Mathis/CUDANet
1
0
Fork 0
mirror of https://github.com/lordmathis/CUDANet.git synced 2025-11-06 01:34:22 +00:00
Code Issues Packages Projects Releases Wiki Activity

Implement padding kernel

Browse Source
This commit is contained in:
Matúš Námešný
2024-03-03 15:24:39 +01:00
parent 019ccc33d9
commit 7e4460cc5e
2 changed files with 40 additions and 27 deletions
Download Patch File Download Diff File Expand all files Collapse all files

19
test/kernels/test_padding.cu
View File

@@ -27,9 +27,7 @@ TEST_F(PaddingTest, SimplePaddingTest) {
cudaStatus = cudaMalloc((void**)&d_input, sizeof(float) * inputSize);
EXPECT_EQ(cudaStatus, cudaSuccess);
cudaStatus = cudaMalloc(
(void**)&d_padded, sizeof(float) * paddedSize
);
cudaStatus = cudaMalloc((void**)&d_padded, sizeof(float) * paddedSize);
EXPECT_EQ(cudaStatus, cudaSuccess);
/*
@@ -54,7 +52,12 @@ TEST_F(PaddingTest, SimplePaddingTest) {
cublasSetVector(inputSize, sizeof(float), input.data(), 1, d_input, 1);
EXPECT_EQ(cublasStatus, CUBLAS_STATUS_SUCCESS);
pad_matrix_kernel<<<1, 1>>>(d_input, d_padded, w, h, n, p);
int THREADS_PER_BLOCK = 64;
int BLOCKS = paddedSize / THREADS_PER_BLOCK + 1;
pad_matrix_kernel<<<BLOCKS, THREADS_PER_BLOCK>>>(
d_input, d_padded, w, h, n, p
);
cudaStatus = cudaDeviceSynchronize();
EXPECT_EQ(cudaStatus, cudaSuccess);
@@ -70,13 +73,7 @@ TEST_F(PaddingTest, SimplePaddingTest) {
paddedSize, sizeof(float), d_padded, 1, output.data(), 1
);
std::cout << "Actual output: " << std::endl;
for (int i = 0; i < paddedSize; i++) {
std::cout << output[i] << " ";
}
std::cout << std::endl;
for (int i = 0; i < paddedSize; i++) {
EXPECT_NEAR(expectedOutput[i], output[i], 1e-5);
}
}
}