[llvm] [AMDGPU] Extend promotion of alloca to vectors (PR #127973)

[Carl Ritson via llvm-commits]

================
@@ -310,12 +322,17 @@ bool AMDGPUPromoteAllocaImpl::run(Function &F, bool PromoteToLDS) {
 
   bool SufficientLDS = PromoteToLDS ? hasSufficientLocalMem(F) : false;
 
-  // Use up to 1/4 of available register budget for vectorization.
-  // FIXME: Increase the limit for whole function budgets? Perhaps x2?
+  const unsigned VGPRRatio =
+      PromoteAllocaToVectorVGPRRatio.getNumOccurrences()
+          ? PromoteAllocaToVectorVGPRRatio
+          : F.getFnAttributeAsParsedInteger(
----------------
perlfu wrote:

Sorry, I don't understand -- this is all variables?

Given Matt's comment about pass parameters, this will likely move to being a variable in the pass class definition.

https://github.com/llvm/llvm-project/pull/127973