r338899 - [OpenCL] Always emit alloca in entry block for enqueue_kernel builtin

Mon Aug 6 09:51:24 PDT 2018

Have you tried setting detect_stack_use_after_return in ASAN_OPTIONS? The
ASan buildbot sets the following ASAN_OPTIONS prior to running tests:
export
ASAN_OPTIONS="check_initialization_order=true:detect_stack_use_after_return=1:detect_leaks=1"

On Mon, Aug 6, 2018 at 7:34 AM <scott at scottlinder.com> wrote:

> I can't seem to reproduce the ASan failure locally, even after building
> a clang with the latest compiler-rt, and then rebuilding my patch with
> LLVM_USE_SANITIZER=Address
>
> I am pretty confident the problem should be fixed with a one-line change
> to my patch:
>
> -    auto CreateArrayForSizeVar = [=](unsigned First) {
> +    auto CreateArrayForSizeVar = [=](unsigned First)
> +        -> std::tuple<llvm::Value *, llvm::Value *, llvm::Value *> {
>
> I don't want to commit something and then immediately have to revert,
> though. Can you think of anything I might be missing locally to
> reproduce the ASan failure?
>
> Thanks,
> Scott
>
> On 2018-08-03 13:48, Vlad Tsyrklevich wrote:
> > This change is causing ASan failures on the sanitizer bots:
> >
> http://lab.llvm.org:8011/builders/sanitizer-x86_64-linux-fast/builds/21898/steps/check-clang%20asan/logs/stdio
> > [9]
> >
> > I've reverted it in r338904.
> >
> > On Fri, Aug 3, 2018 at 8:51 AM Scott Linder via cfe-commits
> > <cfe-commits at lists.llvm.org> wrote:
> >
> >> Author: scott.linder
> >> Date: Fri Aug 3 08:50:52 2018
> >> New Revision: 338899
> >>
> >> URL: http://llvm.org/viewvc/llvm-project?rev=338899&view=rev [1]
> >> Log:
> >> [OpenCL] Always emit alloca in entry block for enqueue_kernel
> >> builtin
> >>
> >> Ensures the statically sized alloca is not converted to
> >> DYNAMIC_STACKALLOC
> >> later because it is not in the entry block.
> >>
> >> Differential Revision: https://reviews.llvm.org/D50104 [2]
> >>
> >> Added:
> >> cfe/trunk/test/CodeGenOpenCL/enqueue-kernel-non-entry-block.cl
> >> [3]
> >> Modified:
> >> cfe/trunk/lib/CodeGen/CGBuiltin.cpp
> >> cfe/trunk/test/CodeGenOpenCL/cl20-device-side-enqueue.cl [4]
> >>
> >> Modified: cfe/trunk/lib/CodeGen/CGBuiltin.cpp
> >> URL:
> >>
> >
> http://llvm.org/viewvc/llvm-project/cfe/trunk/lib/CodeGen/CGBuiltin.cpp?rev=338899&r1=338898&r2=338899&view=diff
> >> [5]
> >>
> >
> ==============================================================================
> >> --- cfe/trunk/lib/CodeGen/CGBuiltin.cpp (original)
> >> +++ cfe/trunk/lib/CodeGen/CGBuiltin.cpp Fri Aug 3 08:50:52 2018
> >> @@ -3338,23 +3338,29 @@ RValue CodeGenFunction::EmitBuiltinExpr(
> >> // Create a temporary array to hold the sizes of local pointer
> >> arguments
> >> // for the block. \p First is the position of the first size
> >> argument.
> >> auto CreateArrayForSizeVar = [=](unsigned First) {
> >> - auto *AT = llvm::ArrayType::get(SizeTy, NumArgs - First);
> >> - auto *Arr = Builder.CreateAlloca(AT);
> >> - llvm::Value *Ptr;
> >> + llvm::APInt ArraySize(32, NumArgs - First);
> >> + QualType SizeArrayTy = getContext().getConstantArrayType(
> >> + getContext().getSizeType(), ArraySize, ArrayType::Normal,
> >> + /*IndexTypeQuals=*/0);
> >> + auto Tmp = CreateMemTemp(SizeArrayTy, "block_sizes");
> >> + llvm::Value *TmpPtr = Tmp.getPointer();
> >> + llvm::Value *TmpSize = EmitLifetimeStart(
> >> +
> >> CGM.getDataLayout().getTypeAllocSize(Tmp.getElementType()), TmpPtr);
> >> + llvm::Value *ElemPtr;
> >> // Each of the following arguments specifies the size of the
> >> corresponding
> >> // argument passed to the enqueued block.
> >> auto *Zero = llvm::ConstantInt::get(IntTy, 0);
> >> for (unsigned I = First; I < NumArgs; ++I) {
> >> auto *Index = llvm::ConstantInt::get(IntTy, I - First);
> >> - auto *GEP = Builder.CreateGEP(Arr, {Zero, Index});
> >> + auto *GEP = Builder.CreateGEP(TmpPtr, {Zero, Index});
> >> if (I == First)
> >> - Ptr = GEP;
> >> + ElemPtr = GEP;
> >> auto *V =
> >> Builder.CreateZExtOrTrunc(EmitScalarExpr(E->getArg(I)),
> >> SizeTy);
> >> Builder.CreateAlignedStore(
> >> V, GEP,
> >> CGM.getDataLayout().getPrefTypeAlignment(SizeTy));
> >> }
> >> - return Ptr;
> >> + return std::tie(ElemPtr, TmpSize, TmpPtr);
> >> };
> >>
> >> // Could have events and/or varargs.
> >> @@ -3366,24 +3372,27 @@ RValue CodeGenFunction::EmitBuiltinExpr(
> >> llvm::Value *Kernel =
> >> Builder.CreatePointerCast(Info.Kernel, GenericVoidPtrTy);
> >> auto *Block = Builder.CreatePointerCast(Info.BlockArg,
> >> GenericVoidPtrTy);
> >> - auto *PtrToSizeArray = CreateArrayForSizeVar(4);
> >> + llvm::Value *ElemPtr, *TmpSize, *TmpPtr;
> >> + std::tie(ElemPtr, TmpSize, TmpPtr) =
> >> CreateArrayForSizeVar(4);
> >>
> >> // Create a vector of the arguments, as well as a constant
> >> value to
> >> // express to the runtime the number of variadic arguments.
