23c2f2e6b2
This patch marks the induction increment of the main induction variable of the vector loop as NUW when not folding the tail. If the tail is not folded, we know that End - Start >= Step (either statically or through the minimum iteration checks). We also know that both Start % Step == 0 and End % Step == 0. We exit the vector loop if %IV + %Step == %End. Hence we must exit the loop before %IV + %Step unsigned overflows and we can mark the induction increment as NUW. This should make SCEV return more precise bounds for the created vector loops, used by later optimizations, like late unrolling. At the moment quite a few tests still need to be updated, but before doing so I'd like to get initial feedback to make sure I am not missing anything. Note that this could probably be further improved by using information from the original IV. Attempt of modeling of the assumption in Alive2: https://alive2.llvm.org/ce/z/H_DL_g Part of a set of fixes required for PR50412. Reviewed By: mkazantsev Differential Revision: https://reviews.llvm.org/D103255
125 lines
6.6 KiB
LLVM
125 lines
6.6 KiB
LLVM
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
|
|
; RUN: opt -enable-vplan-native-path -loop-vectorize -S %s | FileCheck %s
|
|
|
|
; Make sure phi nodes are generated correctly, even if the use list order of
|
|
; the predecessors in the scalar code does not match the order in the generated
|
|
; vector blocks.
|
|
|
|
; Test from PR45958.
|
|
|
|
define void @test([2000 x i32]* %src, i64 %n) {
|
|
; CHECK-LABEL: @test(
|
|
; CHECK-NEXT: entry:
|
|
; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 4
|
|
; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_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: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i64> poison, i64 [[N]], i32 0
|
|
; CHECK-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i64> [[BROADCAST_SPLATINSERT]], <4 x i64> poison, <4 x i32> zeroinitializer
|
|
; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]
|
|
; CHECK: vector.body:
|
|
; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[LOOP_1_LATCH5:%.*]] ]
|
|
; CHECK-NEXT: [[VEC_IND:%.*]] = phi <4 x i64> [ <i64 0, i64 1, i64 2, i64 3>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[LOOP_1_LATCH5]] ]
|
|
; CHECK-NEXT: br label [[LOOP_2_HEADER1:%.*]]
|
|
; CHECK: loop.2.header1:
|
|
; CHECK-NEXT: [[VEC_PHI:%.*]] = phi <4 x i64> [ zeroinitializer, [[VECTOR_BODY]] ], [ [[TMP5:%.*]], [[LOOP_2_LATCH4:%.*]] ]
|
|
; CHECK-NEXT: br label [[LOOP_32:%.*]]
|
|
; CHECK: loop.32:
|
|
; CHECK-NEXT: [[VEC_PHI3:%.*]] = phi <4 x i64> [ zeroinitializer, [[LOOP_2_HEADER1]] ], [ [[TMP2:%.*]], [[LOOP_32]] ]
|
|
; CHECK-NEXT: [[TMP0:%.*]] = getelementptr inbounds [2000 x i32], [2000 x i32]* [[SRC:%.*]], <4 x i64> [[VEC_IND]], <4 x i64> [[VEC_PHI3]]
|
|
; CHECK-NEXT: [[WIDE_MASKED_GATHER:%.*]] = call <4 x i32> @llvm.masked.gather.v4i32.v4p0i32(<4 x i32*> [[TMP0]], i32 4, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x i32> undef)
|
|
; CHECK-NEXT: [[TMP1:%.*]] = mul nsw <4 x i32> [[WIDE_MASKED_GATHER]], <i32 10, i32 10, i32 10, i32 10>
|
|
; CHECK-NEXT: call void @llvm.masked.scatter.v4i32.v4p0i32(<4 x i32> [[TMP1]], <4 x i32*> [[TMP0]], i32 4, <4 x i1> <i1 true, i1 true, i1 true, i1 true>)
|
|
; CHECK-NEXT: [[TMP2]] = add nuw nsw <4 x i64> [[VEC_PHI3]], <i64 1, i64 1, i64 1, i64 1>
|
|
; CHECK-NEXT: [[TMP3:%.*]] = icmp eq <4 x i64> [[TMP2]], [[BROADCAST_SPLAT]]
|
|
; CHECK-NEXT: [[TMP4:%.*]] = extractelement <4 x i1> [[TMP3]], i32 0
|
|
; CHECK-NEXT: br i1 [[TMP4]], label [[LOOP_2_LATCH4]], label [[LOOP_32]]
|
|
; CHECK: loop.2.latch4:
|
|
; CHECK-NEXT: [[TMP5]] = add nuw nsw <4 x i64> [[VEC_PHI]], <i64 1, i64 1, i64 1, i64 1>
|
|
; CHECK-NEXT: [[TMP6:%.*]] = icmp eq <4 x i64> [[TMP5]], [[BROADCAST_SPLAT]]
|
|
; CHECK-NEXT: [[TMP7:%.*]] = extractelement <4 x i1> [[TMP6]], i32 0
