[Mlir-commits] [mlir] [mlir] Compute ceildivs consistently for INT_MIN operands (PR #215696)
Hung Kuan Tseng
llvmlistbot at llvm.org
Wed Aug 12 07:47:37 PDT 2026
Tim096 wrote:
All three addressed, amended into the first commit; the stack is still three commits and the new head is `7a0d9f0ea804`.
Re-ran the oracle sweep on the folder after the switch to `RoundingSDiv`, over 7140 operand pairs (every pair from -40..40 plus `INT_MAX`, `INT_MIN`, `1000000007` and `-1000000007`, zero divisors excluded), comparing against exact ceiling division computed outside LLVM: no mismatches, and the only pair that does not fold is `(-2147483648, -1)`.
`ninja check-mlir` is 3848 passed / 0 failed and `git clang-format` is clean.
https://github.com/llvm/llvm-project/pull/215696
More information about the Mlir-commits
mailing list