Home | History | Annotate | Download | only in X86
      1 ; RUN: opt < %s -loop-vectorize -force-vector-interleave=1 -force-vector-width=4 -dce -instcombine -S | FileCheck %s
      2 
      3 target 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"
      4 target triple = "x86_64-unknown-linux-gnu"
      5 
      6 ; The parallel loop has been invalidated by the new memory accesses introduced
      7 ; by reg2mem (Loop::isParallel() starts to return false). Ensure the loop is
      8 ; now non-vectorizable.
      9 
     10 ;CHECK-NOT: <4 x i32>
     11 define void @parallel_loop(i32* nocapture %a, i32* nocapture %b) nounwind uwtable {
     12 entry:
     13   %indvars.iv.next.reg2mem = alloca i64
     14   %indvars.iv.reg2mem = alloca i64
     15   %"reg2mem alloca point" = bitcast i32 0 to i32
     16   store i64 0, i64* %indvars.iv.reg2mem
     17   br label %for.body
     18 
     19 for.body:                                         ; preds = %for.body.for.body_crit_edge, %entry
     20   %indvars.iv.reload = load i64, i64* %indvars.iv.reg2mem
     21   %arrayidx = getelementptr inbounds i32, i32* %b, i64 %indvars.iv.reload
     22   %0 = load i32, i32* %arrayidx, align 4, !llvm.mem.parallel_loop_access !3
     23   %arrayidx2 = getelementptr inbounds i32, i32* %a, i64 %indvars.iv.reload
     24   %1 = load i32, i32* %arrayidx2, align 4, !llvm.mem.parallel_loop_access !3
     25   %idxprom3 = sext i32 %1 to i64
     26   %arrayidx4 = getelementptr inbounds i32, i32* %a, i64 %idxprom3
     27   store i32 %0, i32* %arrayidx4, align 4, !llvm.mem.parallel_loop_access !3
     28   %indvars.iv.next = add i64 %indvars.iv.reload, 1
     29   ; A new store without the parallel metadata here:
     30   store i64 %indvars.iv.next, i64* %indvars.iv.next.reg2mem
     31   %indvars.iv.next.reload1 = load i64, i64* %indvars.iv.next.reg2mem
     32   %arrayidx6 = getelementptr inbounds i32, i32* %b, i64 %indvars.iv.next.reload1
     33   %2 = load i32, i32* %arrayidx6, align 4, !llvm.mem.parallel_loop_access !3
     34   store i32 %2, i32* %arrayidx2, align 4, !llvm.mem.parallel_loop_access !3
     35   %indvars.iv.next.reload = load i64, i64* %indvars.iv.next.reg2mem
     36   %lftr.wideiv = trunc i64 %indvars.iv.next.reload to i32
     37   %exitcond = icmp eq i32 %lftr.wideiv, 512
     38   br i1 %exitcond, label %for.end, label %for.body.for.body_crit_edge, !llvm.loop !3
     39 
     40 for.body.for.body_crit_edge:                      ; preds = %for.body
     41   %indvars.iv.next.reload2 = load i64, i64* %indvars.iv.next.reg2mem
     42   store i64 %indvars.iv.next.reload2, i64* %indvars.iv.reg2mem
     43   br label %for.body
     44 
     45 for.end:                                          ; preds = %for.body
     46   ret void
     47 }
     48 
     49 !3 = !{!3}
     50