> >> std::vector<llvm::Value *> Args = {
> >> Queue, Flags, Range,
> >> Kernel, Block, ConstantInt::get(IntTy, NumArgs - 4),
> >> - PtrToSizeArray};
> >> + ElemPtr};
> >> std::vector<llvm::Type *> ArgTys = {
> >> - QueueTy, IntTy, RangeTy,
> >> - GenericVoidPtrTy, GenericVoidPtrTy, IntTy,
> >> - PtrToSizeArray->getType()};
> >> + QueueTy, IntTy, RangeTy,
> >> GenericVoidPtrTy,
> >> + GenericVoidPtrTy, IntTy, ElemPtr->getType()};
> >>
> >> llvm::FunctionType *FTy = llvm::FunctionType::get(
> >> Int32Ty, llvm::ArrayRef<llvm::Type *>(ArgTys), false);
> >> - return RValue::get(
> >> - Builder.CreateCall(CGM.CreateRuntimeFunction(FTy, Name),
> >> - llvm::ArrayRef<llvm::Value *>(Args)));
> >> + auto Call =
> >> +
> >> RValue::get(Builder.CreateCall(CGM.CreateRuntimeFunction(FTy, Name),
> >> + llvm::ArrayRef<llvm::Value
> >> *>(Args)));
> >> + if (TmpSize)
> >> + EmitLifetimeEnd(TmpSize, TmpPtr);
> >> + return Call;
> >> }
> >> // Any calls now have event arguments passed.
> >> if (NumArgs >= 7) {
> >> @@ -3430,15 +3439,19 @@ RValue CodeGenFunction::EmitBuiltinExpr(
> >> ArgTys.push_back(Int32Ty);
> >> Name = "__enqueue_kernel_events_varargs";
> >>
> >> - auto *PtrToSizeArray = CreateArrayForSizeVar(7);
> >> - Args.push_back(PtrToSizeArray);
> >> - ArgTys.push_back(PtrToSizeArray->getType());
> >> + llvm::Value *ElemPtr, *TmpSize, *TmpPtr;
> >> + std::tie(ElemPtr, TmpSize, TmpPtr) =
> >> CreateArrayForSizeVar(7);
> >> + Args.push_back(ElemPtr);
> >> + ArgTys.push_back(ElemPtr->getType());
> >>
> >> llvm::FunctionType *FTy = llvm::FunctionType::get(
> >> Int32Ty, llvm::ArrayRef<llvm::Type *>(ArgTys), false);
> >> - return RValue::get(
> >> - Builder.CreateCall(CGM.CreateRuntimeFunction(FTy, Name),
> >> - llvm::ArrayRef<llvm::Value *>(Args)));
> >> + auto Call =
> >> +
> >> RValue::get(Builder.CreateCall(CGM.CreateRuntimeFunction(FTy, Name),
> >> + llvm::ArrayRef<llvm::Value
> >> *>(Args)));
> >> + if (TmpSize)
> >> + EmitLifetimeEnd(TmpSize, TmpPtr);
> >> + return Call;
> >> }
> >> LLVM_FALLTHROUGH;
> >> }
> >>
> >> Modified: cfe/trunk/test/CodeGenOpenCL/cl20-device-side-enqueue.cl
> >> [4]
> >> URL:
> >>
> >
> http://llvm.org/viewvc/llvm-project/cfe/trunk/test/CodeGenOpenCL/cl20-device-side-enqueue.cl?rev=338899&r1=338898&r2=338899&view=diff
> >> [6]
> >>
> >
> ==============================================================================
> >> --- cfe/trunk/test/CodeGenOpenCL/cl20-device-side-enqueue.cl [4]
> >> (original)
> >> +++ cfe/trunk/test/CodeGenOpenCL/cl20-device-side-enqueue.cl [4] Fri
> >> Aug 3 08:50:52 2018
> >> @@ -1,5 +1,6 @@
> >> // RUN: %clang_cc1 %s -cl-std=CL2.0 -ffake-address-space-map -O0
> >> -emit-llvm -o - -triple "spir-unknown-unknown" | FileCheck %s
> >> --check-prefix=COMMON --check-prefix=B32
> >> // RUN: %clang_cc1 %s -cl-std=CL2.0 -ffake-address-space-map -O0
> >> -emit-llvm -o - -triple "spir64-unknown-unknown" | FileCheck %s
> >> --check-prefix=COMMON --check-prefix=B64
> >> +// RUN: %clang_cc1 %s -cl-std=CL2.0 -ffake-address-space-map -O1
> >> -emit-llvm -o - -triple "spir64-unknown-unknown" | FileCheck %s
> >> --check-prefix=CHECK-LIFETIMES
> >>
> >> #pragma OPENCL EXTENSION cl_khr_subgroups : enable
> >>
> >> @@ -46,8 +47,31 @@ kernel void device_side_enqueue(global i
> >> // COMMON: %event_wait_list2 = alloca [1 x %opencl.clk_event_t*]
> >> clk_event_t event_wait_list2[] = {clk_event};
> >>
> >> - // Emits block literal on stack and block kernel [[INVLK1]].
