Avoid send EntryAttr as input parameter for EntryProps constructor.

Remove hlsl attr in DXILTranslateMetadata.

In D144309#4159431, @bob80905 wrote:

• ushort is now unsigned

In D144309#4156971, @bob80905 wrote:

add int support for max/min.

max/min does work on integers in HLSL.
And __builtin_elementwise_max/min support both integers and floats.

Fix format.

LGTM.
Just fix the newline at end of file for the tests.

Only keep debug related module flags which are used by DXIL.
Also let DXILPrepare pass preserve DXILResourceWrapper which is used in asm dump pass for test.

Update to fix issues mentioned in comments.

 void getAnalysisUsage(AnalysisUsage &AU) const override {
  AU.addPreserved<DXILResourceWrapper>();
}

to DXILPrepareModule resolved the issue.

Add log2 together?
They're in same category.

LGTM.
Will next PR add resource support?

rebase.

Move test to correct place.

Fix wrong history.

Move test to vector-transforms.mlir

ping.

add space for warning message.

Fix windows test fail.

rename to requiresValidation.

switch to getDefaultToolChain.

cast ToolChain to HLSLToolChain.

add DX_CONTAINER Driver type.
Code cleanup.

In D141705#4082935, @jhuber6 wrote:

I'm not overly familiar with HLSL or DirectX here. Most of the changes are purely mechanical, but I don't see anywhere we create the tool. Does that come later? Normally you'd test these with -ccc-print-bindings, -ccc-print-bindings, and -###.

Only create dxv job when 'dxv' exist.
Test ccc-print-bindings.
Code cleanup.

ping

Fix clang-format missed by arc lint

Do RemoveConstantIfCondition in ExpandIfCondition for constant if cond.

In D142286#4071304, @python3kgae wrote:

In D142286#4071254, @mehdi_amini wrote:

Could just inverse the condition and return failure()?

Not sure which path you're talking about :(

change assert(op.getIfCond() && "expected Op with IfCond"); to

if (!op.getIfCond())
  return failure();

or
add back the check ifCond as constant like
if (op.getIfCond().template getDefiningOp<arith::ConstantOp>())

return failure();

Or both?

In D142286#4071254, @mehdi_amini wrote:

Could just inverse the condition and return failure()?

In D142286#4071215, @clementval wrote:

Why creating an op if we know we don't need it?

add comments about the transform.

rebase to get stable_sort fix

Create https://reviews.llvm.org/D142110 for the non-determinism between windows and linux.

In D141987#4064561, @ftynse wrote:

Does one of the patterns actually has a higher benefit value? Or do they have the default value and one just happens to be selected instead of the other? In the latter case, it sounds like there is some non-determinism involved (e.g., something stored in a hashmap) that is exposed by platform differences.

Both VectorStoreToMemrefStoreLowering and VectorLoadStoreConversion<mlir::vector::StoreOp, mlir::vector::StoreOpAdaptor> are added to legalize vector.store.
On windows, VectorLoadStoreConversion has higher benefit, so it does not crash.
And on linux, VectorStoreToMemrefStoreLowering has higher benefit, and memref.store is not resisted, so it will crash.

Rebase and update test.

In D136031#4062756, @beanz wrote:

@python3kgae, this change introduced a bunch of warning spew because it is using an API that was deprecated shortly before the change merged. Can you please address this?