Re: 2.2.7_andrea2.bz2

Ben McCann (bmccann@indusriver.com)
Fri, 30 Apr 1999 12:59:51 +0000

I occasionally benchmark Andrea's patches using EGCS 1.0.1 compiling
a complex C++ application. The compiler routinely grows to 80 or even
90MB of VM and I only have 64MB of physical memory. Needless to say,
the machines swaps like crazy.

Here are the results of my latest benchmark between stock 2.2.7 and
Andrea's newest patch:

Baseline 2.2.7. Clean boot followed by:

cd src/tms
make clean
time make -i

546.770u 70.180s 18:35.24 55.3% 0+0k 0+0io 277446pf+223232w

------------------------------------------------------------------------

2.2.7 patched with 2.2.7-andrea2.bz2. Same test, reboot, clean, compile:

546.600u 70.870s 21:37.93 47.5% 0+0k 0+0io 347787pf+200958w

Running _only_ the compiler, the patched kernel is about 10% slower
(in wall-clock time). But, the modified kernel is _much_ better with
interactive response while under this heavy load. I'm typing this message
while this compilation runs and the interactive feel of the keyboard
and echo on the screen is as good as an idle system. I never see
good interactive response with this compilation and the stock 2.2.7
kernel. (In fact, its so bad that I just start the compilation and
walk away from the machine for 20 minutes).

Great Job, Andrea!

Linus (if you're reading this), would you consider Andrea's VM tuning
changes for an up-coming point release of 2.2?

-Ben McCann

-- 

---
Ben McCann                              Indus River Networks
                                        31 Nagog Park
                                        Acton, MA, 01720
email: bmccann@indusriver.com           web: www.indusriver.com 
phone: (978) 266-8140                   fax: (978) 266-8111


-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/