[CIR][ThroughMLIR] Lower structs and GetMemberOp

[terapines-osc-2]

Structs are implemented as memref<size x i8>. It is not feasible to represent them as tuples, for tuples can't be put in memref (i.e. pointers to structs would break if we did).

We use memref::ViewOp for this. Unlike PtrStrideOp, the reinterpret cast operation doesn't work here, as the result type is potentially different from i8.

[bcardosolopes]

Overall looks good to me. @keryell might have some input here

[bcardosolopes]

Looks like some tests are failing

[terapines-osc-2]

Seems the failed tests are because of the rename of StructType. Now it's fixed.

[bcardosolopes]

now that I landed your other PR, gh might automatically run the tests, gave it another click!

[keryell]

Yes you can do that.
But there are some philosophical questions about what is the meaning of MLIR standard dialect lowering, along some points discussed in #1219
There is a trade-off between lowering to MLIR standard dialect as is today whatever low-level the generated code is and waiting for MLIR standard dialect to improve so that the code can be high-level.
Probably your approach of having something upstream today is good compared to what I have tried with "what if we had a better tuple in MLIR standard dialect in the future" in #1334

[keryell]

In #1334 I use a tuple-like type to keep the high-level data structure and being able to have for example struct of struct or struct of arrays which have a value semantics which cannot be represented with memref but with the same trick as you to emulate GetMemberOp.

[bcardosolopes]

Fails on windows for some reason!

[terapines-osc-2]

It seems #1569 calculates wrong offsets for struct types, so the test cases will fail. For a non-packed struct { char a; int b; }, it thinks the offset of b is 1, but it should be 4. Currently CIR doesn't have test cases for these, so it's left unnoticed.
I'll try to solve it in another PR and leave this open for a while.

Still don't know why there's failure on windows though.

[bcardosolopes]

I'll try to solve it in another PR and leave this open for a while.

Great!

Still don't know why there's failure on windows though.

Seems to be failing all across the board

[keryell]

To rebase?

[terapines-osc-2]

Now the struct offset issue is fixed.
As it's hard to write an independent test case for struct types, I choose to include the change here (it's small anyway).

[bcardosolopes]

One alternative is to build your clang with ASAN enabled by the host compiler (using CMAKE's -DLLVM_USE_SANITIZER="Address"), that usually helps spotting windows only crashes