[Mlir-commits] [mlir] [mlir] Adopt `ConvertToLLVMPatternInterface` GpuToLLVMConversionPass to align with `convert-to-llvm` (PR #73761)

[Mehdi Amini]

================
@@ -605,11 +621,9 @@ void GpuToLLVMConversionPass::runOnOperation() {
                 !module.getTargetsAttr().empty());
       });
 
-  mlir::arith::populateArithToLLVMConversionPatterns(converter, patterns);
-  mlir::cf::populateControlFlowToLLVMConversionPatterns(converter, patterns);
+  // These aren't covered by the ConvertToLLVMPatternInterface right now.
----------------
joker-eph wrote:

Sorry I didn't follow what you mean by deleted?

The `populateXXXXToLLVMConversionPatterns` that I didn't remove are the one for which we haven't implemented the `ConvertToLLVMPatternInterface`.
The one I removed (`populateControlFlowToLLVMConversionPatterns` and  `populateFuncToLLVMConversionPatterns`) will be handled line 604 above (and only when these dialects are loaded: we only pay what we may need :) )

https://github.com/llvm/llvm-project/pull/73761