shylie/llvm-project
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
llvm-project/llvm/test/Transforms/LoopVectorize/reduction.ll
Florian Hahn 51afb10174
[LV] Create block in mask up-front if needed. (#76635) 
At the moment, block and edge masks are created on demand, which means
that they are inserted at the point where they are demanded and then
cached. It is possible that the mask for a block is looked up later at a
point that's not dominated by the point where the mask has been
inserted.

To avoid this, create masks up front on entry to the corresponding basic
block and leave it to VPlan simplification to remove unneeded masks.

Note that we need to create masks for all blocks, if any of the blocks
in the loop needs predication, as computing the mask of a block depends
on the masks of its predecessor.

Needed for #76090.

https://github.com/llvm/llvm-project/pull/76635
2024-01-09 10:50:08 +00:00

1353 lines
73 KiB
LLVM
Raw Blame History

; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 3
; RUN: opt < %s -passes=loop-vectorize,dce,instcombine -force-vector-interleave=1 -force-vector-width=4 -S | FileCheck %s
define i32 @reduction_sum(i32 %n, ptr %A, ptr %B) {
; CHECK-LABEL: define i32 @reduction_sum(
; CHECK-SAME: i32 [[N:%.*]], ptr [[A:%.*]], ptr [[B:%.*]]) {
; CHECK-NEXT: [[TMP1:%.*]] = icmp sgt i32 [[N]], 0
; CHECK-NEXT: br i1 [[TMP1]], label [[DOTLR_PH_PREHEADER:%.*]], label [[DOT_CRIT_EDGE:%.*]]
; CHECK: .lr.ph.preheader:
; CHECK-NEXT: [[TMP2:%.*]] = zext nneg i32 [[N]] to i64
; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[N]], 4
; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; CHECK: vector.ph:
; CHECK-NEXT: [[N_VEC:%.*]] = and i64 [[TMP2]], 2147483644
; 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 i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP7:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[VEC_IND:%.*]] = phi <4 x i32> [ <i32 0, i32 1, i32 2, i32 3>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDEX]]
; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP3]], align 4
; CHECK-NEXT: [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[INDEX]]
; CHECK-NEXT: [[WIDE_LOAD1:%.*]] = load <4 x i32>, ptr [[TMP4]], align 4
; CHECK-NEXT: [[TMP5:%.*]] = add <4 x i32> [[VEC_PHI]], [[VEC_IND]]
; CHECK-NEXT: [[TMP6:%.*]] = add <4 x i32> [[TMP5]], [[WIDE_LOAD]]
; CHECK-NEXT: [[TMP7]] = add <4 x i32> [[TMP6]], [[WIDE_LOAD1]]
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; CHECK-NEXT: [[VEC_IND_NEXT]] = add <4 x i32> [[VEC_IND]], <i32 4, i32 4, i32 4, i32 4>
; CHECK-NEXT: [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-NEXT: br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
; CHECK: middle.block:
; CHECK-NEXT: [[TMP9:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[TMP7]])
; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N_VEC]], [[TMP2]]
; CHECK-NEXT: br i1 [[CMP_N]], label [[DOT_CRIT_EDGE_LOOPEXIT:%.*]], label [[SCALAR_PH]]
; CHECK: scalar.ph:
; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[DOTLR_PH_PREHEADER]] ]
; CHECK-NEXT: [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP9]], [[MIDDLE_BLOCK]] ], [ 0, [[DOTLR_PH_PREHEADER]] ]
; CHECK-NEXT: br label [[DOTLR_PH:%.*]]
; CHECK: .lr.ph:
; CHECK-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[DOTLR_PH]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
; CHECK-NEXT: [[SUM_02:%.*]] = phi i32 [ [[TMP17:%.*]], [[DOTLR_PH]] ], [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ]
; CHECK-NEXT: [[TMP10:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDVARS_IV]]
; CHECK-NEXT: [[TMP11:%.*]] = load i32, ptr [[TMP10]], align 4
; CHECK-NEXT: [[TMP12:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[INDVARS_IV]]
; CHECK-NEXT: [[TMP13:%.*]] = load i32, ptr [[TMP12]], align 4
; CHECK-NEXT: [[TMP14:%.*]] = trunc i64 [[INDVARS_IV]] to i32
; CHECK-NEXT: [[TMP15:%.*]] = add i32 [[SUM_02]], [[TMP14]]
; CHECK-NEXT: [[TMP16:%.*]] = add i32 [[TMP15]], [[TMP11]]
; CHECK-NEXT: [[TMP17]] = add i32 [[TMP16]], [[TMP13]]
; CHECK-NEXT: [[INDVARS_IV_NEXT]] = add i64 [[INDVARS_IV]], 1
; CHECK-NEXT: [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
; CHECK-NEXT: [[EXITCOND:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], [[N]]
; CHECK-NEXT: br i1 [[EXITCOND]], label [[DOT_CRIT_EDGE_LOOPEXIT]], label [[DOTLR_PH]], !llvm.loop [[LOOP3:![0-9]+]]
; CHECK: ._crit_edge.loopexit:
; CHECK-NEXT: [[DOTLCSSA:%.*]] = phi i32 [ [[TMP17]], [[DOTLR_PH]] ], [ [[TMP9]], [[MIDDLE_BLOCK]] ]
; CHECK-NEXT: br label [[DOT_CRIT_EDGE]]
; CHECK: ._crit_edge:
; CHECK-NEXT: [[SUM_0_LCSSA:%.*]] = phi i32 [ 0, [[TMP0:%.*]] ], [ [[DOTLCSSA]], [[DOT_CRIT_EDGE_LOOPEXIT]] ]
; CHECK-NEXT: ret i32 [[SUM_0_LCSSA]]
;
%1 = icmp sgt i32 %n, 0
br i1 %1, label %.lr.ph, label %._crit_edge
.lr.ph: ; preds = %0, %.lr.ph
%indvars.iv = phi i64 [ %indvars.iv.next, %.lr.ph ], [ 0, %0 ]
%sum.02 = phi i32 [ %9, %.lr.ph ], [ 0, %0 ]
%2 = getelementptr inbounds i32, ptr %A, i64 %indvars.iv
%3 = load i32, ptr %2, align 4
%4 = getelementptr inbounds i32, ptr %B, i64 %indvars.iv
%5 = load i32, ptr %4, align 4
%6 = trunc i64 %indvars.iv to i32
%7 = add i32 %sum.02, %6
%8 = add i32 %7, %3
%9 = add i32 %8, %5
%indvars.iv.next = add i64 %indvars.iv, 1
%lftr.wideiv = trunc i64 %indvars.iv.next to i32
%exitcond = icmp eq i32 %lftr.wideiv, %n
br i1 %exitcond, label %._crit_edge, label %.lr.ph
._crit_edge: ; preds = %.lr.ph, %0
%sum.0.lcssa = phi i32 [ 0, %0 ], [ %9, %.lr.ph ]
ret i32 %sum.0.lcssa
}
define i32 @reduction_prod(i32 %n, ptr %A, ptr %B) {
; CHECK-LABEL: define i32 @reduction_prod(
; CHECK-SAME: i32 [[N:%.*]], ptr [[A:%.*]], ptr [[B:%.*]]) {
; CHECK-NEXT: [[TMP1:%.*]] = icmp sgt i32 [[N]], 0
; CHECK-NEXT: br i1 [[TMP1]], label [[DOTLR_PH_PREHEADER:%.*]], label [[DOT_CRIT_EDGE:%.*]]
; CHECK: .lr.ph.preheader:
; CHECK-NEXT: [[TMP2:%.*]] = zext nneg i32 [[N]] to i64
; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[N]], 4
; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; CHECK: vector.ph:
; CHECK-NEXT: [[N_VEC:%.*]] = and i64 [[TMP2]], 2147483644
; 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 i32> [ <i32 1, i32 1, i32 1, i32 1>, [[VECTOR_PH]] ], [ [[TMP7:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[VEC_IND:%.*]] = phi <4 x i32> [ <i32 0, i32 1, i32 2, i32 3>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDEX]]
; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP3]], align 4
; CHECK-NEXT: [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[INDEX]]
; CHECK-NEXT: [[WIDE_LOAD1:%.*]] = load <4 x i32>, ptr [[TMP4]], align 4
; CHECK-NEXT: [[TMP5:%.*]] = mul <4 x i32> [[VEC_PHI]], [[VEC_IND]]
; CHECK-NEXT: [[TMP6:%.*]] = mul <4 x i32> [[TMP5]], [[WIDE_LOAD]]
; CHECK-NEXT: [[TMP7]] = mul <4 x i32> [[TMP6]], [[WIDE_LOAD1]]
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; CHECK-NEXT: [[VEC_IND_NEXT]] = add <4 x i32> [[VEC_IND]], <i32 4, i32 4, i32 4, i32 4>
; CHECK-NEXT: [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-NEXT: br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
; CHECK: middle.block:
; CHECK-NEXT: [[TMP9:%.*]] = call i32 @llvm.vector.reduce.mul.v4i32(<4 x i32> [[TMP7]])
; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N_VEC]], [[TMP2]]
; CHECK-NEXT: br i1 [[CMP_N]], label [[DOT_CRIT_EDGE_LOOPEXIT:%.*]], label [[SCALAR_PH]]
; CHECK: scalar.ph:
; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[DOTLR_PH_PREHEADER]] ]
; CHECK-NEXT: [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP9]], [[MIDDLE_BLOCK]] ], [ 1, [[DOTLR_PH_PREHEADER]] ]
; CHECK-NEXT: br label [[DOTLR_PH:%.*]]
; CHECK: .lr.ph:
; CHECK-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[DOTLR_PH]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
; CHECK-NEXT: [[PROD_02:%.*]] = phi i32 [ [[TMP17:%.*]], [[DOTLR_PH]] ], [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ]
; CHECK-NEXT: [[TMP10:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDVARS_IV]]
; CHECK-NEXT: [[TMP11:%.*]] = load i32, ptr [[TMP10]], align 4
; CHECK-NEXT: [[TMP12:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[INDVARS_IV]]
; CHECK-NEXT: [[TMP13:%.*]] = load i32, ptr [[TMP12]], align 4
; CHECK-NEXT: [[TMP14:%.*]] = trunc i64 [[INDVARS_IV]] to i32
; CHECK-NEXT: [[TMP15:%.*]] = mul i32 [[PROD_02]], [[TMP14]]
; CHECK-NEXT: [[TMP16:%.*]] = mul i32 [[TMP15]], [[TMP11]]
; CHECK-NEXT: [[TMP17]] = mul i32 [[TMP16]], [[TMP13]]
; CHECK-NEXT: [[INDVARS_IV_NEXT]] = add i64 [[INDVARS_IV]], 1
; CHECK-NEXT: [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
; CHECK-NEXT: [[EXITCOND:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], [[N]]
