[llvm-branch-commits] [clang] [llvm] [mlir] [OMPIRBuilder] Add support for explicit deallocation points (PR #154752)

[Tom Eccles via llvm-branch-commits]

tblah wrote:

> I don't know about this. However, I wasn't aware mixing allocas and heap allocations in the same block could cause issues, could you share an example where that would be problematic? I see the benefit of splitting allocations from initializations and other instructions, because it makes sure we don't accidentally try to access memory before it has been defined, but that's not something that would happen by interleaving heap allocation instructions (they're not supposed to access any potential sibling allocas).

Yes I guess that's a point. It might be that the improvement here outweighs the disadvantage.

As I understand it, it is best to create all allocas together because then it can all be reduced to a single addition/subtraction on the stack pointer instead of a series of them. Putting executable code (e.g. calls to heap allocators) might lead to there being several separate adjustments to the stack pointer. Also as I understand it, collecting allocas together (ideally at the start of the entry block, after we have done all of the outlining) is considered a more canonical form of LLVMIR (it is better than it used to be but we definitely don't achieve this in all cases currently).

https://github.com/llvm/llvm-project/pull/154752