> >> // COMMON: [[NDR:%[a-z0-9]+]] = alloca %struct.ndrange_t, align 4
> >> +
> >> + // B32: %[[BLOCK_SIZES1:.*]] = alloca [1 x i32]
> >> + // B64: %[[BLOCK_SIZES1:.*]] = alloca [1 x i64]
> >> + // CHECK-LIFETIMES: %[[BLOCK_SIZES1:.*]] = alloca [1 x i64]
> >> + // B32: %[[BLOCK_SIZES2:.*]] = alloca [1 x i32]
> >> + // B64: %[[BLOCK_SIZES2:.*]] = alloca [1 x i64]
> >> + // CHECK-LIFETIMES: %[[BLOCK_SIZES2:.*]] = alloca [1 x i64]
> >> + // B32: %[[BLOCK_SIZES3:.*]] = alloca [1 x i32]
> >> + // B64: %[[BLOCK_SIZES3:.*]] = alloca [1 x i64]
> >> + // CHECK-LIFETIMES: %[[BLOCK_SIZES3:.*]] = alloca [1 x i64]
> >> + // B32: %[[BLOCK_SIZES4:.*]] = alloca [1 x i32]
> >> + // B64: %[[BLOCK_SIZES4:.*]] = alloca [1 x i64]
> >> + // CHECK-LIFETIMES: %[[BLOCK_SIZES4:.*]] = alloca [1 x i64]
> >> + // B32: %[[BLOCK_SIZES5:.*]] = alloca [1 x i32]
> >> + // B64: %[[BLOCK_SIZES5:.*]] = alloca [1 x i64]
> >> + // CHECK-LIFETIMES: %[[BLOCK_SIZES5:.*]] = alloca [1 x i64]
> >> + // B32: %[[BLOCK_SIZES6:.*]] = alloca [3 x i32]
> >> + // B64: %[[BLOCK_SIZES6:.*]] = alloca [3 x i64]
> >> + // CHECK-LIFETIMES: %[[BLOCK_SIZES6:.*]] = alloca [3 x i64]
> >> + // B32: %[[BLOCK_SIZES7:.*]] = alloca [1 x i32]
> >> + // B64: %[[BLOCK_SIZES7:.*]] = alloca [1 x i64]
> >> + // CHECK-LIFETIMES: %[[BLOCK_SIZES7:.*]] = alloca [1 x i64]
> >> +
> >> + // Emits block literal on stack and block kernel [[INVLK1]].
> >> // COMMON: [[DEF_Q:%[0-9]+]] = load %opencl.queue_t{{.*}}*,
> >> %opencl.queue_t{{.*}}** %default_queue
> >> // COMMON: [[FLAGS:%[0-9]+]] = load i32, i32* %flags
> >> // B32: [[BL:%[0-9]+]] = bitcast <{ i32, i32, i32 addrspace(1)*,
> >> i32, i32 addrspace(1)* }>* %block to void ()*
> >> @@ -73,7 +97,6 @@ kernel void device_side_enqueue(global i
> >> // COMMON-SAME: (%opencl.queue_t{{.*}}* [[DEF_Q]], i32 [[FLAGS]],
> >> %struct.ndrange_t* {{.*}}, i32 2, %opencl.clk_event_t{{.*}}*
> >> addrspace(4)* [[WAIT_EVNT]], %opencl.clk_event_t{{.*}}*
> >> addrspace(4)* [[EVNT]],
> >> // COMMON-SAME: i8 addrspace(4)* addrspacecast (i8* bitcast
> >> ({{.*}} [[INVLK2:[^ ]+_kernel]] to i8*) to i8 addrspace(4)*),
> >> // COMMON-SAME: i8 addrspace(4)* [[BL_I8]])
> >> -
> >> enqueue_kernel(default_queue, flags, ndrange, 2,
> >> &event_wait_list, &clk_event,
> >> ^(void) {
> >> a[i] = b[i];
> >> @@ -82,39 +105,46 @@ kernel void device_side_enqueue(global i
> >> // Emits global block literal [[BLG1]] and block kernel
> >> [[INVGK1]].
> >> // COMMON: [[DEF_Q:%[0-9]+]] = load %opencl.queue_t{{.*}}*,
> >> %opencl.queue_t{{.*}}** %default_queue
> >> // COMMON: [[FLAGS:%[0-9]+]] = load i32, i32* %flags
> >> - // B32: %[[TMP:.*]] = alloca [1 x i32]
> >> - // B32: %[[TMP1:.*]] = getelementptr [1 x i32], [1 x i32]*
> >> %[[TMP]], i32 0, i32 0
> >> - // B32: store i32 256, i32* %[[TMP1]], align 4
> >> - // B64: %[[TMP:.*]] = alloca [1 x i64]
> >> - // B64: %[[TMP1:.*]] = getelementptr [1 x i64], [1 x i64]*
> >> %[[TMP]], i32 0, i32 0
> >> - // B64: store i64 256, i64* %[[TMP1]], align 8
> >> + // CHECK-LIFETIMES: [[LIFETIME_PTR:%[0-9]+]] = bitcast [1 x i64]*
> >> %[[BLOCK_SIZES1]] to i8*
> >> + // CHECK-LIFETIMES-NEXT: call void @llvm.lifetime.start.p0i8(i64
> >> 8, i8* nonnull [[LIFETIME_PTR]])
> >> + // CHECK-LIFETIMES-NEXT: getelementptr inbounds [1 x i64], [1 x
> >> i64]* %[[BLOCK_SIZES1]], i64 0, i64 0
> >> + // CHECK-LIFETIMES-LABEL: call i32 @__enqueue_kernel_varargs(
> >> + // CHECK-LIFETIMES-NEXT: call void @llvm.lifetime.end.p0i8(i64 8,
> >> i8* nonnull [[LIFETIME_PTR]])
> >> + // B32: %[[TMP:.*]] = getelementptr [1 x i32], [1 x i32]*
> >> %[[BLOCK_SIZES1]], i32 0, i32 0
> >> + // B32: store i32 256, i32* %[[TMP]], align 4
> >> + // B64: %[[TMP:.*]] = getelementptr [1 x i64], [1 x i64]*
> >> %[[BLOCK_SIZES1]], i32 0, i32 0
> >> + // B64: store i64 256, i64* %[[TMP]], align 8
> >> // COMMON-LABEL: call i32 @__enqueue_kernel_varargs(
> >> // COMMON-SAME: %opencl.queue_t{{.*}}* [[DEF_Q]], i32 [[FLAGS]],
> >> %struct.ndrange_t* [[NDR]]{{([0-9]+)?}},
> >> // COMMON-SAME: i8 addrspace(4)* addrspacecast (i8* bitcast
> >> ({{.*}} [[INVGK1:[^ ]+_kernel]] to i8*) to i8 addrspace(4)*),
> >> // COMMON-SAME: i8 addrspace(4)* addrspacecast (i8 addrspace(1)*
> >> bitcast ({ i32, i32 } addrspace(1)* [[BLG1]] to i8 addrspace(1)*) to
> >> i8 addrspace(4)*), i32 1,
> >> - // B32-SAME: i32* %[[TMP1]])
> >> - // B64-SAME: i64* %[[TMP1]])
> >> + // B32-SAME: i32* %[[TMP]])
> >> + // B64-SAME: i64* %[[TMP]])
> >> enqueue_kernel(default_queue, flags, ndrange,
> >> ^(local void *p) {
> >> return;
> >> },
> >> 256);
> >> +
> >> char c;
> >> // Emits global block literal [[BLG2]] and block kernel
> >> [[INVGK2]].
