Add standard deviation to output of ave-time benchmarking word

db4
Aaron Schaefer 2008-07-27 17:46:43 -04:00
parent c9a1c0a02f
commit d986b30eb6
2 changed files with 25 additions and 12 deletions

View File

@ -1,22 +1,31 @@
USING: arrays help.markup help.syntax math memory quotations sequences system tools.time ;
USING: arrays help.markup help.syntax math math.parser memory quotations
sequences system tools.time ;
IN: project-euler.ave-time
HELP: collect-benchmarks
{ $values { "quot" quotation } { "n" integer } { "seq" sequence } }
{ $description "Runs a quotation " { $snippet "n" } " times, collecting the wall clock time and the time spent in the garbage collector into pairs inside of a sequence." }
{ $notes "The stack effect of " { $snippet "quot" } " is inferred and only one set of outputs will remain on the stack no matter how many trials are run."
{ $description "Runs a quotation " { $snippet "n" } " times, collecting the wall clock time inside of a sequence." }
{ $notes "The stack effect of " { $snippet "quot" } " is accounted for and only one set of outputs will remain on the stack no matter how many trials are run."
$nl
"A nicer word for interactive use is " { $link ave-time } "." } ;
HELP: nth-place
{ $values { "x" float } { "n" integer } { "y" float } }
{ $description "Rounds a floating point number to " { $snippet "n" } " decimal places." }
{ $examples
"This word is useful for display purposes when showing 15 decimal places is not desired:"
{ $unchecked-example "3.141592653589793 3 nth-place number>string" "\"3.142\"" }
} ;
HELP: ave-time
{ $values { "quot" quotation } { "n" integer } }
{ $description "Runs a quotation " { $snippet "n" } " times, then prints the average run time and the average time spent in the garbage collector." }
{ $notes "The stack effect of " { $snippet "quot" } " is inferred and only one set of outputs will remain on the stack no matter how many trials are run." }
{ $description "Runs a quotation " { $snippet "n" } " times, then prints the average run time and standard deviation." }
{ $notes "The stack effect of " { $snippet "quot" } " is accounted for and only one set of outputs will remain on the stack no matter how many trials are run." }
{ $examples
"This word can be used to compare performance of the non-optimizing and optimizing compilers."
$nl
"First, we time a quotation directly; quotations are compiled by the non-optimizing quotation compiler:"
{ $unchecked-example "[ 1000000 0 [ + ] reduce drop ] 10 ave-time" "1116 ms run time - 10 trials" }
"Now we define a word and compile it with the optimizing word compiler. This results is faster execution:"
{ $unchecked-example ": foo 1000000 0 [ + ] reduce ;" "\\ foo compile" "[ foo drop ] 10 ave-time" "202 ms run time - 10 trials" }
{ $unchecked-example "[ 1000000 0 [ + ] reduce drop ] 10 ave-time" "465 ms ave run time - 13.37 SD (10 trials)" }
"Now we define a word and compile it with the optimizing word compiler. This results in faster execution:"
{ $unchecked-example ": foo 1000000 0 [ + ] reduce ;" "\\ foo compile" "[ foo drop ] 10 ave-time" "202 ms ave run time - 22.73 SD (10 trials)" }
} ;

View File

@ -1,7 +1,7 @@
! Copyright (c) 2007 Aaron Schaefer.
! See http://factorcode.org/license.txt for BSD license.
USING: arrays combinators io kernel math math.functions math.parser
math.statistics namespaces sequences tools.time ;
USING: combinators io kernel math math.functions math.parser math.statistics
namespaces tools.time ;
IN: project-euler.ave-time
: collect-benchmarks ( quot n -- seq )
@ -10,7 +10,11 @@ IN: project-euler.ave-time
[ with-datastack drop ] 2curry r> swap times call
] { } make ;
: nth-place ( x n -- y )
10 swap ^ [ * round ] keep / ;
: ave-time ( quot n -- )
[ collect-benchmarks ] keep swap mean round [
# " ms run time - " % # " trials" %
[ collect-benchmarks ] keep
swap [ std 2 nth-place ] [ mean round ] bi [
# " ms ave run time - " % # " SD (" % # " trials)" %
] "" make print flush ; inline