[Mlir-commits] [mlir] [mlir][ArmSME] Use liveness information in the tile allocator (PR #90448)

Andrzej WarzyƄski llvmlistbot at llvm.org
Tue Apr 30 07:49:43 PDT 2024


================
@@ -6,7 +6,7 @@
 //
 //===----------------------------------------------------------------------===//
 //
-// This pass allocates SME tiles at the 'func.func' op level for ArmSME
+// This transform allocates SME tiles at the 'func.func' op level for ArmSME
 // operations. It does this using a 16-bit tile mask that has a bit for each
 // 128-bit element tile (ZA0.Q-ZA15.Q), the smallest ZA tile granule.
----------------
banach-space wrote:

Please, could we have more documentation in this file? For example, a high level overview of how tile allocation works (live range calculation, followed by transformation). With some info for every section. 

https://github.com/llvm/llvm-project/pull/90448


More information about the Mlir-commits mailing list