NVIDIA/cutlass
1
0
Fork 0
mirror of https://github.com/NVIDIA/cutlass.git synced 2026-04-20 06:48:59 +00:00
Code Issues Packages Projects Releases Wiki Activity

Fix CUDA version checking in examples (#2894)

Browse Source 
* examples: update CUDA version requirements in Blackwell examples

* examples: fix comments to specify the correct CUDA version requirement
This commit is contained in:
Andrew Yooeun Chun
2026-01-07 14:20:37 +09:00
committed by GitHub
parent 670480df3a
commit eb61c91147
11 changed files with 57 additions and 26 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
examples/68_hopper_fp8_warp_specialized_grouped_gemm_with_blockwise_scaling/68_hopper_fp8_warp_specialized_grouped_gemm_with_blockwise_scaling.cu
View File

@@ -738,7 +738,7 @@ int run(OptionType &options, bool host_problem_shapes_available = true)
int main(int argc, char const **args) {
// CUTLASS must be compiled with CUDA 12.0 Toolkit to run this example
// CUTLASS must be compiled with CUDA 12.3 Toolkit to run this example
// and must have compute capability at least 90.
if (__CUDACC_VER_MAJOR__ < 12 || (__CUDACC_VER_MAJOR__ == 12 && __CUDACC_VER_MINOR__ < 3)) {
std::cerr << "This example requires CUDA 12.3 or newer.\n";

2
examples/68_hopper_fp8_warp_specialized_grouped_gemm_with_blockwise_scaling/68_hopper_fp8_warp_specialized_grouped_gemm_with_blockwise_scaling_with_sparse_groups.cu
View File

@@ -751,7 +751,7 @@ int run(OptionType &options, bool host_problem_shapes_available = true)
int main(int argc, char const **args) {
// CUTLASS must be compiled with CUDA 12.0 Toolkit to run this example
// CUTLASS must be compiled with CUDA 12.3 Toolkit to run this example
// and must have compute capability at least 90.
if (__CUDACC_VER_MAJOR__ < 12 || (__CUDACC_VER_MAJOR__ == 12 && __CUDACC_VER_MINOR__ < 3)) {
std::cerr << "This example requires CUDA 12.3 or newer.\n";

2
examples/70_blackwell_gemm/70_blackwell_fp16_gemm.cu
View File

@@ -440,7 +440,7 @@ int run(Options &options)
int main(int argc, char const **args) {
// CUTLASS must be compiled with CUDA 12.0 Toolkit to run this example
// CUTLASS must be compiled with CUDA 12.8 Toolkit to run this example
// and must have compute capability at least 100a.
if (__CUDACC_VER_MAJOR__ < 12 || (__CUDACC_VER_MAJOR__ == 12 && __CUDACC_VER_MINOR__ < 8)) {

8
examples/70_blackwell_gemm/70_blackwell_fp8_gemm.cu
View File

@@ -626,11 +626,11 @@ int run(Options &options)
int main(int argc, char const **args) {
// CUTLASS must be compiled with CUDA 12.0 Toolkit to run this example
// CUTLASS must be compiled with CUDA 12.8 Toolkit to run this example
// and must have compute capability at least sm100a.
if (__CUDACC_VER_MAJOR__ < 12) {
std::cerr << "This example requires CUDA 12 or newer.\n";
if (__CUDACC_VER_MAJOR__ < 12 || (__CUDACC_VER_MAJOR__ == 12 && __CUDACC_VER_MINOR__ < 8)) {
std::cerr << "This example requires CUDA 12.8 or newer." << std::endl;
// Returning zero so this test passes on older Toolkits. Its actions are no-op.
return 0;
}

4
examples/76_blackwell_conv/76_blackwell_conv_wgrad.cu
View File

@@ -486,8 +486,8 @@ int run(Options &options)
int main(int argc, char const **args) {
// CUTLASS must be compiled with CUDA 12.0 Toolkit to run this example
// and must have compute capability at least 90.
// CUTLASS must be compiled with CUDA 12.8 Toolkit to run this example
// and must have compute capability at least sm100a.
if (__CUDACC_VER_MAJOR__ < 12 || (__CUDACC_VER_MAJOR__ == 12 && __CUDACC_VER_MINOR__ < 8)) {
std::cerr << "This example requires CUDA 12.8 or newer." << std::endl;
// Returning zero so this test passes on older Toolkits. Its actions are no-op.

