca { [100 x float]*, float }
; CHECK: %omp.userContext = alloca { float }

View File

@ -55,12 +55,10 @@ entry:
declare void @llvm.memset.p0i8.i32(i8* nocapture, i8, i32, i32, i1) nounwind
; CHECK: %omp.userContext = alloca { i32, [10 x double]* }
; CHECK: getelementptr inbounds { i32, [10 x double]* }* %omp.userContext, i32 0, i32 0
; CHECK: %omp.userContext = alloca { i32 }
; CHECK: getelementptr inbounds { i32 }* %omp.userContext, i32 0, i32 0
; CHECK: store i32 %polly.loopiv, i32* %1
; CHECK: getelementptr inbounds { i32, [10 x double]* }* %omp.userContext, i32 0, i32 1
; CHECK: store [10 x double]* @A, [10 x double]** %2
; CHECK: %omp_data = bitcast { i32, [10 x double]* }* %omp.userContext to i8*
; CHECK: %omp_data = bitcast { i32 }* %omp.userContext to i8*
; CHECK: call void @GOMP_parallel_loop_runtime_start(void (i8*)* @loop_openmp.omp_subfn, i8* %omp_data, i32 0, i32 0, i32 10, i32 1)
; CHECK: call void @loop_openmp.omp_subfn(i8* %omp_data)
; CHECK: call void @GOMP_parallel_end()

View File

@ -72,6 +72,6 @@ entry:
declare void @llvm.memset.p0i8.i32(i8* nocapture, i8, i32, i32, i1) nounwind
; CHECK: %omp.userContext = alloca { [5 x float]*, [5 x float]* }
; CHECK: %omp.userContext1 = alloca { i32, [5 x float]*, [5 x float]* }
; CHECK: %omp.userContext = alloca {}
; CHECK: %omp.userContext1 = alloca { i32 }