This is an automated email from the ASF dual-hosted git repository.
guanmingchiu pushed a commit to branch dev-qdp
in repository https://gitbox.apache.org/repos/asf/mahout.git
The following commit(s) were added to refs/heads/dev-qdp by this push:
new 328eeccf2 [QDP] support turing architecture GPU (#731)
328eeccf2 is described below
commit 328eeccf23a6a606a220d66e59e1b505aae648be
Author: Nary Yeh <[email protected]>
AuthorDate: Tue Dec 16 05:35:54 2025 -0800
[QDP] support turing architecture GPU (#731)
---
qdp/qdp-kernels/build.rs | 5 ++++-
1 file changed, 4 insertions(+), 1 deletion(-)
diff --git a/qdp/qdp-kernels/build.rs b/qdp/qdp-kernels/build.rs
index c60d27c4a..55d19f683 100644
--- a/qdp/qdp-kernels/build.rs
+++ b/qdp/qdp-kernels/build.rs
@@ -67,11 +67,14 @@ fn main() {
.flag("-cudart=shared") // Use shared CUDA runtime
.flag("-std=c++17") // C++17 for modern CUDA features
// GPU architecture targets
+ // SM 75 = Turing (T4, RTX 2000 series)
// SM 80 = Ampere (A100, RTX 3000 series)
// SM 86 = Ampere (RTX 3090, A40)
// SM 89 = Ada Lovelace (RTX 4000 series)
// SM 90 = Hopper (H100)
- // For MVP, we target SM 80 as baseline
+ // Support both Turing (sm_75) and Ampere+ architectures
+ .flag("-gencode")
+ .flag("arch=compute_75,code=sm_75")
.flag("-gencode")
.flag("arch=compute_80,code=sm_80")
// Optional: Add more architectures for production
