Files
clang-p2996/llvm/test/Transforms/LoopVectorize/X86/induction-step.ll
Florian Hahn 4277c21059 [VPlan] Introduce explicit broadcasts for live-ins. (#124644)
Add a new VPInstruction::Broadcast opcode and use it to materialize
explicit broadcasts of live-ins. The initial patch only materlizes the
broadcasts if the vector preheader dominates all uses that need it.
Later patches will pick the best valid insert point, thus retiring
implicit hoisting of broadcasts from VPTransformsState::get().

PR: https://github.com/llvm/llvm-project/pull/124644
2025-02-26 13:57:51 +00:00

155 lines
8.9 KiB
LLVM

; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
; RUN: opt -passes=loop-vectorize -force-vector-width=4 -S %s | FileCheck %s
target triple = "x86_64-unknown-linux-gnu"
define i16 @wide_add_induction_step_live_in(ptr %dst, i64 %N, i16 %off) {
; CHECK-LABEL: @wide_add_induction_step_live_in(
; CHECK-NEXT: entry:
; CHECK-NEXT: [[O_1:%.*]] = add i16 [[OFF:%.*]], 2
; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 8
; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; CHECK: vector.ph:
; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N]], 8
; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
; CHECK-NEXT: [[DOTCAST:%.*]] = trunc i64 [[N_VEC]] to i16
; CHECK-NEXT: [[TMP0:%.*]] = mul i16 [[DOTCAST]], [[O_1]]
; CHECK-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i16> poison, i16 [[O_1]], i64 0
; CHECK-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i16> [[BROADCAST_SPLATINSERT]], <4 x i16> poison, <4 x i32> zeroinitializer
; CHECK-NEXT: [[TMP1:%.*]] = mul <4 x i16> splat (i16 4), [[BROADCAST_SPLAT]]
; CHECK-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <4 x i16> poison, i16 [[O_1]], i64 0
; CHECK-NEXT: [[DOTSPLAT:%.*]] = shufflevector <4 x i16> [[DOTSPLATINSERT]], <4 x i16> poison, <4 x i32> zeroinitializer
; CHECK-NEXT: [[TMP2:%.*]] = mul <4 x i16> <i16 0, i16 1, i16 2, i16 3>, [[DOTSPLAT]]
; CHECK-NEXT: [[INDUCTION:%.*]] = add <4 x i16> zeroinitializer, [[TMP2]]
; 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 i16> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[STEP_ADD:%.*]] = add <4 x i16> [[VEC_IND]], [[TMP1]]
; CHECK-NEXT: [[TMP3:%.*]] = add i64 [[INDEX]], 0
; CHECK-NEXT: [[TMP4:%.*]] = add <4 x i16> [[VEC_IND]], [[BROADCAST_SPLAT]]
; CHECK-NEXT: [[TMP9:%.*]] = add <4 x i16> [[STEP_ADD]], [[BROADCAST_SPLAT]]
; CHECK-NEXT: [[TMP5:%.*]] = getelementptr inbounds i16, ptr [[DST:%.*]], i64 [[TMP3]]
; CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds i16, ptr [[TMP5]], i32 0
; CHECK-NEXT: [[TMP8:%.*]] = getelementptr inbounds i16, ptr [[TMP5]], i32 4
; CHECK-NEXT: store <4 x i16> [[TMP4]], ptr [[TMP6]], align 2
; CHECK-NEXT: store <4 x i16> [[TMP9]], ptr [[TMP8]], align 2
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
; CHECK-NEXT: [[VEC_IND_NEXT]] = add <4 x i16> [[STEP_ADD]], [[TMP1]]
; CHECK-NEXT: [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-NEXT: br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
; 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: [[BC_RESUME_VAL1:%.*]] = phi i16 [ [[TMP0]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
; CHECK-NEXT: br label [[LOOP:%.*]]
; CHECK: loop:
; CHECK-NEXT: [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
; CHECK-NEXT: [[IV_2:%.*]] = phi i16 [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ], [ [[ADD:%.*]], [[LOOP]] ]
; CHECK-NEXT: [[ADD]] = add i16 [[IV_2]], [[O_1]]
; CHECK-NEXT: [[GEP_DST:%.*]] = getelementptr inbounds i16, ptr [[DST]], i64 [[IV]]
; CHECK-NEXT: store i16 [[ADD]], ptr [[GEP_DST]], align 2
; CHECK-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; CHECK-NEXT: [[EC:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
; CHECK-NEXT: br i1 [[EC]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP3:![0-9]+]]
; CHECK: exit:
; CHECK-NEXT: [[ADD_LCSSA:%.*]] = phi i16 [ [[ADD]], [[LOOP]] ], [ [[TMP0]], [[MIDDLE_BLOCK]] ]
; CHECK-NEXT: ret i16 [[ADD_LCSSA]]
;
entry:
%o.1 = add i16 %off, 2
br label %loop
loop:
%iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
%iv.2 = phi i16 [ 0, %entry ], [ %add, %loop ]
