bf9473b2d8
Regions with one affine loop can be profitable if the loop is distributable. To this end we will allow them to be treated as profitable if they contain at least two non-trivial basic blocks. llvm-svn: 269064
168 lines
6.2 KiB
LLVM
168 lines
6.2 KiB
LLVM
; RUN: opt %loadPolly -polly-detect -polly-process-unprofitable=false -analyze \
|
|
; RUN: -polly-allow-nonaffine-loops < %s | FileCheck %s
|
|
;
|
|
; Even if we allow non-affine loops we can only model the outermost loop, all
|
|
; other loops are boxed in non-affine regions. However, the inner loops can be
|
|
; distributed as black-boxes, thus we will recognize the outer loop as profitable.
|
|
;
|
|
; CHECK: Valid Region for Scop: for.cond => for.end.51
|
|
;
|
|
; void f(int *A) {
|
|
; for (int i = 0; i < 100; i++) {
|
|
; // Non-affine
|
|
; for (int j = 0; j < i * i; j++)
|
|
; for (int k = 0; k < i; k++)
|
|
; A[i]++;
|
|
; // Non-affine
|
|
; for (int j = 0; j < i * i; j++)
|
|
; // Non-affine
|
|
; for (int k = 0; k < j; k++)
|
|
; A[i]++;
|
|
; // Non-affine
|
|
; if (A[i])
|
|
; for (int j = 0; j < 100; j++)
|
|
; for (int k = 0; k < j * j; k++)
|
|
; A[i]++;
|
|
; }
|
|
; }
|
|
;
|
|
target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
|
|
|
|
define void @f(i32* %A) {
|
|
entry:
|
|
br label %for.cond
|
|
|
|
for.cond: ; preds = %for.inc.49, %entry
|
|
%indvars.iv5 = phi i64 [ %indvars.iv.next6, %for.inc.49 ], [ 0, %entry ]
|
|
%indvars.iv = phi i32 [ %indvars.iv.next, %for.inc.49 ], [ 0, %entry ]
|
|
%exitcond9 = icmp ne i64 %indvars.iv5, 100
|
|
br i1 %exitcond9, label %for.body, label %for.end.51
|
|
|
|
for.body: ; preds = %for.cond
|
|
br label %for.cond.1
|
|
|
|
for.cond.1: ; preds = %for.inc.8, %for.body
|
|
%j.0 = phi i32 [ 0, %for.body ], [ %inc9, %for.inc.8 ]
|
|
%tmp = mul nsw i64 %indvars.iv5, %indvars.iv5
|
|
%tmp10 = sext i32 %j.0 to i64
|
|
%cmp2 = icmp slt i64 %tmp10, %tmp
|
|
br i1 %cmp2, label %for.body.3, label %for.end.10
|
|
|
|
for.body.3: ; preds = %for.cond.1
|
|
br label %for.cond.4
|
|
|
|
for.cond.4: ; preds = %for.inc, %for.body.3
|
|
%k.0 = phi i32 [ 0, %for.body.3 ], [ %inc7, %for.inc ]
|
|
%exitcond = icmp ne i32 %k.0, %indvars.iv
|
|
br i1 %exitcond, label %for.body.6, label %for.end
|
|
|
|
for.body.6: ; preds = %for.cond.4
|
|
%arrayidx = getelementptr inbounds i32, i32* %A, i64 %indvars.iv5
|
|
%tmp11 = load i32, i32* %arrayidx, align 4
|
|
%inc = add nsw i32 %tmp11, 1
|
|
store i32 %inc, i32* %arrayidx, align 4
|
|
br label %for.inc
|
|
|
|
for.inc: ; preds = %for.body.6
|
|
%inc7 = add nuw nsw i32 %k.0, 1
|
|
br label %for.cond.4
|
|
|
|
for.end: ; preds = %for.cond.4
|
|
br label %for.inc.8
|
|
|
|
for.inc.8: ; preds = %for.end
|
|
%inc9 = add nuw nsw i32 %j.0, 1
|
|
br label %for.cond.1
|
|
|
|
for.end.10: ; preds = %for.cond.1
|
|
br label %for.cond.12
|
|
|
|
for.cond.12: ; preds = %for.inc.26, %for.end.10
|
|