> >> // COMMON: [[DEF_Q:%[0-9]+]] = load %opencl.queue_t{{.*}}*,
> >> %opencl.queue_t{{.*}}** %default_queue
> >> // COMMON: [[FLAGS:%[0-9]+]] = load i32, i32* %flags
> >> - // B32: %[[TMP:.*]] = alloca [1 x i32]
> >> - // B32: %[[TMP1:.*]] = getelementptr [1 x i32], [1 x i32]*
> >> %[[TMP]], i32 0, i32 0
> >> - // B32: store i32 %{{.*}}, i32* %[[TMP1]], align 4
> >> - // B64: %[[TMP:.*]] = alloca [1 x i64]
> >> - // B64: %[[TMP1:.*]] = getelementptr [1 x i64], [1 x i64]*
> >> %[[TMP]], i32 0, i32 0
> >> - // B64: store i64 %{{.*}}, i64* %[[TMP1]], align 8
> >> + // CHECK-LIFETIMES: [[LIFETIME_PTR:%[0-9]+]] = bitcast [1 x i64]*
> >> %[[BLOCK_SIZES2]] to i8*
> >> + // CHECK-LIFETIMES-NEXT: call void @llvm.lifetime.start.p0i8(i64
> >> 8, i8* nonnull [[LIFETIME_PTR]])
> >> + // CHECK-LIFETIMES-NEXT: getelementptr inbounds [1 x i64], [1 x
> >> i64]* %[[BLOCK_SIZES2]], i64 0, i64 0
> >> + // CHECK-LIFETIMES-LABEL: call i32 @__enqueue_kernel_varargs(
> >> + // CHECK-LIFETIMES-NEXT: call void @llvm.lifetime.end.p0i8(i64 8,
> >> i8* nonnull [[LIFETIME_PTR]])
> >> + // B32: %[[TMP:.*]] = getelementptr [1 x i32], [1 x i32]*
> >> %[[BLOCK_SIZES2]], i32 0, i32 0
> >> + // B32: store i32 %{{.*}}, i32* %[[TMP]], align 4
> >> + // B64: %[[TMP:.*]] = getelementptr [1 x i64], [1 x i64]*
> >> %[[BLOCK_SIZES2]], i32 0, i32 0
> >> + // B64: store i64 %{{.*}}, i64* %[[TMP]], align 8
> >> // COMMON-LABEL: call i32 @__enqueue_kernel_varargs(
> >> // COMMON-SAME: %opencl.queue_t{{.*}}* [[DEF_Q]], i32 [[FLAGS]],
> >> %struct.ndrange_t* [[NDR]]{{([0-9]+)?}},
> >> // COMMON-SAME: i8 addrspace(4)* addrspacecast (i8* bitcast
> >> ({{.*}} [[INVGK2:[^ ]+_kernel]] to i8*) to i8 addrspace(4)*),
> >> // COMMON-SAME: i8 addrspace(4)* addrspacecast (i8 addrspace(1)*
> >> bitcast ({ i32, i32 } addrspace(1)* [[BLG2]] to i8 addrspace(1)*) to
> >> i8 addrspace(4)*), i32 1,
> >> - // B32-SAME: i32* %[[TMP1]])
> >> - // B64-SAME: i64* %[[TMP1]])
> >> + // B32-SAME: i32* %[[TMP]])
> >> + // B64-SAME: i64* %[[TMP]])
> >> enqueue_kernel(default_queue, flags, ndrange,
> >> ^(local void *p) {
> >> return;
> >> @@ -127,18 +157,21 @@ kernel void device_side_enqueue(global i
> >> // COMMON: [[AD:%arraydecay[0-9]*]] = getelementptr inbounds [1 x
> >> %opencl.clk_event_t*], [1 x %opencl.clk_event_t*]*
> >> %event_wait_list2, i32 0, i32 0
> >> // COMMON: [[WAIT_EVNT:%[0-9]+]] = addrspacecast
> >> %opencl.clk_event_t{{.*}}** [[AD]] to %opencl.clk_event_t{{.*}}*
> >> addrspace(4)*
> >> // COMMON: [[EVNT:%[0-9]+]] = addrspacecast
> >> %opencl.clk_event_t{{.*}}** %clk_event to %opencl.clk_event_t{{.*}}*
> >> addrspace(4)*
> >> - // B32: %[[TMP:.*]] = alloca [1 x i32]
> >> - // B32: %[[TMP1:.*]] = getelementptr [1 x i32], [1 x i32]*
> >> %[[TMP]], i32 0, i32 0
> >> - // B32: store i32 256, i32* %[[TMP1]], align 4
> >> - // B64: %[[TMP:.*]] = alloca [1 x i64]
> >> - // B64: %[[TMP1:.*]] = getelementptr [1 x i64], [1 x i64]*
> >> %[[TMP]], i32 0, i32 0
> >> - // B64: store i64 256, i64* %[[TMP1]], align 8
> >> + // CHECK-LIFETIMES: [[LIFETIME_PTR:%[0-9]+]] = bitcast [1 x i64]*
> >> %[[BLOCK_SIZES3]] to i8*
> >> + // CHECK-LIFETIMES-NEXT: call void @llvm.lifetime.start.p0i8(i64
> >> 8, i8* nonnull [[LIFETIME_PTR]])
> >> + // CHECK-LIFETIMES-NEXT: getelementptr inbounds [1 x i64], [1 x
> >> i64]* %[[BLOCK_SIZES3]], i64 0, i64 0
> >> + // CHECK-LIFETIMES-LABEL: call i32
> >> @__enqueue_kernel_events_varargs(
> >> + // CHECK-LIFETIMES-NEXT: call void @llvm.lifetime.end.p0i8(i64 8,
> >> i8* nonnull [[LIFETIME_PTR]])
> >> + // B32: %[[TMP:.*]] = getelementptr [1 x i32], [1 x i32]*
> >> %[[BLOCK_SIZES3]], i32 0, i32 0
> >> + // B32: store i32 256, i32* %[[TMP]], align 4
> >> + // B64: %[[TMP:.*]] = getelementptr [1 x i64], [1 x i64]*