; CHECK-NEXT: br i1 [[EXITCOND]], label [[DOT_CRIT_EDGE_LOOPEXIT]], label [[DOTLR_PH]], !llvm.loop [[LOOP5:![0-9]+]]
; CHECK: ._crit_edge.loopexit:
; CHECK-NEXT: [[DOTLCSSA:%.*]] = phi i32 [ [[TMP17]], [[DOTLR_PH]] ], [ [[TMP9]], [[MIDDLE_BLOCK]] ]
; CHECK-NEXT: br label [[DOT_CRIT_EDGE]]
; CHECK: ._crit_edge:
; CHECK-NEXT: [[PROD_0_LCSSA:%.*]] = phi i32 [ 1, [[TMP0:%.*]] ], [ [[DOTLCSSA]], [[DOT_CRIT_EDGE_LOOPEXIT]] ]
; CHECK-NEXT: ret i32 [[PROD_0_LCSSA]]
;
%1 = icmp sgt i32 %n, 0
br i1 %1, label %.lr.ph, label %._crit_edge
.lr.ph: ; preds = %0, %.lr.ph
%indvars.iv = phi i64 [ %indvars.iv.next, %.lr.ph ], [ 0, %0 ]
%prod.02 = phi i32 [ %9, %.lr.ph ], [ 1, %0 ]
%2 = getelementptr inbounds i32, ptr %A, i64 %indvars.iv
%3 = load i32, ptr %2, align 4
%4 = getelementptr inbounds i32, ptr %B, i64 %indvars.iv
%5 = load i32, ptr %4, align 4
%6 = trunc i64 %indvars.iv to i32
%7 = mul i32 %prod.02, %6
%8 = mul i32 %7, %3
%9 = mul i32 %8, %5
%indvars.iv.next = add i64 %indvars.iv, 1
%lftr.wideiv = trunc i64 %indvars.iv.next to i32
%exitcond = icmp eq i32 %lftr.wideiv, %n
br i1 %exitcond, label %._crit_edge, label %.lr.ph
._crit_edge: ; preds = %.lr.ph, %0
%prod.0.lcssa = phi i32 [ 1, %0 ], [ %9, %.lr.ph ]
ret i32 %prod.0.lcssa
}
define i32 @reduction_mix(i32 %n, ptr %A, ptr %B) {
; CHECK-LABEL: define i32 @reduction_mix(
; CHECK-SAME: i32 [[N:%.*]], ptr [[A:%.*]], ptr [[B:%.*]]) {
; CHECK-NEXT: [[TMP1:%.*]] = icmp sgt i32 [[N]], 0
; CHECK-NEXT: br i1 [[TMP1]], label [[DOTLR_PH_PREHEADER:%.*]], label [[DOT_CRIT_EDGE:%.*]]
; CHECK: .lr.ph.preheader:
; CHECK-NEXT: [[TMP2:%.*]] = zext nneg i32 [[N]] to i64
; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[N]], 4
; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; CHECK: vector.ph:
; CHECK-NEXT: [[N_VEC:%.*]] = and i64 [[TMP2]], 2147483644
; 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 i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP7:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[VEC_IND:%.*]] = phi <4 x i32> [ <i32 0, i32 1, i32 2, i32 3>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDEX]]
; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP3]], align 4
; CHECK-NEXT: [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[INDEX]]
; CHECK-NEXT: [[WIDE_LOAD1:%.*]] = load <4 x i32>, ptr [[TMP4]], align 4
; CHECK-NEXT: [[TMP5:%.*]] = mul nsw <4 x i32> [[WIDE_LOAD1]], [[WIDE_LOAD]]
; CHECK-NEXT: [[TMP6:%.*]] = add <4 x i32> [[VEC_PHI]], [[VEC_IND]]
; CHECK-NEXT: [[TMP7]] = add <4 x i32> [[TMP6]], [[TMP5]]
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; CHECK-NEXT: [[VEC_IND_NEXT]] = add <4 x i32> [[VEC_IND]], <i32 4, i32 4, i32 4, i32 4>
; CHECK-NEXT: [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-NEXT: br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
; CHECK: middle.block:
; CHECK-NEXT: [[TMP9:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[TMP7]])
; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N_VEC]], [[TMP2]]
; CHECK-NEXT: br i1 [[CMP_N]], label [[DOT_CRIT_EDGE_LOOPEXIT:%.*]], label [[SCALAR_PH]]
; CHECK: scalar.ph:
; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[DOTLR_PH_PREHEADER]] ]
; CHECK-NEXT: [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP9]], [[MIDDLE_BLOCK]] ], [ 0, [[DOTLR_PH_PREHEADER]] ]
; CHECK-NEXT: br label [[DOTLR_PH:%.*]]
; CHECK: .lr.ph:
; CHECK-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[DOTLR_PH]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
; CHECK-NEXT: [[SUM_02:%.*]] = phi i32 [ [[TMP17:%.*]], [[DOTLR_PH]] ], [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ]
; CHECK-NEXT: [[TMP10:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDVARS_IV]]
; CHECK-NEXT: [[TMP11:%.*]] = load i32, ptr [[TMP10]], align 4
; CHECK-NEXT: [[TMP12:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[INDVARS_IV]]
; CHECK-NEXT: [[TMP13:%.*]] = load i32, ptr [[TMP12]], align 4
; CHECK-NEXT: [[TMP14:%.*]] = mul nsw i32 [[TMP13]], [[TMP11]]
; CHECK-NEXT: [[TMP15:%.*]] = trunc i64 [[INDVARS_IV]] to i32
; CHECK-NEXT: [[TMP16:%.*]] = add i32 [[SUM_02]], [[TMP15]]
; CHECK-NEXT: [[TMP17]] = add i32 [[TMP16]], [[TMP14]]
; CHECK-NEXT: [[INDVARS_IV_NEXT]] = add i64 [[INDVARS_IV]], 1
; CHECK-NEXT: [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
; CHECK-NEXT: [[EXITCOND:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], [[N]]
; CHECK-NEXT: br i1 [[EXITCOND]], label [[DOT_CRIT_EDGE_LOOPEXIT]], label [[DOTLR_PH]], !llvm.loop [[LOOP7:![0-9]+]]
; CHECK: ._crit_edge.loopexit:
; CHECK-NEXT: [[DOTLCSSA:%.*]] = phi i32 [ [[TMP17]], [[DOTLR_PH]] ], [ [[TMP9]], [[MIDDLE_BLOCK]] ]
; CHECK-NEXT: br label [[DOT_CRIT_EDGE]]
; CHECK: ._crit_edge:
; CHECK-NEXT: [[SUM_0_LCSSA:%.*]] = phi i32 [ 0, [[TMP0:%.*]] ], [ [[DOTLCSSA]], [[DOT_CRIT_EDGE_LOOPEXIT]] ]
; CHECK-NEXT: ret i32 [[SUM_0_LCSSA]]
;
%1 = icmp sgt i32 %n, 0
br i1 %1, label %.lr.ph, label %._crit_edge
.lr.ph: ; preds = %0, %.lr.ph
%indvars.iv = phi i64 [ %indvars.iv.next, %.lr.ph ], [ 0, %0 ]
%sum.02 = phi i32 [ %9, %.lr.ph ], [ 0, %0 ]
%2 = getelementptr inbounds i32, ptr %A, i64 %indvars.iv
%3 = load i32, ptr %2, align 4
%4 = getelementptr inbounds i32, ptr %B, i64 %indvars.iv
%5 = load i32, ptr %4, align 4
%6 = mul nsw i32 %5, %3
%7 = trunc i64 %indvars.iv to i32
%8 = add i32 %sum.02, %7
%9 = add i32 %8, %6
%indvars.iv.next = add i64 %indvars.iv, 1
%lftr.wideiv = trunc i64 %indvars.iv.next to i32
%exitcond = icmp eq i32 %lftr.wideiv, %n
br i1 %exitcond, label %._crit_edge, label %.lr.ph
._crit_edge: ; preds = %.lr.ph, %0
%sum.0.lcssa = phi i32 [ 0, %0 ], [ %9, %.lr.ph ]
ret i32 %sum.0.lcssa
}
define i32 @reduction_mul(i32 %n, ptr %A, ptr %B) {
; CHECK-LABEL: define i32 @reduction_mul(
; CHECK-SAME: i32 [[N:%.*]], ptr [[A:%.*]], ptr [[B:%.*]]) {
; CHECK-NEXT: [[TMP1:%.*]] = icmp sgt i32 [[N]], 0
; CHECK-NEXT: br i1 [[TMP1]], label [[DOTLR_PH_PREHEADER:%.*]], label [[DOT_CRIT_EDGE:%.*]]
; CHECK: .lr.ph.preheader:
; CHECK-NEXT: [[TMP2:%.*]] = zext nneg i32 [[N]] to i64
; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[N]], 4
; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; CHECK: vector.ph:
; CHECK-NEXT: [[N_VEC:%.*]] = and i64 [[TMP2]], 2147483644
; 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 i32> [ <i32 19, i32 1, i32 1, i32 1>, [[VECTOR_PH]] ], [ [[TMP7:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[VEC_IND:%.*]] = phi <4 x i32> [ <i32 0, i32 1, i32 2, i32 3>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDEX]]
; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP3]], align 4
; CHECK-NEXT: [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[INDEX]]
; CHECK-NEXT: [[WIDE_LOAD1:%.*]] = load <4 x i32>, ptr [[TMP4]], align 4
; CHECK-NEXT: [[TMP5:%.*]] = add <4 x i32> [[WIDE_LOAD]], [[VEC_IND]]
; CHECK-NEXT: [[TMP6:%.*]] = add <4 x i32> [[TMP5]], [[WIDE_LOAD1]]
; CHECK-NEXT: [[TMP7]] = mul <4 x i32> [[TMP6]], [[VEC_PHI]]
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; CHECK-NEXT: [[VEC_IND_NEXT]] = add <4 x i32> [[VEC_IND]], <i32 4, i32 4, i32 4, i32 4>
; CHECK-NEXT: [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-NEXT: br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
; CHECK: middle.block:
; CHECK-NEXT: [[TMP9:%.*]] = call i32 @llvm.vector.reduce.mul.v4i32(<4 x i32> [[TMP7]])
; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N_VEC]], [[TMP2]]
; CHECK-NEXT: br i1 [[CMP_N]], label [[DOT_CRIT_EDGE_LOOPEXIT:%.*]], label [[SCALAR_PH]]
; CHECK: scalar.ph:
; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[DOTLR_PH_PREHEADER]] ]
; CHECK-NEXT: [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP9]], [[MIDDLE_BLOCK]] ], [ 19, [[DOTLR_PH_PREHEADER]] ]
; CHECK-NEXT: br label [[DOTLR_PH:%.*]]
; CHECK: .lr.ph:
; CHECK-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[DOTLR_PH]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
; CHECK-NEXT: [[SUM_02:%.*]] = phi i32 [ [[TMP17:%.*]], [[DOTLR_PH]] ], [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ]
; CHECK-NEXT: [[TMP10:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDVARS_IV]]
; CHECK-NEXT: [[TMP11:%.*]] = load i32, ptr [[TMP10]], align 4
; CHECK-NEXT: [[TMP12:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[INDVARS_IV]]
; CHECK-NEXT: [[TMP13:%.*]] = load i32, ptr [[TMP12]], align 4
; CHECK-NEXT: [[TMP14:%.*]] = trunc i64 [[INDVARS_IV]] to i32
; CHECK-NEXT: [[TMP15:%.*]] = add i32 [[TMP11]], [[TMP14]]
; CHECK-NEXT: [[TMP16:%.*]] = add i32 [[TMP15]], [[TMP13]]
; CHECK-NEXT: [[TMP17]] = mul i32 [[TMP16]], [[SUM_02]]
; CHECK-NEXT: [[INDVARS_IV_NEXT]] = add i64 [[INDVARS_IV]], 1
; CHECK-NEXT: [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
; CHECK-NEXT: [[EXITCOND:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], [[N]]
