[llvm] SimplifyLibCalls: Use the correct address space when computing integer widths. (PR #118586)

[Eli Friedman via llvm-commits]

================
@@ -16,7 +16,7 @@ define arm_aapcscc void @test_simplify1() {
 
 
   call arm_aapcscc ptr @strcpy(ptr @a, ptr @hello)
-; CHECK: @llvm.memcpy.p0.p0.i32
+; CHECK: @llvm.memcpy.p0.p0.i64
----------------
efriedma-quic wrote:

The CreateMemCpy overload in question just calls getInt64().

I'm not sure it's much of a practical issue: backend legalization will truncate the integer to the appropriate size.  But maybe we should canonicalize somewhere.

https://github.com/llvm/llvm-project/pull/118586