[mlir][linalg][nfc] map using SmallVector
#154981
Conversation
elamc-2
requested review from
dcaballe,
nicolasvasilache and
rengolin
as code owners
|
Thank you for submitting a Pull Request (PR) to the LLVM Project! This PR will be automatically labeled and the relevant teams will be notified. If you wish to, you can add reviewers by using the "Reviewers" section on this page. If this is not working for you, it is probably because you do not have write permissions for the repository. In which case you can instead tag reviewers by name in a comment by using If you have received no comments on your PR for a week, you can request a review by "ping"ing the PR by adding a comment “Ping”. The common courtesy "ping" rate is once a week. Please remember that you are asking for valuable time from other developers. If you have further questions, they may be answered by the LLVM GitHub User Guide. You can also ask questions in a comment on this PR, on the LLVM Discord or on the forums. |
|
@llvm/pr-subscribers-mlir @llvm/pr-subscribers-mlir-linalg Author: Elam Cohavi (elamc-2) ChangesThis change swaps Full diff: https://github.com/llvm/llvm-project/pull/154981.diff 1 Files Affected:
diff --git a/mlir/lib/Dialect/Linalg/Transforms/DecomposeGenericByUnfoldingPermutation.cpp b/mlir/lib/Dialect/Linalg/Transforms/DecomposeGenericByUnfoldingPermutation.cpp
index b7da20c108f3e..9b5c14887b253 100644
--- a/mlir/lib/Dialect/Linalg/Transforms/DecomposeGenericByUnfoldingPermutation.cpp
+++ b/mlir/lib/Dialect/Linalg/Transforms/DecomposeGenericByUnfoldingPermutation.cpp
@@ -8,7 +8,6 @@
//
#include "mlir/Dialect/Linalg/IR/Linalg.h"
#include "mlir/Dialect/Linalg/Transforms/Transforms.h"
-#include <map>
#include <utility>
using namespace mlir;
@@ -125,7 +124,7 @@ computeTransposeBroadcast(AffineMap &map) {
// refering to d3, d4. Therefore, re-base the transpose dimensions so
// that they start from d0.
permutation.resize(minorSize);
- std::map<int64_t, int64_t> minorMap;
+ llvm::DenseMap<int64_t, int64_t> minorMap;
for (int64_t i = 0; i < minorSize; ++i)
minorMap.insert({sortedResMap[i], i});
|
| @@ -125,7 +124,7 @@ computeTransposeBroadcast(AffineMap &map) { | |||
| // refering to d3, d4. Therefore, re-base the transpose dimensions so | |||
| // that they start from d0. | |||
| permutation.resize(minorSize); | |||
| std::map<int64_t, int64_t> minorMap; | |||
| llvm::DenseMap<int64_t, int64_t> minorMap; | |||
There was a problem hiding this comment.
Seems to me that this is all a map of positions right? So why can't we use a SmallVector<int> here?
I think something like:
| llvm::DenseMap<int64_t, int64_t> minorMap; | |
| llvm::DenseMap<int> minorMap; | |
| minorMap.resize(map.getNumResults()); |
There was a problem hiding this comment.
minorMap maps transposed : sorted dimensions, so to construct remappedResult using SmallVector<pair<int,int>> might make more sense.
There was a problem hiding this comment.
I don’t quite follow: we have key-value pair, the natural translation when you have keys that are small integers is an array of values, I don’t see where the pair would come from (the position in the array is the key)
There was a problem hiding this comment.
That makes sense. I wonder in that case if the original intention to maintain a map is for cases such as:
[1,2,4]
.. where a dimension isn't present and you wouldn't be able to rely on the index being equivalent to the key.
There was a problem hiding this comment.
Possibly, but we never iterate the map, so having unused entries does not matter I believe.
elamc-2
changed the title
llvm::DenseMapSmallVector
|
@joker-eph Thanks for the suggestions, I added the changes |
This change swaps
std::mapfor a vector, and uses its LLVM equivalent type.