[llvm] [LV] Vectorize selecting last IV of min/max element. (PR #141431)

[via llvm-commits]

================
@@ -0,0 +1,132 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --check-globals none --version 5
+; RUN: opt -passes=loop-vectorize -force-vector-width=4 -force-vector-interleave=1 -S %s | FileCheck %s
+
+define i64 @test_vectorize_select_fmaxnum_idx(ptr %src, i64 %n) {
+; CHECK-LABEL: define i64 @test_vectorize_select_fmaxnum_idx(
+; CHECK-SAME: ptr [[SRC:%.*]], i64 [[N:%.*]]) {
+; CHECK-NEXT:  [[ENTRY:.*]]:
+; CHECK-NEXT:    br label %[[LOOP:.*]]
+; CHECK:       [[LOOP]]:
+; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
+; CHECK-NEXT:    [[MIN_IDX:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[MIN_IDX_NEXT:%.*]], %[[LOOP]] ]
+; CHECK-NEXT:    [[MAX_VAL:%.*]] = phi float [ -3.000000e+00, %[[ENTRY]] ], [ [[MAX_VAL_NEXT:%.*]], %[[LOOP]] ]
+; CHECK-NEXT:    [[GEP:%.*]] = getelementptr float, ptr [[SRC]], i64 [[IV]]
+; CHECK-NEXT:    [[L:%.*]] = load float, ptr [[GEP]], align 4
+; CHECK-NEXT:    [[CMP:%.*]] = fcmp uge float [[MAX_VAL]], [[L]]
+; CHECK-NEXT:    [[MAX_VAL_NEXT]] = call float @llvm.maxnum.f32(float [[L]], float [[MAX_VAL]])
+; CHECK-NEXT:    [[MIN_IDX_NEXT]] = select i1 [[CMP]], i64 [[IV]], i64 [[MIN_IDX]]
+; CHECK-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
+; CHECK-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
+; CHECK-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT:.*]], label %[[LOOP]]
+; CHECK:       [[EXIT]]:
+; CHECK-NEXT:    [[RES:%.*]] = phi i64 [ [[MIN_IDX_NEXT]], %[[LOOP]] ]
+; CHECK-NEXT:    ret i64 [[RES]]
+;
+entry:
+  br label %loop
+
+loop:
+  %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
+  %min.idx = phi i64 [ 0, %entry ], [ %min.idx.next, %loop ]
+  %max.val = phi float [ -3.0000, %entry ], [ %max.val.next, %loop ]
+  %gep = getelementptr float, ptr %src, i64 %iv
+  %l = load float, ptr %gep
+  %cmp = fcmp uge float %max.val, %l
+  %max.val.next = call float @llvm.maxnum.f32(float %l, float %max.val)
+  %min.idx.next = select i1 %cmp, i64 %iv, i64 %min.idx
+  %iv.next = add nuw nsw i64 %iv, 1
+  %exitcond.not = icmp eq i64 %iv.next, %n
+  br i1 %exitcond.not, label %exit, label %loop
+
+exit:
+  %res = phi i64 [ %min.idx.next, %loop ]
+  ret i64 %res
+}
+
+define i64 @test_vectorize_select_fmax_idx_without_fmfs(ptr %src, i64 %n) {
+; CHECK-LABEL: define i64 @test_vectorize_select_fmax_idx_without_fmfs(
+; CHECK-SAME: ptr [[SRC:%.*]], i64 [[N:%.*]]) {
+; CHECK-NEXT:  [[ENTRY:.*]]:
+; CHECK-NEXT:    br label %[[LOOP:.*]]
+; CHECK:       [[LOOP]]:
+; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
+; CHECK-NEXT:    [[MIN_IDX:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[MIN_IDX_NEXT:%.*]], %[[LOOP]] ]
+; CHECK-NEXT:    [[MAX_VAL:%.*]] = phi float [ -3.000000e+00, %[[ENTRY]] ], [ [[MAX_VAL_NEXT:%.*]], %[[LOOP]] ]
+; CHECK-NEXT:    [[GEP:%.*]] = getelementptr float, ptr [[SRC]], i64 [[IV]]
+; CHECK-NEXT:    [[L:%.*]] = load float, ptr [[GEP]], align 4
+; CHECK-NEXT:    [[CMP:%.*]] = fcmp uge float [[MAX_VAL]], [[L]]
+; CHECK-NEXT:    [[MAX_CMP:%.*]] = fcmp ogt float [[L]], [[MAX_VAL]]
+; CHECK-NEXT:    [[MAX_VAL_NEXT]] = select i1 [[MAX_CMP]], float [[L]], float [[MAX_VAL]]
+; CHECK-NEXT:    [[MIN_IDX_NEXT]] = select i1 [[CMP]], i64 [[IV]], i64 [[MIN_IDX]]
+; CHECK-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
+; CHECK-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
+; CHECK-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT:.*]], label %[[LOOP]]
+; CHECK:       [[EXIT]]:
+; CHECK-NEXT:    [[RES:%.*]] = phi i64 [ [[MIN_IDX_NEXT]], %[[LOOP]] ]
+; CHECK-NEXT:    ret i64 [[RES]]
+;
+entry:
+  br label %loop
+
+loop:
+  %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
+  %min.idx = phi i64 [ 0, %entry ], [ %min.idx.next, %loop ]
+  %max.val = phi float [ -3.0000, %entry ], [ %max.val.next, %loop ]
+  %gep = getelementptr float, ptr %src, i64 %iv
+  %l = load float, ptr %gep
+  %cmp = fcmp uge float %max.val, %l
+  %max.cmp = fcmp ogt float %l, %max.val
----------------
ayalz wrote:

Are these intentionally distinct compares.

https://github.com/llvm/llvm-project/pull/141431