Klimi's new dotfiles with stow.
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
 
 
 
 
 
 

1213 lines
52 KiB

;;; -*- Mode: LISP; Package: monitor; Syntax: Common-lisp; Base: 10.; -*-
;;; Tue Jan 25 18:32:28 1994 by Mark Kantrowitz <mkant@GLINDA.OZ.CS.CMU.EDU>
;;; ****************************************************************
;;; Metering System ************************************************
;;; ****************************************************************
;;;
;;; The Metering System is a portable Common Lisp code profiling tool.
;;; It gathers timing and consing statistics for specified functions
;;; while a program is running.
;;;
;;; The Metering System is a combination of
;;; o the Monitor package written by Chris McConnell
;;; o the Profile package written by Skef Wholey and Rob MacLachlan
;;; The two systems were merged and extended by Mark Kantrowitz.
;;;
;;; Address: Carnegie Mellon University
;;; School of Computer Science
;;; Pittsburgh, PA 15213
;;;
;;; This code is in the public domain and is distributed without warranty
;;; of any kind.
;;;
;;; This copy is from SLIME, http://www.common-lisp.net/project/slime/
;;;
;;;
;;; ********************************
;;; Change Log *********************
;;; ********************************
;;;
;;; 26-JUN-90 mk Merged functionality of Monitor and Profile packages.
;;; 26-JUN-90 mk Now handles both inclusive and exclusive statistics
;;; with respect to nested calls. (Allows it to subtract
;;; total monitoring overhead for each function, not just
;;; the time spent monitoring the function itself.)
;;; 26-JUN-90 mk The table is now saved so that one may manipulate
;;; the data (sorting it, etc.) even after the original
;;; source of the data has been cleared.
;;; 25-SEP-90 mk Added get-cons functions for Lucid 3.0, MACL 1.3.2
;;; required-arguments functions for Lucid 3.0,
;;; Franz Allegro CL, and MACL 1.3.2.
;;; 25-JAN-91 mk Now uses fdefinition if available.
;;; 25-JAN-91 mk Replaced (and :allegro (not :coral)) with :excl.
;;; Much better solution for the fact that both call
;;; themselves :allegro.
;;; 5-JUL-91 mk Fixed warning to occur only when file is loaded
;;; uncompiled.
;;; 5-JUL-91 mk When many unmonitored functions, print out number
;;; instead of whole list.
;;; 24-MAR-92 mk Updated for CLtL2 compatibility. space measuring
;;; doesn't work in MCL, but fixed so that timing
;;; statistics do.
;;; 26-MAR-92 mk Updated for Lispworks. Replaced :ccl with
;;; (and :ccl (not :lispworks)).
;;; 27-MAR-92 mk Added get-cons for Allegro-V4.0.
;;; 01-JAN-93 mk v2.0 Support for MCL 2.0, CMU CL 16d, Allegro V3.1/4.0/4.1,
;;; Lucid 4.0, ibcl
;;; 25-JAN-94 mk v2.1 Patches for CLISP from Bruno Haible.
;;; 01-APR-05 lgorrie Removed support for all Lisps except CLISP and OpenMCL.
;;; Purely to cut down on stale code (e.g. #+cltl2) in this
;;; version that is bundled with SLIME.
;;; 22-Aug-08 stas Define TIME-TYPE for Clozure CL.
;;; 07-Aug-12 heller Break lines at 80 columns
;;;
;;; ********************************
;;; To Do **************************
;;; ********************************
;;;
;;; - Need get-cons for Allegro, AKCL.
;;; - Speed up monitoring code. Replace use of hash tables with an embedded
;;; offset in an array so that it will be faster than using gethash.
;;; (i.e., svref/closure reference is usually faster than gethash).
;;; - Beware of (get-internal-run-time) overflowing. Yikes!
;;; - Check robustness with respect to profiled functions.
;;; - Check logic of computing inclusive and exclusive time and consing.
;;; Especially wrt incf/setf comment below. Should be incf, so we
;;; sum recursive calls.
;;; - Add option to record caller statistics -- this would list who
;;; called which functions and how often.
;;; - switches to turn timing/CONSING statistics collection on/off.
;;; ********************************
;;; Notes **************************
;;; ********************************
;;;
;;; METERING has been tested (successfully) in the following lisps:
;;; CMU Common Lisp (16d, Python Compiler 1.0 ) :new-compiler
;;; CMU Common Lisp (M2.9 15-Aug-90, Compiler M1.8 15-Aug-90)
;;; Macintosh Allegro Common Lisp (1.3.2)
;;; Macintosh Common Lisp (2.0)
;;; ExCL (Franz Allegro CL 3.1.12 [DEC 3100] 11/19/90) :allegro-v3.1
;;; ExCL (Franz Allegro CL 4.0.1 [Sun4] 2/8/91) :allegro-v4.0
;;; ExCL (Franz Allegro CL 4.1 [SPARC R1] 8/28/92 14:06) :allegro-v4.1
;;; ExCL (Franz ACL 5.0.1 [Linux/X86] 6/29/99 16:11) :allegro-v5.0.1
;;; Lucid CL (Version 2.1 6-DEC-87)
;;; Lucid Common Lisp (3.0)
;;; Lucid Common Lisp (4.0.1 HP-700 12-Aug-91)
;;; AKCL (1.86, June 30, 1987 or later)
;;; Ibuki Common Lisp (Version 2, release 01.027)
;;; CLISP (January 1994)
;;;
;;; METERING needs to be tested in the following lisps:
;;; Symbolics Common Lisp (8.0)
;;; KCL (June 3, 1987 or later)
;;; TI (Release 4.1 or later)
;;; Golden Common Lisp (3.1 IBM-PC)
;;; VAXLisp (2.0, 3.1)
;;; Procyon Common Lisp
;;; ****************************************************************
;;; Documentation **************************************************
