You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I am using your default ini files in my simulations (./DRAMSim -t traces/k6_aoe_02_short.trc -s system.ini.example -d ini/DDR3_micron_32M_8B_x8_sg15.ini -c 1000000). Average latency values are 280-300 ns. However, people use 50-100 ns as memory latency values in Gem5. It seems like my latency values are much higher than usual. I was wondering what the reason is for this difference between two latency values?
Thanks in advance,
Ahmet
The text was updated successfully, but these errors were encountered:
In answering your question, I think the reason is that k6_aoe_02_short.trc is only an example trace. The average latency depends heavily on your modeled system configurations and the work load (memory access patterns). In some system setup you could indeed get latency in the smaller range.
Thanks for your answer. Actually, I understood the problem and tried to solve it by writing a program called "remover.py". Basically, it randomly removes 'x' percentage of lines of the given memory trace to obtain a relatively less memory intensive trace. I posted it to the project's repo in case people may want to use it.
Hi,
I am using your default ini files in my simulations (./DRAMSim -t traces/k6_aoe_02_short.trc -s system.ini.example -d ini/DDR3_micron_32M_8B_x8_sg15.ini -c 1000000). Average latency values are 280-300 ns. However, people use 50-100 ns as memory latency values in Gem5. It seems like my latency values are much higher than usual. I was wondering what the reason is for this difference between two latency values?
Thanks in advance,
Ahmet
The text was updated successfully, but these errors were encountered: