Skip to content
Snippets Groups Projects
Commit e9894714 authored by Abdul Rafae Noor's avatar Abdul Rafae Noor
Browse files

FuseHPVMTensorNodes: Bugfix for multiple launches of the same DFG

parent 0bba65cc
No related branches found
No related tags found
No related merge requests found
Hide whitespace changes
Inline Side-by-side
hpvm/lib/Transforms/FuseHPVMTensorNodes/FuseHPVMTensorNodes.cpp
+ 11
− 0
View file @ e9894714
......@@ -1001,13 +1001,24 @@ bool FuseHPVMTensorNodesWrapper::runOnModule(Module &M) {
FindFusionTargetsTraversal *FTTVisitor =
new FindFusionTargetsTraversal(M, DFG);
// Visit each DFG only once
std::set<Function*> Visited;
errs() << "Find targets\n";
// Iterate over all the DFGs and produce code for each one of them
for (auto rootNode: Roots) {
Function* rootFunc = rootNode->getFuncPointer();
if(Visited.find(rootFunc) != Visited.end()) continue;
// Initiate code generation for root DFNode
FTTVisitor->visit(rootNode);
Visited.insert(rootFunc);
}
errs()<<"Finished visiting DFGs ...\n";
FuseHPVMTensorNodes::FusionTargets &FTs = FTTVisitor->getFusionTargets();
FuseHPVMTensorNodes Fuse;
......
hpvm/test/regressionTests/FuseHPVMTensorNodes/multiLaunch.hpvm.ll 0 → 100644
+ 184
− 0
View file @ e9894714
; RUN: opt -load LLVMBuildDFG.so -load LLVMInPlaceDFGAnalysis.so -load LLVMFuseHPVMTensorNodes.so -S -inplace -hpvm-fuse < %s | FileCheck %s
; ModuleID = 'multiLaunch.ll'
source_filename = "multiLaunch.cpp"
target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
target triple = "x86_64-unknown-linux-gnu"
%struct.out._Z9conv_nodePvmS_m = type <{ i8*, i64 }>
%struct.out._Z13bias_add_nodePvmS_m = type <{ i8*, i64 }>
%struct.out._Z9tanh_nodePvm = type <{ i8*, i64 }>
%struct.out._Z13pool_max_nodePvm = type <{ i8*, i64 }>
%struct.out._Z4rootPvmS_mS_m = type <{ i8*, i64 }>
; CHECK-LABEL: @_Z4rootPvmS_mS_m_cloned(
; CHECK: call i8* @llvm.hpvm.createNode(
; CHECK-NEXT: call void @llvm.hpvm.bind.input(
; CHECK-NEXT: call void @llvm.hpvm.bind.input(
; CHECK-NEXT: call void @llvm.hpvm.bind.input(
; CHECK-NEXT: call void @llvm.hpvm.bind.input(
; CHECK-NEXT: call void @llvm.hpvm.bind.input(
; CHECK-NEXT: call void @llvm.hpvm.bind.input(
; CHECK: call void @llvm.hpvm.bind.output(
; CHECK: call void @llvm.hpvm.bind.output(
; CHECK: ret
; CHECK-LABEL: @_Z9conv_nodePvmS_m_cloned__Z13bias_add_nodePvmS_m_cloned__Z9tanh_nodePvm_cloned__Z13pool_max_nodePvm_cloned(
; CHECK: call i8* @llvm.hpvm.tensor.convolution(
; CHECK-NEXT: call i8* @llvm.hpvm.tensor.add(
; CHECK-NEXT: call i8* @llvm.hpvm.tensor.tanh(
; CHECK-NEXT: call i8* @llvm.hpvm.tensor.pool.max(
; CHECK: ret
; Function Attrs: norecurse uwtable
define dso_local i32 @main() local_unnamed_addr #0 {
entry:
call void @llvm.hpvm.init()
%call = tail call noalias i8* @malloc(i64 64) #3
%call1 = tail call noalias i8* @malloc(i64 64) #3
%graphID = call i8* @llvm.hpvm.launch(i8* bitcast (%struct.out._Z4rootPvmS_mS_m (i8*, i64, i8*, i64, i8*, i64)* @_Z4rootPvmS_mS_m_cloned to i8*), i8* %call, i1 false)
