aboutsummaryrefslogtreecommitdiff
path: root/libc/src/math/amdgpu/nearbyintf.cpp
diff options
context:
space:
mode:
authorJeremy Kun <jkun@google.com>2025-10-02 10:56:40 -0700
committerGitHub <noreply@github.com>2025-10-02 10:56:40 -0700
commit99d85906c542c3801a24137ba6d6f2c367308563 (patch)
tree5f0b26fd9651bdf2fb2bccfdd5467464c5d2cfd4 /libc/src/math/amdgpu/nearbyintf.cpp
parent67c000efb7bb7858f9fb6e577f2c9d1f24291ba0 (diff)
downloadllvm-main.zip
llvm-main.tar.gz
llvm-main.tar.bz2
[mlir] [irdl] Add support for regions in irdl-to-cpp (#158540)HEADmain
Fixes https://github.com/llvm/llvm-project/issues/158034 For the input ```mlir irdl.dialect @conditional_dialect { // A conditional operation with regions irdl.operation @conditional { // Create region constraints %r0 = irdl.region // Unconstrained region %r1 = irdl.region() // Region with no entry block arguments %v0 = irdl.any %r2 = irdl.region(%v0) // Region with one i1 entry block argument irdl.regions(cond: %r2, then: %r0, else: %r1) } } ``` This produces the following cpp: https://gist.github.com/j2kun/d2095f108efbd8d403576d5c460e0c00 Summary of changes: - The op class and adaptor get named accessors to the regions `Region &get<RegionName>()` and `getRegions()` - The op now gets `OpTrait::NRegions<3>` and `OpInvariants` to trigger the region verification - Support for region block argument constraints is added, but not working for all constraints until codegen for `irdl.is` is added (filed https://github.com/llvm/llvm-project/issues/161018 and left a TODO). - Helper functions for the individual verification steps are added, following mlir-tblgen's format (in the above gist, `__mlir_irdl_local_region_constraint_ConditionalOp_cond` and similar), and `verifyInvariantsImpl` that calls them. - Regions are added in the builder ## Questions for the reviewer ### What is the "correct" interface for verification? I used `mlir-tblgen` on an analogous version of the example `ConditionalOp` in this PR, and I see an `::mlir::OpTrait::OpInvariants` trait as well as ```cpp ::llvm::LogicalResult ConditionalOp::verifyInvariantsImpl() { { unsigned index = 0; (void)index; for (auto &region : ::llvm::MutableArrayRef((*this)->getRegion(0))) if (::mlir::failed(__mlir_ods_local_region_constraint_test1(*this, region, "cond", index++))) return ::mlir::failure(); for (auto &region : ::llvm::MutableArrayRef((*this)->getRegion(1))) if (::mlir::failed(__mlir_ods_local_region_constraint_test1(*this, region, "then", index++))) return ::mlir::failure(); for (auto &region : ::llvm::MutableArrayRef((*this)->getRegion(2))) if (::mlir::failed(__mlir_ods_local_region_constraint_test1(*this, region, "else", index++))) return ::mlir::failure(); } return ::mlir::success(); } ::llvm::LogicalResult ConditionalOp::verifyInvariants() { if(::mlir::succeeded(verifyInvariantsImpl()) && ::mlir::succeeded(verify())) return ::mlir::success(); return ::mlir::failure(); } ``` However, `OpInvariants` only seems to need `verifyInvariantsImpl`, so it's not clear to me what is the purpose of the `verifyInvariants` function, or, if I leave out `verifyInvariants`, whether I need to call `verify()` in my implementation of `verifyInvariantsImpl`. In this PR, I omitted `verifyInvariants` and generated `verifyInvariantsImpl`. ### Is testing sufficient? I am not certain I implemented the builders properly, and it's unclear to me to what extent the existing tests check this (which look like they compile the generated cpp, but don't actually use it). Did I omit some standard function or overload? --------- Co-authored-by: Jeremy Kun <j2kun@users.noreply.github.com>
Diffstat (limited to 'libc/src/math/amdgpu/nearbyintf.cpp')
0 files changed, 0 insertions, 0 deletions