-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 On 06/10/2010 05:06 PM, David Abrahams wrote:
I assume you wanted to use stack-based memory for speed, but due to the above, it would likely be slower than the equivalent heap-based integer. Especially with a good caching allocator.
In my experience, stack allocated memory easily beat heap allocation even for large buffer sizes. In the past I have used a fixed string type with a ridiculously large size for some specialized application and it outperformed the heap based string type it replaced (I didn't try to manually optimize copies away).
I assume that we're going to be using the infamous “small string optimization” at some point, which amounts to something like
boost::variant< unique_ptr<char[]>, boost::array<char, 6> >
right?
I had something like that in earlier versions of the library (the "quickdigits"). I've eliminated them from more recent versions; I figure I can re-add them later if they prove to be useful. Until then, with all the rewriting I've been doing on it, it was just one more place things could go wrong. - -- Chad Nelson Oak Circle Software, Inc. * * * -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.10 (GNU/Linux) Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/ iEYEARECAAYFAkwRV7AACgkQp9x9jeZ9/wT8GwCgr7CjwiycgCV8ZmL/mDZmA1i8 KF4Anj+2HYo8lk89lXjKX3CbPLY4hwuT =4N4Q -----END PGP SIGNATURE-----