Implemented commented posted by Aaron (Thanks)

Implemented comments posted until 02/27 (Thanks Aaron and Erich)

Implemented comments posted until 02/26 (Thanks Aaron)

I added a comment ignoring nocf_check attribute in case -fcf-protection is not set.
Now LLVM is identical to GCC.

Implemented comments posted until 02/22 (Thanks Aaron).

Implemented comments posted until 02/15 (Thanks Aaron)

Implemented comments posted until 2/14 (Thanks Aaron and Craig)

ping

ping

Implemented comments posted until 01/30 (Thanks Craig)

Implemented comments posted until 01/28 (Thanks Craig)

I will add "notrack" to X86 ASM Parser (and also add symbolic test)

Thanks Craig for your comments, i made the following changes:

• I print notrack prefix instead of ds prefix
• I used same mechanism as lock/rep and created unified SDNode for the prefix and the jmp/call

In D41879#978046, @craig.topper wrote:

Should we be printing the DS_PREFIX as "notrack" like gcc?

I added encoding test in the following review:
https://reviews.llvm.org/D40223

Implemented comments posted until 01/04 (Thanks Craig)

Reverted clang-format whitespaces updates

Implemented comments posted until 12/24 (Thanks Craig)

Implemented comments posted until 12/24 (Thanks Craig)
Moved cf-protection attributes from function attributes to module attributes.

In D40478#962348, @craig.topper wrote:

Are we sure we want a different command line option name from gcc? From our internal conversations with the gcc folks I thought they were suggesting that -fcf-protection could imply a software mechanism if a hardware mechanism was not available thorugh -mibt or -march?

Should we emit an error to the user if -mibt isn't available? We should be able to add virtual methods on TargetInfo that X86 can customize to check for ibt and shstk.

Can you provide more information about the miscompile on MSVC? I think we should do more to understand that, this sounds like it could be a time bomb waiting to fail somewhere else.

ping

ping

Updated attribute name

Implemented comments posted until 12/19 (Thanks Craig)

-mibt is currently in discussions with other compilers, any change will be uploaded to a different review.
If you have more comments i will appreciate it.

Implemented all comments posted until 12/14 (Thanks Peter)

Implemented all comments posted until 12/14 (Thanks Peter).

Implemented some of the comments posted until 12/12 (Thanks Peter)

Implemented some of the comments posted until 12/12 (Thanks Peter)

Implemented comments posted until 11/30 (Thank you Craig and Peter)

Implemented comments posted till 11/29 (Thanks Craig)

Implemented comments posted until 11/29 (Thanks Craig)

Implemented comments posted until 11/21 (Thanks Craig)

Implemented comments posted until 11/21 (Thanks Craig)

Removed redundant FeatureCET flag.

Implemented comments posted until 11/20 (Thanks Craig and Erich).

Implemented comments posted until 11/19 (Thanks Craig)

In D37885#893264, @craig.topper wrote:

I imagine it would be possible, but we'd also need a new target hook to indicate that MOV32rr is coalescable. It just seemed easier to use the existing stuff rather than implement a second way to do it.

Is it possible to change InstrEmitter/peephole pass to coalesce even in cases of SUBREG_TO_REG?

LGTM

LGTM

LGTM

Implemented comments posted until 05/25 (Thanks Simon and Craig)

In D33169#763254, @RKSimon wrote:

Any idea why phabricator is showing so many unchanged lines from X86InstrInfo.cpp? Have you changed the line endings or something? They aren't appearing in the the downloaded diff FWIW.

I will appreciate any additional comments.
Please help me finish the review.

Implemented comments posted until 05/20 (Thanks Craig)

Updated the tests (Thanks Simon)

Removed TB_NO_REVERSE flag (Thanks Simon)

Implemented comments posted until 05/16 (Thanks again Simon and Craig)

In D33169#755488, @RKSimon wrote:

A possible addition would be to custom lower i8/i16 vectors with a trunc(popcnt(zext))) pattern.

I agree with you, Will it be OK to create a separate patch for it?

Implemented comments posted until 05/16 (Thanks Craig)

Reverted clang-format for lib/Target/X86/X86InstrInfo.cpp and removed NoVLX predicate (Thanks Simon)

Implemented comments posted until 05/15 (Thanks Craig and Simon)

Please notice that clang-format reformatted some lists that I modified in the file lib/Target/X86/X86InstrInfo.cpp.
It caused major cosmetic changes which introduce many diffs in that file.

Implemented comments posted until 05/15 (Thanks Simon and Craig)

Removed files properties

Added the fix to all SystemV ELF based operating systems (Thanks Rafael and David)