Add conditional check to prevent the freezing of a load with freeze_bits metadata

@nikic I added this patch to a work in progress (WIP) "stack" of patches. D158352 and D158353 show what the intended progression looks like for mem2reg. I will add WIP SROA changes in the next few days.

@nikic Thanks for responding. I will get a "work in progress" patch up in the next three days.

Ping

Ping

Refactor AMDGPUPromoteAlloca to use getInitialValueOfAllocation.

Remove reliance on TLI objects where only alloca instructions are processed.

@nikic as per my GSoC project I am trying, under the guidance of @nlopes, a load attribute based approach to migrating uninitialized load to poison. The first attribute that I am working on is !freeze which effectively inserts a freeze poison if the load is uninitialized. Obviously this can only be done in
optimizations that allow instruction creation. In the future I am looking to add a load instruction parameter to getInitialValueOfAllocation to allow modification of the returned constant based on the allocation function/instruction and the presence of a load attribute.

Abandoning this patch as it will be superseded in functionality by D148216.

@jdoerfert thanks for the LGTM!

Regenerating test using D148216

@goldstein.w.n thanks for taking the time to review.

@goldstein.w.n thanks for the reply.

@nikic this patch is for #54856, which originated from @erikdesjardins.

@goldstein.w.n thanks for taking the time to review.

Add foldPowerOf2AndShiftedMask(...) and call from InstCombinerImpl::foldAndOrOfICmps(...)

Move tests from icmp-logical.ll to icmp-power2-and-icmp-shifted-mask.ll

Add postfix of "fail" to the names of tests that intentionally do not change

Add postfix of "fail" to the names of tests that intentionally do not change

Refactor scalar and vector tests to emit icmp from one location

Refactor scalar and vector tests to emit icmp from one location

Resolving feedback items identified by @goldstein.w.n

Add undef based splat vector tests

@RKSimon thanks for taking the time to review the patch!

@RKSimon thanks for taking the time to review the patch!

Thanks @RKSimon for reviewing! Can you please review the predecessor D143031?

@RKSimon Thanks for taking the time to review!

Rebase onto latest main.

@RKSimon I added Alive2 examples exercising llvm::decomposeBitTestICmp using splat/mask vectors for all supported comparisons.

Added Alive2 URLs to patch summary.

I have added Alive2 links to the summary as per @RKSimon request.

Rebase onto latest main.

Rebase onto latest main.

Rebase to latest main.

Rebase onto latest main.

Rebase onto latest main.

Add comments to each non nullptr conditional.

Rebase to latest main.

I am closing this ticket as we are working on alternative solutions to poison based load semantics.

I am closing this ticket as we are working on alternative solutions to poison based load semantics.

I am closing this ticket as we are working on alternative solutions to poison based load semantics.

ping

Update icmp-logical.ll test names as per recommendation in D143044

Remove sequential test names in favor of pattern details

Fix no change test case updated in D143044

Fix vector of two elements test case

Add support for vectors:

As per @RKSimon's feedback I have discovered a number of vector deficiencies in my implementation.

The Discourse proposal is available: RFC Load Instruction: Uninitialized Memory Semantics

As Nuno mentioned we are targeting the proposal for next week. I will update the ticket with the Discourse link once it becomes available.

Updating D139378: [clang][CodeGen] Add noundef metadata to load instructions (preliminary 4 or 4)

Updating D138983: [clang][CodeGen] Add noundef metadata to load instructions (preliminary 3 or 3)

Updating D134410: [clang][CodeGen] Add noundef metadata to load instructions (preliminary 1 or 3)

Updating D134410: [clang][CodeGen] Add noundef metadata to load instructions (preliminary 1 or 2)