[PATCH] D72878: [X86][BranchAlign] Suppress branch alignment for {,_}__tls_get_addr

[Fangrui Song via Phabricator via llvm-commits]

MaskRay marked 2 inline comments as done.
MaskRay added inline comments.


================
Comment at: llvm/test/CodeGen/X86/align-branch-boundary-suppressions-tls.ll:21
+; CHECK: #noautopadding
+; 32: leal ld at TLSLDM(%ebx), %eax
+; 32: calll ___tls_get_addr at PLT
----------------
LuoYuanke wrote:
> Obviously disable auto padding for TLS is safety, but I am curious what's the problem if we don't  disable it? In our previous patch, we check if there is any variant symbol in needAlignInst(...).
This only affects x86-64 General Dynamic TLS model, which emits some prefixes.

```
(DATA16_PREFIX)
(LEA64r)
<--- Diff 21 can place cs (0x2e) here and break the General Dynamic TLS code sequence --->
<MCInst 851> (DATA16_PREFIX)
<MCInst 851> (DATA16_PREFIX)
<MCInst 2450> (REX64_PREFIX)
<MCInst 602 <MCOperand Expr:(__tls_get_addr at PLT)>> CALL64pcrel32
```

D72225 can prepend another prefix and cause a failure.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D72878/new/

https://reviews.llvm.org/D72878