On Dec 4, 2007 5:40 PM, Tony Mechelynck <[EMAIL PROTECTED]> wrote: > skelker wrote:
> > I have noticed that swap file writes are done in 4K blocks, but file > > reads are done in 64K blocks. If it isn't possible to adjust this > > behavior with configuration, then I suggest opening this up as a > > performance issue that could be addressed in future releases of gvim? That's not quite true, is it? The number of pages in a block times the page size (hopefully the page size of the file system) bytes are written at a time. > Swap file writes cannot be delayed, or the swap file would lose its utility: > thus swap file writes happen more often (sometimes much more often) than edit > file read/writes or swap file reads. OTOH, swap file writes are usually > limited to a rather small area at the end of the file. Thus it is essential > for performance (and also to minimize the risk of getting a power-fail halfway > through a swap file write) to keep the swap file buffer as small as reasonably > possible, while the "ideal" size for the edit file (which is always written > /in toto/, but much less often) is quite bigger. Bullshit. Unless you also fsync [1] there's no guarantee that anything you've written will be on disk at power-failure time. So the number of bytes written at a time is not relevant. In fact, writing fewer bytes is going to increase your chances of data loss, as you're making more system calls, which wastes a lot of time. Regardless, I wonder how much of an issue this really is. It seems to me that the code is quite optimized [1] fsync is a slow operation, especially on a reiser4 file-system. --~--~---------~--~----~------------~-------~--~----~ You received this message from the "vim_dev" maillist. For more information, visit http://www.vim.org/maillist.php -~----------~----~----~----~------~----~------~--~---
