p35189024
/
InfiniTensor
forked from jiuyuan/InfiniTensor
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

code format fix

This commit is contained in:
wanghailu 2023-01-16 14:08:14 +08:00
parent 7167badbb7
commit f2f149861a
8 changed files with 28 additions and 27 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

12
src/kernels/bang/element_wise.cc
View File

@ -553,14 +553,14 @@ class SquaredDifferenceCnnl : public BangKernelWithoutConfig {
CNNL_DTYPE_FLOAT, 4, dim_array));
size_t wsSize;
cnnlGetSquaredDifferenceWorkspaceSize(context->cnnlHandle(), aDesc, bDesc, cDesc,
&wsSize);
cnnlGetSquaredDifferenceWorkspaceSize(context->cnnlHandle(), aDesc,
bDesc, cDesc, &wsSize);
BangPtr wsData = context->getWorkspace(wsSize);
cnnlStatus_t stat =
cnnlSquaredDifference(context->cnnlHandle(), aDesc, aData, bDesc, bData,
cDesc, cData, wsData, wsSize);
cnnlSquaredDifference(context->cnnlHandle(), aDesc, aData, bDesc,
bData, cDesc, cData, wsData, wsSize);
if (stat != CNNL_STATUS_SUCCESS)
return;
@ -675,8 +675,8 @@ REGISTER_KERNEL(Device::BANG, OpType::FloorDivTrunc, DataType::Float32,
FloorDivTruncCnnl, "FloorDivTrunc_cnnl_BANG_Float32");
REGISTER_KERNEL(Device::BANG, OpType::FloorMod, DataType::Float32, FloorModCnnl,
"FloorMod_cnnl_BANG_Float32");
REGISTER_KERNEL(Device::BANG, OpType::SquaredDifference, DataType::Float32, SquaredDifferenceCnnl,
"SquaredDifference_cnnl_BANG_Float32");
REGISTER_KERNEL(Device::BANG, OpType::SquaredDifference, DataType::Float32,
SquaredDifferenceCnnl, "SquaredDifference_cnnl_BANG_Float32");
// REGISTER_KERNEL(Device::BANG, OpType::FloorModTrunc, DataType::Float32,
// FloorModTruncCnnl,
// "FloorModTrunc_cnnl_BANG_Float32");

14
src/kernels/bang/pooling.cc
View File

@ -16,26 +16,26 @@ class poolingCnnl : public BangKernelWithoutConfig {
const auto [ph, pw, sh, sw, dh, dw] = op->getPadStrideDilation();
// get inputs
int inArray[4] = {n,c,h,w};
int inArray[4] = {n, c, h, w};
cnnlTensorDescriptor_t inDesc;
checkCnnlError(cnnlCreateTensorDescriptor(&inDesc));
checkCnnlError(cnnlSetTensorDescriptor(
inDesc, CNNL_LAYOUT_NCHW, CNNL_DTYPE_FLOAT, 4, inArray));
checkCnnlError(cnnlSetTensorDescriptor(inDesc, CNNL_LAYOUT_NCHW,
CNNL_DTYPE_FLOAT, 4, inArray));
// get maxpool descriptor
cnnlPoolingDescriptor_t poolingDesc;
checkCnnlError(cnnlCreatePoolingDescriptor(&poolingDesc));
checkCnnlError(cnnlSetPooling2dDescriptor_v2(
poolingDesc, getPoolingMode(), CNNL_NOT_PROPAGATE_NAN, kh, kw, ph, ph,
pw, pw, sh, sw, dh, dw, false));
poolingDesc, getPoolingMode(), CNNL_NOT_PROPAGATE_NAN, kh, kw, ph,
ph, pw, pw, sh, sw, dh, dw, false));
// get outputs
auto outVec = op->getOutput()->getDims();
int outArray[4] = {outVec[0], outVec[1],outVec[2], outVec[3]};
int outArray[4] = {outVec[0], outVec[1], outVec[2], outVec[3]};
cnnlTensorDescriptor_t outDesc;
checkCnnlError(cnnlCreateTensorDescriptor(&outDesc));
checkCnnlError(cnnlSetTensorDescriptor(outDesc, CNNL_LAYOUT_NCHW,
CNNL_DTYPE_FLOAT, 4, outArray));
CNNL_DTYPE_FLOAT, 4, outArray));
size_t wsSize;
cnnlGetPoolingWorkspaceSize(context->cnnlHandle(), getPoolingMode(),
outVec[3], outVec[2], &wsSize);

