Skip to content

[Test] Optimize the speed of tests #4

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
ConvolutedDog merged 2 commits into from
Sep 29, 2025
Merged

[Test] Optimize the speed of tests #4

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
dbd2bda
[Test] Optimize the speed of tests
ConvolutedDog Sep 29, 2025
c5114af
Speedup
ConvolutedDog Sep 29, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
25 changes: 15 additions & 10 deletions tests/short-time-LatestTVM.sh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -15,22 +15,27 @@ run_benchmarks() {

export CUDA_VISIBLE_DEVICES=$gpu_id

# Run conv benchmark
echo "Running fused_conv_expr_S1D1P1 (float16)..."
python -u test_op.py --gen_check_code --backend tvm --topk 1 --code_dir generated_source/conv --smem_tiling --reg_tiling --codegen_input_reg_tiling --shared_fetch_vectorize --data_type "float16" --op fused_conv_expr_S1D1P1 --shape 128 128 28 28 128 3 3
# Only test the compiling functionality on GV100, because its profiling implementation is faster than that on RTX 4090.
if [[ "$device_name" == "NVIDIA GV100" ]]; then
# Run conv benchmark
echo "Running fused_conv_expr_S1D1P1 (float16)..."
python -u test_op.py --gen_check_code --backend tvm --topk 1 --code_dir generated_source/conv --smem_tiling --reg_tiling --codegen_input_reg_tiling --shared_fetch_vectorize --data_type "float16" --op fused_conv_expr_S1D1P1 --shape 128 128 28 28 128 3 3

# Run matmul benchmarks
echo "Running matmul_expr (float32)..."
python -u test_op.py --gen_check_code --backend tvm --topk 1 --code_dir generated_source/matmul --smem_tiling --reg_tiling --codegen_input_reg_tiling --shared_fetch_vectorize --data_type "float32" --op matmul_expr --shape 1024 1024 1024
fi

# Run matmul benchmarks
echo "Running matmul_expr (float16) with tensor cores..."
python -u test_op.py --gen_check_code --backend tvm --topk 1 --code_dir generated_source/matmul --smem_tiling --reg_tiling --codegen_input_reg_tiling --shared_fetch_vectorize --use_tc --data_type "float16" --op matmul_expr --shape 4096 4096 4096

echo "Running matmul_expr (float32)..."
python -u test_op.py --gen_check_code --backend tvm --topk 1 --code_dir generated_source/matmul --smem_tiling --reg_tiling --codegen_input_reg_tiling --shared_fetch_vectorize --data_type "float32" --op matmul_expr --shape 4096 4096 4096
# Only test the profiling functionality on RTX 4090, the rightness of the matmul_expr has been tested on GV100.
if [[ "$device_name" == "NVIDIA RTX 4090" ]]; then
echo "Running matmul_expr (float16) with tensor cores..."
python -u test_op.py --gen_check_code --backend tvm --topk 1 --code_dir generated_source/matmul --smem_tiling --reg_tiling --codegen_input_reg_tiling --shared_fetch_vectorize --use_tc --data_type "float16" --op matmul_expr --shape 1024 1024 1024
fi

echo ""
}

# Run benchmarks on all GPUs
run_benchmarks 0 "NVIDIA H100"
run_benchmarks 5 "NVIDIA RTX 4090"
run_benchmarks 4 "NVIDIA GV100"

Expand Down
25 changes: 15 additions & 10 deletions tests/short-time-OldTVM.sh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -15,22 +15,27 @@ run_benchmarks() {

export CUDA_VISIBLE_DEVICES=$gpu_id

# Run conv benchmark
echo "Running fused_conv_expr_S1D1P1 (float16)..."
python -u test_op.py --gen_check_code --backend tvm --topk 1 --code_dir generated_source/conv --smem_tiling --reg_tiling --codegen_input_reg_tiling --shared_fetch_vectorize --data_type "float16" --op fused_conv_expr_S1D1P1 --shape 128 128 28 28 128 3 3
# Only test the compiling functionality on GV100, because its profiling implementation is faster than that on RTX 4090.
if [[ "$device_name" == "NVIDIA GV100" ]]; then
# Run conv benchmark
echo "Running fused_conv_expr_S1D1P1 (float16)..."
python -u test_op.py --gen_check_code --backend tvm --topk 1 --code_dir generated_source/conv --smem_tiling --reg_tiling --codegen_input_reg_tiling --shared_fetch_vectorize --data_type "float16" --op fused_conv_expr_S1D1P1 --shape 128 128 28 28 128 3 3

# Run matmul benchmarks
echo "Running matmul_expr (float32)..."
python -u test_op.py --gen_check_code --backend tvm --topk 1 --code_dir generated_source/matmul --smem_tiling --reg_tiling --codegen_input_reg_tiling --shared_fetch_vectorize --data_type "float32" --op matmul_expr --shape 1024 1024 1024
fi

# Run matmul benchmarks
echo "Running matmul_expr (float16) with tensor cores..."
python -u test_op.py --gen_check_code --backend tvm --topk 1 --code_dir generated_source/matmul --smem_tiling --reg_tiling --codegen_input_reg_tiling --shared_fetch_vectorize --use_tc --data_type "float16" --op matmul_expr --shape 4096 4096 4096

echo "Running matmul_expr (float32)..."
python -u test_op.py --gen_check_code --backend tvm --topk 1 --code_dir generated_source/matmul --smem_tiling --reg_tiling --codegen_input_reg_tiling --shared_fetch_vectorize --data_type "float32" --op matmul_expr --shape 4096 4096 4096
# Only test the profiling functionality on RTX 4090, the rightness of the matmul_expr has been tested on GV100.
if [[ "$device_name" == "NVIDIA RTX 4090" ]]; then
echo "Running matmul_expr (float16) with tensor cores..."
python -u test_op.py --gen_check_code --backend tvm --topk 1 --code_dir generated_source/matmul --smem_tiling --reg_tiling --codegen_input_reg_tiling --shared_fetch_vectorize --use_tc --data_type "float16" --op matmul_expr --shape 1024 1024 1024
fi

echo ""
}

# Run benchmarks on all GPUs
run_benchmarks 0 "NVIDIA H100"
run_benchmarks 5 "NVIDIA RTX 4090"
run_benchmarks 4 "NVIDIA GV100"

Expand Down