llvm dev - Aug 2008 - [LLVMdev] A case where llvm created different cfg for same code

Hi,

The following two segments of code are actually the same,
but llvm created different cfg for them.

Form1:

1 #define N 10
2 int test(int A[N][N])
3 {
4 int i, j;
5 int result =0;
6 for(j=0; j+2<N; ++j) {
7 //for(i=0; i<j && i+j+1<N; i++) {
8 for(i=0; i<j && i<N-j-1; i++) {
9 A[i+j+1][j] = A[j + 2][j-i] + i;
10 }
11 }
12
13 for (i=0; i<N-2; ++i)
14 for (j=0; j<N; ++j)
15 result ^= A[i][j];
16 return result;
17 }

Form2:

1 #define N 10
2 int test(int A[N][N])
3 {
4 int i, j;
5 int result =0;
6 for(j=0; j+2<N; ++j) {
7 for(i=0; i<j && i+j+1<N; i++) {
8 //for(i=0; i<j && i<N-j-1; i++) {
9 A[i+j+1][j] = A[j + 2][j-i] + i;
10 }
11 }
12
13 for (i=0; i<N-2; ++i)
14 for (j=0; j<N; ++j)
15 result ^= A[i][j];
16 return result;
17 }

The prime difference is that: cfg of form2 has additional basic block
which has a back edge to a non-header-block
I think the loop in that cfg is not canonical.

I tried -loopsimplify and -indvars , but no improvement.

Any comments for this? Thanks in advance.


Sheng.

On Tue, Aug 12, 2008 at 1:24 AM, Sheng Zhou <zhousheng at autoesl.com>
wrote:
> Hi,
>
> The following two segments of code are actually the same,
> but llvm created different cfg for them.
>
...
>
> The prime difference is that: cfg of form2 has additional basic block
> which has a back edge to a non-header-block
> I think the loop in that cfg is not canonical.
>
> I tried -loopsimplify and -indvars , but no improvement.
>
> Any comments for this? Thanks in advance.
>
The code is not the same. 7 is commented out in the first and 8 in the
second. Why would you expect the CFGs to match? Is there something
wrong (inefficient) in the resulting output of these two functions
when compiled?

-bw