Bugfix: Change Syntax so that Parentheses are Required for llvm.mlir.constant

[AtticusKuhn]

Here is a reference in the LLVM manual:

llvm.mlir.constant (LLVM::ConstantOp)

operation ::= `llvm.mlir.constant` `(` $value `)` attr-dict `:` type($res)

// Integer constant, internal i32 is mandatory
%0 = llvm.mlir.constant(42 : i32) : i32

// It's okay to omit i64.
%1 = llvm.mlir.constant(42) : i64

// Floating point constant.
%2 = llvm.mlir.constant(42.0 : f32) : f32

// Splat dense vector constant.
%3 = llvm.mlir.constant(dense<1.0> : vector<4xf32>) : vector<4xf32>

In the llvm reference manual, parentheses are required after llvm.mlir.constant, but the syntax did not require parentheses. Thus, I changed the syntax.

It's my intuition that the Lean version of LLVM should stay as close to the real version of LLVM as possible.

This closes #440

[tobiasgrosser]

Nice. There are some build failures, but this looks overall great. @alexkeizer might be able to comment on the lean code more.

[github-actions]

Alive Statistics: 58 / 93 (35 failed)

[alexkeizer]

Integer constant, internal i32 is mandatory
Interesting, I think we do allow the internal i32 to be omitted, as in

%0 = llvm.mlir.constant(42) : i32

Like you say, we should stick with the official grammar, so that's something we should probably address too (at some point, not necessarily in this PR)

[alexkeizer]

The build fails because there is another file, AliveAutoGenerated, which has a collection of mlir programs in the DSL too.
It's a bit slow, so we don't build it by default, you have to run lake build AliveExamples to explicitly build them.

As the name suggests, though, these examples are (mostly) automatically generated. Let's talk a bit more about how to fix this tomorrow

[tobiasgrosser]

I like this direction. @AtticusKuhn, it might be nice to close this on Monday.

[AtticusKuhn]

I like this direction. @AtticusKuhn, it might be nice to close this on Monday.

I realised that I don't actually need this PR to do
#452
because I can use alternate syntax.
But still, I agree that it would be good to close this on Monday.

[tobiasgrosser]

Finishing this PR would be great, but don't hesitate to just close it if this is not relevant anymore.

[github-actions]

Alive Statistics: 63 / 93 (30 failed)

[github-actions]

Alive Statistics: 63 / 93 (30 failed)

[alexkeizer]

I'm a bit hesitant to change AliveAutoGenerated manually. I think the best course of action at the moment is to just allow both the old syntax (without parens) alongside the proper LLVM syntax (with parens). That way, we can leave AliveAutoGenerated as is.

[AtticusKuhn]

I'm a bit hesitant to change AliveAutoGenerated manually. I think the best course of action at the moment is to just allow both the old syntax (without parens) alongside the proper LLVM syntax (with parens). That way, we can leave AliveAutoGenerated as is.

That's probably a good idea. I could change the script that generates AliveAutoGenerated, but it's no extra work just to allow both syntaxes.

[github-actions]

Alive Statistics: 63 / 93 (30 failed)

[AtticusKuhn]

Anyway, I've made the parentheses syntax optional, so both

    %c0 = llvm.mlir.constant 0

and

    %c0 = llvm.mlir.constant(0 : i32) : i32

are allowed

[github-actions]

Alive Statistics: 63 / 93 (30 failed)

[alexkeizer]

LGTM, a few minor comments left. Please address those and feel free to merge

[github-actions]

Alive Statistics: 59 / 93 (34 failed)

[github-actions]

Alive Statistics: 59 / 93 (34 failed)