%indvars.iv1 = phi i32 [ %indvars.iv.next2, %for.inc.26 ], [ 0, %for.end.10 ]
|
|
%tmp12 = mul nsw i64 %indvars.iv5, %indvars.iv5
|
|
%tmp13 = sext i32 %indvars.iv1 to i64
|
|
%cmp14 = icmp slt i64 %tmp13, %tmp12
|
|
br i1 %cmp14, label %for.body.15, label %for.end.28
|
|
|
|
for.body.15: ; preds = %for.cond.12
|
|
br label %for.cond.17
|
|
|
|
for.cond.17: ; preds = %for.inc.23, %for.body.15
|
|
%k16.0 = phi i32 [ 0, %for.body.15 ], [ %inc24, %for.inc.23 ]
|
|
%exitcond3 = icmp ne i32 %k16.0, %indvars.iv1
|
|
br i1 %exitcond3, label %for.body.19, label %for.end.25
|
|
|
|
for.body.19: ; preds = %for.cond.17
|
|
%arrayidx21 = getelementptr inbounds i32, i32* %A, i64 %indvars.iv5
|
|
%tmp14 = load i32, i32* %arrayidx21, align 4
|
|
%inc22 = add nsw i32 %tmp14, 1
|
|
store i32 %inc22, i32* %arrayidx21, align 4
|
|
br label %for.inc.23
|
|
|
|
for.inc.23: ; preds = %for.body.19
|
|
%inc24 = add nuw nsw i32 %k16.0, 1
|
|
br label %for.cond.17
|
|
|
|
for.end.25: ; preds = %for.cond.17
|
|
br label %for.inc.26
|
|
|
|
for.inc.26: ; preds = %for.end.25
|
|
%indvars.iv.next2 = add nuw nsw i32 %indvars.iv1, 1
|
|
br label %for.cond.12
|
|
|
|
for.end.28: ; preds = %for.cond.12
|
|
%arrayidx30 = getelementptr inbounds i32, i32* %A, i64 %indvars.iv5
|
|
%tmp15 = load i32, i32* %arrayidx30, align 4
|
|
%tobool = icmp eq i32 %tmp15, 0
|
|
br i1 %tobool, label %if.end, label %if.then
|
|
|
|
if.then: ; preds = %for.end.28
|
|
br label %for.cond.32
|
|
|
|
for.cond.32: ; preds = %for.inc.46, %if.then
|
|
%j31.0 = phi i32 [ 0, %if.then ], [ %inc47, %for.inc.46 ]
|
|
%exitcond4 = icmp ne i32 %j31.0, 100
|
|
br i1 %exitcond4, label %for.body.34, label %for.end.48
|
|
|
|
for.body.34: ; preds = %for.cond.32
|
|
br label %for.cond.36
|
|
|
|
for.cond.36: ; preds = %for.inc.43, %for.body.34
|
|
%k35.0 = phi i32 [ 0, %for.body.34 ], [ %inc44, %for.inc.43 ]
|
|
%mul37 = mul nsw i32 %j31.0, %j31.0
|
|
%cmp38 = icmp slt i32 %k35.0, %mul37
|
|
br i1 %cmp38, label %for.body.39, label %for.end.45
|
|
|
|
for.body.39: ; preds = %for.cond.36
|
|
%arrayidx41 = getelementptr inbounds i32, i32* %A, i64 %indvars.iv5
|
|
%tmp16 = load i32, i32* %arrayidx41, align 4
|
|
%inc42 = add nsw i32 %tmp16, 1
|
|
store i32 %inc42, i32* %arrayidx41, align 4
|
|
br label %for.inc.43
|
|
|
|
for.inc.43: ; preds = %for.body.39
|
|
%inc44 = add nuw nsw i32 %k35.0, 1
|
|
br label %for.cond.36
|
|
|
|
for.end.45: ; preds = %for.cond.36
|
|
br label %for.inc.46
|
|
|
|
for.inc.46: ; preds = %for.end.45
|
|
%inc47 = add nuw nsw i32 %j31.0, 1
|
|
br label %for.cond.32
|
|
|
|
for.end.48: ; preds = %for.cond.32
|
|
br label %if.end
|
|
|
|
if.end: ; preds = %for.end.28, %for.end.48
|
|
br label %for.inc.49
|
|
|
|
for.inc.49: ; preds = %if.end
|
|
%indvars.iv.next6 = add nuw nsw i64 %indvars.iv5, 1
|
|
%indvars.iv.next = add nuw nsw i32 %indvars.iv, 1
|
|
br label %for.cond
|
|
|
|
for.end.51: ; preds = %for.cond
|
|
ret void
|
|
}
|