[PATCH] D102588: [UpdateTestChecks] Default --x86_scrub_rip to False

[Harald van Dijk via Phabricator via llvm-commits]

hvdijk added inline comments.


================
Comment at: llvm/test/CodeGen/X86/abi-isel.ll:1615
 ; LINUX-64-STATIC:       # %bb.0: # %entry
-; LINUX-64-STATIC-NEXT:    movl dsrc+{{.*}}(%rip), %eax
-; LINUX-64-STATIC-NEXT:    movl %eax, ddst+{{.*}}(%rip)
+; LINUX-64-STATIC-NEXT:    movl dsrc+64(%rip), %eax
+; LINUX-64-STATIC-NEXT:    movl %eax, ddst+64(%rip)
----------------
pengfei wrote:
> Not sure how many testing scenarios like this. But I think this is the origin of this knob. Can we change it to something like
> `SCRUB_X86_RIP_RE = re.compile(r'[.\d]+\(%rip\)')` instead?
The `LINUX-32-STATIC` case already tested for `movl dsrc+64, %eax`. It seems a bit silly to me to allow a fixed offset there but to not allow it in the `(%rip)` case, so I personally do like the test being updated this way.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D102588/new/

https://reviews.llvm.org/D102588