[LV] Add tests with canonical widen IV, reductions in different order.

Add missing test coverage for re-using the resume value from the main
vector loop for the canonical IV in the epilogue.

Author: fhahn

llvm/test/Transforms/LoopVectorize/epilog-vectorization-reductions.ll

@@ -477,3 +477,192 @@ for.cond:
 for.end:
   ret i32 %sub
 }
+
+define i64 @test_reduction_with_widen_induction_order_1(ptr %A, i64 %N) {
+; CHECK-LABEL: @test_reduction_with_widen_induction_order_1(
+; CHECK-NEXT:  iter.check:
+; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 4
+; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[VEC_EPILOG_SCALAR_PH:%.*]], label [[VECTOR_MAIN_LOOP_ITER_CHECK:%.*]]
+; CHECK:       vector.main.loop.iter.check:
+; CHECK-NEXT:    [[MIN_ITERS_CHECK1:%.*]] = icmp ult i64 [[N]], 4
+; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK1]], label [[VEC_EPILOG_PH:%.*]], label [[VECTOR_PH:%.*]]
+; CHECK:       vector.ph:
+; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], 4
+; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
+; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
+; CHECK:       vector.body:
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <4 x i64> [ <i64 0, i64 1, i64 2, i64 3>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP1:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[TMP0:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], i64 [[INDEX]]
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP0]], align 4
+; CHECK-NEXT:    [[TMP1]] = add <4 x i64> [[VEC_PHI]], [[WIDE_LOAD]]
+; CHECK-NEXT:    store <4 x i64> [[VEC_IND]], ptr [[TMP0]], align 4
+; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
+; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <4 x i64> [[VEC_IND]], splat (i64 4)
+; CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP2]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP17:![0-9]+]]
+; CHECK:       middle.block:
+; CHECK-NEXT:    [[TMP3:%.*]] = call i64 @llvm.vector.reduce.add.v4i64(<4 x i64> [[TMP1]])
+; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[CMP_N]], label [[EXIT:%.*]], label [[VEC_EPILOG_ITER_CHECK:%.*]]
+; CHECK:       vec.epilog.iter.check:
+; CHECK-NEXT:    [[N_VEC_REMAINING:%.*]] = sub i64 [[N]], [[N_VEC]]
+; CHECK-NEXT:    [[MIN_EPILOG_ITERS_CHECK:%.*]] = icmp ult i64 [[N_VEC_REMAINING]], 4
+; CHECK-NEXT:    br i1 [[MIN_EPILOG_ITERS_CHECK]], label [[VEC_EPILOG_SCALAR_PH]], label [[VEC_EPILOG_PH]]
+; CHECK:       vec.epilog.ph:
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[VECTOR_MAIN_LOOP_ITER_CHECK]] ]
+; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i64 [ [[TMP3]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[VECTOR_MAIN_LOOP_ITER_CHECK]] ]
+; CHECK-NEXT:    [[N_MOD_VF2:%.*]] = urem i64 [[N]], 4
+; CHECK-NEXT:    [[N_VEC3:%.*]] = sub i64 [[N]], [[N_MOD_VF2]]
+; CHECK-NEXT:    [[TMP4:%.*]] = insertelement <4 x i64> zeroinitializer, i64 [[BC_MERGE_RDX]], i32 0
+; CHECK-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i64> poison, i64 [[BC_RESUME_VAL]], i64 0
+; CHECK-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i64> [[BROADCAST_SPLATINSERT]], <4 x i64> poison, <4 x i32> zeroinitializer
+; CHECK-NEXT:    [[INDUCTION:%.*]] = add <4 x i64> [[BROADCAST_SPLAT]], <i64 0, i64 1, i64 2, i64 3>
+; CHECK-NEXT:    br label [[VEC_EPILOG_VECTOR_BODY:%.*]]
