brintos

brintos / llvm-project-archived public Read only

0
0
Text · 4.3 KiB · 0b58bb6 Raw
143 lines · plain
1; RUN: opt -passes=loop-vectorize -force-vector-interleave=1 -force-vector-width=4 -S %s | FileCheck --check-prefix=VF4 %s2; RUN: opt -passes=loop-vectorize -force-vector-interleave=2 -force-vector-width=1 -S %s | FileCheck --check-prefix=IC2 %s3 4; rdar://problem/128481625 6target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"7 8@a = common global [2048 x i32] zeroinitializer, align 169 10define void @example12() {11; VF4-LABEL: @example12(12; VF4-LABEL: vector.body:13; VF4: [[VEC_IND:%.+]] = phi <4 x i32>14; VF4: store <4 x i32> [[VEC_IND]]15; VF4: middle.block:16;17; IC2-LABEL: @example12(18; IC2-LABEL: vector.body:19; IC2-NEXT:   [[INDEX:%.+]] = phi i64 [ 0, %vector.ph ]20; IC2:        [[TRUNC:%.+]] = trunc i64 [[INDEX]] to i3221; IC2-NEXT:   [[TRUNC1:%.+]] = add i32 [[TRUNC]], 122; IC2:        store i32 [[TRUNC]],23; IC2-NEXT:   store i32 [[TRUNC1]],24;25entry:26  br label %loop27 28loop:29  %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]30  %gep = getelementptr inbounds [2048 x i32], ptr @a, i64 0, i64 %iv31  %iv.trunc = trunc i64 %iv to i3232  store i32 %iv.trunc, ptr %gep, align 433  %iv.next = add i64 %iv, 134  %iv.next.trunc = trunc i64 %iv.next to i3235  %exitcond = icmp eq i32 %iv.next.trunc, 102436  br i1 %exitcond, label %exit, label %loop37 38exit:39  ret void40}41 42define void @redundant_iv_cast(ptr %dst) {43; VF4-LABEL: @redundant_iv_cast44; VF4: vector.body:45; VF4:   [[VEC_IND:%.+]] = phi <4 x i16> [ <i16 0, i16 1, i16 2, i16 3>, %vector.ph ], [ [[VEC_IND_NEXT:%.+]], %vector.body ]46; VF4:  store <4 x i16> [[VEC_IND]]47; VF4:  [[VEC_IND_NEXT]] = add <4 x i16> [[VEC_IND]], splat (i16 4)48;49; IC2-LABEL: @redundant_iv_cast50; IC2:      vector.body:51; IC2-NEXT:  [[CAN_IV:%.+]] = phi i32 [ 0, %vector.ph ], [ [[CAN_IV_NEXT:%.+]], %vector.body ]52; IC2-NEXT:  [[OFFSET_IDX:%.+]] = trunc i32 [[CAN_IV]] to i1653; IC2-NEXT:  [[P1:%.+]] = add i16 [[OFFSET_IDX]], 154; IC2:       store i16 [[OFFSET_IDX]]55; IC2-NEXT:  store i16 [[P1]]56;57entry:58  br label %loop59 60loop:61  %j.0 = phi i16 [ 0, %entry ], [ %inc, %loop ]62  %ext = zext i16 %j.0 to i3263  %trunc = trunc i32 %ext to i1664  %gep = getelementptr inbounds i16, ptr %dst, i16 %j.065  store i16 %trunc, ptr %gep66  %0 = icmp eq i16 10000, %j.067  %inc = add i16 %j.0, 168  br i1 %0, label %exit, label %loop69 70 71exit:72  ret void73}74 75 76define void @cast_variable_step(i64 %step) {77; VF4-LABEL: @cast_variable_step(78; VF4-LABEL: vector.body:79; VF4: [[VEC_IND:%.+]] = phi <4 x i32>80; VF4: store <4 x i32> [[VEC_IND]]81; VF4: middle.block:82;83; IC2-LABEL: @cast_variable_step(84; IC2:   [[TRUNC_STEP:%.+]] = trunc i64 %step to i3285; IC2:   br label %vector.body86 87; IC2-LABEL: vector.body:88; IC2-NEXT:   [[INDEX:%.+]] = phi i64 [ 0, %vector.ph ]89; IC2:        [[MUL:%.+]] = mul i64 %index, %step90; IC2-NEXT:   [[OFFSET_IDX:%.+]] = add i64 10, [[MUL]]91; IC2-NEXT:   [[TRUNC_OFF:%.+]] = trunc i64 [[OFFSET_IDX]] to i3292; IC2-NEXT:   [[STEP1:%.+]] = mul i32 1, [[TRUNC_STEP]]93; IC2-NEXT:   [[T1:%.+]] = add i32 [[TRUNC_OFF]], [[STEP1]]94; IC2:        store i32 [[TRUNC_OFF]],95; IC2-NEXT:   store i32 [[T1]],96;97entry:98  br label %loop99 100loop:101  %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]102  %iv.2 = phi i64 [ 10, %entry ], [ %iv.2.next, %loop ]103  %gep = getelementptr inbounds [2048 x i32], ptr @a, i64 0, i64 %iv104  %iv.2.trunc = trunc i64 %iv.2 to i32105  store i32 %iv.2.trunc, ptr %gep, align 4106  %iv.next = add i64 %iv, 1107  %iv.2.next = add i64 %iv.2, %step108  %exitcond = icmp eq i64 %iv.next, 1024109  br i1 %exitcond, label %exit, label %loop110 111exit:112  ret void113}114 115define void @cast_induction_tail_folding(ptr %A) {116; VF4-LABEL: @cast_induction_tail_folding(117; VF4-LABEL: vector.body:118; VF4-NEXT:   br i1 true, label %pred.store.if, label %pred.store.continue119 120; IC2-LABEL: @cast_induction_tail_folding(121; IC2:      [[INDEX:%.+]] = phi i32 [ 0, %vector.ph ]122; IC2-NEXT: [[INDEX0:%.+]] = add i32 [[INDEX]], 0123; IC2-NEXT: [[INDEX1:%.+]] = add i32 [[INDEX]], 1124; IC2-NEXT: = icmp ule i32 [[INDEX0]], 2125; IC2-NEXT: = icmp ule i32 [[INDEX1]], 2126;127entry:128  br label %loop129 130loop:131  %iv = phi i32 [ 0, %entry ], [ %iv.next, %loop ]132  %iv.ext = sext i32 %iv to i64133  %iv.trunc  = trunc i64 %iv.ext to i32134  %gep = getelementptr inbounds i32, ptr %A, i64 %iv.ext135  store i32 %iv.trunc, ptr %gep136  %iv.next = add i32 %iv, 1137  %c = icmp slt i32 %iv.next, 3138  br i1 %c, label %loop, label %exit139 140exit:141  ret void142}143