Misc. updates/fixes to analysis utils used for DMA generation; update DMA

generation pass to make it drop certain assumptions, complete TODOs.

- multiple fixes for getMemoryFootprintBytes
  - pass loopDepth correctly from getMemoryFootprintBytes()
  - use union while computing memory footprints

- bug fixes for addAffineForOpDomain
  - take into account loop step
  - add domains of other loop IVs in turn that might have been used in the bounds

- dma-generate: drop assumption of "non-unit stride loops being tile space loops
  and skipping those and recursing to inner depths"; DMA generation is now purely
  based on available fast mem capacity and memory footprint's calculated

- handle memory region compute failures/bailouts correctly from dma-generate

- loop tiling cleanup/NFC

- update some debug and error messages to use emitNote/emitError in
  pipeline-data-transfer pass - NFC

PiperOrigin-RevId: 234245969

1 files changed, 2 insertions, 2 deletions

diff --git a/mlir/lib/Transforms/PipelineDataTransfer.cpp b/mlir/lib/Transforms/PipelineDataTransfer.cpp
index a943ac280d1..a85f428bde6 100644
--- a/mlir/lib/Transforms/PipelineDataTransfer.cpp
+++ b/mlir/lib/Transforms/PipelineDataTransfer.cpp
@@ -261,7 +261,7 @@ PassResult
 PipelineDataTransfer::runOnAffineForOp(OpPointer<AffineForOp> forOp) {
   auto mayBeConstTripCount = getConstantTripCount(forOp);
   if (!mayBeConstTripCount.hasValue()) {
-    LLVM_DEBUG(llvm::dbgs() << "unknown trip count loop\n");
+    LLVM_DEBUG(forOp->emitNote("unknown trip count loop"));
     return success();
   }
 
@@ -269,7 +269,7 @@ PipelineDataTransfer::runOnAffineForOp(OpPointer<AffineForOp> forOp) {
   findMatchingStartFinishInsts(forOp, startWaitPairs);
 
   if (startWaitPairs.empty()) {
-    LLVM_DEBUG(llvm::dbgs() << "No dma start/finish pairs\n";);
+    LLVM_DEBUG(forOp->emitNote("No dma start/finish pairs\n"));
     return success();
   }