12
examples/77_blackwell_fmha/77_blackwell_fmha.cu
View File

@@ -1036,10 +1036,16 @@ int main_single(int argc, char const **args) {
return -1;
}
if (__CUDACC_VER_MAJOR__ < 12 || props.major != 10) {
std::cout
if (__CUDACC_VER_MAJOR__ < 12 || (__CUDACC_VER_MAJOR__ == 12 && __CUDACC_VER_MINOR__ < 8)) {
std::cerr << "This example requires CUDA 12.8 or newer." << std::endl;
// Returning zero so this test passes on older Toolkits. Its actions are no-op.
return 0;
}
if (props.major != 10) {
std::cerr
<< "This example requires a GPU of NVIDIA's Blackwell Architecture "
<< "(compute capability major 10) and CUDA 12.8 or greater.\n";
<< "(compute capability 100a)." << std::endl;
return 0;
}

12
examples/77_blackwell_fmha/77_blackwell_fmha_bwd.cu
View File

@@ -939,10 +939,16 @@ int main_single(int argc, char const **args) {
return -1;
}
if (__CUDACC_VER_MAJOR__ < 12 || props.major != 10) {
std::cout
if (__CUDACC_VER_MAJOR__ < 12 || (__CUDACC_VER_MAJOR__ == 12 && __CUDACC_VER_MINOR__ < 8)) {
std::cerr << "This example requires CUDA 12.8 or newer." << std::endl;
// Returning zero so this test passes on older Toolkits. Its actions are no-op.
return 0;
}
if (props.major != 10) {
std::cerr
<< "This example requires a GPU of NVIDIA's Blackwell Architecture "
<< "(compute capability 100a) and CUDA 12.8 or greater.\n";
<< "(compute capability 100a)." << std::endl;
return 0;
}

15
examples/77_blackwell_fmha/77_blackwell_fmha_gen.cu
View File

@@ -722,12 +722,19 @@ int main_single(int argc, char const **args) {
return -1;
}
if (__CUDACC_VER_MAJOR__ < 12 || props.major < 10) {
std::cout
<< "This example requires a GPU of NVIDIA's Blackwell Architecture or "
<< "later (compute capability 90 or greater) and CUDA 12.0 or greater.\n";
if (__CUDACC_VER_MAJOR__ < 12 || (__CUDACC_VER_MAJOR__ == 12 && __CUDACC_VER_MINOR__ < 8)) {
std::cerr << "This example requires CUDA 12.8 or newer." << std::endl;
// Returning zero so this test passes on older Toolkits. Its actions are no-op.
return 0;
}
if (props.major != 10) {
std::cerr
<< "This example requires a GPU of NVIDIA's Blackwell Architecture "
<< "(compute capability 100a)." << std::endl;
return 0;
}
//
// Parse options
//

12
examples/77_blackwell_fmha/77_blackwell_mla.cu
View File

@@ -766,10 +766,16 @@ int main_single(int argc, char const **args) {
return -1;
}
if (__CUDACC_VER_MAJOR__ < 12 || props.major != 10) {
std::cout
if (__CUDACC_VER_MAJOR__ < 12 || (__CUDACC_VER_MAJOR__ == 12 && __CUDACC_VER_MINOR__ < 8)) {
std::cerr << "This example requires CUDA 12.8 or newer." << std::endl;
// Returning zero so this test passes on older Toolkits. Its actions are no-op.
return 0;
}
if (props.major != 10) {
std::cerr
<< "This example requires a GPU of NVIDIA's Blackwell Architecture "
<< "(compute capability major 10) and CUDA 12.8 or greater.\n";
<< "(compute capability 100a)." << std::endl;
return 0;
}

12
examples/77_blackwell_fmha/77_blackwell_mla_fwd.cu
View File

@@ -972,10 +972,16 @@ int main_single(int argc, char const **args) {
return -1;
}
if (__CUDACC_VER_MAJOR__ < 12 || props.major != 10) {
std::cout
if (__CUDACC_VER_MAJOR__ < 12 || (__CUDACC_VER_MAJOR__ == 12 && __CUDACC_VER_MINOR__ < 8)) {
std::cerr << "This example requires CUDA 12.8 or newer." << std::endl;
// Returning zero so this test passes on older Toolkits. Its actions are no-op.
return 0;
}
if (props.major != 10) {
std::cerr
<< "This example requires a GPU of NVIDIA's Blackwell Architecture "
<< "(compute capability major 10) and CUDA 12.8 or greater.\n";
<< "(compute capability 100a)." << std::endl;
return 0;
}

2
examples/78_blackwell_emulated_bf16x9_gemm/78_blackwell_emulated_bf16x9_gemm.cu
View File

@@ -429,7 +429,7 @@ int run(Options &options)
int main(int argc, char const **args) {
// CUTLASS must be compiled with CUDA 12.0 Toolkit to run this example
// CUTLASS must be compiled with CUDA 12.8 Toolkit to run this example
// and must have compute capability at least 100a.
if (__CUDACC_VER_MAJOR__ < 12 || (__CUDACC_VER_MAJOR__ == 12 && __CUDACC_VER_MINOR__ < 8)) {
std::cerr << "This example requires CUDA 12.8 or newer." << std::endl;