minor update
This commit is contained in:
@@ -73,7 +73,7 @@ OBJS := $(addsuffix .o, $(notdir $(SRCS)))
|
||||
all: $(PROJECT) kernel.pocl
|
||||
|
||||
kernel.pocl: kernel.cl
|
||||
LLVM_PREFIX=$(LLVM_VORTEX) POCL_DEBUG=all LD_LIBRARY_PATH=$(LLVM_POCL)/lib:$(POCL_CC_PATH)/lib:$(LLVM_VORTEX)/lib POCL_VORTEX_CFLAGS="$(K_CFLAGS)" POCL_VORTEX_LDFLAGS="$(K_LDFLAGS)" $(POCL_CC_PATH)/bin/poclcc -o kernel.pocl kernel.cl
|
||||
LD_LIBRARY_PATH=$(LLVM_POCL)/lib:$(POCL_CC_PATH)/lib:$(LLVM_VORTEX)/lib:$(LD_LIBRARY_PATH) LLVM_PREFIX=$(LLVM_VORTEX) POCL_DEBUG=all POCL_VORTEX_CFLAGS="$(K_CFLAGS)" POCL_VORTEX_LDFLAGS="$(K_LDFLAGS)" $(POCL_CC_PATH)/bin/poclcc -o kernel.pocl kernel.cl
|
||||
|
||||
%.cc.o: %.cc
|
||||
$(CXX) $(CXXFLAGS) -c $< -o $@
|
||||
@@ -87,6 +87,9 @@ kernel.pocl: kernel.cl
|
||||
$(PROJECT): $(OBJS)
|
||||
$(CXX) $(CXXFLAGS) $^ $(LDFLAGS) -o $@
|
||||
|
||||
run-hostgpu: $(PROJECT) kernel.pocl
|
||||
./$(PROJECT) $(OPTS)
|
||||
|
||||
run-simx: $(PROJECT) kernel.pocl
|
||||
LD_LIBRARY_PATH=$(POCL_RT_PATH)/lib:$(VORTEX_RT_PATH)/simx:$(LD_LIBRARY_PATH) ./$(PROJECT) $(OPTS)
|
||||
|
||||
|
||||
@@ -28,6 +28,9 @@ __kernel void matmul(__global float *A,
|
||||
for (int j = 0; j < localSize; j++) {
|
||||
sum += localA[localRow * localSize + j] * localB[j * localSize + localCol];
|
||||
}
|
||||
|
||||
// Ensure computation is done before loading next block
|
||||
barrier(CLK_LOCAL_MEM_FENCE);
|
||||
}
|
||||
|
||||
C[globalRow * N + globalCol] = sum;
|
||||
|
||||
Reference in New Issue
Block a user