[PATCH] D135760: [AArch64]SME2 single-multi and multi-multi INT dot product instructions[part2]

[Paul Walker via Phabricator via llvm-commits]

paulwalker-arm accepted this revision.
paulwalker-arm added a comment.
This revision is now accepted and ready to land.

I have an observation regarding the quantity of tests but that's just something to consider for later patches (or later refactoring).  The patch itself looks good to me.

I've not mentioned this before because the tests can be easily changed later but this patch with its handful of instructions best highlights my next point.  There seems to be far more tests than the variability of the instruction encoding would suggest is necessary. Taking `udot    za.d[w#, #, vgx4], {z#.t - z#.t}, z#.t` as an example, this one variant has 40+ tests.  Tests are good and all but for something that reuses much of decoding operands and classes of other instructions it seems excessive. Another example is checking `{ z0.b - z1.b }` is synonymous with `{ z0.b, z1.b }`.  This seems like something we can test in isolation rather than duplicating every equivalent test across all opcodes.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D135760/new/

https://reviews.llvm.org/D135760