[polly] r185254 - TempScop: (Partial) Implement the printDetail function.

[Tobias Grosser]

On 06/29/2013 12:00 AM, Hongbin Zheng wrote:
> Author: ether
> Date: Sat Jun 29 02:00:14 2013
> New Revision: 185254
>
> URL: http://llvm.org/viewvc/llvm-project?rev=185254&view=rev
> Log:
> TempScop: (Partial) Implement the printDetail function.

Thanks!

Also, it would be nice to check the changes with clang-format before you 
commit them.

If clang-format is in your path, you can just run 'make 
polly-check-format' (or 'make polly-update-format' to fix the formatting)

Some comments inline.

Cheers
Tobias

>   void TempScop::printDetail(raw_ostream &OS, ScalarEvolution *SE,
>                              LoopInfo *LI, const Region *CurR,
> -                           unsigned ind) const {}
> +                           unsigned ind) const {
> +
> +  // FIXME: Print other details rather than memory accesses.
> +  typedef Region::const_block_iterator bb_iterator;
> +  for (bb_iterator I = CurR->block_begin(), E = CurR->block_end(); I != E; ++I){
> +    BasicBlock *CurBlock = *I;
> +
> +    AccFuncMapType::const_iterator AccSetIt = AccFuncMap.find(CurBlock);
> +
> +    // Ignore trivial blocks that do not contain any memory access.
> +    if (AccSetIt == AccFuncMap.end()) continue;

It would be nice to move this into a separate function that is also used 
in place of the Scop::isTrivialBB() function in ScopInfo.cpp.

Such a patch is approved for commit. ;-)


>   void TempScopInfo::buildScalarDependences(Instruction *Inst, Region *R) {
>     // No need to translate these scalar dependences into polyhedral form, because
>
> Added: polly/trunk/test/TempScop/tempscop-printing.ll
> URL: http://llvm.org/viewvc/llvm-project/polly/trunk/test/TempScop/tempscop-printing.ll?rev=185254&view=auto
> ==============================================================================
> --- polly/trunk/test/TempScop/tempscop-printing.ll (added)
> +++ polly/trunk/test/TempScop/tempscop-printing.ll Sat Jun 29 02:00:14 2013
> @@ -0,0 +1,51 @@
> +; RUN: opt %loadPolly -basicaa -polly-analyze-ir -analyze < %s | FileCheck %s

It would be nice to run this also with -polly-codegen-scev enabled.

> +
> +; void f(long A[], int N, int *init_ptr) {
> +;   long i, j;
> +;
> +;   for (i = 0; i < N; ++i) {
> +;     init = *init_ptr;
> +;     for (i = 0; i < N; ++i) {
> +;       A[i] = init + 2;
> +;     }
> +;   }
> +; }
> +
> +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"
> +target triple = "x86_64-unknown-linux-gnu"
> +
> +define void @f(i64* noalias %A, i64 %N, i64* noalias %init_ptr) nounwind {
> +entry:
> +  br label %for.i
> +
> +for.i:
> +  %indvar.i = phi i64 [ 0, %entry ], [ %indvar.i.next, %for.i.end ]
> +  %indvar.i.next = add nsw i64 %indvar.i, 1
> +  br label %entry.next
> +
> +entry.next:
> +; CHECK: BB: entry.next
> +  %init = load i64* %init_ptr
> +; CHECK:  Read init_ptr[0]
> +; CHECK:  Write init.s2a[0]
> +  br label %for.j
> +
> +for.j:
> +  %indvar.j = phi i64 [ 0, %entry.next ], [ %indvar.j.next, %for.j ]
> +; CHECK: BB: for.j
> +; CHECK: Read init.s2a[0]
> +; CHECK: Write A[{0,+,8}<%for.j>]
> +  %init_plus_two = add i64 %init, 2
> +  %scevgep = getelementptr i64* %A, i64 %indvar.j
> +  store i64 %init_plus_two, i64* %scevgep
> +  %indvar.j.next = add nsw i64 %indvar.j, 1
> +  %exitcond.j = icmp eq i64 %indvar.j.next, %N
> +  br i1 %exitcond.j, label %for.i.end, label %for.j
> +
> +for.i.end:
> +  %exitcond.i = icmp eq i64 %indvar.i.next, %N
> +  br i1 %exitcond.i, label %return, label %for.i
> +
> +return:
> +  ret void
> +}

Overall a very nice well targeted test case!

Thanks,
Tobi