; CHECK-NEXT: br i1 [[EXITCOND]], label [[DOT_CRIT_EDGE_LOOPEXIT]], label [[DOTLR_PH]], !llvm.loop [[LOOP9:![0-9]+]]
; CHECK: ._crit_edge.loopexit:
; CHECK-NEXT: [[DOTLCSSA:%.*]] = phi i32 [ [[TMP17]], [[DOTLR_PH]] ], [ [[TMP9]], [[MIDDLE_BLOCK]] ]
; CHECK-NEXT: br label [[DOT_CRIT_EDGE]]
; CHECK: ._crit_edge:
; CHECK-NEXT: [[SUM_0_LCSSA:%.*]] = phi i32 [ 0, [[TMP0:%.*]] ], [ [[DOTLCSSA]], [[DOT_CRIT_EDGE_LOOPEXIT]] ]
; CHECK-NEXT: ret i32 [[SUM_0_LCSSA]]
;
%1 = icmp sgt i32 %n, 0
br i1 %1, label %.lr.ph, label %._crit_edge
.lr.ph: ; preds = %0, %.lr.ph
%indvars.iv = phi i64 [ %indvars.iv.next, %.lr.ph ], [ 0, %0 ]
%sum.02 = phi i32 [ %9, %.lr.ph ], [ 19, %0 ]
%2 = getelementptr inbounds i32, ptr %A, i64 %indvars.iv
%3 = load i32, ptr %2, align 4
%4 = getelementptr inbounds i32, ptr %B, i64 %indvars.iv
%5 = load i32, ptr %4, align 4
%6 = trunc i64 %indvars.iv to i32
%7 = add i32 %3, %6
%8 = add i32 %7, %5
%9 = mul i32 %8, %sum.02
%indvars.iv.next = add i64 %indvars.iv, 1
%lftr.wideiv = trunc i64 %indvars.iv.next to i32
%exitcond = icmp eq i32 %lftr.wideiv, %n
br i1 %exitcond, label %._crit_edge, label %.lr.ph
._crit_edge: ; preds = %.lr.ph, %0
%sum.0.lcssa = phi i32 [ 0, %0 ], [ %9, %.lr.ph ]
ret i32 %sum.0.lcssa
}
define i32 @start_at_non_zero(ptr %in, ptr %coeff, ptr %out, i32 %n) {
; CHECK-LABEL: define i32 @start_at_non_zero(
; CHECK-SAME: ptr [[IN:%.*]], ptr [[COEFF:%.*]], ptr [[OUT:%.*]], i32 [[N:%.*]]) {
; CHECK-NEXT: entry:
; CHECK-NEXT: [[CMP7:%.*]] = icmp sgt i32 [[N]], 0
; CHECK-NEXT: br i1 [[CMP7]], label [[FOR_BODY_PREHEADER:%.*]], label [[FOR_END:%.*]]
; CHECK: for.body.preheader:
; CHECK-NEXT: [[TMP0:%.*]] = zext nneg i32 [[N]] to i64
; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[N]], 4
; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; CHECK: vector.ph:
; CHECK-NEXT: [[N_VEC:%.*]] = and i64 [[TMP0]], 2147483644
; 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 i32> [ <i32 120, i32 0, i32 0, i32 0>, [[VECTOR_PH]] ], [ [[TMP4:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[IN]], i64 [[INDEX]]
; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP1]], align 4
; CHECK-NEXT: [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[COEFF]], i64 [[INDEX]]
; CHECK-NEXT: [[WIDE_LOAD1:%.*]] = load <4 x i32>, ptr [[TMP2]], align 4
; CHECK-NEXT: [[TMP3:%.*]] = mul nsw <4 x i32> [[WIDE_LOAD1]], [[WIDE_LOAD]]
; CHECK-NEXT: [[TMP4]] = add <4 x i32> [[TMP3]], [[VEC_PHI]]
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; CHECK-NEXT: [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-NEXT: br i1 [[TMP5]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
; CHECK: middle.block:
; CHECK-NEXT: [[TMP6:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[TMP4]])
; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N_VEC]], [[TMP0]]
; CHECK-NEXT: br i1 [[CMP_N]], label [[FOR_END_LOOPEXIT:%.*]], label [[SCALAR_PH]]
; CHECK: scalar.ph:
; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_PREHEADER]] ]
; CHECK-NEXT: [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP6]], [[MIDDLE_BLOCK]] ], [ 120, [[FOR_BODY_PREHEADER]] ]
; CHECK-NEXT: br label [[FOR_BODY:%.*]]
; CHECK: for.body:
; CHECK-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
; CHECK-NEXT: [[SUM_09:%.*]] = phi i32 [ [[ADD:%.*]], [[FOR_BODY]] ], [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ]
; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[IN]], i64 [[INDVARS_IV]]
; CHECK-NEXT: [[TMP7:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
; CHECK-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds i32, ptr [[COEFF]], i64 [[INDVARS_IV]]
; CHECK-NEXT: [[TMP8:%.*]] = load i32, ptr [[ARRAYIDX2]], align 4
; CHECK-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP8]], [[TMP7]]
; CHECK-NEXT: [[ADD]] = add nsw i32 [[MUL]], [[SUM_09]]
; CHECK-NEXT: [[INDVARS_IV_NEXT]] = add i64 [[INDVARS_IV]], 1
; CHECK-NEXT: [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
; CHECK-NEXT: [[EXITCOND:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], [[N]]
; CHECK-NEXT: br i1 [[EXITCOND]], label [[FOR_END_LOOPEXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP11:![0-9]+]]
; CHECK: for.end.loopexit:
; CHECK-NEXT: [[ADD_LCSSA:%.*]] = phi i32 [ [[ADD]], [[FOR_BODY]] ], [ [[TMP6]], [[MIDDLE_BLOCK]] ]
; CHECK-NEXT: br label [[FOR_END]]
; CHECK: for.end:
; CHECK-NEXT: [[SUM_0_LCSSA:%.*]] = phi i32 [ 120, [[ENTRY:%.*]] ], [ [[ADD_LCSSA]], [[FOR_END_LOOPEXIT]] ]
; CHECK-NEXT: ret i32 [[SUM_0_LCSSA]]
;
entry:
%cmp7 = icmp sgt i32 %n, 0
br i1 %cmp7, label %for.body, label %for.end
for.body: ; preds = %entry, %for.body
%indvars.iv = phi i64 [ %indvars.iv.next, %for.body ], [ 0, %entry ]
%sum.09 = phi i32 [ %add, %for.body ], [ 120, %entry ]
%arrayidx = getelementptr inbounds i32, ptr %in, i64 %indvars.iv
%0 = load i32, ptr %arrayidx, align 4
%arrayidx2 = getelementptr inbounds i32, ptr %coeff, i64 %indvars.iv
%1 = load i32, ptr %arrayidx2, align 4
%mul = mul nsw i32 %1, %0
%add = add nsw i32 %mul, %sum.09
%indvars.iv.next = add i64 %indvars.iv, 1
%lftr.wideiv = trunc i64 %indvars.iv.next to i32
%exitcond = icmp eq i32 %lftr.wideiv, %n
br i1 %exitcond, label %for.end, label %for.body
for.end: ; preds = %for.body, %entry
%sum.0.lcssa = phi i32 [ 120, %entry ], [ %add, %for.body ]
ret i32 %sum.0.lcssa
}
define i32 @reduction_and(i32 %n, ptr %A, ptr %B) {
; CHECK-LABEL: define i32 @reduction_and(
; CHECK-SAME: i32 [[N:%.*]], ptr [[A:%.*]], ptr [[B:%.*]]) {
; CHECK-NEXT: entry:
; CHECK-NEXT: [[CMP7:%.*]] = icmp sgt i32 [[N]], 0
; CHECK-NEXT: br i1 [[CMP7]], label [[FOR_BODY_PREHEADER:%.*]], label [[FOR_END:%.*]]
; CHECK: for.body.preheader:
; CHECK-NEXT: [[TMP0:%.*]] = zext nneg i32 [[N]] to i64
; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[N]], 4
; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; CHECK: vector.ph:
; CHECK-NEXT: [[N_VEC:%.*]] = and i64 [[TMP0]], 2147483644
; 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 i32> [ <i32 -1, i32 -1, i32 -1, i32 -1>, [[VECTOR_PH]] ], [ [[TMP4:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDEX]]
; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP1]], align 4
; CHECK-NEXT: [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[INDEX]]
; CHECK-NEXT: [[WIDE_LOAD1:%.*]] = load <4 x i32>, ptr [[TMP2]], align 4
; CHECK-NEXT: [[TMP3:%.*]] = add nsw <4 x i32> [[WIDE_LOAD1]], [[WIDE_LOAD]]
; CHECK-NEXT: [[TMP4]] = and <4 x i32> [[TMP3]], [[VEC_PHI]]
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; CHECK-NEXT: [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-NEXT: br i1 [[TMP5]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
; CHECK: middle.block:
; CHECK-NEXT: [[TMP6:%.*]] = call i32 @llvm.vector.reduce.and.v4i32(<4 x i32> [[TMP4]])
; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N_VEC]], [[TMP0]]
; CHECK-NEXT: br i1 [[CMP_N]], label [[FOR_END_LOOPEXIT:%.*]], label [[SCALAR_PH]]
; CHECK: scalar.ph:
; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_PREHEADER]] ]
; CHECK-NEXT: [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP6]], [[MIDDLE_BLOCK]] ], [ -1, [[FOR_BODY_PREHEADER]] ]
; CHECK-NEXT: br label [[FOR_BODY:%.*]]
; CHECK: for.body:
; CHECK-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
; CHECK-NEXT: [[RESULT_08:%.*]] = phi i32 [ [[AND:%.*]], [[FOR_BODY]] ], [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ]
; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDVARS_IV]]
; CHECK-NEXT: [[TMP7:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
; CHECK-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[INDVARS_IV]]
; CHECK-NEXT: [[TMP8:%.*]] = load i32, ptr [[ARRAYIDX2]], align 4
; CHECK-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP8]], [[TMP7]]
; CHECK-NEXT: [[AND]] = and i32 [[ADD]], [[RESULT_08]]
; CHECK-NEXT: [[INDVARS_IV_NEXT]] = add i64 [[INDVARS_IV]], 1
; CHECK-NEXT: [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
; CHECK-NEXT: [[EXITCOND:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], [[N]]
; CHECK-NEXT: br i1 [[EXITCOND]], label [[FOR_END_LOOPEXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP13:![0-9]+]]
; CHECK: for.end.loopexit:
; CHECK-NEXT: [[AND_LCSSA:%.*]] = phi i32 [ [[AND]], [[FOR_BODY]] ], [ [[TMP6]], [[MIDDLE_BLOCK]] ]
; CHECK-NEXT: br label [[FOR_END]]
; CHECK: for.end:
; CHECK-NEXT: [[RESULT_0_LCSSA:%.*]] = phi i32 [ -1, [[ENTRY:%.*]] ], [ [[AND_LCSSA]], [[FOR_END_LOOPEXIT]] ]
; CHECK-NEXT: ret i32 [[RESULT_0_LCSSA]]
;
entry:
%cmp7 = icmp sgt i32 %n, 0
br i1 %cmp7, label %for.body, label %for.end
for.body: ; preds = %entry, %for.body
%indvars.iv = phi i64 [ %indvars.iv.next, %for.body ], [ 0, %entry ]
