[llvm] [LV] Add early-exit-with-store tests (PR #140899)

[David Sherwood via llvm-commits]

================
@@ -0,0 +1,383 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 4
+; REQUIRES: asserts
+; RUN: opt -S < %s -p loop-vectorize -debug-only=loop-vectorize -enable-early-exit-vectorization -force-vector-width=4 -disable-output 2>&1 | FileCheck %s
+
+define i64 @loop_contains_store(ptr %dest) {
+; CHECK-LABEL: LV: Checking a loop in 'loop_contains_store'
+; CHECK:       LV: Not vectorizing: Writes to memory unsupported in early exit loops
+entry:
+  %p1 = alloca [1024 x i8]
+  call void @init_mem(ptr %p1, i64 1024)
+  br label %loop
+
+loop:
+  %index = phi i64 [ %index.next, %loop.inc ], [ 3, %entry ]
+  %arrayidx = getelementptr inbounds i32, ptr %p1, i64 %index
+  %ld1 = load i32, ptr %arrayidx, align 1
+  %arrayidx2 = getelementptr inbounds i32, ptr %dest, i64 %index
+  store i32 %ld1, ptr %arrayidx2, align 4
+  %cmp = icmp eq i32 %ld1, 1
+  br i1 %cmp, label %loop.inc, label %loop.end
+
+loop.inc:
+  %index.next = add i64 %index, 1
+  %exitcond = icmp ne i64 %index.next, 67
+  br i1 %exitcond, label %loop, label %loop.end
+
+loop.end:
+  %retval = phi i64 [ %index, %loop ], [ 67, %loop.inc ]
+  ret i64 %retval
+}
+
+define void @loop_contains_store_condition_load_has_single_user(ptr dereferenceable(40) noalias %array, ptr align 2 dereferenceable(40) readonly %pred) {
+; CHECK-LABEL: LV: Checking a loop in 'loop_contains_store_condition_load_has_single_user'
+; CHECK:       LV: Not vectorizing: Writes to memory unsupported in early exit loops.
+entry:
+  br label %for.body
+
+for.body:
+  %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.inc ]
+  %st.addr = getelementptr inbounds nuw i16, ptr %array, i64 %iv
+  %data = load i16, ptr %st.addr, align 2
+  %inc = add nsw i16 %data, 1
+  store i16 %inc, ptr %st.addr, align 2
+  %ee.addr = getelementptr inbounds nuw i16, ptr %pred, i64 %iv
+  %ee.val = load i16, ptr %ee.addr, align 2
+  %ee.cond = icmp sgt i16 %ee.val, 500
+  br i1 %ee.cond, label %exit, label %for.inc
+
+for.inc:
+  %iv.next = add nuw nsw i64 %iv, 1
+  %counted.cond = icmp eq i64 %iv.next, 20
+  br i1 %counted.cond, label %exit, label %for.body
+
+exit:
+  ret void
+}
+
+define void @loop_contains_store_ee_condition_is_invariant(ptr dereferenceable(40) noalias %array, i16 %ee.val) {
+; CHECK-LABEL: LV: Checking a loop in 'loop_contains_store_ee_condition_is_invariant'
+; CHECK:       LV: Not vectorizing: Writes to memory unsupported in early exit loops.
+entry:
+  br label %for.body
+
+for.body:
+  %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.inc ]
+  %st.addr = getelementptr inbounds nuw i16, ptr %array, i64 %iv
+  %data = load i16, ptr %st.addr, align 2
+  %inc = add nsw i16 %data, 1
+  store i16 %inc, ptr %st.addr, align 2
+  %ee.cond = icmp sgt i16 %ee.val, 500
+  br i1 %ee.cond, label %exit, label %for.inc
+
+for.inc:
+  %iv.next = add nuw nsw i64 %iv, 1
+  %counted.cond = icmp eq i64 %iv.next, 20
+  br i1 %counted.cond, label %exit, label %for.body
+
+exit:
+  ret void
+}
+
+define void @loop_contains_store_fcmp_condition(ptr dereferenceable(40) noalias %array, ptr align 2 dereferenceable(40) readonly %pred) {
+; CHECK-LABEL: LV: Checking a loop in 'loop_contains_store_fcmp_condition'
+; CHECK:       LV: Not vectorizing: Writes to memory unsupported in early exit loops.
+entry:
+  br label %for.body
+
+for.body:
+  %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.inc ]
+  %st.addr = getelementptr inbounds nuw i16, ptr %array, i64 %iv
+  %data = load i16, ptr %st.addr, align 2
+  %inc = add nsw i16 %data, 1
+  store i16 %inc, ptr %st.addr, align 2
+  %ee.addr = getelementptr inbounds nuw half, ptr %pred, i64 %iv
+  %ee.val = load half, ptr %ee.addr, align 2
+  %ee.cond = fcmp ugt half %ee.val, 500.0
+  br i1 %ee.cond, label %exit, label %for.inc
+
+for.inc:
+  %iv.next = add nuw nsw i64 %iv, 1
+  %counted.cond = icmp eq i64 %iv.next, 20
+  br i1 %counted.cond, label %exit, label %for.body
+
+exit:
+  ret void
+}
+
+define void @loop_contains_store_safe_dependency(ptr dereferenceable(40) noalias %array, ptr align 2 dereferenceable(80) readonly %pred) {
+; CHECK-LABEL: LV: Checking a loop in 'loop_contains_store_safe_dependency'
+; CHECK:       LV: Not vectorizing: Writes to memory unsupported in early exit loops.
+entry:
+  %forward = getelementptr i16, ptr %pred, i64 -8
+  br label %for.body
+
+for.body:
+  %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.inc ]
+  %st.addr = getelementptr inbounds nuw i16, ptr %array, i64 %iv
+  %data = load i16, ptr %st.addr, align 2
+  %inc = add nsw i16 %data, 1
+  store i16 %inc, ptr %st.addr, align 2
+  %ee.addr = getelementptr inbounds nuw i16, ptr %pred, i64 %iv
+  %ee.val = load i16, ptr %ee.addr, align 2
+  %ee.cond = icmp sgt i16 %ee.val, 500
+  %some.addr = getelementptr inbounds nuw i16, ptr %forward, i64 %iv
----------------
david-arm wrote:

This cannot possibly ever get vectorised, right? Even if we can prove the loop is legal to vectorise with the store, the first 4 stores will be out of bounds of `%pred` due to

```
  %forward = getelementptr i16, ptr %pred, i64 -8
```

I'd expect the dereferenceability checks to fail here. Also, `%pred` is marked as `readonly`, which doesn't seem right.

You should be able to achieve the same effect by adding an offset to `%iv` before doing the load and have the stores trailing, right?

https://github.com/llvm/llvm-project/pull/140899