search

HewlettPackard / cacti

An integrated cache and memory access time, cycle time, area, leakage, and dynamic power model
http://www.hpl.hp.com/research/cacti/
384 stars 131 forks source link

3D dram config parameters #2

Open joaopaulocl opened 7 years ago

joaopaulocl commented 7 years ago

Hi, I trying to use the 3D dram config file, but I getting some errors. At first, I ran as it is in the repository and it outputs a float pointing exception. I found the source of exception by debugging it and I included the following parameters (mem_data_width = 8 and num_clk = 2) as a workaround. Then, I ran it again and found some weird values for timing (tRCD, tRAS, tCAS), power and TSV components. Some of them presents values up to e+107 ns and e+54 nJ. When I run with stacked-die-count equals to 1, I get some reasonable values for timing and energy. Has anyone came across this problem? I am also interested in the parameters used in the cacti-3DD paper. Thanks in advance.

ghost commented 6 years ago

Hi, Even I am facing these issues @joaopaulocl were you able to fix these issue?

Thanks.

joaopaulocl commented 6 years ago

@splurgein Unfortunately no.

ghost commented 6 years ago

@joaopaulocl thanks for the quick reply. It's quite strange that a lot of people are making use of cacti for modeling 3D memory and no one has yet posted a fix for this issue.

iamcksong commented 6 years ago

@splurgein and @joaopaulocl There is a code bug in the file of parameter.cc. The line of "tsv_length *= g_ip->num_die_3d;" should be moved to line 846, which is only multiplied once when the line includes an option of -tsv_length.

joaopaulocl commented 6 years ago

Thanks @iamcksong, it worked for me. Are you planning to pull this amendment?

nicmcd commented 6 years ago

Please submit a pull request for this.

rajsng3737 commented 1 month ago

@joaopaulocl @iamcksong it doesn't seems to be working for me.