Date: Wed, 9 Oct 2013 18:29:06 +0000 (UTC) From: Alexander Motin <mav@FreeBSD.org> To: src-committers@freebsd.org, svn-src-all@freebsd.org, svn-src-stable@freebsd.org, svn-src-stable-9@freebsd.org Subject: svn commit: r256208 - stable/9/sys/kern Message-ID: <201310091829.r99IT6Q6044609@svn.freebsd.org>
next in thread | raw e-mail | index | archive | help
Author: mav Date: Wed Oct 9 18:29:06 2013 New Revision: 256208 URL: http://svnweb.freebsd.org/changeset/base/256208 Log: MFC r255363: Micro-optimize cpu_search(), allowing compiler to use more efficient inline ffsl() implementation, when it is available, instead of homegrown iteration. On dual-E5645 amd64 system (2x6x2 cores) under heavy I/O load that reduces time spent inside cpu_search() from 19% to 13%, while IOPS increased by 5%. Modified: stable/9/sys/kern/sched_ule.c Modified: stable/9/sys/kern/sched_ule.c ============================================================================== --- stable/9/sys/kern/sched_ule.c Wed Oct 9 18:23:30 2013 (r256207) +++ stable/9/sys/kern/sched_ule.c Wed Oct 9 18:29:06 2013 (r256208) @@ -632,10 +632,14 @@ cpu_search(const struct cpu_group *cg, s } /* Iterate through the child CPU groups and then remaining CPUs. */ - for (i = cg->cg_children, cpu = mp_maxid; i >= 0; ) { + for (i = cg->cg_children, cpu = mp_maxid; ; ) { if (i == 0) { +#ifdef HAVE_INLINE_FFSL + cpu = CPU_FFS(&cpumask) - 1; +#else while (cpu >= 0 && !CPU_ISSET(cpu, &cpumask)) cpu--; +#endif if (cpu < 0) break; child = NULL; @@ -660,6 +664,7 @@ cpu_search(const struct cpu_group *cg, s break; } } else { /* Handle child CPU. */ + CPU_CLR(cpu, &cpumask); tdq = TDQ_CPU(cpu); load = tdq->tdq_load * 256; rndptr = DPCPU_PTR(randomval); @@ -707,8 +712,11 @@ cpu_search(const struct cpu_group *cg, s i--; if (i == 0 && CPU_EMPTY(&cpumask)) break; - } else + } +#ifndef HAVE_INLINE_FFSL + else cpu--; +#endif } return (total); }
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?201310091829.r99IT6Q6044609>