%result.08 = phi i32 [ %and, %for.body ], [ -1, %entry ]
%arrayidx = getelementptr inbounds i32, ptr %A, i64 %indvars.iv
%0 = load i32, ptr %arrayidx, align 4
%arrayidx2 = getelementptr inbounds i32, ptr %B, i64 %indvars.iv
%1 = load i32, ptr %arrayidx2, align 4
%add = add nsw i32 %1, %0
%and = and i32 %add, %result.08
%indvars.iv.next = add i64 %indvars.iv, 1
%lftr.wideiv = trunc i64 %indvars.iv.next to i32
%exitcond = icmp eq i32 %lftr.wideiv, %n
br i1 %exitcond, label %for.end, label %for.body
for.end: ; preds = %for.body, %entry
%result.0.lcssa = phi i32 [ -1, %entry ], [ %and, %for.body ]
ret i32 %result.0.lcssa
}
define i32 @reduction_or(i32 %n, ptr %A, ptr %B) {
; CHECK-LABEL: define i32 @reduction_or(
; CHECK-SAME: i32 [[N:%.*]], ptr [[A:%.*]], ptr [[B:%.*]]) {
; CHECK-NEXT: entry:
; CHECK-NEXT: [[CMP7:%.*]] = icmp sgt i32 [[N]], 0
; CHECK-NEXT: br i1 [[CMP7]], label [[FOR_BODY_PREHEADER:%.*]], label [[FOR_END:%.*]]
; CHECK: for.body.preheader:
; CHECK-NEXT: [[TMP0:%.*]] = zext nneg i32 [[N]] to i64
; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[N]], 4
; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; CHECK: vector.ph:
; CHECK-NEXT: [[N_VEC:%.*]] = and i64 [[TMP0]], 2147483644
; 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 i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP4:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDEX]]
; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP1]], align 4
; CHECK-NEXT: [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[INDEX]]
; CHECK-NEXT: [[WIDE_LOAD1:%.*]] = load <4 x i32>, ptr [[TMP2]], align 4
; CHECK-NEXT: [[TMP3:%.*]] = add nsw <4 x i32> [[WIDE_LOAD1]], [[WIDE_LOAD]]
; CHECK-NEXT: [[TMP4]] = or <4 x i32> [[TMP3]], [[VEC_PHI]]
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; CHECK-NEXT: [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-NEXT: br i1 [[TMP5]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP14:![0-9]+]]
; CHECK: middle.block:
; CHECK-NEXT: [[TMP6:%.*]] = call i32 @llvm.vector.reduce.or.v4i32(<4 x i32> [[TMP4]])
; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N_VEC]], [[TMP0]]
; CHECK-NEXT: br i1 [[CMP_N]], label [[FOR_END_LOOPEXIT:%.*]], label [[SCALAR_PH]]
; CHECK: scalar.ph:
; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_PREHEADER]] ]
; CHECK-NEXT: [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP6]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_PREHEADER]] ]
; CHECK-NEXT: br label [[FOR_BODY:%.*]]
; CHECK: for.body:
; CHECK-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
; CHECK-NEXT: [[RESULT_08:%.*]] = phi i32 [ [[OR:%.*]], [[FOR_BODY]] ], [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ]
; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDVARS_IV]]
; CHECK-NEXT: [[TMP7:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
; CHECK-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[INDVARS_IV]]
; CHECK-NEXT: [[TMP8:%.*]] = load i32, ptr [[ARRAYIDX2]], align 4
; CHECK-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP8]], [[TMP7]]
; CHECK-NEXT: [[OR]] = or i32 [[ADD]], [[RESULT_08]]
; CHECK-NEXT: [[INDVARS_IV_NEXT]] = add i64 [[INDVARS_IV]], 1
; CHECK-NEXT: [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
; CHECK-NEXT: [[EXITCOND:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], [[N]]
; CHECK-NEXT: br i1 [[EXITCOND]], label [[FOR_END_LOOPEXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP15:![0-9]+]]
; CHECK: for.end.loopexit:
; CHECK-NEXT: [[OR_LCSSA:%.*]] = phi i32 [ [[OR]], [[FOR_BODY]] ], [ [[TMP6]], [[MIDDLE_BLOCK]] ]
; CHECK-NEXT: br label [[FOR_END]]
; CHECK: for.end:
; CHECK-NEXT: [[RESULT_0_LCSSA:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[OR_LCSSA]], [[FOR_END_LOOPEXIT]] ]
; CHECK-NEXT: ret i32 [[RESULT_0_LCSSA]]
;
entry:
%cmp7 = icmp sgt i32 %n, 0
br i1 %cmp7, label %for.body, label %for.end
for.body: ; preds = %entry, %for.body
%indvars.iv = phi i64 [ %indvars.iv.next, %for.body ], [ 0, %entry ]
%result.08 = phi i32 [ %or, %for.body ], [ 0, %entry ]
%arrayidx = getelementptr inbounds i32, ptr %A, i64 %indvars.iv
%0 = load i32, ptr %arrayidx, align 4
%arrayidx2 = getelementptr inbounds i32, ptr %B, i64 %indvars.iv
%1 = load i32, ptr %arrayidx2, align 4
%add = add nsw i32 %1, %0
%or = or i32 %add, %result.08
%indvars.iv.next = add i64 %indvars.iv, 1
%lftr.wideiv = trunc i64 %indvars.iv.next to i32
%exitcond = icmp eq i32 %lftr.wideiv, %n
br i1 %exitcond, label %for.end, label %for.body
for.end: ; preds = %for.body, %entry
%result.0.lcssa = phi i32 [ 0, %entry ], [ %or, %for.body ]
ret i32 %result.0.lcssa
}
define i32 @reduction_xor(i32 %n, ptr %A, ptr %B) {
; CHECK-LABEL: define i32 @reduction_xor(
; CHECK-SAME: i32 [[N:%.*]], ptr [[A:%.*]], ptr [[B:%.*]]) {
; CHECK-NEXT: entry:
; CHECK-NEXT: [[CMP7:%.*]] = icmp sgt i32 [[N]], 0
; CHECK-NEXT: br i1 [[CMP7]], label [[FOR_BODY_PREHEADER:%.*]], label [[FOR_END:%.*]]
; CHECK: for.body.preheader:
; CHECK-NEXT: [[TMP0:%.*]] = zext nneg i32 [[N]] to i64
; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[N]], 4
; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; CHECK: vector.ph:
; CHECK-NEXT: [[N_VEC:%.*]] = and i64 [[TMP0]], 2147483644
; 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 i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP4:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDEX]]
; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP1]], align 4
; CHECK-NEXT: [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[INDEX]]
; CHECK-NEXT: [[WIDE_LOAD1:%.*]] = load <4 x i32>, ptr [[TMP2]], align 4
; CHECK-NEXT: [[TMP3:%.*]] = add nsw <4 x i32> [[WIDE_LOAD1]], [[WIDE_LOAD]]
; CHECK-NEXT: [[TMP4]] = xor <4 x i32> [[TMP3]], [[VEC_PHI]]
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; CHECK-NEXT: [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-NEXT: br i1 [[TMP5]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]]
; CHECK: middle.block:
; CHECK-NEXT: [[TMP6:%.*]] = call i32 @llvm.vector.reduce.xor.v4i32(<4 x i32> [[TMP4]])
; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N_VEC]], [[TMP0]]
; CHECK-NEXT: br i1 [[CMP_N]], label [[FOR_END_LOOPEXIT:%.*]], label [[SCALAR_PH]]
; CHECK: scalar.ph:
; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_PREHEADER]] ]
; CHECK-NEXT: [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP6]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_PREHEADER]] ]
; CHECK-NEXT: br label [[FOR_BODY:%.*]]
; CHECK: for.body:
; CHECK-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
; CHECK-NEXT: [[RESULT_08:%.*]] = phi i32 [ [[XOR:%.*]], [[FOR_BODY]] ], [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ]
; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDVARS_IV]]
; CHECK-NEXT: [[TMP7:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
; CHECK-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[INDVARS_IV]]
; CHECK-NEXT: [[TMP8:%.*]] = load i32, ptr [[ARRAYIDX2]], align 4
; CHECK-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP8]], [[TMP7]]
; CHECK-NEXT: [[XOR]] = xor i32 [[ADD]], [[RESULT_08]]
; CHECK-NEXT: [[INDVARS_IV_NEXT]] = add i64 [[INDVARS_IV]], 1
; CHECK-NEXT: [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
; CHECK-NEXT: [[EXITCOND:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], [[N]]
; CHECK-NEXT: br i1 [[EXITCOND]], label [[FOR_END_LOOPEXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP17:![0-9]+]]
; CHECK: for.end.loopexit:
; CHECK-NEXT: [[XOR_LCSSA:%.*]] = phi i32 [ [[XOR]], [[FOR_BODY]] ], [ [[TMP6]], [[MIDDLE_BLOCK]] ]
; CHECK-NEXT: br label [[FOR_END]]
; CHECK: for.end:
; CHECK-NEXT: [[RESULT_0_LCSSA:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[XOR_LCSSA]], [[FOR_END_LOOPEXIT]] ]
; CHECK-NEXT: ret i32 [[RESULT_0_LCSSA]]
;
entry:
%cmp7 = icmp sgt i32 %n, 0
br i1 %cmp7, label %for.body, label %for.end
for.body: ; preds = %entry, %for.body
%indvars.iv = phi i64 [ %indvars.iv.next, %for.body ], [ 0, %entry ]
%result.08 = phi i32 [ %xor, %for.body ], [ 0, %entry ]
%arrayidx = getelementptr inbounds i32, ptr %A, i64 %indvars.iv
%0 = load i32, ptr %arrayidx, align 4
%arrayidx2 = getelementptr inbounds i32, ptr %B, i64 %indvars.iv
%1 = load i32, ptr %arrayidx2, align 4
%add = add nsw i32 %1, %0
%xor = xor i32 %add, %result.08
%indvars.iv.next = add i64 %indvars.iv, 1
%lftr.wideiv = trunc i64 %indvars.iv.next to i32
%exitcond = icmp eq i32 %lftr.wideiv, %n
br i1 %exitcond, label %for.end, label %for.body
for.end: ; preds = %for.body, %entry
%result.0.lcssa = phi i32 [ 0, %entry ], [ %xor, %for.body ]
ret i32 %result.0.lcssa
}
; In this code the subtracted variable is on the RHS and this is not an induction variable.
