[PATCH] D21449: Target independent codesize heuristics for Loop Idiom Recognition

[Sunita_Marathe via llvm-commits]

Sunita_Marathe added inline comments.

================
Comment at: test/Transforms/LoopIdiom/lir-heurs-multi-block-loop.ll:15
@@ +14,3 @@
+; CHECK: memset
+;
+define void @LoopMemset([2048 x i8]* noalias nocapture %D, i32 %N) optsize {
----------------
Sunita_Marathe wrote:
> mcrosier wrote:
> > I applied you patch, ran this lit test and it fails.  The input IR and the output IR after running the command line was unchanged.
> I need to look into this because the lit test passes for me ...
When I run the lit test in the workspace from which the uploaded patch was created, I get the following IR.   Note that 
- For the first two functions which are intended to check the exempted cases for avoiding LIR in multi-block loops:  LIR has been done and memset calls inserted in the loop preheader 
- For the third function which is intended to check that LIR is avoided for a multi-block top level loop: the pre and post IR is the same.

; ModuleID = '<stdin>'
source_filename = "<stdin>"

declare void @GF()

; Function Attrs: argmemonly nounwind
declare void @llvm.memset.p0i8.i64(i8* nocapture, i8, i64, i32, i1) #0

; Function Attrs: optsize
define void @LoopMemset([2048 x i8]* noalias nocapture %D, i32 %N) #1 {
entry:
  %D1 = bitcast [2048 x i8]* %D to i8*
  %cmp7 = icmp eq i32 %N, 0
  br i1 %cmp7, label %for.end, label %for.body.preheader

for.body.preheader:                               ; preds = %entry
  %0 = add i32 %N, -1
  %1 = zext i32 %0 to i64
  %2 = shl i64 %1, 11
  %3 = add i64 %2, 2048
  call void @llvm.memset.p0i8.i64(i8* %D1, i8 -1, i64 %3, i32 1, i1 false)
  br label %for.body

for.body:                                         ; preds = %for.inc, %for.body.preheader
  %i.08 = phi i32 [ %inc, %for.inc ], [ 0, %for.body.preheader ]
  %cmp1 = icmp ugt i32 %i.08, 15
  br i1 %cmp1, label %if.then, label %for.inc

if.then:                                          ; preds = %for.body
  tail call void @GF()
  br label %for.inc

for.inc:                                          ; preds = %if.then, %for.body
  %inc = add i32 %i.08, 1
  %cmp = icmp ult i32 %inc, %N
  br i1 %cmp, label %for.body, label %for.end.loopexit

for.end.loopexit:                                 ; preds = %for.inc
  br label %for.end

for.end:                                          ; preds = %for.end.loopexit, %entry
  ret void
}


; Function Attrs: optsize
define void @NestedMemset_LoopMemset([2046 x i8]* noalias nocapture %D, i32 %N) #1 {
entry:
  %D1 = bitcast [2046 x i8]* %D to i8*
  %cmp18 = icmp eq i32 %N, 0
  br i1 %cmp18, label %for.end9, label %for.cond1.preheader.preheader

for.cond1.preheader.preheader:                    ; preds = %entry
  %0 = add i32 %N, -1
  %1 = zext i32 %0 to i64
  %2 = mul i64 %1, 2046
  %3 = add i64 %2, 2046
  call void @llvm.memset.p0i8.i64(i8* %D1, i8 0, i64 %3, i32 1, i1 false)
  br label %for.cond1.preheader

for.cond1.preheader:                              ; preds = %for.inc7, %for.cond1.preheader.preheader
  %indvar = phi i64 [ %indvar.next, %for.inc7 ], [ 0, %for.cond1.preheader.preheader ]
  %i.019 = phi i32 [ %inc8, %for.inc7 ], [ 0, %for.cond1.preheader.preheader ]
  br label %for.body3

for.body3:                                        ; preds = %for.inc, %for.cond1.preheader
  %j.017 = phi i32 [ 0, %for.cond1.preheader ], [ %inc, %for.inc ]
  %cmp6 = icmp ugt i32 %j.017, 15
  br i1 %cmp6, label %if.then, label %for.inc

if.then:                                          ; preds = %for.body3
  tail call void @GF()
  br label %for.inc

for.inc:                                          ; preds = %if.then, %for.body3
  %inc = add i32 %j.017, 1
  %cmp2 = icmp ult i32 %inc, 2046
  br i1 %cmp2, label %for.body3, label %for.inc7

for.inc7:                                         ; preds = %for.inc
  %inc8 = add i32 %i.019, 1
  %cmp = icmp ult i32 %inc8, %N
  %indvar.next = add i64 %indvar, 1
  br i1 %cmp, label %for.cond1.preheader, label %for.end9.loopexit

for.end9.loopexit:                                ; preds = %for.inc7
  br label %for.end9

for.end9:                                         ; preds = %for.end9.loopexit, %entry
  ret void
}

; Function Attrs: optsize
define void @Non_NestedMemset(i8* noalias nocapture %D, i32 %N) #1 {
entry:
  %cmp6 = icmp eq i32 %N, 0
  br i1 %cmp6, label %for.end, label %for.body.preheader

for.body.preheader:                               ; preds = %entry
  br label %for.body

for.body:                                         ; preds = %for.inc, %for.body.preheader
  %i.07 = phi i32 [ %inc, %for.inc ], [ 0, %for.body.preheader ]
  %idxprom = zext i32 %i.07 to i64
  %arrayidx = getelementptr inbounds i8, i8* %D, i64 %idxprom
  store i8 0, i8* %arrayidx, align 1
  %cmp1 = icmp ugt i32 %i.07, 15
  br i1 %cmp1, label %if.then, label %for.inc

if.then:                                          ; preds = %for.body
  tail call void @GF()
  br label %for.inc

for.inc:                                          ; preds = %if.then, %for.body
  %inc = add i32 %i.07, 1
  %cmp = icmp ult i32 %inc, %N
  br i1 %cmp, label %for.body, label %for.end.loopexit

for.end.loopexit:                                 ; preds = %for.inc
  br label %for.end

for.end:                                          ; preds = %for.end.loopexit, %entry
  ret void
}




https://reviews.llvm.org/D21449