Committed revision r354576.

Add the changes to isAddExpression instead.

Sorry i should have provided more context, this is the original crash:

opt: ../lib/Transforms/Scalar/InferAddressSpaces.cpp:316: void {anonymous}::InferAddressSpaces::appendsFlatAddressExpressionToPostorderStack(llvm::Value*, std::vector<std::pair<llvm::Value*, bool> >&, llvm::DenseSet<llvm::Value*>&) const: Assertion `V->getType()->isPointerTy()' failed.

I re-worked where the builtins are actually inserted, now it's in a similar place to the "normal" clang builtins.

Re-uploading the patch. The previous version of the patch was missing the include and lib prefixes.

LGTM

LGTM!

Some guidance on how to fix this in the SLPVectorzier would be helpful. Or if it's small enough that someone else can fix it, that's fine with me too.

Here is the (reduced) motivating example:

In D50840#1202436, @lebedev.ri wrote:

1. Why is this limited to extensions? Why can't this also be done for trunc? Or more generally, why don't we want to do this if the same operation is applied for all the elements?
2. Do the vectorizer passes handle this? (Especially in light of the last question of 1.)

I committed all the parts separately: r309567 (with r309571 to fix a test), r309678 and r310477.

Updated all the comments you made and rebased.

In D34948#798332, @bader wrote:

LGTM.
Do you have another built-in in mind which can use this diagnostic message?
If so, it would make sense to re-use it in the same patch.

Making that change, going to re-run the tests and then I'll commit.

Committed this as r304784.

LGTM!

Thanks for that Craig, I had planned on looking at that today so thanks for doing it!

Constant fold inside SimplifyGEPInst instead.

After talking to Matthias and Matt, this seems like the wrong solution.

In D31194#725318, @craig.topper wrote:

I wonder if you should just call ConstantFoldInstOperands where we currently call ConstantExpr::getElementPtr? I think that will try to constant fold it before falling back to ConstantExpr::getElementPtr.

Ping!

Ping!

In D31194#713113, @majnemer wrote:

Do you have insight into how this differs from what we already do? My understanding is that each of the simplify routines immediately try to constant fold.

Moved the getLimitedValue call into the helper function.

Created a helper function with some comments.

Added a test case using the 'Big' number in the unit test.
Added a test case for a 7-bit number.
Added cases where the previous code would assert due to RHS being 0.

Updated the code to use 'urem' to get the remainder, rather than calling getZExtValue.

Committed in r288332. Thanks Yaron!

Pipe types cannot be merged by ASTContext::mergeTypes.

Committed as r287343.

Committed as r287100.

Fixed a latent infinite loop bug in 'getImageAccess', it was dereferencing Attrs, instead of Next.

Two minor comments, but otherwise LGTM!

Committed r278234.

LGTM!

LGTM! (Apart from those tiny changes)

Drive-by comments.

Thanks Alp!

Thanks guys, r209006!

Fixed the comment! Ok now?

Thanks Chandler, r208705!

Thanks Tim! r207917

LGTM.

LGTM.