Re: [patch-2.4.0-test10-pre3] logic of __alloc_pages_limit(
On Mon, 16 Oct 2000, Petr Vandrovec wrote: > On 16 Oct 00 at 22:50, Tigran Aivazian wrote: > > + struct page *page; > > /* If possible, reclaim a page directly. */ > > - if (direct_reclaim && z->free_pages < z->pages_min + 8) > > + if (direct_reclaim && z->free_pages < z->pages_min + 8) { > > page = reclaim_page(z); > > - /* If that fails, fall back to rmqueue. */ > > - if (!page) > > - page = rmqueue(z, order); > > - if (page) > > - return page; > > + /* If that fails, fall back to rmqueue. */ > > + if (!page) { > > + page = rmqueue(z, order); > > + if (page) > > + return page; > > + } > > Old code returned page from both reclaim_page() or rmqueue(), while new > returns pages only from rmqueue... What happens with page grabbed by > rmqueue, BTW ? Or is there something out of picture I do not see? > Petr You are absolutely right, sorry. I will keep my eyes open a bit wider next time. I put back Linus and linux-kernel to admit my mistake. Regards, Tigran - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] Please read the FAQ at http://www.tux.org/lkml/
[patch-2.4.0-test10-pre3] logic of __alloc_pages_limit().
Hi Linus and Rik, The same analysis I did for __alloc_pages() applies to __alloc_pages_limit(), namely it can be optimized by looking at the logic of 'page == NULL'. In both cases, of course, I checked the assembly listing to make sure that my patch didn't make a worse code. It was always a few instructions shorter and therefore worth it. (I didn't check whether gcc produced fewer branches, I assume he did, otherwise he would be totally braindead...). Regards, Tigran --- linux/mm/page_alloc.c Sun Oct 15 20:40:38 2000 +++ work/mm/page_alloc.cMon Oct 16 22:39:13 2000 @@ -262,15 +262,17 @@ } if (z->free_pages + z->inactive_clean_pages >= water_mark) { - struct page *page = NULL; + struct page *page; /* If possible, reclaim a page directly. */ - if (direct_reclaim && z->free_pages < z->pages_min + 8) + if (direct_reclaim && z->free_pages < z->pages_min + 8) { page = reclaim_page(z); - /* If that fails, fall back to rmqueue. */ - if (!page) - page = rmqueue(z, order); - if (page) - return page; + /* If that fails, fall back to rmqueue. */ + if (!page) { + page = rmqueue(z, order); + if (page) + return page; + } + } } } - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] Please read the FAQ at http://www.tux.org/lkml/
[patch-2.4.0-test10-pre3] logic of __alloc_pages_limit().
Hi Linus and Rik, The same analysis I did for __alloc_pages() applies to __alloc_pages_limit(), namely it can be optimized by looking at the logic of 'page == NULL'. In both cases, of course, I checked the assembly listing to make sure that my patch didn't make a worse code. It was always a few instructions shorter and therefore worth it. (I didn't check whether gcc produced fewer branches, I assume he did, otherwise he would be totally braindead...). Regards, Tigran --- linux/mm/page_alloc.c Sun Oct 15 20:40:38 2000 +++ work/mm/page_alloc.cMon Oct 16 22:39:13 2000 @@ -262,15 +262,17 @@ } if (z-free_pages + z-inactive_clean_pages = water_mark) { - struct page *page = NULL; + struct page *page; /* If possible, reclaim a page directly. */ - if (direct_reclaim z-free_pages z-pages_min + 8) + if (direct_reclaim z-free_pages z-pages_min + 8) { page = reclaim_page(z); - /* If that fails, fall back to rmqueue. */ - if (!page) - page = rmqueue(z, order); - if (page) - return page; + /* If that fails, fall back to rmqueue. */ + if (!page) { + page = rmqueue(z, order); + if (page) + return page; + } + } } } - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] Please read the FAQ at http://www.tux.org/lkml/
Re: [patch-2.4.0-test10-pre3] logic of __alloc_pages_limit(
On Mon, 16 Oct 2000, Petr Vandrovec wrote: On 16 Oct 00 at 22:50, Tigran Aivazian wrote: + struct page *page; /* If possible, reclaim a page directly. */ - if (direct_reclaim z-free_pages z-pages_min + 8) + if (direct_reclaim z-free_pages z-pages_min + 8) { page = reclaim_page(z); - /* If that fails, fall back to rmqueue. */ - if (!page) - page = rmqueue(z, order); - if (page) - return page; + /* If that fails, fall back to rmqueue. */ + if (!page) { + page = rmqueue(z, order); + if (page) + return page; + } Old code returned page from both reclaim_page() or rmqueue(), while new returns pages only from rmqueue... What happens with page grabbed by rmqueue, BTW ? Or is there something out of picture I do not see? Petr You are absolutely right, sorry. I will keep my eyes open a bit wider next time. I put back Linus and linux-kernel to admit my mistake. Regards, Tigran - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] Please read the FAQ at http://www.tux.org/lkml/