The patch below improves performance when the system is under memory pressure, by avoiding a TLB purge (broadcast on all processors) just to purge a single page. This case is frequently seen when the system tries to evict buffer cache memory to make room for something else. An easy reproducer is to run 'find /usr -type f | xargs cat > /dev/null' to fill the buffer cache on a more or less continuous basis, while doing a gmake -j50 install on a 4-way box to keep allocating memory. Instrumentation showed that the number of global TLB purges went from a few millions down to about 170 over a 12 hours run. Opinions? Regards, Christophe diff -Naur linux-2.6.20.3/arch/ia64/mm/tlb.c linux-2.6.20.3-ptce-avoidance/arch/ia64/mm/tlb.c --- linux-2.6.20.3/arch/ia64/mm/tlb.c 2007-03-13 19:27:08.000000000 +0100 +++ linux-2.6.20.3-ptce-avoidance/arch/ia64/mm/tlb.c 2007-03-14 23:32:59.000000000 +0100 @@ -10,6 +10,7 @@ * IPI based ptc implementation and A-step IPI implementation. * Rohit Seth <rohit.seth@intel.com> * Ken Chen <kenneth.w.chen@intel.com> + * Christophe de Dinechin <ddd@hp.com>: Avoid ptc.e on memory allocation */ #include <linux/module.h> #include <linux/init.h> @@ -89,9 +90,16 @@ { static DEFINE_SPINLOCK(ptcg_lock); - if (mm != current->active_mm || !current->mm) { - flush_tlb_all(); - return; + struct mm_struct *active_mm = current->active_mm; + + if (mm != active_mm) { + /* Restore region IDs for mm */ + if (mm && active_mm) { + activate_context(mm); + } else { + flush_tlb_all(); + return; + } } /* HW requires global serialization of ptc.ga. */ @@ -107,6 +115,10 @@ } while (start < end); } spin_unlock(&ptcg_lock); + + if (mm != active_mm) { + activate_context(active_mm); + } } void - To unsubscribe from this list: send the line "unsubscribe linux-ia64" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.htmlReceived on Thu Mar 15 04:07:15 2007
This archive was generated by hypermail 2.1.8 : 2007-03-15 04:07:29 EST