> >> %[[BLOCK_SIZES3]], i32 0, i32 0
> >> + // B64: store i64 256, i64* %[[TMP]], align 8
> >> // COMMON-LABEL: call i32 @__enqueue_kernel_events_varargs
> >> // COMMON-SAME: (%opencl.queue_t{{.*}}* [[DEF_Q]], i32 [[FLAGS]],
> >> %struct.ndrange_t* {{.*}}, i32 2, %opencl.clk_event_t{{.*}}
> >> [[WAIT_EVNT]], %opencl.clk_event_t{{.*}} [[EVNT]],
> >> // COMMON-SAME: i8 addrspace(4)* addrspacecast (i8* bitcast
> >> ({{.*}} [[INVGK3:[^ ]+_kernel]] to i8*) to i8 addrspace(4)*),
> >> // COMMON-SAME: i8 addrspace(4)* addrspacecast (i8 addrspace(1)*
> >> bitcast ({ i32, i32 } addrspace(1)* [[BLG3]] to i8 addrspace(1)*) to
> >> i8 addrspace(4)*), i32 1,
> >> - // B32-SAME: i32* %[[TMP1]])
> >> - // B64-SAME: i64* %[[TMP1]])
> >> + // B32-SAME: i32* %[[TMP]])
> >> + // B64-SAME: i64* %[[TMP]])
> >> enqueue_kernel(default_queue, flags, ndrange, 2,
> >> event_wait_list2, &clk_event,
> >> ^(local void *p) {
> >> return;
> >> @@ -151,18 +184,21 @@ kernel void device_side_enqueue(global i
> >> // COMMON: [[AD:%arraydecay[0-9]*]] = getelementptr inbounds [1 x
> >> %opencl.clk_event_t*], [1 x %opencl.clk_event_t*]*
> >> %event_wait_list2, i32 0, i32 0
> >> // COMMON: [[WAIT_EVNT:%[0-9]+]] = addrspacecast
> >> %opencl.clk_event_t{{.*}}** [[AD]] to %opencl.clk_event_t{{.*}}*
> >> addrspace(4)*
> >> // COMMON: [[EVNT:%[0-9]+]] = addrspacecast
> >> %opencl.clk_event_t{{.*}}** %clk_event to %opencl.clk_event_t{{.*}}*
> >> addrspace(4)*
> >> - // B32: %[[TMP:.*]] = alloca [1 x i32]
> >> - // B32: %[[TMP1:.*]] = getelementptr [1 x i32], [1 x i32]*
> >> %[[TMP]], i32 0, i32 0
> >> - // B32: store i32 %{{.*}}, i32* %[[TMP1]], align 4
> >> - // B64: %[[TMP:.*]] = alloca [1 x i64]
> >> - // B64: %[[TMP1:.*]] = getelementptr [1 x i64], [1 x i64]*
> >> %[[TMP]], i32 0, i32 0
> >> - // B64: store i64 %{{.*}}, i64* %[[TMP1]], align 8
> >> + // CHECK-LIFETIMES: [[LIFETIME_PTR:%[0-9]+]] = bitcast [1 x i64]*
> >> %[[BLOCK_SIZES4]] to i8*
> >> + // CHECK-LIFETIMES-NEXT: call void @llvm.lifetime.start.p0i8(i64
> >> 8, i8* nonnull [[LIFETIME_PTR]])
> >> + // CHECK-LIFETIMES-NEXT: getelementptr inbounds [1 x i64], [1 x
> >> i64]* %[[BLOCK_SIZES4]], i64 0, i64 0
> >> + // CHECK-LIFETIMES-LABEL: call i32
> >> @__enqueue_kernel_events_varargs(
> >> + // CHECK-LIFETIMES-NEXT: call void @llvm.lifetime.end.p0i8(i64 8,
> >> i8* nonnull [[LIFETIME_PTR]])
> >> + // B32: %[[TMP:.*]] = getelementptr [1 x i32], [1 x i32]*
> >> %[[BLOCK_SIZES4]], i32 0, i32 0
> >> + // B32: store i32 %{{.*}}, i32* %[[TMP]], align 4
> >> + // B64: %[[TMP:.*]] = getelementptr [1 x i64], [1 x i64]*
> >> %[[BLOCK_SIZES4]], i32 0, i32 0
> >> + // B64: store i64 %{{.*}}, i64* %[[TMP]], align 8
> >> // COMMON-LABEL: call i32 @__enqueue_kernel_events_varargs
> >> // COMMON-SAME: (%opencl.queue_t{{.*}}* [[DEF_Q]], i32 [[FLAGS]],
> >> %struct.ndrange_t* {{.*}}, i32 2, %opencl.clk_event_t{{.*}}*
> >> addrspace(4)* [[WAIT_EVNT]], %opencl.clk_event_t{{.*}}*
> >> addrspace(4)* [[EVNT]],
> >> // COMMON-SAME: i8 addrspace(4)* addrspacecast (i8* bitcast
> >> ({{.*}} [[INVGK4:[^ ]+_kernel]] to i8*) to i8 addrspace(4)*),
> >> // COMMON-SAME: i8 addrspace(4)* addrspacecast (i8 addrspace(1)*
> >> bitcast ({ i32, i32 } addrspace(1)* [[BLG4]] to i8 addrspace(1)*) to
> >> i8 addrspace(4)*), i32 1,
> >> - // B32-SAME: i32* %[[TMP1]])
> >> - // B64-SAME: i64* %[[TMP1]])
> >> + // B32-SAME: i32* %[[TMP]])
> >> + // B64-SAME: i64* %[[TMP]])
> >> enqueue_kernel(default_queue, flags, ndrange, 2,
> >> event_wait_list2, &clk_event,
> >> ^(local void *p) {
> >> return;
> >> @@ -173,18 +209,21 @@ kernel void device_side_enqueue(global i
> >> // Emits global block literal [[BLG5]] and block kernel
> >> [[INVGK5]].
