maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity

Add AutomaticAllocationScope to memref.alloca_scope 

This change adds `AutomaticAllocationScope` to the
memref.alloca_scope op. Additionally, it also clarifies
that alloca_scope is is conceptually a passthrough operation.

Reviewed By: ftynse, bondhugula

Differential Revision: https://reviews.llvm.org/D104227
This commit is contained in:
Denys Shabalin 2021-06-14 16:07:40 +02:00 committed by Alex Zinenko
parent b88fa0e39f
commit c83e696732
1 changed files with 5 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

7
mlir/include/mlir/Dialect/MemRef/IR/MemRefOps.td
View File

@ -204,7 +204,8 @@ def MemRef_AllocaOp : AllocLikeOp<"alloca", AutomaticAllocationScopeResource> {
//===----------------------------------------------------------------------===//
def MemRef_AllocaScopeOp : MemRef_Op<"alloca_scope",
[DeclareOpInterfaceMethods<RegionBranchOpInterface>,
[AutomaticAllocationScope,
DeclareOpInterfaceMethods<RegionBranchOpInterface>,
SingleBlockImplicitTerminator<"AllocaScopeReturnOp">,
RecursiveSideEffects,
NoRegionArguments]> {
@ -223,7 +224,9 @@ def MemRef_AllocaScopeOp : MemRef_Op<"alloca_scope",
```
Here, `%myalloca` memref is valid within the explicitly delimited scope
and is automatically deallocated at the end of the given region.
and is automatically deallocated at the end of the given region. Conceptually,
`memref.alloca_scope` is a passthrough operation with
`AutomaticAllocationScope` that spans the body of the region within the operation.
`memref.alloca_scope` may also return results that are defined in the nested
region. To return a value, one should use `memref.alloca_scope.return`