================
@@ -524,9 +524,7 @@ define i64 @utest_f64i64(double %x) {
; CHECK-NEXT: local.get 0
; CHECK-NEXT: call __fixunsdfti
; CHECK-NEXT: local.get 1
-; CHECK-NEXT: i32.const 8
-; CHECK-NEXT: i32.add
-; CHECK-NEXT: i64.load 0
+; CHECK-NEXT: i64.load 8
----------------
workingjubilee wrote:
hmm. in general, more folds are valid with naturally-aligned types, because if
you see two random 4-byte-aligned types, using an 8-byte-aligned load is only
valid if they're on the same boundary. LLVM tends to allow these sorts of
things that are only relevant for some machines to creep into its global
reasoning about optimization, and not with no reason: it's always safer to do
the more conservative optimization that is valid on all machines.
so wasm's model might suggest it shouldn't be affected by this, but in practice
LLVM reasons substantially about this. so I am not surprised a machine model
that _should_ care less still optimizes better.
https://github.com/llvm/llvm-project/pull/119204
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
