Dark Silicon Web Based Model, version 0.1

Based on models described in:

Dark Silicon and the End of Multicore Scaling
Hadi Esmaeilzadeh, Emily Blem, Renee St. Amant, Karu Sankaralingam, and Doug Burger
ISCA 2011, IEEE Top Picks 2012

Interactive tool is still under development. Upcoming features include:

User defined technology scaling factors (area, power, and frequency)
Input your own Pareto frontiers (at 45 nm)
Area and power budget parameters
Ability to pick cache miss rate functions and cache budgets
Baseline chips other than a quad-core Nehalem
If you would like to see additional features, please send us your suggestions!

If you would like a copy of the source code, have technical questions/difficulties, or have suggestions, please email blem@cs.wisc.edu.

Select input parameters from list below:

Technology Scaling: Conservative ITRS

Multicore Topology: Symmetric Asymmetric Dynamic Fused
Core style: CPU-like GPU-like

Benchmark:

If you picked "Arbitrary" above, give benchmark parameters here:

Fraction of code that can parallelized (f_parallel):

Fraction of instructions that are loads or stores (r_ls):

L1 cache miss rate (m_L1):

L2 cache miss rate (m_L2):

Notes:
All values should be in the range [0,1].
Support for this feature is under development; for now, functions for m_L1 and m_L2 are not supported.

Technology Scaling:	Conservative	ITRS
Multicore Topology:	Symmetric	Asymmetric	Dynamic	Fused
Core style:	CPU-like	GPU-like
Benchmark:
	If you picked "Arbitrary" above, give benchmark parameters here:
	Fraction of code that can parallelized (f_parallel):
	Fraction of instructions that are loads or stores (r_ls):
	L1 cache miss rate (m_L1):
	L2 cache miss rate (m_L2):
	Notes: All values should be in the range [0,1]. Support for this feature is under development; for now, functions for m_L1 and m_L2 are not supported.

1){ echo "

ERROR: f_parallel must be between 0 and 1

"; $success = 0; } if(is_null($rls) || !is_numeric($rls) || $rls <0 || $rls > 1){ echo "

ERROR: r_ls must be between 0 and 1

"; $success = 0; } if(is_null($mL1) || !is_numeric($mL1) || $mL1 <=0 || $mL1 >= 1){ echo "

ERROR: m_l1 must be between 0 and 1

"; $success = 0; } if(is_null($mL2) || !is_numeric($mL2) || $mL2 <0 || $mL2 > 1){ echo "

ERROR: m_l2 must be between 0 and 1

"; $success = 0; } # TODO: generate inputs to plug into model if($success==1){ # put chars into a file $fp = fopen($dir."/temp/chars.txt","w"); fwrite($fp, $rls."\n"); fwrite($fp, $mL1."\n"); fwrite($fp, $mL2."\n"); fwrite($fp, "0\n"); fwrite($fp, "0\n"); fwrite($fp, $fparallel."\n"); fclose($fp); # echo "java -classpath /u/b/l/blem/public/html/web_model/model complete_model_arb_findCPI sym v CPU ITRS 20 0 125 ".$benchmark." > /u/b/l/blem/public/html/web_model/gen_ipc.txt"; echo exec("java -classpath ".$dir."model complete_model_arb_findCPI sym v CPU ITRS 20 0 125 ".$benchmark." ".$dir." > ".$dir."/temp/gen_ipc.txt", $output, $return); # echo "java -classpath /u/b/l/blem/public/html/web_model/model complete_model_arb ".$top." v ".$core." ".$scaling." 20 0 125 ".$benchmark." > /u/b/l/blem/public/html/web_model/temp.txt"; echo exec("java -classpath ".$dir."model complete_model_arb ".$top." v ".$core." ".$scaling." 20 0 125 ".$benchmark." ".$dir." > ".$dir."/temp/output.txt", $output, $return); } }else{ if(!is_null($fparallel) || !is_null($rls) || !is_null($mL1) || !is_null($mL2)){ echo "

Warning: pre-defined benchmark chosen; text box inputs ignored

"; } echo exec("java -classpath ".$dir."model complete_model_allconfigs ".$top." v ".$core." ".$scaling." 20 0 125 ".$benchmark." ".$dir." > ".$dir."/temp/output.txt", $output, $return); } #echo "java -classpath /u/b/l/blem/public/html/web_model/model complete_model_allconfigs ".$top." v ".$core." ".$scaling." 20 0 125 ".$benchmark."
"; if($success==1){ echo exec("python ".$dir."graph/generate_graphs.py ".$dir, $output, $return); ?>

Projected results:

ALT="projected speedup" TITLE = "Speedup">

ALT="projected number of cores" TITLE = "Cores">

ALT="projected amount of dark silicon" TITLE = "Dark Silicon"> "; } } #"; //echo exec('whoami'); #} #?>

Technology scaling:
Multicore topology:
Core Style:
Benchmark:
	f_parallel:
	r_ls:
	m_L1:
	m_L2: