sounds like a good explanation but, this is happening at any combinations. I seems as L1 alone is always giving better performance that L1/L2 combination even after I changed icache from 32kB to 256B. am I doing something wrong here? my command is:
build/X86/gem5.opt configs/example/se.py --cpu-type="detailed" --caches --l1d_size="8kB" --l1i_size="256B" --l2cache --maxinsts=10000000 --cmd=bzip2/bzip2 --options="bzip2/input.program" On Mon, Jul 2, 2012 at 8:18 PM, Ali Saidi <[email protected]> wrote: > ** > > Likely the entire instruction footprint fits in the L1 i cache and there > is very little re-use on the data side. Removing the L2 probably lowers the > access time to main memory by a substantial amount which improve > performance. > > > > Ali > > > > On 02.07.2012 20:06, Nyunyi Tshibangu wrote: > > I am having hardtime understanding the number of cycles from SE mode > statistcs > > 1)running with L1 only > build/X86/gem5.opt configs/example/se.py --cpu-type="detailed" --caches > --maxinsts=10000000 --cmd=bzip2/bzip2 --options="bzip2/input.program" > I am getting number of cycles as: system.cpu.numCycles > 54101855 > 2)running with L1/L2 caches > build/X86/gem5.opt configs/example/se.py --cpu-type="detailed" --caches > --l2cache --maxinsts=10000000 --cmd=bzip2/bzip2 > --options="bzip2/input.program" > I am getting number of cycles as: system.cpu.numCycles > 82713679 > does it make sense that performance get worse when adding L2 cache? why > is ipc worse? am I looking at wrong info to determine instruction/cycles? > any idea? > Thanks > Marcus > > > > > _______________________________________________ > gem5-users mailing list > [email protected] > http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users >
_______________________________________________ gem5-users mailing list [email protected] http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