+; CHECK:       vec.epilog.vector.body:
+; CHECK-NEXT:    [[INDEX4:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT8:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VEC_IND5:%.*]] = phi <4 x i64> [ [[INDUCTION]], [[VEC_EPILOG_PH]] ], [ [[VEC_IND_NEXT9:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VEC_PHI6:%.*]] = phi <4 x i64> [ [[TMP4]], [[VEC_EPILOG_PH]] ], [ [[TMP6:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
+; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[INDEX4]]
+; CHECK-NEXT:    [[WIDE_LOAD7:%.*]] = load <4 x i64>, ptr [[TMP5]], align 4
+; CHECK-NEXT:    [[TMP6]] = add <4 x i64> [[VEC_PHI6]], [[WIDE_LOAD7]]
+; CHECK-NEXT:    store <4 x i64> [[VEC_IND5]], ptr [[TMP5]], align 4
+; CHECK-NEXT:    [[INDEX_NEXT8]] = add nuw i64 [[INDEX4]], 4
+; CHECK-NEXT:    [[VEC_IND_NEXT9]] = add <4 x i64> [[VEC_IND5]], splat (i64 4)
+; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT8]], [[N_VEC3]]
+; CHECK-NEXT:    br i1 [[TMP7]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP18:![0-9]+]]
+; CHECK:       vec.epilog.middle.block:
+; CHECK-NEXT:    [[TMP8:%.*]] = call i64 @llvm.vector.reduce.add.v4i64(<4 x i64> [[TMP6]])
+; CHECK-NEXT:    [[CMP_N10:%.*]] = icmp eq i64 [[N]], [[N_VEC3]]
+; CHECK-NEXT:    br i1 [[CMP_N10]], label [[EXIT]], label [[VEC_EPILOG_SCALAR_PH]]
+; CHECK:       vec.epilog.scalar.ph:
+; CHECK-NEXT:    [[BC_RESUME_VAL11:%.*]] = phi i64 [ [[N_VEC3]], [[VEC_EPILOG_MIDDLE_BLOCK]] ], [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[ITER_CHECK:%.*]] ]
+; CHECK-NEXT:    [[BC_MERGE_RDX12:%.*]] = phi i64 [ [[TMP8]], [[VEC_EPILOG_MIDDLE_BLOCK]] ], [ [[TMP3]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[ITER_CHECK]] ]
+; CHECK-NEXT:    br label [[LOOP:%.*]]
+; CHECK:       loop:
+; CHECK-NEXT:    [[IV_1:%.*]] = phi i64 [ [[BC_RESUME_VAL11]], [[VEC_EPILOG_SCALAR_PH]] ], [ [[IV_1_NEXT:%.*]], [[LOOP]] ]
+; CHECK-NEXT:    [[RED:%.*]] = phi i64 [ [[BC_MERGE_RDX12]], [[VEC_EPILOG_SCALAR_PH]] ], [ [[RED_NEXT:%.*]], [[LOOP]] ]
+; CHECK-NEXT:    [[GEP_A:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[IV_1]]
+; CHECK-NEXT:    [[L:%.*]] = load i64, ptr [[GEP_A]], align 4
+; CHECK-NEXT:    [[RED_NEXT]] = add i64 [[RED]], [[L]]
+; CHECK-NEXT:    store i64 [[IV_1]], ptr [[GEP_A]], align 4
+; CHECK-NEXT:    [[IV_1_NEXT]] = add nuw nsw i64 [[IV_1]], 1
+; CHECK-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[IV_1_NEXT]], [[N]]
+; CHECK-NEXT:    br i1 [[EXITCOND]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP19:![0-9]+]]
+; CHECK:       exit:
+; CHECK-NEXT:    [[RED_NEXT_LCSSA:%.*]] = phi i64 [ [[RED_NEXT]], [[LOOP]] ], [ [[TMP3]], [[MIDDLE_BLOCK]] ], [ [[TMP8]], [[VEC_EPILOG_MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    ret i64 [[RED_NEXT_LCSSA]]
+;
+entry:
+  br label %loop
+
+loop:
+  %iv.1 = phi i64 [ 0, %entry ], [ %iv.1.next, %loop ]
+  %red = phi i64 [ 0, %entry ], [ %red.next, %loop ]
+  %gep.A = getelementptr inbounds i64, ptr %A, i64 %iv.1
+  %l = load i64, ptr %gep.A
+  %red.next = add i64 %red, %l
+  store i64 %iv.1, ptr %gep.A, align 4
+  %iv.1.next = add nuw nsw i64 %iv.1, 1
+  %exitcond = icmp eq i64 %iv.1.next, %N
+  br i1 %exitcond, label %exit, label %loop
+
+exit:
+  ret i64 %red.next
+}
+
+; Same as @test_reduction_with_widen_induction_order_1, but with phi order flipped.