> >> // COMMON: [[DEF_Q:%[0-9]+]] = load %opencl.queue_t{{.*}}*,
> >> %opencl.queue_t{{.*}}** %default_queue
> >> // COMMON: [[FLAGS:%[0-9]+]] = load i32, i32* %flags
> >> - // B32: %[[TMP:.*]] = alloca [1 x i32]
> >> - // B32: %[[TMP1:.*]] = getelementptr [1 x i32], [1 x i32]*
> >> %[[TMP]], i32 0, i32 0
> >> - // B32: store i32 %{{.*}}, i32* %[[TMP1]], align 4
> >> - // B64: %[[TMP:.*]] = alloca [1 x i64]
> >> - // B64: %[[TMP1:.*]] = getelementptr [1 x i64], [1 x i64]*
> >> %[[TMP]], i32 0, i32 0
> >> - // B64: store i64 %{{.*}}, i64* %[[TMP1]], align 8
> >> + // CHECK-LIFETIMES: [[LIFETIME_PTR:%[0-9]+]] = bitcast [1 x i64]*
> >> %[[BLOCK_SIZES5]] to i8*
> >> + // CHECK-LIFETIMES-NEXT: call void @llvm.lifetime.start.p0i8(i64
> >> 8, i8* nonnull [[LIFETIME_PTR]])
> >> + // CHECK-LIFETIMES-NEXT: getelementptr inbounds [1 x i64], [1 x
> >> i64]* %[[BLOCK_SIZES5]], i64 0, i64 0
> >> + // CHECK-LIFETIMES-LABEL: call i32 @__enqueue_kernel_varargs(
> >> + // CHECK-LIFETIMES-NEXT: call void @llvm.lifetime.end.p0i8(i64 8,
> >> i8* nonnull [[LIFETIME_PTR]])
> >> + // B32: %[[TMP:.*]] = getelementptr [1 x i32], [1 x i32]*
> >> %[[BLOCK_SIZES5]], i32 0, i32 0
> >> + // B32: store i32 %{{.*}}, i32* %[[TMP]], align 4
> >> + // B64: %[[TMP:.*]] = getelementptr [1 x i64], [1 x i64]*
> >> %[[BLOCK_SIZES5]], i32 0, i32 0
> >> + // B64: store i64 %{{.*}}, i64* %[[TMP]], align 8
> >> // COMMON-LABEL: call i32 @__enqueue_kernel_varargs
> >> // COMMON-SAME: (%opencl.queue_t{{.*}}* [[DEF_Q]], i32 [[FLAGS]],
> >> %struct.ndrange_t* [[NDR]]{{([0-9]+)?}},
> >> // COMMON-SAME: i8 addrspace(4)* addrspacecast (i8* bitcast
> >> ({{.*}} [[INVGK5:[^ ]+_kernel]] to i8*) to i8 addrspace(4)*),
> >> // COMMON-SAME: i8 addrspace(4)* addrspacecast (i8 addrspace(1)*
> >> bitcast ({ i32, i32 } addrspace(1)* [[BLG5]] to i8 addrspace(1)*) to
> >> i8 addrspace(4)*), i32 1,
> >> - // B32-SAME: i32* %[[TMP1]])
> >> - // B64-SAME: i64* %[[TMP1]])
> >> + // B32-SAME: i32* %[[TMP]])
> >> + // B64-SAME: i64* %[[TMP]])
> >> enqueue_kernel(default_queue, flags, ndrange,
> >> ^(local void *p) {
> >> return;
> >> @@ -194,26 +233,29 @@ kernel void device_side_enqueue(global i
> >> // Emits global block literal [[BLG6]] and block kernel
> >> [[INVGK6]].
