ping...

In D145032#4160993, @RKSimon wrote:

Out of curiosity - what motivated this change? Do you have dependent patches you're developing or did a linter recommend it?

rebase and address comment

In D132439#3745305, @RKSimon wrote:

Thanks - please can you update the title and summary

address comment

In D132349#3738804, @sdesmalen wrote:

Hi @Jimerlife is there a specific reason why you're refactoring this? This patch may reduce the number of lines of code, but it is arguable whether the new version is more readable or better. Code-refactoring creates merge-problems for people needing to maintain downstream branches, so unless there is a clear benefit to the code it is preferable to keep this to a minimum.

In D132206#3734342, @dmgreen wrote:

XOR will be 0 or val depending on if there are an even number of vector elements.

Can you add some tests for both cases?

In D132206#3734343, @craig.topper wrote:

Do you have real world examples of C code where this occurs?

Would this be better to do in InstCombine?

Where are the test deltas?

rebase test

If this test is OK, could someone help me to accept this test patch?

In D132115#3731453, @RKSimon wrote:

test case?

In D132107#3731229, @sdesmalen wrote:

Can you update the commit message to be a little more descriptive, e.g. [DAGCombiner][NFC] Tidy up unnecessary brackets in visitADD ?

address comment

address comment

update and add const keyword

Only use getConstantOperandVal(Num) API instead of cast<ConstantSDNode>(N->getOperand(Num))->getZExtValue(). If We already have Idx, still keeping origin cast form. Thanks your suggestions @craig.topper @RKSimon.

address comment

address comment

rebase to use new ISD name

Move VP_SITOFP and VP_UITOFP to line 2728

ping...

rebase main

In D128778#3637651, @jrtc27 wrote:

From https://llvm.org/docs/DeveloperPolicy.html:

• The commit message for the reverting commit should explain why patch is being reverted.