diff options
| -rw-r--r-- | lib/Target/R600/AMDGPUISelLowering.cpp | 14 | ||||
| -rw-r--r-- | lib/Target/R600/AMDGPUMachineFunction.h | 4 | ||||
| -rw-r--r-- | test/CodeGen/R600/lds-size.ll | 26 |
3 files changed, 40 insertions, 4 deletions
diff --git a/lib/Target/R600/AMDGPUISelLowering.cpp b/lib/Target/R600/AMDGPUISelLowering.cpp index 1237323..d6b7cbe 100644 --- a/lib/Target/R600/AMDGPUISelLowering.cpp +++ b/lib/Target/R600/AMDGPUISelLowering.cpp @@ -246,12 +246,18 @@ SDValue AMDGPUTargetLowering::LowerGlobalAddress(AMDGPUMachineFunction* MFI, assert(G->getOffset() == 0 && "Do not know what to do with an non-zero offset"); - unsigned Offset = MFI->LDSSize; const GlobalValue *GV = G->getGlobal(); - uint64_t Size = TD->getTypeAllocSize(GV->getType()->getElementType()); - // XXX: Account for alignment? - MFI->LDSSize += Size; + unsigned Offset; + if (MFI->LocalMemoryObjects.count(GV) == 0) { + uint64_t Size = TD->getTypeAllocSize(GV->getType()->getElementType()); + Offset = MFI->LDSSize; + MFI->LocalMemoryObjects[GV] = Offset; + // XXX: Account for alignment? + MFI->LDSSize += Size; + } else { + Offset = MFI->LocalMemoryObjects[GV]; + } return DAG.getConstant(Offset, getPointerTy(G->getAddressSpace())); } diff --git a/lib/Target/R600/AMDGPUMachineFunction.h b/lib/Target/R600/AMDGPUMachineFunction.h index 789b96a..fe80ce3 100644 --- a/lib/Target/R600/AMDGPUMachineFunction.h +++ b/lib/Target/R600/AMDGPUMachineFunction.h @@ -14,6 +14,7 @@ #define AMDGPUMACHINEFUNCTION_H #include "llvm/CodeGen/MachineFunction.h" +#include <map> namespace llvm { @@ -21,6 +22,9 @@ class AMDGPUMachineFunction : public MachineFunctionInfo { public: AMDGPUMachineFunction(const MachineFunction &MF); unsigned ShaderType; + /// A map to keep track of local memory objects and their offsets within + /// the local memory space. + std::map<const GlobalValue *, unsigned> LocalMemoryObjects; /// Number of bytes in the LDS that are being used. unsigned LDSSize; }; diff --git a/test/CodeGen/R600/lds-size.ll b/test/CodeGen/R600/lds-size.ll new file mode 100644 index 0000000..2185180 --- /dev/null +++ b/test/CodeGen/R600/lds-size.ll @@ -0,0 +1,26 @@ +; RUN: llc < %s -march=r600 -mcpu=redwood | FileCheck %s + +; This test makes sure we do not double count global values when they are +; used in different basic blocks. + +; CHECK-LABEL: @test +; CHECK: .long 166120 +; CHECK-NEXT: .long 1 +@lds = internal addrspace(3) unnamed_addr global i32 zeroinitializer, align 4 + +define void @test(i32 addrspace(1)* %out, i32 %cond) { +entry: + %0 = icmp eq i32 %cond, 0 + br i1 %0, label %if, label %else + +if: + store i32 1, i32 addrspace(3)* @lds + br label %endif + +else: + store i32 2, i32 addrspace(3)* @lds + br label %endif + +endif: + ret void +} |