[PATCH] D11393: [X86] Allow X86::COND_NE_OR_P and X86::COND_NP_OR_E to be reversed.

[Cong Hou via llvm-commits]

congh added a comment.

Thanks for the review, Chandler!


================
Comment at: test/CodeGen/X86/block-placement.ll:470
@@ -468,2 +469,3 @@
+; fall-through.
 ; CHECK: fpcmp_unanalyzable_branch
 ; CHECK: %entry
----------------
chandlerc wrote:
> Here and elsewhere, when updating a test, it would be good to convert it to use CHECK-LABEL at least, and then in the specific test cases, actually write narrow checks. For cases where we have very microscopic functions testing single instruction sequences, using update_llc_test_checks.py is incredibly useful. You can also look at the style of tests in generates and generate comparably structured checks for more complex test cases.
OK. I have update this test according to the results from running update_llc_test_checks.py.

================
Comment at: test/CodeGen/X86/fp-une-cmp.ll:79-83
@@ -79,3 +78,7 @@
 
-!1 = !{!"branch_weights", i32 1, i32 1000}
+; Test if the negation of the non-equality check between floating points are
+; translated to jnp followed by jne.
 
+; CHECK: jne
+; CHECK-NEXT: jnp
+define void @foo(float %f) {
----------------
spatel wrote:
> chandlerc wrote:
> > These checks don't really make sense to me. Why are they above the function, but the function has a CHECK-LABEL adn seemingly similar checks?
> Those checks are the old lines from before the update check script was run. They should be deleted.
Yes. I have deleted them.


http://reviews.llvm.org/D11393