8
src/kernels/bang/split.cc
View File

@ -35,8 +35,8 @@ class SplitCnnl : public BangKernelWithoutConfig {
}
int dim_array[4] = {dim[0], dim[1], dim[2], dim[3]};
checkCnnlError(cnnlCreateTensorDescriptor(&desc));
checkCnnlError(cnnlSetTensorDescriptor(
desc, CNNL_LAYOUT_NCHW, CNNL_DTYPE_FLOAT, 4, dim_array));
checkCnnlError(cnnlSetTensorDescriptor(desc, CNNL_LAYOUT_NCHW,
CNNL_DTYPE_FLOAT, 4, dim_array));
cnnlTensorDescriptor_t descArray[num];
for (int i = 0; i < num; ++i) {
checkCnnlError(cnnlCreateTensorDescriptor(&descArray[i]));
@ -50,8 +50,8 @@ class SplitCnnl : public BangKernelWithoutConfig {
BangPtr wsData = context->getWorkspace(wsSize);
cnnlStatus_t stat =
cnnlSplit(context->cnnlHandle(), num, axis, desc, inputData,
wsData, wsSize, descArray, argv);
cnnlSplit(context->cnnlHandle(), num, axis, desc, inputData, wsData,
wsSize, descArray, argv);
if (stat != CNNL_STATUS_SUCCESS)
return;

4
test/kernels/bang/test_bang_pooling.cc
View File

@ -10,7 +10,7 @@ namespace infini {
template <class T>
void testPooling(const std::function<void(void *, size_t, DataType)> &generator,
const Shape &shape) {
const Shape &shape) {
// Runtime
Runtime cpuRuntime = CpuRuntimeObj::getInstance();
auto bangRuntime = make_ref<BangRuntimeObj>();
@ -23,7 +23,7 @@ void testPooling(const std::function<void(void *, size_t, DataType)> &generator,
// GPU
Graph bangGraph = make_ref<GraphObj>(bangRuntime);
auto inputGpu = bangGraph->cloneTensor(inputCpu);
auto gpuOp = bangGraph->addOp<T>(inputGpu, nullptr, 3,3,1,1,1,1,2,2);
auto gpuOp = bangGraph->addOp<T>(inputGpu, nullptr, 3, 3, 1, 1, 1, 1, 2, 2);
bangGraph->dataMalloc();
bangRuntime->run(bangGraph);
auto outputGpu = gpuOp->getOutput();

2
test/kernels/bang/test_bang_round.cc
View File

@ -10,7 +10,7 @@ namespace infini {
template <class T>
void testRound(const std::function<void(void *, size_t, DataType)> &generator,
const Shape &shape) {
const Shape &shape) {
// Runtime
Runtime cpuRuntime = CpuRuntimeObj::getInstance();
auto bangRuntime = make_ref<BangRuntimeObj>();

5
test/kernels/bang/test_bang_split.cc
View File

@ -10,7 +10,7 @@ namespace infini {
template <class T>
void testSplit(const std::function<void(void *, size_t, DataType)> &generator,
const Shape &shape) {
const Shape &shape) {
// Runtime
Runtime cpuRuntime = CpuRuntimeObj::getInstance();
auto bangRuntime = make_ref<BangRuntimeObj>();
@ -23,8 +23,7 @@ void testSplit(const std::function<void(void *, size_t, DataType)> &generator,
// GPU
Graph bangGraph = make_ref<GraphObj>(bangRuntime);
auto inputGpu1 = bangGraph->cloneTensor(inputCpu1);
auto gpuOp =
bangGraph->addOp<T>(inputGpu1, std::nullopt, 3, 3);
auto gpuOp = bangGraph->addOp<T>(inputGpu1, std::nullopt, 3, 3);
bangGraph->dataMalloc();
bangRuntime->run(bangGraph);
auto o0Cpu = gpuOp->getOutput(0)->clone(cpuRuntime);

2
test/kernels/bang/test_bang_square.cc
View File

@ -10,7 +10,7 @@ namespace infini {
template <class T>
void testSquare(const std::function<void(void *, size_t, DataType)> &generator,
const Shape &shape) {
const Shape &shape) {
// Runtime
Runtime cpuRuntime = CpuRuntimeObj::getInstance();
auto bangRuntime = make_ref<BangRuntimeObj>();

8
test/kernels/bang/test_bang_squaredDifference.cc
View File

@ -9,8 +9,9 @@
namespace infini {
template <class T>
void testSquaredDifference(const std::function<void(void *, size_t, DataType)> &generator,
const Shape &shape) {
void testSquaredDifference(
const std::function<void(void *, size_t, DataType)> &generator,
const Shape &shape) {
// Runtime
Runtime cpuRuntime = CpuRuntimeObj::getInstance();
auto bangRuntime = make_ref<BangRuntimeObj>();
@ -40,7 +41,8 @@ void testSquaredDifference(const std::function<void(void *, size_t, DataType)> &
}
TEST(cnnl_SquaredDifference, run) {
testSquaredDifference<SquaredDifferenceObj>(IncrementalGenerator(), Shape{1, 2, 2, 3});
testSquaredDifference<SquaredDifferenceObj>(IncrementalGenerator(),
Shape{1, 2, 2, 3});
}
} // namespace infini