call void @llvm.hpvm.wait(i8* %graphID)
%input = bitcast i8* %call to i8**
%0 = load i8*, i8** %input, align 1, !tbaa !7
tail call void @hpvm_request_tensor(i8* %0, i32 1)
%graphID1 = call i8* @llvm.hpvm.launch(i8* bitcast (%struct.out._Z4rootPvmS_mS_m (i8*, i64, i8*, i64, i8*, i64)* @_Z4rootPvmS_mS_m_cloned to i8*), i8* %call1, i1 false)
call void @llvm.hpvm.wait(i8* %graphID1)
%input4 = bitcast i8* %call1 to i8**
%1 = load i8*, i8** %input4, align 1, !tbaa !7
tail call void @hpvm_request_tensor(i8* %1, i32 1)
call void @llvm.hpvm.cleanup()
ret i32 0
}
; Function Attrs: nofree nounwind
declare dso_local noalias i8* @malloc(i64) local_unnamed_addr #1
declare dso_local void @hpvm_request_tensor(i8*, i32) local_unnamed_addr #2
; Function Attrs: nounwind
declare i8* @llvm.hpvm.tensor.convolution(i8*, i8*, i32, i32, i32, i32) #3
; Function Attrs: uwtable
define dso_local %struct.out._Z9conv_nodePvmS_m @_Z9conv_nodePvmS_m_cloned(i8* in %t1, i64 %bytes_t1, i8* in %t2, i64 %bytes_t2) #4 {
entry:
%call1 = call i8* @llvm.hpvm.tensor.convolution(i8* %t1, i8* %t2, i32 2, i32 2, i32 1, i32 1)
%returnStruct = insertvalue %struct.out._Z9conv_nodePvmS_m undef, i8* %call1, 0
%returnStruct2 = insertvalue %struct.out._Z9conv_nodePvmS_m %returnStruct, i64 0, 1
ret %struct.out._Z9conv_nodePvmS_m %returnStruct2
}
; Function Attrs: nounwind
declare i8* @llvm.hpvm.tensor.add(i8*, i8*) #3
; Function Attrs: uwtable
define dso_local %struct.out._Z13bias_add_nodePvmS_m @_Z13bias_add_nodePvmS_m_cloned(i8* in %t1, i64 %bytes_t1, i8* in %t2, i64 %bytes_t2) #4 {
entry:
%call1 = call i8* @llvm.hpvm.tensor.add(i8* %t1, i8* %t2)
%returnStruct = insertvalue %struct.out._Z13bias_add_nodePvmS_m undef, i8* %call1, 0
%returnStruct2 = insertvalue %struct.out._Z13bias_add_nodePvmS_m %returnStruct, i64 0, 1
ret %struct.out._Z13bias_add_nodePvmS_m %returnStruct2
}
; Function Attrs: nounwind
declare i8* @llvm.hpvm.tensor.tanh(i8*) #3
; Function Attrs: uwtable
define dso_local %struct.out._Z9tanh_nodePvm @_Z9tanh_nodePvm_cloned(i8* in %t1, i64 %bytes_t1) #4 {
entry:
%call1 = call i8* @llvm.hpvm.tensor.tanh(i8* %t1)
%returnStruct = insertvalue %struct.out._Z9tanh_nodePvm undef, i8* %call1, 0
%returnStruct2 = insertvalue %struct.out._Z9tanh_nodePvm %returnStruct, i64 0, 1
ret %struct.out._Z9tanh_nodePvm %returnStruct2
}
; Function Attrs: nounwind
declare i8* @llvm.hpvm.tensor.pool.max(i8*, i32, i32, i32, i32, i32, i32) #3
; Function Attrs: uwtable
define dso_local %struct.out._Z13pool_max_nodePvm @_Z13pool_max_nodePvm_cloned(i8* in %t1, i64 %bytes_t1) #4 {
entry:
%call1 = call i8* @llvm.hpvm.tensor.pool.max(i8* %t1, i32 2, i32 2, i32 0, i32 0, i32 2, i32 2)
%returnStruct = insertvalue %struct.out._Z13pool_max_nodePvm undef, i8* %call1, 0
%returnStruct2 = insertvalue %struct.out._Z13pool_max_nodePvm %returnStruct, i64 0, 1