define i32 @reduction_sub_rhs(i32 %n, ptr %A) {
; CHECK-LABEL: define i32 @reduction_sub_rhs(
; CHECK-SAME: i32 [[N:%.*]], ptr [[A:%.*]]) {
; CHECK-NEXT: entry:
; CHECK-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[N]], 0
; CHECK-NEXT: br i1 [[CMP4]], label [[FOR_BODY_PREHEADER:%.*]], label [[FOR_END:%.*]]
; CHECK: for.body.preheader:
; CHECK-NEXT: br label [[FOR_BODY:%.*]]
; CHECK: for.body:
; CHECK-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ 0, [[FOR_BODY_PREHEADER]] ]
; CHECK-NEXT: [[X_05:%.*]] = phi i32 [ [[SUB:%.*]], [[FOR_BODY]] ], [ 0, [[FOR_BODY_PREHEADER]] ]
; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDVARS_IV]]
; CHECK-NEXT: [[TMP0:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
; CHECK-NEXT: [[SUB]] = sub nsw i32 [[TMP0]], [[X_05]]
; CHECK-NEXT: [[INDVARS_IV_NEXT]] = add i64 [[INDVARS_IV]], 1
; CHECK-NEXT: [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
; CHECK-NEXT: [[EXITCOND:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], [[N]]
; CHECK-NEXT: br i1 [[EXITCOND]], label [[FOR_END_LOOPEXIT:%.*]], label [[FOR_BODY]]
; CHECK: for.end.loopexit:
; CHECK-NEXT: br label [[FOR_END]]
; CHECK: for.end:
; CHECK-NEXT: [[X_0_LCSSA:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[SUB]], [[FOR_END_LOOPEXIT]] ]
; CHECK-NEXT: ret i32 [[X_0_LCSSA]]
;
entry:
%cmp4 = icmp sgt i32 %n, 0
br i1 %cmp4, label %for.body, label %for.end
for.body: ; preds = %entry, %for.body
%indvars.iv = phi i64 [ %indvars.iv.next, %for.body ], [ 0, %entry ]
%x.05 = phi i32 [ %sub, %for.body ], [ 0, %entry ]
%arrayidx = getelementptr inbounds i32, ptr %A, i64 %indvars.iv
%0 = load i32, ptr %arrayidx, align 4
%sub = sub nsw i32 %0, %x.05
%indvars.iv.next = add i64 %indvars.iv, 1
%lftr.wideiv = trunc i64 %indvars.iv.next to i32
%exitcond = icmp eq i32 %lftr.wideiv, %n
br i1 %exitcond, label %for.end, label %for.body
for.end: ; preds = %for.body, %entry
%x.0.lcssa = phi i32 [ 0, %entry ], [ %sub, %for.body ]
ret i32 %x.0.lcssa
}
; In this test the reduction variable is on the LHS and we can vectorize it.
define i32 @reduction_sub_lhs(i32 %n, ptr %A) {
; CHECK-LABEL: define i32 @reduction_sub_lhs(
; CHECK-SAME: i32 [[N:%.*]], ptr [[A:%.*]]) {
; CHECK-NEXT: entry:
; CHECK-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[N]], 0
; CHECK-NEXT: br i1 [[CMP4]], label [[FOR_BODY_PREHEADER:%.*]], label [[FOR_END:%.*]]
; CHECK: for.body.preheader:
; CHECK-NEXT: [[TMP0:%.*]] = zext nneg i32 [[N]] to i64
; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[N]], 4
; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; CHECK: vector.ph:
; CHECK-NEXT: [[N_VEC:%.*]] = and i64 [[TMP0]], 2147483644
; 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 i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP2:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDEX]]
; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP1]], align 4
; CHECK-NEXT: [[TMP2]] = sub <4 x i32> [[VEC_PHI]], [[WIDE_LOAD]]
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; CHECK-NEXT: [[TMP3:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-NEXT: br i1 [[TMP3]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP18:![0-9]+]]
; CHECK: middle.block:
; CHECK-NEXT: [[TMP4:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[TMP2]])
; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N_VEC]], [[TMP0]]
; CHECK-NEXT: br i1 [[CMP_N]], label [[FOR_END_LOOPEXIT:%.*]], label [[SCALAR_PH]]
; CHECK: scalar.ph:
; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_PREHEADER]] ]
; CHECK-NEXT: [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP4]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_PREHEADER]] ]
; CHECK-NEXT: br label [[FOR_BODY:%.*]]
; CHECK: for.body:
; CHECK-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
; CHECK-NEXT: [[X_05:%.*]] = phi i32 [ [[SUB:%.*]], [[FOR_BODY]] ], [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ]
; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDVARS_IV]]
; CHECK-NEXT: [[TMP5:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
; CHECK-NEXT: [[SUB]] = sub nsw i32 [[X_05]], [[TMP5]]
; CHECK-NEXT: [[INDVARS_IV_NEXT]] = add i64 [[INDVARS_IV]], 1
; CHECK-NEXT: [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
; CHECK-NEXT: [[EXITCOND:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], [[N]]
; CHECK-NEXT: br i1 [[EXITCOND]], label [[FOR_END_LOOPEXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP19:![0-9]+]]
; CHECK: for.end.loopexit:
; CHECK-NEXT: [[SUB_LCSSA:%.*]] = phi i32 [ [[SUB]], [[FOR_BODY]] ], [ [[TMP4]], [[MIDDLE_BLOCK]] ]
; CHECK-NEXT: br label [[FOR_END]]
; CHECK: for.end:
; CHECK-NEXT: [[X_0_LCSSA:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[SUB_LCSSA]], [[FOR_END_LOOPEXIT]] ]
; CHECK-NEXT: ret i32 [[X_0_LCSSA]]
;
entry:
%cmp4 = icmp sgt i32 %n, 0
br i1 %cmp4, label %for.body, label %for.end
for.body: ; preds = %entry, %for.body
%indvars.iv = phi i64 [ %indvars.iv.next, %for.body ], [ 0, %entry ]
%x.05 = phi i32 [ %sub, %for.body ], [ 0, %entry ]
%arrayidx = getelementptr inbounds i32, ptr %A, i64 %indvars.iv
%0 = load i32, ptr %arrayidx, align 4
%sub = sub nsw i32 %x.05, %0
%indvars.iv.next = add i64 %indvars.iv, 1
%lftr.wideiv = trunc i64 %indvars.iv.next to i32
%exitcond = icmp eq i32 %lftr.wideiv, %n
br i1 %exitcond, label %for.end, label %for.body
for.end: ; preds = %for.body, %entry
%x.0.lcssa = phi i32 [ 0, %entry ], [ %sub, %for.body ]
ret i32 %x.0.lcssa
}
; We can vectorize conditional reductions with multi-input phis.
define float @reduction_conditional(ptr %A, ptr %B, ptr %C, float %S) {
; CHECK-LABEL: define float @reduction_conditional(
; CHECK-SAME: ptr [[A:%.*]], ptr [[B:%.*]], ptr [[C:%.*]], float [[S:%.*]]) {
; CHECK-NEXT: entry:
; CHECK-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; CHECK: vector.ph:
; CHECK-NEXT: [[TMP0:%.*]] = insertelement <4 x float> <float poison, float 0.000000e+00, float 0.000000e+00, float 0.000000e+00>, float [[S]], i64 0
; 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 float> [ [[TMP0]], [[VECTOR_PH]] ], [ [[PREDPHI3:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[TMP1:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[INDEX]]
; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP1]], align 4
; CHECK-NEXT: [[TMP2:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[INDEX]]
; CHECK-NEXT: [[WIDE_LOAD1:%.*]] = load <4 x float>, ptr [[TMP2]], align 4
; CHECK-NEXT: [[TMP3:%.*]] = fcmp ogt <4 x float> [[WIDE_LOAD]], [[WIDE_LOAD1]]
; CHECK-NEXT: [[TMP4:%.*]] = fcmp ule <4 x float> [[WIDE_LOAD1]], <float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00>
; CHECK-NEXT: [[TMP6:%.*]] = and <4 x i1> [[TMP3]], [[TMP4]]
; CHECK-NEXT: [[TMP5:%.*]] = fcmp ogt <4 x float> [[WIDE_LOAD]], <float 2.000000e+00, float 2.000000e+00, float 2.000000e+00, float 2.000000e+00>
; CHECK-NEXT: [[TMP7:%.*]] = and <4 x i1> [[TMP6]], [[TMP5]]
; CHECK-NEXT: [[TMP8:%.*]] = xor <4 x i1> [[TMP5]], <i1 true, i1 true, i1 true, i1 true>
; CHECK-NEXT: [[TMP9:%.*]] = and <4 x i1> [[TMP6]], [[TMP8]]
; CHECK-NEXT: [[TMP10:%.*]] = xor <4 x i1> [[TMP3]], <i1 true, i1 true, i1 true, i1 true>
; CHECK-NEXT: [[PREDPHI_V:%.*]] = select <4 x i1> [[TMP7]], <4 x float> [[WIDE_LOAD1]], <4 x float> [[WIDE_LOAD]]
; CHECK-NEXT: [[TMP11:%.*]] = select <4 x i1> [[TMP10]], <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x i1> [[TMP9]]
; CHECK-NEXT: [[PREDPHI2:%.*]] = select <4 x i1> [[TMP11]], <4 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, <4 x float> [[PREDPHI_V]]
; CHECK-NEXT: [[PREDPHI3]] = fadd fast <4 x float> [[VEC_PHI]], [[PREDPHI2]]
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; CHECK-NEXT: [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], 128
; CHECK-NEXT: br i1 [[TMP12]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP20:![0-9]+]]
; CHECK: middle.block:
; CHECK-NEXT: [[TMP13:%.*]] = call fast float @llvm.vector.reduce.fadd.v4f32(float -0.000000e+00, <4 x float> [[PREDPHI3]])
; CHECK-NEXT: br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
; CHECK: scalar.ph:
; CHECK-NEXT: br label [[FOR_BODY:%.*]]
; CHECK: for.body:
; CHECK-NEXT: br i1 poison, label [[IF_THEN:%.*]], label [[FOR_INC:%.*]]
; CHECK: if.then:
; CHECK-NEXT: br i1 poison, label [[IF_THEN8:%.*]], label [[IF_ELSE:%.*]]
; CHECK: if.then8:
; CHECK-NEXT: br label [[FOR_INC]]
; CHECK: if.else:
; CHECK-NEXT: br i1 poison, label [[IF_THEN16:%.*]], label [[FOR_INC]]
; CHECK: if.then16:
; CHECK-NEXT: br label [[FOR_INC]]
; CHECK: for.inc:
; CHECK-NEXT: br i1 poison, label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP21:![0-9]+]]
; CHECK: for.end:
; CHECK-NEXT: [[SUM_1_LCSSA:%.*]] = phi float [ poison, [[FOR_INC]] ], [ [[TMP13]], [[MIDDLE_BLOCK]] ]
; CHECK-NEXT: ret float [[SUM_1_LCSSA]]
;
entry:
br label %for.body
for.body:
%indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %for.inc ]
%sum.033 = phi float [ %S, %entry ], [ %sum.1, %for.inc ]
%arrayidx = getelementptr inbounds float, ptr %A, i64 %indvars.iv
%0 = load float, ptr %arrayidx, align 4
%arrayidx2 = getelementptr inbounds float, ptr %B, i64 %indvars.iv
%1 = load float, ptr %arrayidx2, align 4
%cmp3 = fcmp ogt float %0, %1
br i1 %cmp3, label %if.then, label %for.inc
if.then:
%cmp6 = fcmp ogt float %1, 1.000000e+00
br i1 %cmp6, label %if.then8, label %if.else
if.then8:
%add = fadd fast float %sum.033, %0
br label %for.inc
if.else:
%cmp14 = fcmp ogt float %0, 2.000000e+00
br i1 %cmp14, label %if.then16, label %for.inc
if.then16:
%add19 = fadd fast float %sum.033, %1
br label %for.inc
for.inc:
%sum.1 = phi float [ %add, %if.then8 ], [ %add19, %if.then16 ], [ %sum.033, %if.else ], [ %sum.033, %for.body ]
%indvars.iv.next = add i64 %indvars.iv, 1
%lftr.wideiv = trunc i64 %indvars.iv.next to i32
%exitcond = icmp ne i32 %lftr.wideiv, 128
br i1 %exitcond, label %for.body, label %for.end
for.end:
%sum.1.lcssa = phi float [ %sum.1, %for.inc ]
ret float %sum.1.lcssa
}
; We can't vectorize reductions with phi inputs from outside the reduction.
