Messages in this thread | | | Date | Mon, 14 Nov 2011 19:47:02 +0400 | From | Ilya Zykov <> | Subject | Re: [PATCH v3 2/2] TTY: tty flip buffer optimisation. |
| |
Alan Cox wrote:
>> Hi, the results are indeed nice. However is there any *real* load other >> than this tailor-made microbenchmark where the added code complexity is >> worth it? > > I'm wondering if we need the complexity in the first place. Certainly 256 > does seem a bit small for pty/tty traffic. A 'real world' benchmark would > be an ls -lR / on a machine with a fast graphics card or in console mode > > ie > > ls -lR / # prime cache > time ls -lR / > > and there are cases where people do a lot of traffic over a pty like this > so I don't think it's entirely fake. > > I don't like the complexity but we could certainly go from using 256 byte > buffers to "tty->buf.bufsize" and make it configurable without > that complexity. > > Alan >
For avoid complexity we need remove free buffer at all. And use kmalloc()-kfree() for every chunk. Don't need tty_buffer_find(). It will be fast and easy. Ilya.
| |