ret %struct.out._Z13pool_max_nodePvm %returnStruct2
}
; Function Attrs: nounwind
declare i8* @llvm.hpvm.createNode(i8*) #3
; Function Attrs: nounwind
declare void @llvm.hpvm.bind.input(i8*, i32, i32, i1) #3
; Function Attrs: nounwind
declare i8* @llvm.hpvm.createEdge(i8*, i8*, i1, i32, i32, i1) #3
; Function Attrs: nounwind
declare void @llvm.hpvm.bind.output(i8*, i32, i32, i1) #3
; Function Attrs: uwtable
define dso_local %struct.out._Z4rootPvmS_mS_m @_Z4rootPvmS_mS_m_cloned(i8* in %input, i64 %input_bytes, i8* in %conv2d_1_w, i64 %conv2d_1_w_bytes, i8* in %conv2d_1_b, i64 %conv2d_1_b_bytes) #4 {
entry:
%_Z9conv_nodePvmS_m_cloned.node = call i8* @llvm.hpvm.createNode(i8* bitcast (%struct.out._Z9conv_nodePvmS_m (i8*, i64, i8*, i64)* @_Z9conv_nodePvmS_m_cloned to i8*))
call void @llvm.hpvm.bind.input(i8* %_Z9conv_nodePvmS_m_cloned.node, i32 0, i32 0, i1 false)
call void @llvm.hpvm.bind.input(i8* %_Z9conv_nodePvmS_m_cloned.node, i32 1, i32 1, i1 false)
call void @llvm.hpvm.bind.input(i8* %_Z9conv_nodePvmS_m_cloned.node, i32 2, i32 2, i1 false)
call void @llvm.hpvm.bind.input(i8* %_Z9conv_nodePvmS_m_cloned.node, i32 3, i32 3, i1 false)
%_Z13bias_add_nodePvmS_m_cloned.node = call i8* @llvm.hpvm.createNode(i8* bitcast (%struct.out._Z13bias_add_nodePvmS_m (i8*, i64, i8*, i64)* @_Z13bias_add_nodePvmS_m_cloned to i8*))
%output = call i8* @llvm.hpvm.createEdge(i8* %_Z9conv_nodePvmS_m_cloned.node, i8* %_Z13bias_add_nodePvmS_m_cloned.node, i1 true, i32 0, i32 0, i1 false)
%output1 = call i8* @llvm.hpvm.createEdge(i8* %_Z9conv_nodePvmS_m_cloned.node, i8* %_Z13bias_add_nodePvmS_m_cloned.node, i1 true, i32 1, i32 1, i1 false)
call void @llvm.hpvm.bind.input(i8* %_Z13bias_add_nodePvmS_m_cloned.node, i32 4, i32 2, i1 false)
call void @llvm.hpvm.bind.input(i8* %_Z13bias_add_nodePvmS_m_cloned.node, i32 5, i32 3, i1 false)
%_Z9tanh_nodePvm_cloned.node = call i8* @llvm.hpvm.createNode(i8* bitcast (%struct.out._Z9tanh_nodePvm (i8*, i64)* @_Z9tanh_nodePvm_cloned to i8*))
%output2 = call i8* @llvm.hpvm.createEdge(i8* %_Z13bias_add_nodePvmS_m_cloned.node, i8* %_Z9tanh_nodePvm_cloned.node, i1 true, i32 0, i32 0, i1 false)
%output3 = call i8* @llvm.hpvm.createEdge(i8* %_Z13bias_add_nodePvmS_m_cloned.node, i8* %_Z9tanh_nodePvm_cloned.node, i1 true, i32 1, i32 1, i1 false)
%_Z13pool_max_nodePvm_cloned.node = call i8* @llvm.hpvm.createNode(i8* bitcast (%struct.out._Z13pool_max_nodePvm (i8*, i64)* @_Z13pool_max_nodePvm_cloned to i8*))
%output4 = call i8* @llvm.hpvm.createEdge(i8* %_Z9tanh_nodePvm_cloned.node, i8* %_Z13pool_max_nodePvm_cloned.node, i1 true, i32 0, i32 0, i1 false)
%output5 = call i8* @llvm.hpvm.createEdge(i8* %_Z9tanh_nodePvm_cloned.node, i8* %_Z13pool_max_nodePvm_cloned.node, i1 true, i32 1, i32 1, i1 false)
call void @llvm.hpvm.bind.output(i8* %_Z13pool_max_nodePvm_cloned.node, i32 0, i32 0, i1 false)
call void @llvm.hpvm.bind.output(i8* %_Z13pool_max_nodePvm_cloned.node, i32 1, i32 1, i1 false)
