[PATCH] D71904: [mlir] GPU: introduce utilities for promotion to workgroup memory

[Alex Zinenko via Phabricator via llvm-commits]

ftynse marked 7 inline comments as done.
ftynse added inline comments.


================
Comment at: mlir/lib/Dialect/GPU/Transforms/MemoryPromotion.cpp:73
+/// are dominated by any operation this emitted.
+static void emitLoopBounds(InsertCopiesLoopBounds &bounds, OpBuilder &builder,
+                           Location loc, MemRefType type) {
----------------
nicolasvasilache wrote:
> To tie the previous suggestions together:
> 1. use EDSC `MemRefView` (feel free to rename), to capture all lbs/ubs/steps
> 2. add your dummy dims
> 3. emit loops, they are captured automatically
> 4. apply `mapToProcessorIds` on the k-innermost loops
Done.
The generated code is much uglier, but let's see if it triggers any issues wrt canonicalization between every step.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D71904/new/

https://reviews.llvm.org/D71904