define float @noreduction_phi(ptr %A, ptr %B, ptr %C, float %S) {
; CHECK-LABEL: define float @noreduction_phi(
; CHECK-SAME: ptr [[A:%.*]], ptr [[B:%.*]], ptr [[C:%.*]], float [[S:%.*]]) {
; CHECK-NEXT: entry:
; CHECK-NEXT: br label [[FOR_BODY:%.*]]
; CHECK: for.body:
; CHECK-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_INC:%.*]] ]
; CHECK-NEXT: [[SUM_033:%.*]] = phi float [ [[S]], [[ENTRY]] ], [ [[SUM_1:%.*]], [[FOR_INC]] ]
; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[INDVARS_IV]]
; CHECK-NEXT: [[TMP0:%.*]] = load float, ptr [[ARRAYIDX]], align 4
; CHECK-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[INDVARS_IV]]
; CHECK-NEXT: [[TMP1:%.*]] = load float, ptr [[ARRAYIDX2]], align 4
; CHECK-NEXT: [[CMP3:%.*]] = fcmp ogt float [[TMP0]], [[TMP1]]
; CHECK-NEXT: br i1 [[CMP3]], label [[IF_THEN:%.*]], label [[FOR_INC]]
; CHECK: if.then:
; CHECK-NEXT: [[CMP6:%.*]] = fcmp ogt float [[TMP1]], 1.000000e+00
; CHECK-NEXT: br i1 [[CMP6]], label [[IF_THEN8:%.*]], label [[IF_ELSE:%.*]]
; CHECK: if.then8:
; CHECK-NEXT: [[ADD:%.*]] = fadd fast float [[SUM_033]], [[TMP0]]
; CHECK-NEXT: br label [[FOR_INC]]
; CHECK: if.else:
; CHECK-NEXT: [[CMP14:%.*]] = fcmp ogt float [[TMP0]], 2.000000e+00
; CHECK-NEXT: br i1 [[CMP14]], label [[IF_THEN16:%.*]], label [[FOR_INC]]
; CHECK: if.then16:
; CHECK-NEXT: [[ADD19:%.*]] = fadd fast float [[SUM_033]], [[TMP1]]
; CHECK-NEXT: br label [[FOR_INC]]
; CHECK: for.inc:
; CHECK-NEXT: [[SUM_1]] = phi float [ [[ADD]], [[IF_THEN8]] ], [ [[ADD19]], [[IF_THEN16]] ], [ 0.000000e+00, [[IF_ELSE]] ], [ [[SUM_033]], [[FOR_BODY]] ]
; CHECK-NEXT: [[INDVARS_IV_NEXT]] = add i64 [[INDVARS_IV]], 1
; CHECK-NEXT: [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
; CHECK-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], 128
; CHECK-NEXT: br i1 [[EXITCOND_NOT]], label [[FOR_END:%.*]], label [[FOR_BODY]]
; CHECK: for.end:
; CHECK-NEXT: ret float [[SUM_1]]
;
entry:
br label %for.body
for.body:
%indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %for.inc ]
%sum.033 = phi float [ %S, %entry ], [ %sum.1, %for.inc ]
%arrayidx = getelementptr inbounds float, ptr %A, i64 %indvars.iv
%0 = load float, ptr %arrayidx, align 4
%arrayidx2 = getelementptr inbounds float, ptr %B, i64 %indvars.iv
%1 = load float, ptr %arrayidx2, align 4
%cmp3 = fcmp ogt float %0, %1
br i1 %cmp3, label %if.then, label %for.inc
if.then:
%cmp6 = fcmp ogt float %1, 1.000000e+00
br i1 %cmp6, label %if.then8, label %if.else
if.then8:
%add = fadd fast float %sum.033, %0
br label %for.inc
if.else:
%cmp14 = fcmp ogt float %0, 2.000000e+00
br i1 %cmp14, label %if.then16, label %for.inc
if.then16:
%add19 = fadd fast float %sum.033, %1
br label %for.inc
for.inc:
%sum.1 = phi float [ %add, %if.then8 ], [ %add19, %if.then16 ], [ 0.000000e+00, %if.else ], [ %sum.033, %for.body ]
%indvars.iv.next = add i64 %indvars.iv, 1
%lftr.wideiv = trunc i64 %indvars.iv.next to i32
%exitcond = icmp ne i32 %lftr.wideiv, 128
br i1 %exitcond, label %for.body, label %for.end
for.end:
%sum.1.lcssa = phi float [ %sum.1, %for.inc ]
ret float %sum.1.lcssa
}
; We can't vectorize reductions that feed another header PHI.
define float @noredux_header_phi(ptr %A, ptr %B, ptr %C, float %S) {
; CHECK-LABEL: define float @noredux_header_phi(
; CHECK-SAME: ptr [[A:%.*]], ptr [[B:%.*]], ptr [[C:%.*]], float [[S:%.*]]) {
; CHECK-NEXT: entry:
; CHECK-NEXT: br label [[FOR_BODY:%.*]]
; CHECK: for.body:
; CHECK-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
; CHECK-NEXT: [[SUM2_09:%.*]] = phi float [ 0.000000e+00, [[ENTRY]] ], [ [[ADD1:%.*]], [[FOR_BODY]] ]
; CHECK-NEXT: [[SUM_08:%.*]] = phi float [ [[S]], [[ENTRY]] ], [ [[ADD:%.*]], [[FOR_BODY]] ]
; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[INDVARS_IV]]
; CHECK-NEXT: [[TMP0:%.*]] = load float, ptr [[ARRAYIDX]], align 4
; CHECK-NEXT: [[ADD]] = fadd fast float [[SUM_08]], [[TMP0]]
; CHECK-NEXT: [[ADD1]] = fadd fast float [[SUM2_09]], [[ADD]]
; CHECK-NEXT: [[INDVARS_IV_NEXT]] = add i64 [[INDVARS_IV]], 1
; CHECK-NEXT: [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
; CHECK-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], 128
; CHECK-NEXT: br i1 [[EXITCOND_NOT]], label [[FOR_END:%.*]], label [[FOR_BODY]]
; CHECK: for.end:
; CHECK-NEXT: [[ADD2:%.*]] = fadd fast float [[ADD]], [[ADD1]]
; CHECK-NEXT: ret float [[ADD2]]
;
entry:
br label %for.body
for.body:
%indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %for.body ]
%sum2.09 = phi float [ 0.000000e+00, %entry ], [ %add1, %for.body ]
%sum.08 = phi float [ %S, %entry ], [ %add, %for.body ]
%arrayidx = getelementptr inbounds float, ptr %B, i64 %indvars.iv
%0 = load float, ptr %arrayidx, align 4
%add = fadd fast float %sum.08, %0
%add1 = fadd fast float %sum2.09, %add
%indvars.iv.next = add i64 %indvars.iv, 1
%lftr.wideiv = trunc i64 %indvars.iv.next to i32
%exitcond = icmp ne i32 %lftr.wideiv, 128
br i1 %exitcond, label %for.body, label %for.end
for.end:
%add1.lcssa = phi float [ %add1, %for.body ]
%add.lcssa = phi float [ %add, %for.body ]
%add2 = fadd fast float %add.lcssa, %add1.lcssa
ret float %add2
}
; When vectorizing a reduction whose loop header phi value is used outside the
; loop special care must be taken. Otherwise, the reduced value feeding into the
; outside user misses a few iterations (VF-1) of the loop.
; PR16522
define i32 @phivalueredux(i32 %p) {
; CHECK-LABEL: define i32 @phivalueredux(
; CHECK-SAME: i32 [[P:%.*]]) {
; CHECK-NEXT: entry:
; CHECK-NEXT: br label [[FOR_BODY:%.*]]
; CHECK: for.body:
; CHECK-NEXT: [[T_03:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[INC:%.*]], [[FOR_BODY]] ]
; CHECK-NEXT: [[P_ADDR_02:%.*]] = phi i32 [ [[P]], [[ENTRY]] ], [ [[XOR:%.*]], [[FOR_BODY]] ]
; CHECK-NEXT: [[XOR]] = xor i32 [[P_ADDR_02]], -1
; CHECK-NEXT: [[INC]] = add nuw nsw i32 [[T_03]], 1
; CHECK-NEXT: [[EXITCOND:%.*]] = icmp eq i32 [[INC]], 16
; CHECK-NEXT: br i1 [[EXITCOND]], label [[FOR_END:%.*]], label [[FOR_BODY]]
; CHECK: for.end:
; CHECK-NEXT: ret i32 [[P_ADDR_02]]
;
entry:
br label %for.body
for.body:
%t.03 = phi i32 [ 0, %entry ], [ %inc, %for.body ]
%p.addr.02 = phi i32 [ %p, %entry ], [ %xor, %for.body ]
%xor = xor i32 %p.addr.02, -1
%inc = add nsw i32 %t.03, 1
%exitcond = icmp eq i32 %inc, 16
br i1 %exitcond, label %for.end, label %for.body
for.end:
ret i32 %p.addr.02
}
; Don't vectorize a reduction value that is not the last in a reduction cyle. We
; would loose iterations (VF-1) on the operations after that use.