> >> // COMMON: [[DEF_Q:%[0-9]+]] = load %opencl.queue_t{{.*}}*,
> >> %opencl.queue_t{{.*}}** %default_queue
> >> // COMMON: [[FLAGS:%[0-9]+]] = load i32, i32* %flags
> >> - // B32: %[[TMP:.*]] = alloca [3 x i32]
> >> - // B32: %[[TMP1:.*]] = getelementptr [3 x i32], [3 x i32]*
> >> %[[TMP]], i32 0, i32 0
> >> - // B32: store i32 1, i32* %[[TMP1]], align 4
> >> - // B32: %[[TMP2:.*]] = getelementptr [3 x i32], [3 x i32]*
> >> %[[TMP]], i32 0, i32 1
> >> - // B32: store i32 2, i32* %[[TMP2]], align 4
> >> - // B32: %[[TMP3:.*]] = getelementptr [3 x i32], [3 x i32]*
> >> %[[TMP]], i32 0, i32 2
> >> - // B32: store i32 4, i32* %[[TMP3]], align 4
> >> - // B64: %[[TMP:.*]] = alloca [3 x i64]
> >> - // B64: %[[TMP1:.*]] = getelementptr [3 x i64], [3 x i64]*
> >> %[[TMP]], i32 0, i32 0
> >> - // B64: store i64 1, i64* %[[TMP1]], align 8
> >> - // B64: %[[TMP2:.*]] = getelementptr [3 x i64], [3 x i64]*
> >> %[[TMP]], i32 0, i32 1
> >> - // B64: store i64 2, i64* %[[TMP2]], align 8
> >> - // B64: %[[TMP3:.*]] = getelementptr [3 x i64], [3 x i64]*
> >> %[[TMP]], i32 0, i32 2
> >> - // B64: store i64 4, i64* %[[TMP3]], align 8
> >> + // CHECK-LIFETIMES: [[LIFETIME_PTR:%[0-9]+]] = bitcast [3 x i64]*
> >> %[[BLOCK_SIZES6]] to i8*
> >> + // CHECK-LIFETIMES-NEXT: call void @llvm.lifetime.start.p0i8(i64
> >> 24, i8* nonnull [[LIFETIME_PTR]])
> >> + // CHECK-LIFETIMES-NEXT: getelementptr inbounds [3 x i64], [3 x
> >> i64]* %[[BLOCK_SIZES6]], i64 0, i64 0
> >> + // CHECK-LIFETIMES-LABEL: call i32 @__enqueue_kernel_varargs(
> >> + // CHECK-LIFETIMES-NEXT: call void @llvm.lifetime.end.p0i8(i64
> >> 24, i8* nonnull [[LIFETIME_PTR]])
> >> + // B32: %[[TMP:.*]] = getelementptr [3 x i32], [3 x i32]*
> >> %[[BLOCK_SIZES6]], i32 0, i32 0
> >> + // B32: store i32 1, i32* %[[TMP]], align 4
> >> + // B32: %[[BLOCK_SIZES62:.*]] = getelementptr [3 x i32], [3 x
> >> i32]* %[[BLOCK_SIZES6]], i32 0, i32 1
> >> + // B32: store i32 2, i32* %[[BLOCK_SIZES62]], align 4
> >> + // B32: %[[BLOCK_SIZES63:.*]] = getelementptr [3 x i32], [3 x
> >> i32]* %[[BLOCK_SIZES6]], i32 0, i32 2
> >> + // B32: store i32 4, i32* %[[BLOCK_SIZES63]], align 4
> >> + // B64: %[[TMP:.*]] = getelementptr [3 x i64], [3 x i64]*
> >> %[[BLOCK_SIZES6]], i32 0, i32 0
> >> + // B64: store i64 1, i64* %[[TMP]], align 8
> >> + // B64: %[[BLOCK_SIZES62:.*]] = getelementptr [3 x i64], [3 x
> >> i64]* %[[BLOCK_SIZES6]], i32 0, i32 1
> >> + // B64: store i64 2, i64* %[[BLOCK_SIZES62]], align 8
> >> + // B64: %[[BLOCK_SIZES63:.*]] = getelementptr [3 x i64], [3 x
> >> i64]* %[[BLOCK_SIZES6]], i32 0, i32 2
> >> + // B64: store i64 4, i64* %[[BLOCK_SIZES63]], align 8
> >> // COMMON-LABEL: call i32 @__enqueue_kernel_varargs
> >> // COMMON-SAME: (%opencl.queue_t{{.*}}* [[DEF_Q]], i32 [[FLAGS]],
> >> %struct.ndrange_t* [[NDR]]{{([0-9]+)?}},
> >> // COMMON-SAME: i8 addrspace(4)* addrspacecast (i8* bitcast
> >> ({{.*}} [[INVGK6:[^ ]+_kernel]] to i8*) to i8 addrspace(4)*),
> >> // COMMON-SAME: i8 addrspace(4)* addrspacecast (i8 addrspace(1)*
> >> bitcast ({ i32, i32 } addrspace(1)* [[BLG6]] to i8 addrspace(1)*) to
> >> i8 addrspace(4)*), i32 3,
> >> - // B32-SAME: i32* %[[TMP1]])
> >> - // B64-SAME: i64* %[[TMP1]])
> >> + // B32-SAME: i32* %[[TMP]])
> >> + // B64-SAME: i64* %[[TMP]])
> >> enqueue_kernel(default_queue, flags, ndrange,
> >> ^(local void *p1, local void *p2, local void *p3)
> >> {
> >> return;
> >> @@ -223,18 +265,21 @@ kernel void device_side_enqueue(global i
> >> // Emits global block literal [[BLG7]] and block kernel
> >> [[INVGK7]].
> >> // COMMON: [[DEF_Q:%[0-9]+]] = load %opencl.queue_t*,
> >> %opencl.queue_t** %default_queue
> >> // COMMON: [[FLAGS:%[0-9]+]] = load i32, i32* %flags
> >> - // B32: %[[TMP:.*]] = alloca [1 x i32]
> >> - // B32: %[[TMP1:.*]] = getelementptr [1 x i32], [1 x i32]*
> >> %[[TMP]], i32 0, i32 0
> >> - // B32: store i32 0, i32* %[[TMP1]], align 4
> >> - // B64: %[[TMP:.*]] = alloca [1 x i64]
> >> - // B64: %[[TMP1:.*]] = getelementptr [1 x i64], [1 x i64]*
> >> %[[TMP]], i32 0, i32 0
> >> - // B64: store i64 4294967296, i64* %[[TMP1]], align 8
> >> + // CHECK-LIFETIMES: [[LIFETIME_PTR:%[0-9]+]] = bitcast [1 x i64]*
> >> %[[BLOCK_SIZES7]] to i8*
> >> + // CHECK-LIFETIMES-NEXT: call void @llvm.lifetime.start.p0i8(i64
> >> 8, i8* nonnull [[LIFETIME_PTR]])
> >> + // CHECK-LIFETIMES-NEXT: getelementptr inbounds [1 x i64], [1 x
> >> i64]* %[[BLOCK_SIZES7]], i64 0, i64 0
> >> + // CHECK-LIFETIMES-LABEL: call i32 @__enqueue_kernel_varargs(
