zhangyiss/mlx
1
0
Fork 0
mirror of https://github.com/ml-explore/mlx.git synced 2025-12-16 01:49:05 +08:00
Code Issues Actions Packages Projects Releases Wiki Activity

use arch specific targets when possible (#2771)

Browse Source
This commit is contained in:
Awni Hannun
2025-11-14 20:04:18 -08:00
committed by GitHub
parent 3c622ddd1d
commit 1bf605d56d
4 changed files with 34 additions and 5 deletions
Download Patch File Download Diff File Expand all files Collapse all files

9
mlx/backend/cuda/jit_module.cpp
View File

@@ -279,11 +279,14 @@ void compile(
// Compile program.
std::vector<const char*> args;
bool use_sass = compiler_supports_device_sass(device);
auto cc = device.compute_capability_major();
std::string arch_tag = (cc == 90 || cc == 100 || cc == 121) ? "a" : "";
std::string compute = fmt::format(
"--gpu-architecture={}_{}{}",
"--gpu-architecture={}_{}{}{}",
use_sass ? "sm" : "compute",
device.compute_capability_major(),
device.compute_capability_minor());
cc,
device.compute_capability_minor(),
arch_tag);
args.push_back(compute.c_str());
std::string cccl_include = cccl_dir();
if (!cccl_include.empty()) {