[llvm] [LLVM][NVPTX] Add NVPTX codegen support for clusterlaunchcontrol instruction (PR #134568)

via llvm-commits llvm-commits at lists.llvm.org
Mon Apr 7 05:20:33 PDT 2025


================
@@ -112,6 +112,20 @@ class NVPTXSubtarget : public NVPTXGenSubtargetInfo {
     return HasTcgen05 && PTXVersion >= 86;
   }
 
+  bool isBlackwellArchAccelerated() const {
+    auto HasRequiredSM = [&]() {
+      if (FullSmVersion == 1001)
+        return true;
+      if (FullSmVersion == 1011)
+        return true;
+      if (FullSmVersion == 1201)
+        return true;
+      return false;
+    };
+
+    return HasRequiredSM() && PTXVersion >= 86;
----------------
gonzalobg wrote:

The non-`.multicast` instructions being added are not arch conditional, so `sm >= 100 && ptx >= 86` suffice.

For the `.multicast` flavors, they only apply to this one instruction, so calling it "blackwellArchAccelerated" instead of `hasMulticastClusterLaunchControlTryCancel` is not very accurate.

https://github.com/llvm/llvm-project/pull/134568


More information about the llvm-commits mailing list