+define i64 @test_reduction_with_widen_induction_order_2(ptr %A, i64 %N) {
+; CHECK-LABEL: @test_reduction_with_widen_induction_order_2(
+; CHECK-NEXT:  iter.check:
+; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 4
+; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[VEC_EPILOG_SCALAR_PH:%.*]], label [[VECTOR_MAIN_LOOP_ITER_CHECK:%.*]]
+; CHECK:       vector.main.loop.iter.check:
+; CHECK-NEXT:    [[MIN_ITERS_CHECK1:%.*]] = icmp ult i64 [[N]], 4
+; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK1]], label [[VEC_EPILOG_PH:%.*]], label [[VECTOR_PH:%.*]]
+; CHECK:       vector.ph:
+; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], 4
+; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
+; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
+; CHECK:       vector.body:
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP1:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <4 x i64> [ <i64 0, i64 1, i64 2, i64 3>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[TMP0:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], i64 [[INDEX]]
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP0]], align 4
+; CHECK-NEXT:    [[TMP1]] = add <4 x i64> [[VEC_PHI]], [[WIDE_LOAD]]
+; CHECK-NEXT:    store <4 x i64> [[VEC_IND]], ptr [[TMP0]], align 4
+; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
+; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <4 x i64> [[VEC_IND]], splat (i64 4)
+; CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP2]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP20:![0-9]+]]
+; CHECK:       middle.block:
+; CHECK-NEXT:    [[TMP3:%.*]] = call i64 @llvm.vector.reduce.add.v4i64(<4 x i64> [[TMP1]])
+; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[CMP_N]], label [[EXIT:%.*]], label [[VEC_EPILOG_ITER_CHECK:%.*]]
+; CHECK:       vec.epilog.iter.check:
+; CHECK-NEXT:    [[N_VEC_REMAINING:%.*]] = sub i64 [[N]], [[N_VEC]]
+; CHECK-NEXT:    [[MIN_EPILOG_ITERS_CHECK:%.*]] = icmp ult i64 [[N_VEC_REMAINING]], 4
+; CHECK-NEXT:    br i1 [[MIN_EPILOG_ITERS_CHECK]], label [[VEC_EPILOG_SCALAR_PH]], label [[VEC_EPILOG_PH]]
+; CHECK:       vec.epilog.ph:
+; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i64 [ [[TMP3]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[VECTOR_MAIN_LOOP_ITER_CHECK]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[VECTOR_MAIN_LOOP_ITER_CHECK]] ]
+; CHECK-NEXT:    [[N_MOD_VF2:%.*]] = urem i64 [[N]], 4
+; CHECK-NEXT:    [[N_VEC3:%.*]] = sub i64 [[N]], [[N_MOD_VF2]]
+; CHECK-NEXT:    [[TMP4:%.*]] = insertelement <4 x i64> zeroinitializer, i64 [[BC_MERGE_RDX]], i32 0
+; CHECK-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i64> poison, i64 [[BC_RESUME_VAL]], i64 0
+; CHECK-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i64> [[BROADCAST_SPLATINSERT]], <4 x i64> poison, <4 x i32> zeroinitializer
+; CHECK-NEXT:    [[INDUCTION:%.*]] = add <4 x i64> [[BROADCAST_SPLAT]], <i64 0, i64 1, i64 2, i64 3>
+; CHECK-NEXT:    br label [[VEC_EPILOG_VECTOR_BODY:%.*]]
+; CHECK:       vec.epilog.vector.body:
