[bug #18502] Suboptimal cache configuration for IA-32 processors
Update of bug #18502 (project hurd): Status:None = Fixed Open/Closed:Open = Closed ___ Follow-up Comment #3: There is no real basic reason why caching should hit us, so I have commited it. ___ Reply to this item at: http://savannah.gnu.org/bugs/?18502 ___ Message posté via/par Savannah http://savannah.gnu.org/
Re: [bug #18502] Suboptimal cache configuration for IA-32 processors
Barry deFreese, le Thu 14 Dec 2006 21:57:49 -0500, a écrit : Samuel Thibault wrote: Barry deFreese, le Thu 14 Dec 2006 02:03:10 +, a écrit : That is just from running times on a little C program that counts to 10, so it's obviously not scientific. Try compiling something, gnumach for instance ;) Not much of a diff there either: gnumach build: No cache: 12:50 Cache: 12:25 What processor is this? Samuel ___ Bug-hurd mailing list Bug-hurd@gnu.org http://lists.gnu.org/mailman/listinfo/bug-hurd
Re: [bug #18502] Suboptimal cache configuration for IA-32 processors
Samuel Thibault wrote: Barry deFreese, le Thu 14 Dec 2006 21:57:49 -0500, a écrit : Samuel Thibault wrote: Barry deFreese, le Thu 14 Dec 2006 02:03:10 +, a écrit : That is just from running times on a little C program that counts to 10, so it's obviously not scientific. Try compiling something, gnumach for instance ;) Not much of a diff there either: gnumach build: No cache: 12:50 Cache: 12:25 What processor is this? Samuel Pentium II 733Mhz 384Mb Thanks, Barry deFreese (aka bddebian) ___ Bug-hurd mailing list Bug-hurd@gnu.org http://lists.gnu.org/mailman/listinfo/bug-hurd
Re: [bug #18502] Suboptimal cache configuration for IA-32 processors
On 12/15/06, Barry deFreese [EMAIL PROTECTED] wrote: Samuel Thibault wrote: Barry deFreese, le Thu 14 Dec 2006 21:57:49 -0500, a écrit : Samuel Thibault wrote: Barry deFreese, le Thu 14 Dec 2006 02:03:10 +, a écrit : That is just from running times on a little C program that counts to 10, so it's obviously not scientific. Try compiling something, gnumach for instance ;) Are counting to 10 and compiling gnumach both very cache intensive applications? I can't see how the 100 example is, and GCC is mostly sensitive to the cache size, not bandwidth-bound. Try something where the working set is greater than the cache size and you might see some results. Kenneth Østby ___ Bug-hurd mailing list Bug-hurd@gnu.org http://lists.gnu.org/mailman/listinfo/bug-hurd
Re: [bug #18502] Suboptimal cache configuration for IA-32 processors
Kenneth Østby, le Fri 15 Dec 2006 16:12:25 +0100, a écrit : Try something where the working set is greater than the cache size and you might see some results. We precisely want to check the effect of disabling the cache. Using a working set that doesn't fit the cache will behave the same with or without the cache... Samuel ___ Bug-hurd mailing list Bug-hurd@gnu.org http://lists.gnu.org/mailman/listinfo/bug-hurd
Re: [bug #18502] Suboptimal cache configuration for IA-32 processors
Samuel Thibault wrote: Barry deFreese, le Thu 14 Dec 2006 02:03:10 +, a écrit : That is just from running times on a little C program that counts to 10, so it's obviously not scientific. Try compiling something, gnumach for instance ;) Samuel Not much of a diff there either: gnumach build: No cache: 12:50 Cache: 12:25 Barry ___ Bug-hurd mailing list Bug-hurd@gnu.org http://lists.gnu.org/mailman/listinfo/bug-hurd
[bug #18502] Suboptimal cache configuration for IA-32 processors
Follow-up Comment #2, bug #18502 (project hurd): OK, I built gnumach with the following in model_dep.c: set_cr0((get_cr0() | CR0_PG | CR0_WP) ~(CR0_CD | CR0_NW)); And I got worse performance: 0m0.040s 0m0.130s 0m0.000s 0m0.000s Booted back to normal gnumach: 0m0.010s 0m0.020s 0m0.000s 0m0.000s That is just from running times on a little C program that counts to 10, so it's obviously not scientific. ___ Reply to this item at: http://savannah.gnu.org/bugs/?18502 ___ Message sent via/by Savannah http://savannah.gnu.org/ ___ Bug-hurd mailing list Bug-hurd@gnu.org http://lists.gnu.org/mailman/listinfo/bug-hurd
Re: [bug #18502] Suboptimal cache configuration for IA-32 processors
Barry deFreese, le Thu 14 Dec 2006 02:03:10 +, a écrit : That is just from running times on a little C program that counts to 10, so it's obviously not scientific. Try compiling something, gnumach for instance ;) Samuel ___ Bug-hurd mailing list Bug-hurd@gnu.org http://lists.gnu.org/mailman/listinfo/bug-hurd