; PR17498
define i32 @not_last_operation(i32 %p, i32 %val) {
; CHECK-LABEL: define i32 @not_last_operation(
; CHECK-SAME: i32 [[P:%.*]], i32 [[VAL:%.*]]) {
; CHECK-NEXT: entry:
; CHECK-NEXT: [[TOBOOL:%.*]] = icmp ne i32 [[P]], 0
; CHECK-NEXT: br label [[FOR_BODY:%.*]]
; CHECK: for.body:
; CHECK-NEXT: [[INC613_1:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[INC6_1:%.*]], [[FOR_BODY]] ]
; CHECK-NEXT: [[INC511_1:%.*]] = phi i32 [ [[VAL]], [[ENTRY]] ], [ [[INC5_1:%.*]], [[FOR_BODY]] ]
; CHECK-NEXT: [[INC4_1:%.*]] = zext i1 [[TOBOOL]] to i32
; CHECK-NEXT: [[INC511_1_INC4_1:%.*]] = add nsw i32 [[INC511_1]], [[INC4_1]]
; CHECK-NEXT: [[INC5_1]] = add nsw i32 [[INC511_1_INC4_1]], 1
; CHECK-NEXT: [[INC6_1]] = add nuw nsw i32 [[INC613_1]], 1
; CHECK-NEXT: [[EXITCOND_1:%.*]] = icmp eq i32 [[INC6_1]], 22
; CHECK-NEXT: br i1 [[EXITCOND_1]], label [[EXIT:%.*]], label [[FOR_BODY]]
; CHECK: exit:
; CHECK-NEXT: [[INC_2:%.*]] = add nsw i32 [[INC511_1_INC4_1]], 2
; CHECK-NEXT: ret i32 [[INC_2]]
;
entry:
%tobool = icmp eq i32 %p, 0
br label %for.body
for.body:
%inc613.1 = phi i32 [ 0, %entry ], [ %inc6.1, %for.body ]
%inc511.1 = phi i32 [ %val, %entry ], [ %inc5.1, %for.body ]
%0 = zext i1 %tobool to i32
%inc4.1 = xor i32 %0, 1
%inc511.1.inc4.1 = add nsw i32 %inc511.1, %inc4.1
%inc5.1 = add nsw i32 %inc511.1.inc4.1, 1
%inc6.1 = add nsw i32 %inc613.1, 1
%exitcond.1 = icmp eq i32 %inc6.1, 22
br i1 %exitcond.1, label %exit, label %for.body
exit:
%inc.2 = add nsw i32 %inc511.1.inc4.1, 2
ret i32 %inc.2
}
define i32 @reduction_sum_multiuse(i32 %n, ptr %A, ptr %B) {
; CHECK-LABEL: define i32 @reduction_sum_multiuse(
; CHECK-SAME: i32 [[N:%.*]], ptr [[A:%.*]], ptr [[B:%.*]]) {
; CHECK-NEXT: [[TMP1:%.*]] = icmp sgt i32 [[N]], 0
; CHECK-NEXT: br i1 [[TMP1]], label [[DOTLR_PH_PREHEADER:%.*]], label [[END:%.*]]
; CHECK: .lr.ph.preheader:
; CHECK-NEXT: [[TMP2:%.*]] = zext nneg i32 [[N]] to i64
; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[N]], 4
; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; CHECK: vector.ph:
; CHECK-NEXT: [[N_VEC:%.*]] = and i64 [[TMP2]], 2147483644
; 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 i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP7:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[VEC_IND:%.*]] = phi <4 x i32> [ <i32 0, i32 1, i32 2, i32 3>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDEX]]
; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP3]], align 4
; CHECK-NEXT: [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[INDEX]]
; CHECK-NEXT: [[WIDE_LOAD1:%.*]] = load <4 x i32>, ptr [[TMP4]], align 4
; CHECK-NEXT: [[TMP5:%.*]] = add <4 x i32> [[VEC_PHI]], [[VEC_IND]]
; CHECK-NEXT: [[TMP6:%.*]] = add <4 x i32> [[TMP5]], [[WIDE_LOAD]]
; CHECK-NEXT: [[TMP7]] = add <4 x i32> [[TMP6]], [[WIDE_LOAD1]]
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; CHECK-NEXT: [[VEC_IND_NEXT]] = add <4 x i32> [[VEC_IND]], <i32 4, i32 4, i32 4, i32 4>
; CHECK-NEXT: [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-NEXT: br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP22:![0-9]+]]
; CHECK: middle.block:
; CHECK-NEXT: [[TMP9:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[TMP7]])
; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N_VEC]], [[TMP2]]
; CHECK-NEXT: br i1 [[CMP_N]], label [[DOT_CRIT_EDGE:%.*]], label [[SCALAR_PH]]
; CHECK: scalar.ph:
; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[DOTLR_PH_PREHEADER]] ]
; CHECK-NEXT: [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP9]], [[MIDDLE_BLOCK]] ], [ 0, [[DOTLR_PH_PREHEADER]] ]
; CHECK-NEXT: br label [[DOTLR_PH:%.*]]
; CHECK: .lr.ph:
; CHECK-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[DOTLR_PH]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
; CHECK-NEXT: [[SUM_02:%.*]] = phi i32 [ [[TMP17:%.*]], [[DOTLR_PH]] ], [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ]
; CHECK-NEXT: [[TMP10:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDVARS_IV]]
; CHECK-NEXT: [[TMP11:%.*]] = load i32, ptr [[TMP10]], align 4
; CHECK-NEXT: [[TMP12:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[INDVARS_IV]]
; CHECK-NEXT: [[TMP13:%.*]] = load i32, ptr [[TMP12]], align 4
; CHECK-NEXT: [[TMP14:%.*]] = trunc i64 [[INDVARS_IV]] to i32
; CHECK-NEXT: [[TMP15:%.*]] = add i32 [[SUM_02]], [[TMP14]]
; CHECK-NEXT: [[TMP16:%.*]] = add i32 [[TMP15]], [[TMP11]]
; CHECK-NEXT: [[TMP17]] = add i32 [[TMP16]], [[TMP13]]
; CHECK-NEXT: [[INDVARS_IV_NEXT]] = add i64 [[INDVARS_IV]], 1
; CHECK-NEXT: [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
; CHECK-NEXT: [[EXITCOND:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], [[N]]
; CHECK-NEXT: br i1 [[EXITCOND]], label [[DOT_CRIT_EDGE]], label [[DOTLR_PH]], !llvm.loop [[LOOP23:![0-9]+]]
; CHECK: ._crit_edge:
; CHECK-NEXT: [[SUM_COPY:%.*]] = phi i32 [ [[TMP17]], [[DOTLR_PH]] ], [ [[TMP9]], [[MIDDLE_BLOCK]] ]
; CHECK-NEXT: [[TMP18:%.*]] = shl i32 [[SUM_COPY]], 1
; CHECK-NEXT: br label [[END]]
; CHECK: end:
; CHECK-NEXT: [[F2:%.*]] = phi i32 [ 0, [[TMP0:%.*]] ], [ [[TMP18]], [[DOT_CRIT_EDGE]] ]
; CHECK-NEXT: ret i32 [[F2]]
;
%1 = icmp sgt i32 %n, 0
br i1 %1, label %.lr.ph.preheader, label %end
.lr.ph.preheader: ; preds = %0
br label %.lr.ph
.lr.ph: ; preds = %0, %.lr.ph
%indvars.iv = phi i64 [ %indvars.iv.next, %.lr.ph ], [ 0, %.lr.ph.preheader ]
%sum.02 = phi i32 [ %9, %.lr.ph ], [ 0, %.lr.ph.preheader ]
%2 = getelementptr inbounds i32, ptr %A, i64 %indvars.iv
%3 = load i32, ptr %2, align 4
%4 = getelementptr inbounds i32, ptr %B, i64 %indvars.iv
%5 = load i32, ptr %4, align 4
%6 = trunc i64 %indvars.iv to i32
%7 = add i32 %sum.02, %6
%8 = add i32 %7, %3
%9 = add i32 %8, %5
%indvars.iv.next = add i64 %indvars.iv, 1
%lftr.wideiv = trunc i64 %indvars.iv.next to i32
%exitcond = icmp eq i32 %lftr.wideiv, %n
br i1 %exitcond, label %._crit_edge, label %.lr.ph
._crit_edge: ; preds = %.lr.ph, %0
%sum.lcssa = phi i32 [ %9, %.lr.ph ]
%sum.copy = phi i32 [ %9, %.lr.ph ]
br label %end
end:
%f1 = phi i32 [ 0, %0 ], [ %sum.lcssa, %._crit_edge ]
%f2 = phi i32 [ 0, %0 ], [ %sum.copy, %._crit_edge ]
%final = add i32 %f1, %f2
ret i32 %final
}
; This looks like a predicated reduction, but it is a reset of the reduction
; variable. We cannot vectorize this.
define void @reduction_reset(i32 %N, ptr %arrayA, ptr %arrayB) {
; CHECK-LABEL: define void @reduction_reset(
; CHECK-SAME: i32 [[N:%.*]], ptr [[ARRAYA:%.*]], ptr [[ARRAYB:%.*]]) {
; CHECK-NEXT: entry:
; CHECK-NEXT: [[C4:%.*]] = icmp sgt i32 [[N]], 0
; CHECK-NEXT: br i1 [[C4]], label [[DOTLR_PH_PREHEADER:%.*]], label [[DOT_CRIT_EDGE:%.*]]
; CHECK: .lr.ph.preheader:
; CHECK-NEXT: [[C5:%.*]] = add nsw i32 [[N]], -1
; CHECK-NEXT: [[WIDE_TRIP_COUNT:%.*]] = zext nneg i32 [[N]] to i64
; CHECK-NEXT: br label [[DOTLR_PH:%.*]]
; CHECK: .lr.ph:
; CHECK-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ 0, [[DOTLR_PH_PREHEADER]] ], [ [[INDVARS_IV_NEXT:%.*]], [[DOTLR_PH]] ]
; CHECK-NEXT: [[DOT017:%.*]] = phi i32 [ 100, [[DOTLR_PH_PREHEADER]] ], [ [[CSEL:%.*]], [[DOTLR_PH]] ]
; CHECK-NEXT: [[C6:%.*]] = getelementptr inbounds i32, ptr [[ARRAYA]], i64 [[INDVARS_IV]]
; CHECK-NEXT: [[C7:%.*]] = load i32, ptr [[C6]], align 4
; CHECK-NEXT: [[C8:%.*]] = icmp sgt i32 [[C7]], 0
; CHECK-NEXT: [[C9:%.*]] = add nsw i32 [[C7]], [[DOT017]]
; CHECK-NEXT: [[CSEL]] = select i1 [[C8]], i32 [[C9]], i32 0
; CHECK-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
; CHECK-NEXT: [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], [[WIDE_TRIP_COUNT]]
; CHECK-NEXT: br i1 [[EXITCOND]], label [[DOT_CRIT_EDGE_LOOPEXIT:%.*]], label [[DOTLR_PH]]
; CHECK: ._crit_edge.loopexit:
; CHECK-NEXT: [[PHITMP19:%.*]] = sext i32 [[C5]] to i64
; CHECK-NEXT: br label [[DOT_CRIT_EDGE]]
; CHECK: ._crit_edge:
; CHECK-NEXT: [[DOT015_LCSSA:%.*]] = phi i64 [ -1, [[ENTRY:%.*]] ], [ [[PHITMP19]], [[DOT_CRIT_EDGE_LOOPEXIT]] ]
; CHECK-NEXT: [[DOT0_LCSSA:%.*]] = phi i32 [ 100, [[ENTRY]] ], [ [[CSEL]], [[DOT_CRIT_EDGE_LOOPEXIT]] ]
; CHECK-NEXT: [[C10:%.*]] = getelementptr inbounds i32, ptr [[ARRAYB]], i64 [[DOT015_LCSSA]]
; CHECK-NEXT: store i32 [[DOT0_LCSSA]], ptr [[C10]], align 4
; CHECK-NEXT: ret void
;
entry:
%c4 = icmp sgt i32 %N, 0
br i1 %c4, label %.lr.ph.preheader, label %._crit_edge
.lr.ph.preheader: ; preds = %entry
%c5 = add i32 %N, -1
%wide.trip.count = zext i32 %N to i64
br label %.lr.ph
.lr.ph: ; preds = %.lr.ph, %.lr.ph.preheader
%indvars.iv = phi i64 [ 0, %.lr.ph.preheader ], [ %indvars.iv.next, %.lr.ph ]
%.017 = phi i32 [ 100, %.lr.ph.preheader ], [ %csel, %.lr.ph ]
%c6 = getelementptr inbounds i32, ptr %arrayA, i64 %indvars.iv
%c7 = load i32, ptr %c6, align 4
%c8 = icmp sgt i32 %c7, 0
%c9 = add nsw i32 %c7, %.017
%csel = select i1 %c8, i32 %c9, i32 0
%indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
%exitcond = icmp eq i64 %indvars.iv.next, %wide.trip.count
br i1 %exitcond, label %._crit_edge.loopexit, label %.lr.ph
._crit_edge.loopexit: ; preds = %.lr.ph
%csel.lcssa = phi i32 [ %csel, %.lr.ph ]
%phitmp19 = sext i32 %c5 to i64
br label %._crit_edge
._crit_edge: ; preds = %._crit_edge.loopexit, %entry
%.015.lcssa = phi i64 [ -1, %entry ], [ %phitmp19, %._crit_edge.loopexit ]
%.0.lcssa = phi i32 [ 100, %entry ], [ %csel.lcssa, %._crit_edge.loopexit ]
%c10 = getelementptr inbounds i32, ptr %arrayB, i64 %.015.lcssa
store i32 %.0.lcssa, ptr %c10, align 4
ret void
}
; Can vectorize reduction with redundant single-operand phi input.
