[llvm-bugs] [Bug 25729] clang 300x slower to compile recursive constexpr template vs. gcc

[via llvm-bugs]

https://llvm.org/bugs/show_bug.cgi?id=25729

Richard Smith <richard-llvm at metafoo.co.uk> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|---                         |DUPLICATE

--- Comment #2 from Richard Smith <richard-llvm at metafoo.co.uk> ---
The code is obviously exponential-time in Index:

    template <unsigned Index>
    constexpr unsigned crc32 (char const * str) {
        unsigned const a = crc32 <Index - 1> (str) >> 8;
        unsigned const b = crc32 <Index - 1> (str) ^ str [Index - 1];
        return a ^ crc_table [b & 0x000000FF];
    }

Don't compute crc32<Index - 1>(str) twice, compute it once and reuse the
result.

I don't know why people expect to get asymptotic performance improvements by
slapping 'constexpr' on their functions, and I'm not convinced GCC's
memoization is doing anyone any favours here -- if we or GCC choose to evaluate
the initializer at runtime (as we are permitted to do), it will also be
extremely slow (in a -O0 build at least), due to the use of a ridiculous
exponential-time algorithm.

*** This bug has been marked as a duplicate of bug 12850 ***

-- 
You are receiving this mail because:
You are on the CC list for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-bugs/attachments/20151203/6a198031/attachment.html>