[mlir][spirv] Modernize op assembly formats and syntax docs

[kuhar]

The SPIR-V dialect predates the assemblyFormat definitions in TableGen and some of the SPRI-V ops needlessly use handwritten parsers. We should go over the ops in the dialect and update to assemblyFormat wherever possible.

With assemblyFormat in place, we can also drop the handwritten syntax docs, since the latter is redundant in the presence of assemblyFormat. This way the two can't possibly diverge and become confusing. This is similar to the cleanup done in #73343.

[llvmbot]

@llvm/issue-subscribers-mlir-spirv

Author: Jakub Kuderski (kuhar)

The SPIR-V dialect predates the `assemblyFormat` definitions in TableGen and some of the SPRI-V ops needlessly use handwritten parsers. We should go over the ops in the dialect and update to `assemblyFormat` wherever possible.

With assemblyFormat in place, we can also drop the handwritten syntax docs, since the latter is redundant in the presence of assemblyFormat. This way the two can't possibly diverge and become confusing. This is similar to the cleanup done in #73343.

[llvmbot]

Hi!

This issue may be a good introductory issue for people new to working on LLVM. If you would like to work on this issue, your first steps are:

1. Assign the issue to you.
2. Fix the issue locally.
3. Run the test suite locally.
   3.1) Remember that the subdirectories under test/ create fine-grained testing targets, so you can
   e.g. use make check-clang-ast to only run Clang's AST tests.
4. Create a git commit
5. Run git clang-format HEAD~1 to format your changes.
6. Submit the patch to Phabricator.
   6.1) Detailed instructions can be found here

For more instructions on how to submit a patch to LLVM, see our documentation.

If you have any further questions about this issue, don't hesitate to ask via a comment on this Github issue.

[llvmbot]

@llvm/issue-subscribers-good-first-issue

Author: Jakub Kuderski (kuhar)

The SPIR-V dialect predates the `assemblyFormat` definitions in TableGen and some of the SPRI-V ops needlessly use handwritten parsers. We should go over the ops in the dialect and update to `assemblyFormat` wherever possible.

With assemblyFormat in place, we can also drop the handwritten syntax docs, since the latter is redundant in the presence of assemblyFormat. This way the two can't possibly diverge and become confusing. This is similar to the cleanup done in #73343.

[alexbeloi]

Hi, I'd like to work on this as my first issue

[kuhar]

Sure @alexbeloi. Feel free to ping me if you need some pointers. Also happy to have a chat on discord or a call.

[alexbeloi]

I don't see any duplicate Syntax: in the SPIRV dialect documentation, but there are a bunch in Math/IR/MathOps.td like atan and others, so I will fix those.

[kuhar]

I don't see any duplicate Syntax: in the SPIRV dialect documentation, but there are a bunch in Math/IR/MathOps.td like atan and others, so I will fix those.

@alexbeloi These duplicate ones have already been removed from the SPIR-V dialect: #73386. The remaining part of the task is to use let assemblyFormat = tablegen syntax definition where we currently rely on hand-written parsers.

You can find these by running something like grep -rn 'Op::print(' mlir/lib/Dialect/SPIRV/IR. Some of these are complicated enough to justify the hand-written parsers and printers, but we also have a bunch of old ops that have simple syntax and should be upgraded to assemblyFormat (even if it requires minor syntax changes like using the <attribute> syntax instead of "attribute").