define i64 @reduction_with_phi_with_one_incoming_on_backedge(i16 %n, ptr %A) {
; CHECK-LABEL: define i64 @reduction_with_phi_with_one_incoming_on_backedge(
; CHECK-SAME: i16 [[N:%.*]], ptr [[A:%.*]]) {
; CHECK-NEXT: entry:
; CHECK-NEXT: [[SMAX:%.*]] = call i16 @llvm.smax.i16(i16 [[N]], i16 2)
; CHECK-NEXT: [[TMP0:%.*]] = add nsw i16 [[SMAX]], -1
; CHECK-NEXT: [[TMP1:%.*]] = zext nneg i16 [[TMP0]] to i32
; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp slt i16 [[N]], 5
; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; CHECK: vector.ph:
; CHECK-NEXT: [[N_VEC:%.*]] = and i32 [[TMP1]], 32764
; CHECK-NEXT: [[DOTCAST:%.*]] = trunc i32 [[N_VEC]] to i16
; CHECK-NEXT: [[IND_END:%.*]] = or disjoint i16 [[DOTCAST]], 1
; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]
; CHECK: vector.body:
; CHECK-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[VEC_PHI:%.*]] = phi <4 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP4:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[DOTCAST1:%.*]] = trunc i32 [[INDEX]] to i16
; CHECK-NEXT: [[OFFSET_IDX:%.*]] = or disjoint i16 [[DOTCAST1]], 1
; CHECK-NEXT: [[TMP2:%.*]] = sext i16 [[OFFSET_IDX]] to i64
; CHECK-NEXT: [[TMP3:%.*]] = getelementptr i64, ptr [[A]], i64 [[TMP2]]
; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP3]], align 4
; CHECK-NEXT: [[TMP4]] = add <4 x i64> [[VEC_PHI]], [[WIDE_LOAD]]
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
; CHECK-NEXT: [[TMP5:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-NEXT: br i1 [[TMP5]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP24:![0-9]+]]
; CHECK: middle.block:
; CHECK-NEXT: [[TMP6:%.*]] = call i64 @llvm.vector.reduce.add.v4i64(<4 x i64> [[TMP4]])
; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[N_VEC]], [[TMP1]]
; CHECK-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
; CHECK: scalar.ph:
; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i16 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 1, [[ENTRY:%.*]] ]
; CHECK-NEXT: [[BC_MERGE_RDX:%.*]] = phi i64 [ [[TMP6]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
; CHECK-NEXT: br label [[LOOP_HEADER:%.*]]
; CHECK: loop.header:
; CHECK-NEXT: [[IV:%.*]] = phi i16 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP_LATCH:%.*]] ]
; CHECK-NEXT: [[SUM:%.*]] = phi i64 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[SUM_NEXT:%.*]], [[LOOP_LATCH]] ]
; CHECK-NEXT: [[TMP7:%.*]] = sext i16 [[IV]] to i64
; CHECK-NEXT: [[GEP_A:%.*]] = getelementptr i64, ptr [[A]], i64 [[TMP7]]
; CHECK-NEXT: [[LV_A:%.*]] = load i64, ptr [[GEP_A]], align 4
; CHECK-NEXT: [[SUM_NEXT]] = add nsw i64 [[SUM]], [[LV_A]]
; CHECK-NEXT: br label [[LOOP_BB:%.*]]
; CHECK: loop.bb:
; CHECK-NEXT: br label [[LOOP_LATCH]]
; CHECK: loop.latch:
; CHECK-NEXT: [[IV_NEXT]] = add nsw i16 [[IV]], 1
; CHECK-NEXT: [[COND:%.*]] = icmp slt i16 [[IV_NEXT]], [[N]]
; CHECK-NEXT: br i1 [[COND]], label [[LOOP_HEADER]], label [[EXIT]], !llvm.loop [[LOOP25:![0-9]+]]
; CHECK: exit:
; CHECK-NEXT: [[LCSSA_EXIT:%.*]] = phi i64 [ [[SUM_NEXT]], [[LOOP_LATCH]] ], [ [[TMP6]], [[MIDDLE_BLOCK]] ]
; CHECK-NEXT: ret i64 [[LCSSA_EXIT]]
;
entry:
br label %loop.header
loop.header:
%iv = phi i16 [ 1, %entry ], [ %iv.next, %loop.latch ]
%sum = phi i64 [ 0, %entry ], [ %phi.sum.next, %loop.latch ]
%gep.A = getelementptr i64, ptr %A, i16 %iv
%lv.A = load i64, ptr %gep.A
%sum.next = add nsw i64 %sum, %lv.A
br label %loop.bb
loop.bb:
%phi.sum.next = phi i64 [ %sum.next, %loop.header ]
br label %loop.latch
loop.latch:
%iv.next = add nsw i16 %iv, 1
%cond = icmp slt i16 %iv.next, %n
br i1 %cond, label %loop.header, label %exit
exit:
%lcssa.exit = phi i64 [ %phi.sum.next, %loop.latch ]
ret i64 %lcssa.exit
}
; Can vectorize reduction with redundant two-operand phi input.
define i64 @reduction_with_phi_with_two_incoming_on_backedge(i16 %n, ptr %A) {
; CHECK-LABEL: define i64 @reduction_with_phi_with_two_incoming_on_backedge(
; CHECK-SAME: i16 [[N:%.*]], ptr [[A:%.*]]) {
; CHECK-NEXT: entry:
; CHECK-NEXT: [[SMAX:%.*]] = call i16 @llvm.smax.i16(i16 [[N]], i16 2)
; CHECK-NEXT: [[TMP0:%.*]] = add nsw i16 [[SMAX]], -1
; CHECK-NEXT: [[TMP1:%.*]] = zext nneg i16 [[TMP0]] to i32
; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp slt i16 [[N]], 5
; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; CHECK: vector.ph:
; CHECK-NEXT: [[N_VEC:%.*]] = and i32 [[TMP1]], 32764
; CHECK-NEXT: [[DOTCAST:%.*]] = trunc i32 [[N_VEC]] to i16
; CHECK-NEXT: [[IND_END:%.*]] = or disjoint i16 [[DOTCAST]], 1
; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]
; CHECK: vector.body:
; CHECK-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[VEC_PHI:%.*]] = phi <4 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP4:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[DOTCAST1:%.*]] = trunc i32 [[INDEX]] to i16
; CHECK-NEXT: [[OFFSET_IDX:%.*]] = or disjoint i16 [[DOTCAST1]], 1
; CHECK-NEXT: [[TMP2:%.*]] = sext i16 [[OFFSET_IDX]] to i64
; CHECK-NEXT: [[TMP3:%.*]] = getelementptr i64, ptr [[A]], i64 [[TMP2]]
; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP3]], align 4
; CHECK-NEXT: [[TMP4]] = add <4 x i64> [[VEC_PHI]], [[WIDE_LOAD]]
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
; CHECK-NEXT: [[TMP5:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-NEXT: br i1 [[TMP5]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP26:![0-9]+]]
; CHECK: middle.block:
; CHECK-NEXT: [[TMP6:%.*]] = call i64 @llvm.vector.reduce.add.v4i64(<4 x i64> [[TMP4]])
; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[N_VEC]], [[TMP1]]
; CHECK-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
; CHECK: scalar.ph:
; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i16 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 1, [[ENTRY:%.*]] ]
; CHECK-NEXT: [[BC_MERGE_RDX:%.*]] = phi i64 [ [[TMP6]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
; CHECK-NEXT: br label [[LOOP_HEADER:%.*]]
; CHECK: loop.header:
; CHECK-NEXT: [[IV:%.*]] = phi i16 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP_LATCH:%.*]] ]
; CHECK-NEXT: [[SUM:%.*]] = phi i64 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[SUM_NEXT:%.*]], [[LOOP_LATCH]] ]
; CHECK-NEXT: [[TMP7:%.*]] = sext i16 [[IV]] to i64
; CHECK-NEXT: [[GEP_A:%.*]] = getelementptr i64, ptr [[A]], i64 [[TMP7]]
; CHECK-NEXT: [[LV_A:%.*]] = load i64, ptr [[GEP_A]], align 4
; CHECK-NEXT: [[SUM_NEXT]] = add nsw i64 [[SUM]], [[LV_A]]
; CHECK-NEXT: [[CMP_0:%.*]] = icmp eq i64 [[LV_A]], 29
; CHECK-NEXT: br i1 [[CMP_0]], label [[LOOP_BB:%.*]], label [[LOOP_LATCH]]
; CHECK: loop.bb:
; CHECK-NEXT: br label [[LOOP_LATCH]]
; CHECK: loop.latch:
; CHECK-NEXT: [[IV_NEXT]] = add nsw i16 [[IV]], 1
; CHECK-NEXT: [[COND:%.*]] = icmp slt i16 [[IV_NEXT]], [[N]]
; CHECK-NEXT: br i1 [[COND]], label [[LOOP_HEADER]], label [[EXIT]], !llvm.loop [[LOOP27:![0-9]+]]
; CHECK: exit:
; CHECK-NEXT: [[LCSSA_EXIT:%.*]] = phi i64 [ [[SUM_NEXT]], [[LOOP_LATCH]] ], [ [[TMP6]], [[MIDDLE_BLOCK]] ]
; CHECK-NEXT: ret i64 [[LCSSA_EXIT]]
;
entry:
br label %loop.header
loop.header:
%iv = phi i16 [ 1, %entry ], [ %iv.next, %loop.latch ]
%sum = phi i64 [ 0, %entry ], [ %phi.sum.next, %loop.latch ]
%gep.A = getelementptr i64, ptr %A, i16 %iv
%lv.A = load i64, ptr %gep.A
%sum.next = add nsw i64 %sum, %lv.A
%cmp.0 = icmp eq i64 %lv.A, 29
br i1 %cmp.0, label %loop.bb, label %loop.latch
loop.bb:
br label %loop.latch
loop.latch:
%phi.sum.next = phi i64 [ %sum.next, %loop.bb ], [ %sum.next, %loop.header ]
%iv.next = add nsw i16 %iv, 1
%cond = icmp slt i16 %iv.next, %n
br i1 %cond, label %loop.header, label %exit
exit:
%lcssa.exit = phi i64 [ %phi.sum.next, %loop.latch ]
ret i64 %lcssa.exit
}
Reference in New Issue View Git Blame Copy Permalink