|
|
; CHECK-NEXT: br i1 [[TMP7]], label [[LOOP_1_LATCH5]], label [[LOOP_2_HEADER1]]
|
|
; CHECK: loop.1.latch5:
|
|
; CHECK-NEXT: [[TMP8:%.*]] = add nuw nsw <4 x i64> [[VEC_IND]], <i64 1, i64 1, i64 1, i64 1>
|
|
; CHECK-NEXT: [[TMP9:%.*]] = icmp eq <4 x i64> [[TMP8]], [[BROADCAST_SPLAT]]
|
|
; CHECK-NEXT: [[TMP10:%.*]] = extractelement <4 x i1> [[TMP9]], i32 0
|
|
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
|
|
; CHECK-NEXT: [[VEC_IND_NEXT]] = add <4 x i64> [[VEC_IND]], <i64 4, i64 4, i64 4, i64 4>
|
|
; CHECK-NEXT: [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
|
|
; CHECK-NEXT: br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], [[LOOP0:!llvm.loop !.*]]
|
|
; CHECK: middle.block:
|
|
; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
|
|
; CHECK-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
|
|
; CHECK: scalar.ph:
|
|
; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
|
|
; CHECK-NEXT: br label [[LOOP_1_HEADER:%.*]]
|
|
; CHECK: loop.1.header:
|
|
; CHECK-NEXT: [[IV_1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_1_NEXT:%.*]], [[LOOP_1_LATCH:%.*]] ]
|
|
; CHECK-NEXT: br label [[LOOP_2_HEADER:%.*]]
|
|
; CHECK: loop.2.header:
|
|
; CHECK-NEXT: [[IV_2:%.*]] = phi i64 [ 0, [[LOOP_1_HEADER]] ], [ [[IV_2_NEXT:%.*]], [[LOOP_2_LATCH:%.*]] ]
|
|
; CHECK-NEXT: br label [[LOOP_3:%.*]]
|
|
; CHECK: loop.3:
|
|
; CHECK-NEXT: [[IV_3:%.*]] = phi i64 [ 0, [[LOOP_2_HEADER]] ], [ [[IV_3_NEXT:%.*]], [[LOOP_3]] ]
|
|
; CHECK-NEXT: [[GEP_SRC:%.*]] = getelementptr inbounds [2000 x i32], [2000 x i32]* [[SRC]], i64 [[IV_1]], i64 [[IV_3]]
|
|
; CHECK-NEXT: [[L1:%.*]] = load i32, i32* [[GEP_SRC]], align 4
|
|
; CHECK-NEXT: [[MUL:%.*]] = mul nsw i32 [[L1]], 10
|
|
; CHECK-NEXT: store i32 [[MUL]], i32* [[GEP_SRC]], align 4
|
|
; CHECK-NEXT: [[IV_3_NEXT]] = add nuw nsw i64 [[IV_3]], 1
|
|
; CHECK-NEXT: [[EC_3:%.*]] = icmp eq i64 [[IV_3_NEXT]], [[N]]
|
|
; CHECK-NEXT: br i1 [[EC_3]], label [[LOOP_2_LATCH]], label [[LOOP_3]]
|
|
; CHECK: loop.2.latch:
|
|
; CHECK-NEXT: [[IV_2_NEXT]] = add nuw nsw i64 [[IV_2]], 1
|
|
; CHECK-NEXT: [[EC_2:%.*]] = icmp eq i64 [[IV_2_NEXT]], [[N]]
|
|
; CHECK-NEXT: br i1 [[EC_2]], label [[LOOP_1_LATCH]], label [[LOOP_2_HEADER]]
|
|
; CHECK: loop.1.latch:
|
|
; CHECK-NEXT: [[IV_1_NEXT]] = add nuw nsw i64 [[IV_1]], 1
|
|
; CHECK-NEXT: [[EC_1:%.*]] = icmp eq i64 [[IV_1_NEXT]], [[N]]
|
|
; CHECK-NEXT: br i1 [[EC_1]], label [[EXIT]], label [[LOOP_1_HEADER]], [[LOOP2:!llvm.loop !.*]]
|
|
; CHECK: exit:
|
|
; CHECK-NEXT: ret void
|
|
;
|
|
entry:
|
|
br label %loop.1.header
|
|
|
|
loop.1.header:
|
|
%iv.1 = phi i64 [ 0, %entry ], [ %iv.1.next, %loop.1.latch ]
|
|
br label %loop.2.header
|
|
|
|
loop.2.header:
|
|
%iv.2 = phi i64 [ 0, %loop.1.header ], [ %iv.2.next, %loop.2.latch ]
|
|
br label %loop.3
|
|
|
|
loop.3:
|
|
%iv.3 = phi i64 [ 0, %loop.2.header ], [ %iv.3.next, %loop.3 ]
|
|
%gep.src = getelementptr inbounds [2000 x i32], [2000 x i32]* %src, i64 %iv.1, i64 %iv.3
|
|
%l1 = load i32, i32* %gep.src, align 4
|
|
%mul = mul nsw i32 %l1, 10
|
|
store i32 %mul, i32* %gep.src, align 4
|
|
%iv.3.next = add nuw nsw i64 %iv.3, 1
|
|
%ec.3 = icmp eq i64 %iv.3.next, %n
|
|
br i1 %ec.3, label %loop.2.latch, label %loop.3
|
|
|
|
loop.2.latch:
|
|
%iv.2.next = add nuw nsw i64 %iv.2, 1
|
|
%ec.2 = icmp eq i64 %iv.2.next, %n
|
|
br i1 %ec.2, label %loop.1.latch, label %loop.2.header
|
|
|
|
loop.1.latch:
|
|
%iv.1.next = add nuw nsw i64 %iv.1, 1
|
|
%ec.1 = icmp eq i64 %iv.1.next, %n
|
|
br i1 %ec.1, label %exit, label %loop.1.header, !llvm.loop !0
|
|
|
|
exit: ; preds = %loop.1.latch
|
|
ret void
|
|
|
|
; uselistorder directives
|
|
uselistorder label %loop.3, { 1, 0 }
|
|
uselistorder label %loop.2.header, { 1, 0 }
|
|
}
|
|
|
|
!0 = distinct !{!0, !1, !2}
|
|
!1 = !{!"llvm.loop.vectorize.width", i32 4}
|
|
!2 = !{!"llvm.loop.vectorize.enable", i1 true}
|