https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113504

--- Comment #7 from Jonathan Wakely <redi at gcc dot gnu.org> ---
Using GCC 15.2.0 (before the Bug 117276 fix) and GCC 15.2.1 20251211 (after the
fix):

| Executable         | Size        | Time     | Max Resident Memory |
| ------------------ | ----------- | -------- | ------------------- |
| ./pstl_sort.before | 33554432    | 0:00.45  | 366568k             |
| ./tbb_sort.before  | 33554432    | 0:00.81  | 135396k             |
| ./seq_sort.before  | 33554432    | 0:03.54  | 134476k             |
| ./pstl_sort.before | 1073741824  | 0:18.54  | 14171232k           |
| ./tbb_sort.before  | 1073741824  | 0:31.69  | 4197836k            |
| ./seq_sort.before  | 1073741824  | 2:20.32  | 4197840k            |
| ./pstl_sort.after  | 33554432    | 0:00.45  | 228604k             |
| ./tbb_sort.after   | 33554432    | 0:00.82  | 135472k             |
| ./seq_sort.after   | 33554432    | 0:03.56  | 134552k             |
| ./pstl_sort.after  | 1073741824  | 0:17.04  | 8349924k            |
| ./tbb_sort.after   | 1073741824  | 0:29.72  | 4197884k            |
| ./seq_sort.after   | 1073741824  | 2:18.19  | 4197884k            |

So the PSTL versions still use nearly twice as much memory as TBB and the
sequential sorts, but that factor is constant; it doesn't grow with the size of
the input.

So I think this was fixed by fixing the memory leak, and this is a duplicate of
Bug 117276

Reply via email to