[llvm] [LV] Pre-commit test for #151664 (PR #151671)

[David Green via llvm-commits]

================
@@ -0,0 +1,62 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --check-globals none --filter-out-after "^scalar.ph" --version 5
+; RUN: opt -passes=loop-vectorize -mtriple=aarch64 -mattr=+sve -S \
+; RUN:   -debug-only=loop-vectorize %s 2>&1 | FileCheck %s
+
+; FIXME: Hoisted vector code should be costed with scalable cost.
+; In this example, `<vscale x 4 x float> @llvm.minimumnum` has an invalid cost,
+; and hence should not be produced by LoopVectorize. Unfortunately, what is
+; actually costed is `<4 x float> @llvm.minimumnum`.
+
+; CHECK: LV: Found an estimated cost of 3 for VF 1 For instruction:   %res = tail call float @llvm.minimumnum.f32(float 0.000000e+00, float 0.000000e+00)
+define void @cost_hoisted_vector_code(ptr %p) {
+; CHECK-LABEL: define void @cost_hoisted_vector_code(
+; CHECK-SAME: ptr [[P:%.*]]) #[[ATTR0:[0-9]+]] {
+; CHECK-NEXT:  [[ENTRY:.*:]]
+; CHECK-NEXT:    [[TMP0:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP1:%.*]] = mul nuw i64 [[TMP0]], 8
+; CHECK-NEXT:    br i1 false, label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
+; CHECK:       [[VECTOR_PH]]:
+; CHECK-NEXT:    [[TMP2:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 8
+; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 -1, [[TMP3]]
+; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 -1, [[N_MOD_VF]]
+; CHECK-NEXT:    [[TMP4:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP5:%.*]] = mul nuw i64 [[TMP4]], 8
+; CHECK-NEXT:    [[TMP6:%.*]] = add i64 1, [[N_VEC]]
+; CHECK-NEXT:    [[TMP7:%.*]] = call <vscale x 4 x float> @llvm.minimumnum.nxv4f32(<vscale x 4 x float> zeroinitializer, <vscale x 4 x float> zeroinitializer)
+; CHECK-NEXT:    br label %[[VECTOR_BODY:.*]]
+; CHECK:       [[VECTOR_BODY]]:
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr float, ptr [[P]], i64 [[INDEX]]
+; CHECK-NEXT:    [[TMP9:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP10:%.*]] = mul nuw i64 [[TMP9]], 4
+; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr float, ptr [[TMP8]], i64 [[TMP10]]
+; CHECK-NEXT:    store <vscale x 4 x float> [[TMP7]], ptr [[TMP8]], align 4
+; CHECK-NEXT:    store <vscale x 4 x float> [[TMP7]], ptr [[TMP11]], align 4
+; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
+; CHECK-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP12]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
+; CHECK:       [[MIDDLE_BLOCK]]:
+; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 -1, [[N_VEC]]
+; CHECK-NEXT:    br i1 [[CMP_N]], [[EXIT:label %.*]], label %[[SCALAR_PH]]
+; CHECK:       [[SCALAR_PH]]:
+;
+entry:
+  br label %loop
+
+loop:                                              ; preds = %loop, %entry
+  %iv = phi i64 [ 1, %entry ], [ %iv.next, %loop ]
+  %idx = phi i64 [ 0, %entry ], [ %idx.next, %loop ]
+  %res = tail call float @llvm.minimumnum.f32(float 0.0, float 0.0)
----------------
davemgreen wrote:

You could add another variant with @llvm.minimumnum.f32(float %arg1, float 0.0) that is still loop-invariant, so I believe might cause the same issue? It can't be simplified away like this one though.

https://github.com/llvm/llvm-project/pull/151671