Messages in this thread | | | Date | Wed, 24 Jun 1998 03:43:30 -0700 (PDT) | From | Dean Gaudet <> | Subject | Re: Thread implementations... |
| |
On Wed, 24 Jun 1998, David S. Miller wrote:
> Just an idea I had... and it would work. I still think the TLB > thrashing costs of mmap() compared to write() (the kernel has large > PTE's so write is unfortunately almost always faster under Linux) is > going to show up still to some extent, and there isn't much we can do > about that.
Oh yeah, that bites. What about when you use net cards with on-card packet assembly? I suppose then it matters whether you can send them v or p addresses for DMA... the v->p has to be paid somewhere. No wait, you can pre-pay the v->p as long as the p doesn't change... you pay it once the first time the file is sent, but then you can re-use the same list of commands for future attempts.
> The main reason you want mmap() (or sendfile()) over read() is to be able > to perform single-copy and zero-copy TCP. read() with page-flipping is > another way to do it, but I really don't know the difficulty. > > Be careful what you talk about. write() is not nearly as difficult to > get zero-copy out of as read(), read() is very difficult without very > specialized hardware and MTU's > PAGE_SIZE as well.
Hey I'm a webserver guy, I only care about writing to sockets ;)
Dean
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu
| |