> >> + // CHECK-LIFETIMES-NEXT: call void @llvm.lifetime.end.p0i8(i64 8,
> >> i8* nonnull [[LIFETIME_PTR]])
> >> + // B32: %[[TMP:.*]] = getelementptr [1 x i32], [1 x i32]*
> >> %[[BLOCK_SIZES7]], i32 0, i32 0
> >> + // B32: store i32 0, i32* %[[TMP]], align 4
> >> + // B64: %[[TMP:.*]] = getelementptr [1 x i64], [1 x i64]*
> >> %[[BLOCK_SIZES7]], i32 0, i32 0
> >> + // B64: store i64 4294967296, i64* %[[TMP]], align 8
> >> // COMMON-LABEL: call i32 @__enqueue_kernel_varargs
> >> // COMMON-SAME: (%opencl.queue_t{{.*}}* [[DEF_Q]], i32 [[FLAGS]],
> >> %struct.ndrange_t* [[NDR]]{{([0-9]+)?}},
> >> // COMMON-SAME: i8 addrspace(4)* addrspacecast (i8* bitcast
> >> ({{.*}} [[INVGK7:[^ ]+_kernel]] to i8*) to i8 addrspace(4)*),
> >> // COMMON-SAME: i8 addrspace(4)* addrspacecast (i8 addrspace(1)*
> >> bitcast ({ i32, i32 } addrspace(1)* [[BLG7]] to i8 addrspace(1)*) to
> >> i8 addrspace(4)*), i32 1,
> >> - // B32-SAME: i32* %[[TMP1]])
> >> - // B64-SAME: i64* %[[TMP1]])
> >> + // B32-SAME: i32* %[[TMP]])
> >> + // B64-SAME: i64* %[[TMP]])
> >> enqueue_kernel(default_queue, flags, ndrange,
> >> ^(local void *p) {
> >> return;
> >>
> >> Added:
> >> cfe/trunk/test/CodeGenOpenCL/enqueue-kernel-non-entry-block.cl [3]
> >> URL:
> >>
> >
> http://llvm.org/viewvc/llvm-project/cfe/trunk/test/CodeGenOpenCL/enqueue-kernel-non-entry-block.cl?rev=338899&view=auto
> >> [7]
> >>
> >
> ==============================================================================
> >> --- cfe/trunk/test/CodeGenOpenCL/enqueue-kernel-non-entry-block.cl
> >> [3] (added)
> >> +++ cfe/trunk/test/CodeGenOpenCL/enqueue-kernel-non-entry-block.cl
> >> [3] Fri Aug 3 08:50:52 2018
> >> @@ -0,0 +1,31 @@
> >> +// RUN: %clang_cc1 -cl-std=CL2.0 -O0 -emit-llvm -o - -triple amdgcn
> >> < %s | FileCheck %s --check-prefixes=COMMON,AMDGPU
> >> +// RUN: %clang_cc1 -cl-std=CL2.0 -O0 -emit-llvm -o - -triple
> >> "spir-unknown-unknown" < %s | FileCheck %s
> >> --check-prefixes=COMMON,SPIR32
> >> +// RUN: %clang_cc1 -cl-std=CL2.0 -O0 -emit-llvm -o - -triple
> >> "spir64-unknown-unknown" < %s | FileCheck %s
> >> --check-prefixes=COMMON,SPIR64
> >> +// RUN: %clang_cc1 -cl-std=CL2.0 -O0 -debug-info-kind=limited
> >> -emit-llvm -o - -triple amdgcn < %s | FileCheck %s
> >> --check-prefixes=CHECK-DEBUG
> >> +
> >> +// Check that the enqueue_kernel array temporary is in the entry
> >> block to avoid
> >> +// a dynamic alloca
> >> +
> >> +typedef struct {int a;} ndrange_t;
> >> +
> >> +kernel void test(int i) {
> >> +// COMMON-LABEL: define {{.*}} void @test
> >> +// COMMON-LABEL: entry:
> >> +// AMDGPU: %block_sizes = alloca [1 x i64]
> >> +// SPIR32: %block_sizes = alloca [1 x i32]
> >> +// SPIR64: %block_sizes = alloca [1 x i64]
> >> +// COMMON-LABEL: if.then:
> >> +// COMMON-NOT: alloca
> >> +// CHECK-DEBUG: getelementptr {{.*}} %block_sizes, {{.*}} !dbg !34
> >> +// COMMON-LABEL: if.end
> >> + queue_t default_queue;
> >> + unsigned flags = 0;
> >> + ndrange_t ndrange;
> >> + if (i)
> >> + enqueue_kernel(default_queue, flags, ndrange, ^(local void *a)
> >> { }, 32);
> >> +}
> >> +
> >> +// Check that the temporary is scoped to the `if`
> >> +
> >> +// CHECK-DEBUG: !32 = distinct !DILexicalBlock(scope: !7, file: !1,
> >> line: 24)
> >> +// CHECK-DEBUG: !34 = !DILocation(line: 25, scope: !32)
> >>
> >> _______________________________________________
> >> cfe-commits mailing list
> >> cfe-commits at lists.llvm.org
> >> http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits [8]
> >
> >
> > Links:
> > ------
> > [1] http://llvm.org/viewvc/llvm-project?rev=338899&view=rev
> > [2] https://reviews.llvm.org/D50104
> > [3] http://enqueue-kernel-non-entry-block.cl
> > [4] http://cl20-device-side-enqueue.cl
> > [5]
> >
> http://llvm.org/viewvc/llvm-project/cfe/trunk/lib/CodeGen/CGBuiltin.cpp?rev=338899&r1=338898&r2=338899&view=diff
> > [6]
> >
> http://llvm.org/viewvc/llvm-project/cfe/trunk/test/CodeGenOpenCL/cl20-device-side-enqueue.cl?rev=338899&r1=338898&r2=338899&view=diff
> > [7]
> >
> http://llvm.org/viewvc/llvm-project/cfe/trunk/test/CodeGenOpenCL/enqueue-kernel-non-entry-block.cl?rev=338899&view=auto
> > [8] http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
> > [9]
> >
> http://lab.llvm.org:8011/builders/sanitizer-x86_64-linux-fast/builds/21898/steps/check-clang%20asan/logs/stdio
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-commits/attachments/20180806/6f5ab095/attachment-0001.html>