Instead of using SDNodeFlags to store metadata information, this patch directly appends an extra operand for rounding mode during the DAG build phase based on Stats's suggestion. This patch currently implements strict fps for fadd, fsub, fmul, fma, and fsqrt. Thanks a lot for @andrew.w.kaylor and @rampitec comments for this!

Removed -DAG checks completely.

Remove duplicate check lines.

Address code reviews.

Ping.

Address code reivews.

Fixed format issue.

Ping.

Ping.

Ping.

Will create another separate ticket to fix the v_ffbl_sdwa instruction generation.

Ping.

Fix the issues that variables are not capitalized.

Address code reviews.

Upload a full diff.

Changes based on code review feedback.

Ping.

Just add a custom lowering ISD:CTTZ to ISD::CTTZ_ZERO_UNDEF

Ping.

If we don't want to change the original way of implementation, we may want to do a custom lowering for ISD::CTTZ at AMDGPU backend to ISD::CTTZ_ZERO_UNDE?

Address code reviews. Thanks!

In D36335#850574, @andrew.w.kaylor wrote:

Can you revert the white space changes in the places you aren't otherwise modifying? In general, you shouldn't make formatting changes outside of the parts of the file your patch is modifying. It complicates the version control blame process without adding a lot of benefit.

Also, your latest diffs seem not to have full file context (such as you get with the -x -U99999 switch with diff). This isn't important for the current review, but it is something to keep in mind going forward.

I really appreciate your work on this patch, and I hate to seem like I'm nit-picking a lot. I just want to make sure we do things correctly. Thanks!

Fixed a format issue.

Patch update after the patch [X86] Remove X86ISD::FMADD in favor ISD::FMA has been upstreamed.

In D36335#849535, @andrew.w.kaylor wrote:

Can you put this off until the patch that Craig submitted in D36983 either lands or gets rejected? If that change goes through, you should be able to remove your modifications to X86ISelDAGToDAG.cpp.

Address code reviews. Thanks a lot!

Ping.

Ping.

Remove extra period. Thanks!

Update LangRef.rst: put more accurate descriptions into the constrained.fma semantic section.

Update LangRef.rst based on comments.

Address code reviews.

Code changes based on Brian's comments.

Upload correct diff.

Address code reviews.

In D36335#832367, @b-sumner wrote:

An update to docs/LangRef.rst is needed.

Add missing lit tests.

Modified function name.

Ping.

Address code reviews.

Ping.

Upload correct diff file.

Fixed a mistake for align2_i8 lit test.

Address code reviews.

Address code reviews.

Address code reviews.

1. Address code reviews. Looks like adding "isDereferenceableAndAlignedPointer" is too strong to prevent expected code transformations.
2. Modify related LIT tests.

Hi, is it possible to upload a full diff?

Modified lit tests.

Address code reviews.

Address code reviews based on Tony's @t-tye comments

Ping.

Address code reviews.

Add gfx600 and gfx601 definitions.

Add new ISA version definitions & fix incorrect ISA versions.

In D33375#761170, @spatel wrote:

Sorry, but I still don't understand.

%inc = add nuw nsw i32 %myValue.0, 1
%cmp15 = icmp slt i32 %inc, 255

%myValue.0 will be 0 when we enter the loop. Eventually, it will increment up to 255, and %cmp15 will be false. How is that logically different than:
%cmp15 = icmp slt i32 %myValue.0, 254 ?

%inc is not always one more than %myValue.0?

In D33375#761110, @spatel wrote:

I don't understand why this change is needed. Can you explain how the test in this patch can miscompile?

Address code review.

1. Address code reviews.
2. Will create another patch for optimized searching of divergent terminator.

Ping.

Change function name.