Skip to content

Conversation

@s-barannikov
Copy link
Contributor

Glue operand is only present if there are variadic register operands, which makes it optional.
Also, change the number of fixed operands to 1 (the trap ID).

* the node always has one *fixed* argument, which is the trap id
* glue operand is only present if there are variadic register operands
@llvmbot
Copy link
Member

llvmbot commented Nov 24, 2024

@llvm/pr-subscribers-backend-amdgpu

Author: Sergei Barannikov (s-barannikov)

Changes

Glue operand is only present if there are variadic register operands, which makes it optional.
Also, change the number of fixed operands to 1 (the trap ID).


Full diff: https://github.com/llvm/llvm-project/pull/117453.diff

1 Files Affected:

  • (modified) llvm/lib/Target/AMDGPU/AMDGPUInstrInfo.td (+2-2)
diff --git a/llvm/lib/Target/AMDGPU/AMDGPUInstrInfo.td b/llvm/lib/Target/AMDGPU/AMDGPUInstrInfo.td
index 702f6e67c55271..bec294a945d2fe 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUInstrInfo.td
+++ b/llvm/lib/Target/AMDGPU/AMDGPUInstrInfo.td
@@ -100,8 +100,8 @@ def AMDGPUtc_return_chain: SDNode<"AMDGPUISD::TC_RETURN_CHAIN",
 >;
 
 def AMDGPUtrap : SDNode<"AMDGPUISD::TRAP",
-  SDTypeProfile<0, -1, [SDTCisVT<0, i16>]>,
-    [SDNPHasChain, SDNPVariadic, SDNPSideEffect, SDNPInGlue]
+  SDTypeProfile<0, 1, [SDTCisVT<0, i16>]>,
+    [SDNPHasChain, SDNPVariadic, SDNPSideEffect, SDNPOptInGlue]
 >;
 
 def AMDGPUconstdata_ptr : SDNode<

@s-barannikov s-barannikov merged commit 8d65073 into llvm:main Nov 24, 2024
10 checks passed
@s-barannikov s-barannikov deleted the sdag/amdgpu-trap branch November 24, 2024 06:25
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants