3x faster byLine implementation on OSX, probably faster on Linux as well #3089
Conversation
| ptrdiff_t getdelim(char**, size_t*, int, FILE*); | ||
| // getline() always comes together with getdelim() | ||
| ptrdiff_t getline(char**, size_t*, FILE*); | ||
| } |
There was a problem hiding this comment.
Wouldn't that be better to use a port of the actual C header of the plateform ?
There was a problem hiding this comment.
Yes, we should update core.stdc.stdio and core.sys.posix.stdio.
|
What's the bugzilla issue about this? |
|
Please note that readlnImpl is currently broken: #2794 |
|
Apparently getdelim is a Posix function and should be available on many platforms. |
| if (n > 128 * 1024) | ||
| { | ||
| // Bound memory used by readln | ||
| free(lineptr); |
There was a problem hiding this comment.
I'm fine with leaking this until program exit, but valgrind would report this as leak.
Adding a ~static this() {} might run into circular dependency issues though.
There was a problem hiding this comment.
Fortunately we have std.stdiobase for that, where we can include a call to the module destructor as well.
|
Auto-merge toggled on |
3x faster byLine implementation on OSX, probably faster on Linux as well
|
@MartinNowak: yah, though |
|
No idea, MSVC 9 does not seem to be to old, and some of the more obscure platforms like HP-UX lack this function but aren't supported anyhow. OSX10.5 and FBSD6 are really old by now. |
|
Better keep it until we have updated the declarations in druntime. |
3x faster byLine implementation on OSX, probably faster on Linux as well Signed-off-by: Martin Krejcirik <mk@krej.cz>
3x faster byLine implementation on OSX, probably faster on Linux as well Signed-off-by: Martin Krejcirik <mk@krej.cz>
These changes make byLine more than 3x faster on OSX without changing semantics. Given that part of the changes involve paths taken by the other OSs, it's possible there's speedup elsewhere as well. Please measure, thanks!