ret %struct.out._Z4rootPvmS_mS_m undef
}
; Function Attrs: nounwind
declare void @llvm.hpvm.init() #3
; Function Attrs: nounwind
declare i8* @llvm.hpvm.launch(i8*, i8*, i1) #3
; Function Attrs: nounwind
declare void @llvm.hpvm.wait(i8*) #3
; Function Attrs: nounwind
declare void @llvm.hpvm.cleanup() #3
attributes #0 = { norecurse uwtable "correctly-rounded-divide-sqrt-fp-math"="false" "disable-tail-calls"="false" "less-precise-fpmad"="false" "min-legal-vector-width"="0" "no-frame-pointer-elim"="false" "no-infs-fp-math"="false" "no-jump-tables"="false" "no-nans-fp-math"="false" "no-signed-zeros-fp-math"="false" "no-trapping-math"="false" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #1 = { nofree nounwind "correctly-rounded-divide-sqrt-fp-math"="false" "disable-tail-calls"="false" "less-precise-fpmad"="false" "no-frame-pointer-elim"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "no-signed-zeros-fp-math"="false" "no-trapping-math"="false" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #2 = { "correctly-rounded-divide-sqrt-fp-math"="false" "disable-tail-calls"="false" "less-precise-fpmad"="false" "no-frame-pointer-elim"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "no-signed-zeros-fp-math"="false" "no-trapping-math"="false" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #3 = { nounwind }
attributes #4 = { uwtable "correctly-rounded-divide-sqrt-fp-math"="false" "disable-tail-calls"="false" "less-precise-fpmad"="false" "min-legal-vector-width"="0" "no-frame-pointer-elim"="false" "no-infs-fp-math"="false" "no-jump-tables"="false" "no-nans-fp-math"="false" "no-signed-zeros-fp-math"="false" "no-trapping-math"="false" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "unsafe-fp-math"="false" "use-soft-float"="false" }
!llvm.module.flags = !{!0}
!llvm.ident = !{!1}
!hpvm_hint_promise = !{!2, !3, !4, !5, !6}
!hpvm_hint_gpu = !{}
!hpvm_hint_cpu = !{}
!hpvm_hint_cpu_gpu = !{}
!hpvm_hint_cudnn = !{}
!0 = !{i32 1, !"wchar_size", i32 4}
!1 = !{!"clang version 9.0.0 (https://gitlab.engr.illinois.edu/llvm/hpvm.git 0bba65ccd19f0ff92a84eaadbc0450fa7a0f0ccc)"}
!2 = !{%struct.out._Z9conv_nodePvmS_m (i8*, i64, i8*, i64)* @_Z9conv_nodePvmS_m_cloned}
!3 = !{%struct.out._Z13bias_add_nodePvmS_m (i8*, i64, i8*, i64)* @_Z13bias_add_nodePvmS_m_cloned}
!4 = !{%struct.out._Z9tanh_nodePvm (i8*, i64)* @_Z9tanh_nodePvm_cloned}
!5 = !{%struct.out._Z13pool_max_nodePvm (i8*, i64)* @_Z13pool_max_nodePvm_cloned}
!6 = !{%struct.out._Z4rootPvmS_mS_m (i8*, i64, i8*, i64, i8*, i64)* @_Z4rootPvmS_mS_m_cloned}
!7 = !{!8, !9, i64 0}
!8 = !{!"_ZTS6RootIn", !9, i64 0, !12, i64 8, !9, i64 16, !12, i64 24, !9, i64 32, !12, i64 40, !13, i64 48}
!9 = !{!"any pointer", !10, i64 0}
!10 = !{!"omnipotent char", !11, i64 0}
!11 = !{!"Simple C++ TBAA"}
!12 = !{!"long", !10, i64 0}
!13 = !{!"_ZTS5ret_t", !9, i64 0, !12, i64 8}
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment