[llvm] llvm: honor VFABI mappings without builtin suppression (PR #193884)

[Graham Hunter via llvm-commits]

================
@@ -0,0 +1,41 @@
+; RUN: opt -passes=loop-vectorize -force-vector-width=2 -S < %s | FileCheck %s
+
+; NOTE: This is a focused reproducer for OpenMP declare-simd style VFABI mapping.
+; The math libcall is marked no-builtin (as clang emits for declare-simd) and
+; memory(none), matching the readnone effects clang gives these libcalls under
+; fast-math. LAA must honor the VFABI mapping and vectorize the loop even though
+; the scalar callee is no-builtin: a readnone call carries no dependence edge.
+
+define void @test_vector_abi(ptr noalias %x, ptr %c) #0 {
+entry:
+  br label %loop
+
+loop:
+  %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
+  %src = getelementptr inbounds double, ptr %c, i64 %iv
+  %v = load double, ptr %src, align 8
+  %r = tail call double @acosh(double noundef %v) #2
+  %dst = getelementptr inbounds double, ptr %x, i64 %iv
+  store double %r, ptr %dst, align 8
+  %iv.next = add nuw nsw i64 %iv, 1
+  %done = icmp eq i64 %iv.next, 1000
+  br i1 %done, label %exit, label %loop, !llvm.loop !0
+
+exit:
+  ret void
+}
+
+; CHECK-LABEL: @test_vector_abi(
----------------
huntergr-arm wrote:

I'd prefer to see the entire result, not just a couple of check lines. Please use the update_test_checks.py script to generate the check lines.

https://github.com/llvm/llvm-project/pull/193884