+; CHECK-NEXT:    [[INDEX4:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT8:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VEC_PHI5:%.*]] = phi <4 x i64> [ [[TMP4]], [[VEC_EPILOG_PH]] ], [ [[TMP6:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VEC_IND6:%.*]] = phi <4 x i64> [ [[INDUCTION]], [[VEC_EPILOG_PH]] ], [ [[VEC_IND_NEXT9:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
+; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[INDEX4]]
+; CHECK-NEXT:    [[WIDE_LOAD7:%.*]] = load <4 x i64>, ptr [[TMP5]], align 4
+; CHECK-NEXT:    [[TMP6]] = add <4 x i64> [[VEC_PHI5]], [[WIDE_LOAD7]]
+; CHECK-NEXT:    store <4 x i64> [[VEC_IND6]], ptr [[TMP5]], align 4
+; CHECK-NEXT:    [[INDEX_NEXT8]] = add nuw i64 [[INDEX4]], 4
+; CHECK-NEXT:    [[VEC_IND_NEXT9]] = add <4 x i64> [[VEC_IND6]], splat (i64 4)
+; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT8]], [[N_VEC3]]
+; CHECK-NEXT:    br i1 [[TMP7]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP21:![0-9]+]]
+; CHECK:       vec.epilog.middle.block:
+; CHECK-NEXT:    [[TMP8:%.*]] = call i64 @llvm.vector.reduce.add.v4i64(<4 x i64> [[TMP6]])
+; CHECK-NEXT:    [[CMP_N10:%.*]] = icmp eq i64 [[N]], [[N_VEC3]]
+; CHECK-NEXT:    br i1 [[CMP_N10]], label [[EXIT]], label [[VEC_EPILOG_SCALAR_PH]]
+; CHECK:       vec.epilog.scalar.ph:
+; CHECK-NEXT:    [[BC_MERGE_RDX11:%.*]] = phi i64 [ [[TMP8]], [[VEC_EPILOG_MIDDLE_BLOCK]] ], [ [[TMP3]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[ITER_CHECK:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL12:%.*]] = phi i64 [ [[N_VEC3]], [[VEC_EPILOG_MIDDLE_BLOCK]] ], [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[ITER_CHECK]] ]
+; CHECK-NEXT:    br label [[LOOP:%.*]]
+; CHECK:       loop:
+; CHECK-NEXT:    [[RED:%.*]] = phi i64 [ [[BC_MERGE_RDX11]], [[VEC_EPILOG_SCALAR_PH]] ], [ [[RED_NEXT:%.*]], [[LOOP]] ]
+; CHECK-NEXT:    [[IV_1:%.*]] = phi i64 [ [[BC_RESUME_VAL12]], [[VEC_EPILOG_SCALAR_PH]] ], [ [[IV_1_NEXT:%.*]], [[LOOP]] ]
+; CHECK-NEXT:    [[GEP_A:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[IV_1]]
+; CHECK-NEXT:    [[L:%.*]] = load i64, ptr [[GEP_A]], align 4
+; CHECK-NEXT:    [[RED_NEXT]] = add i64 [[RED]], [[L]]
+; CHECK-NEXT:    store i64 [[IV_1]], ptr [[GEP_A]], align 4
+; CHECK-NEXT:    [[IV_1_NEXT]] = add nuw nsw i64 [[IV_1]], 1
+; CHECK-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[IV_1_NEXT]], [[N]]
+; CHECK-NEXT:    br i1 [[EXITCOND]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP22:![0-9]+]]
+; CHECK:       exit:
+; CHECK-NEXT:    [[RED_NEXT_LCSSA:%.*]] = phi i64 [ [[RED_NEXT]], [[LOOP]] ], [ [[TMP3]], [[MIDDLE_BLOCK]] ], [ [[TMP8]], [[VEC_EPILOG_MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    ret i64 [[RED_NEXT_LCSSA]]
+;
+entry:
+  br label %loop
+
+loop:
+  %red = phi i64 [ 0, %entry ], [ %red.next, %loop ]
+  %iv.1 = phi i64 [ 0, %entry ], [ %iv.1.next, %loop ]
+  %gep.A = getelementptr inbounds i64, ptr %A, i64 %iv.1
+  %l = load i64, ptr %gep.A
+  %red.next = add i64 %red, %l
+  store i64 %iv.1, ptr %gep.A, align 4
+  %iv.1.next = add nuw nsw i64 %iv.1, 1
+  %exitcond = icmp eq i64 %iv.1.next, %N
+  br i1 %exitcond, label %exit, label %loop
+
+exit:
+  ret i64 %red.next
+}