This is an archive of the discontinued LLVM Phabricator instance.

[mlir] Move min/max ops from Std to Arith.
ClosedPublic

Authored by pifon2a on Nov 15 2021, 3:53 AM.

Details

Summary

The original patch https://reviews.llvm.org/D110540 already intended for these operations to be in the arithmetic dialect.

Diff Detail

Event Timeline

pifon2a created this revision.Nov 15 2021, 3:53 AM
pifon2a requested review of this revision.Nov 15 2021, 3:53 AM
ftynse accepted this revision.Nov 15 2021, 4:16 AM
ftynse edited the summary of this revision. (Show Details)
This revision is now accepted and ready to land.Nov 15 2021, 4:16 AM
This revision was automatically updated to reflect the committed changes.

We've been seeing a failure in our Ubuntu builds since this change - specifically because of an unused function:

FAILED: tools/mlir/lib/Dialect/StandardOps/CMakeFiles/obj.MLIRStandard.dir/IR/Ops.cpp.o 

/mnt/vss/_work/1/s/llvm-project/mlir/lib/Dialect/StandardOps/IR/Ops.cpp:105:13: error: unused function 'printStandardBinaryOp' [-Werror,-Wunused-function]
static void printStandardBinaryOp(Operation *op, OpAsmPrinter &p) {
            ^
1 error generated.

I looked into it and it seems to me that after this change a large number of definitions in Ops.td are now no longer used (e.g. IntBinaryOp and FloatBinaryOp among others). Are they still relevant or should they go as well?