%add = add i16 %iv.2, %o.1
%gep.dst = getelementptr inbounds i16, ptr %dst, i64 %iv
store i16 %add, ptr %gep.dst, align 2
%iv.next = add nuw nsw i64 %iv, 1
%ec = icmp eq i64 %iv.next, %N
br i1 %ec , label %exit, label %loop
exit:
ret i16 %add
}
define i16 @wide_sub_induction_step_live_in(ptr %dst, i64 %N, i16 %off) {
; CHECK-LABEL: @wide_sub_induction_step_live_in(
; CHECK-NEXT: entry:
; CHECK-NEXT: [[O_1:%.*]] = add i16 [[OFF:%.*]], 2
; CHECK-NEXT: [[TMP0:%.*]] = sub i16 -2, [[OFF]]
; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 8
; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; CHECK: vector.ph:
; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N]], 8
; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
; CHECK-NEXT: [[DOTCAST:%.*]] = trunc i64 [[N_VEC]] to i16
; CHECK-NEXT: [[TMP1:%.*]] = mul i16 [[DOTCAST]], [[TMP0]]
; CHECK-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i16> poison, i16 [[O_1]], i64 0
; CHECK-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i16> [[BROADCAST_SPLATINSERT]], <4 x i16> poison, <4 x i32> zeroinitializer
; CHECK-NEXT: [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <4 x i16> poison, i16 [[TMP0]], i64 0
; CHECK-NEXT: [[BROADCAST_SPLAT2:%.*]] = shufflevector <4 x i16> [[BROADCAST_SPLATINSERT1]], <4 x i16> poison, <4 x i32> zeroinitializer
; CHECK-NEXT: [[TMP2:%.*]] = mul <4 x i16> splat (i16 4), [[BROADCAST_SPLAT2]]
; CHECK-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <4 x i16> poison, i16 [[TMP0]], i64 0
; CHECK-NEXT: [[DOTSPLAT:%.*]] = shufflevector <4 x i16> [[DOTSPLATINSERT]], <4 x i16> poison, <4 x i32> zeroinitializer
; CHECK-NEXT: [[TMP3:%.*]] = mul <4 x i16> <i16 0, i16 1, i16 2, i16 3>, [[DOTSPLAT]]
; CHECK-NEXT: [[INDUCTION:%.*]] = add <4 x i16> zeroinitializer, [[TMP3]]
; 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 i16> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[STEP_ADD:%.*]] = add <4 x i16> [[VEC_IND]], [[TMP2]]
; CHECK-NEXT: [[TMP4:%.*]] = add i64 [[INDEX]], 0
; CHECK-NEXT: [[TMP5:%.*]] = sub <4 x i16> [[VEC_IND]], [[BROADCAST_SPLAT]]
; CHECK-NEXT: [[TMP10:%.*]] = sub <4 x i16> [[STEP_ADD]], [[BROADCAST_SPLAT]]
; CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds i16, ptr [[DST:%.*]], i64 [[TMP4]]
; CHECK-NEXT: [[TMP7:%.*]] = getelementptr inbounds i16, ptr [[TMP6]], i32 0
; CHECK-NEXT: [[TMP9:%.*]] = getelementptr inbounds i16, ptr [[TMP6]], i32 4
; CHECK-NEXT: store <4 x i16> [[TMP5]], ptr [[TMP7]], align 2
; CHECK-NEXT: store <4 x i16> [[TMP10]], ptr [[TMP9]], align 2
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
; CHECK-NEXT: [[VEC_IND_NEXT]] = add <4 x i16> [[STEP_ADD]], [[TMP2]]
; 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: [[TMP11:%.*]] = extractelement <4 x i16> [[TMP10]], i32 3
; 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: [[BC_RESUME_VAL3:%.*]] = phi i16 [ [[TMP1]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
; CHECK-NEXT: br label [[LOOP:%.*]]
; CHECK: loop:
; CHECK-NEXT: [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
; CHECK-NEXT: [[IV_2:%.*]] = phi i16 [ [[BC_RESUME_VAL3]], [[SCALAR_PH]] ], [ [[SUB:%.*]], [[LOOP]] ]
; CHECK-NEXT: [[SUB]] = sub i16 [[IV_2]], [[O_1]]
; CHECK-NEXT: [[GEP_DST:%.*]] = getelementptr inbounds i16, ptr [[DST]], i64 [[IV]]
; CHECK-NEXT: store i16 [[SUB]], ptr [[GEP_DST]], align 2
; CHECK-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; CHECK-NEXT: [[EC:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
; CHECK-NEXT: br i1 [[EC]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP5:![0-9]+]]
; CHECK: exit:
; CHECK-NEXT: [[SUB_LCSSA:%.*]] = phi i16 [ [[SUB]], [[LOOP]] ], [ [[TMP11]], [[MIDDLE_BLOCK]] ]
; CHECK-NEXT: ret i16 [[SUB_LCSSA]]
;
entry:
%o.1 = add i16 %off, 2
br label %loop
loop:
%iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
%iv.2 = phi i16 [ 0, %entry ], [ %sub, %loop ]
%sub = sub i16 %iv.2, %o.1
%gep.dst = getelementptr inbounds i16, ptr %dst, i64 %iv
store i16 %sub, ptr %gep.dst, align 2
%iv.next = add nuw nsw i64 %iv, 1
%ec = icmp eq i64 %iv.next, %N
br i1 %ec , label %exit, label %loop
exit:
ret i16 %sub
}