Re: Serious Kernel slowdown with HIMEM (4Gig) in 2.6.7

["Lawrence E. Freil" <lef@freil.com>]

Iteresting idea, here is the /proc/mtrr output:

reg00: base=0x00000000 (   0MB), size= 512MB: write-back, count=1
reg01: base=0x20000000 ( 512MB), size= 256MB: write-back, count=1
reg02: base=0x30000000 ( 768MB), size= 128MB: write-back, count=1
reg03: base=0x38000000 ( 896MB), size=  64MB: write-back, count=1
reg04: base=0x3c000000 ( 960MB), size=  32MB: write-back, count=1
reg05: base=0x3e000000 ( 992MB), size=  16MB: write-back, count=1
reg06: base=0xf0000000 (3840MB), size=   2MB: write-combining, count=1

The memory in question is the block from 896 to 1Gig.  Looks pretty
normal.

At this point I need to recompile "ls" with full profiling and time it
with and without extended memory.  I have a suspicion that it is related
to "mmap" on data in a high memory buffer.  It does not effect actual
file read/write.  Performance on that is pretty much the same.

>On Sat, Aug 14, 2004 at 12:43:11PM -0400, Lawrence E. Freil wrote:
>> Though there is a consistant sys component that is slightly higher.  I ran
>> profiles three times for "fast" and three times for "slow".  Here they are
>
>Try cat /proc/mtrr.  I bet your system isn't marking some of its RAM 
>cachable in the MTRRs.
>
>		-ben

-- 
        Lawrence Freil                      Email:lef@freil.com
        1768 Old Country Place              Phone:(770) 667-9274
        Woodstock, GA 30188