7860c5fe4e
IslAst could mark two nested outer loops as "OutermostParallel". It caused that the code generator tried to OpenMP-parallelize both loops, which it is not prepared loop. It was because the recursive AST build algorithm managed a flag "InParallelFor" to ensure that no nested loop is also marked as "OutermostParallel". Unfortunatetly the same flag was used by nodes marked as SIMD, and reset to false after the SIMD node. Since loops can be marked as SIMD inside "OutermostParallel" loops, the recursive algorithm again tried to mark loops as "OutermostParellel" although still nested inside another "OutermostParallel" loop. The fix exposed another bug: The function "astScheduleDimIsParallel" was only called when a loop was potentially "OutermostParallel" or "InnermostParallel", but as a side-effect also determines the minimum dependence distance. Hence, changing when we need to know whether a loop is "OutermostParallel" also changed which loop was annotated with "#pragma minimal dependence distance". Moreover, some complex condition linked with "InParallelFor" determined whether a loop should be an "InnermostParallel" loop. It missed some situations where it would not use mark as such although being inside an SIMD mark node, and therefore not be annotated using "#pragma simd". The changes in particular: 1. Split the "InParallelFor" flag into an "InParallelFor" and an "InSIMD" flag. 2. Unconditionally call "astScheduleDimIsParallel" for its side-effects and store the result in "InParallel" for later use. 3. Simplify the condition when a loop is "InnermostParallel". Fixes llvm.org/PR33153 and llvm.org/PR38073. llvm-svn: 343212
66 lines
2.6 KiB
LLVM
66 lines
2.6 KiB
LLVM
; RUN: opt %loadPolly -polly-parallel -polly-vectorizer=stripmine -polly-codegen-verify -polly-opt-isl -polly-ast -polly-codegen -analyze < %s | FileCheck %s
|
|
;
|
|
; Check that there are no nested #pragma omp parallel for inside a
|
|
; #pragma omp parallel for loop.
|
|
; See llvm.org/PR38073 and llvm.org/PR33153
|
|
;
|
|
; This test unfortunately is very dependent on the result of the schedule
|
|
; optimizer (-polly-opt-isl).
|
|
;
|
|
target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
|
|
|
|
@b = external dso_local unnamed_addr global [1984 x [1984 x double]], align 16
|
|
@c = external dso_local unnamed_addr global [1984 x [1984 x double]], align 16
|
|
|
|
define dso_local void @main() local_unnamed_addr {
|
|
entry:
|
|
%cond = select i1 undef, i32 undef, i32 1984
|
|
%tmp = zext i32 %cond to i64
|
|
%cond63 = select i1 undef, i32 undef, i32 1984
|
|
%tmp1 = zext i32 %cond63 to i64
|
|
br label %for.cond51.preheader
|
|
|
|
for.cond51.preheader:
|
|
%indvars.iv213 = phi i64 [ 0, %entry ], [ %indvars.iv.next214, %for.inc98 ]
|
|
%cond73 = select i1 undef, i32 undef, i32 1984
|
|
%tmp2 = zext i32 %cond73 to i64
|
|
br label %for.cond56.preheader
|
|
|
|
for.cond56.preheader:
|
|
%indvars.iv223 = phi i64 [ 0, %for.cond51.preheader ], [ %indvars.iv.next224, %for.inc95 ]
|
|
br label %for.cond66.preheader
|
|
|
|
for.cond66.preheader:
|
|
%indvars.iv219 = phi i64 [ %indvars.iv.next220, %for.inc92 ], [ 0, %for.cond56.preheader ]
|
|
br label %for.body75
|
|
|
|
for.body75:
|
|
%indvars.iv215 = phi i64 [ %indvars.iv213, %for.cond66.preheader ], [ %indvars.iv.next216, %for.body75 ]
|
|
%arrayidx83 = getelementptr inbounds [1984 x [1984 x double]], [1984 x [1984 x double]]* @b, i64 0, i64 %indvars.iv219, i64 %indvars.iv215
|
|
%tmp3 = load double, double* %arrayidx83, align 8
|
|
%arrayidx87 = getelementptr inbounds [1984 x [1984 x double]], [1984 x [1984 x double]]* @c, i64 0, i64 %indvars.iv223, i64 %indvars.iv215
|
|
store double undef, double* %arrayidx87, align 8
|
|
%indvars.iv.next216 = add nuw nsw i64 %indvars.iv215, 1
|
|
%cmp74 = icmp ult i64 %indvars.iv.next216, %tmp2
|
|
br i1 %cmp74, label %for.body75, label %for.inc92
|
|
|
|
for.inc92:
|
|
%indvars.iv.next220 = add nuw nsw i64 %indvars.iv219, 1
|
|
%cmp64 = icmp ult i64 %indvars.iv.next220, %tmp1
|
|
br i1 %cmp64, label %for.cond66.preheader, label %for.inc95
|
|
|
|
for.inc95:
|
|
%indvars.iv.next224 = add nuw nsw i64 %indvars.iv223, 1
|
|
%cmp54 = icmp ult i64 %indvars.iv.next224, %tmp
|
|
br i1 %cmp54, label %for.cond56.preheader, label %for.inc98
|
|
|
|
for.inc98:
|
|
%indvars.iv.next214 = add nuw nsw i64 %indvars.iv213, 48
|
|
br label %for.cond51.preheader
|
|
}
|
|
|
|
; No parallel loop except the to outermost.
|
|
; CHECK: #pragma omp parallel for
|
|
; CHECK: #pragma omp parallel for
|
|
; CHECK-NOT: #pragma omp parallel for
|