;;; ****************************************************************
;;;
;;; This system runs in any valid Common Lisp. Four small
;;; implementation-dependent changes can be made to improve performance
;;; and prettiness. In the section labelled "Implementation Dependent
;;; Changes" below, you should tailor the functions REQUIRED-ARGUMENTS,
;;; GET-CONS, GET-TIME, and TIME-UNITS-PER-SECOND to your implementation
;;; for the best results. If GET-CONS is not specified for your
;;; implementation, no consing information will be reported. The other
;;; functions will default to working forms, albeit inefficient, in
;;; non-CMU implementations. If you tailor these functions for a particular
;;; version of Common Lisp, we'd appreciate receiving the code.
;;;
;;; ****************************************************************
;;; Usage Notes ****************************************************
;;; ****************************************************************
;;;
;;; SUGGESTED USAGE:
;;;
;;; Start by monitoring big pieces of the program, then carefully choose
;;; which functions close to, but not in, the inner loop are to be
;;; monitored next. Don't monitor functions that are called by other
;;; monitored functions: you will only confuse yourself.
;;;
;;; If the per-call time reported is less than 1/10th of a second, then
;;; consider the clock resolution and profiling overhead before you believe
;;; the time. It may be that you will need to run your program many times
;;; in order to average out to a higher resolution.
;;;
;;; The easiest way to use this package is to load it and execute either
;;; (swank-monitor:with-monitoring (names*) ()
;;; your-forms*)
;;; or
;;; (swank-monitor:monitor-form your-form)
;;; The former allows you to specify which functions will be monitored; the
;;; latter monitors all functions in the current package. Both automatically
;;; produce a table of statistics. Other variants can be constructed from
;;; the monitoring primitives, which are described below, along with a
;;; fuller description of these two macros.
;;;
;;; For best results, compile this file before using.
;;;
;;;
;;; CLOCK RESOLUTION:
;;;
;;; Unless you are very lucky, the length of your machine's clock "tick" is
;;; probably much longer than the time it takes a simple function to run.
;;; For example, on the IBM RT, the clock resolution is 1/50th of a second.
;;; This means that if a function is only called a few times, then only the
;;; first couple of decimal places are really meaningful.
;;;
;;;
;;; MONITORING OVERHEAD:
;;;
;;; The added monitoring code takes time to run every time that the monitored
;;; function is called, which can disrupt the attempt to collect timing
;;; information. In order to avoid serious inflation of the times for functions
;;; that take little time to run, an estimate of the overhead due to monitoring
;;; is subtracted from the times reported for each function.
;;;
;;; Although this correction works fairly well, it is not totally accurate,
;;; resulting in times that become increasingly meaningless for functions
;;; with short runtimes. For example, subtracting the estimated overhead
;;; may result in negative times for some functions. This is only a concern
;;; when the estimated profiling overhead is many times larger than
;;; reported total CPU time.
;;;
;;; If you monitor functions that are called by monitored functions, in
;;; :inclusive mode the monitoring overhead for the inner function is
;;; subtracted from the CPU time for the outer function. [We do this by
;;; counting for each function not only the number of calls to *this*
;;; function, but also the number of monitored calls while it was running.]
;;; In :exclusive mode this is not necessary, since we subtract the
;;; monitoring time of inner functions, overhead & all.
;;;
;;; Otherwise, the estimated monitoring overhead is not represented in the
;;; reported total CPU time. The sum of total CPU time and the estimated
;;; monitoring overhead should be close to the total CPU time for the
;;; entire monitoring run (as determined by TIME).
;;;
;;; A timing overhead factor is computed at load time. This will be incorrect
;;; if the monitoring code is run in a different environment than this file
;;; was loaded in. For example, saving a core image on a high performance
;;; machine and running it on a low performance one will result in the use
;;; of an erroneously small overhead factor.
;;;
;;;
;;; If your times vary widely, possible causes are:
;;; - Garbage collection. Try turning it off, then running your code.
;;; Be warned that monitoring code will probably cons when it does
;;; (get-internal-run-time).
;;; - Swapping. If you have enough memory, execute your form once
;;; before monitoring so that it will be swapped into memory. Otherwise,
;;; get a bigger machine!
;;; - Resolution of internal-time-units-per-second. If this value is
;;; too low, then the timings become wild. You can try executing more
;;; of whatever your test is, but that will only work if some of your
;;; paths do not match the timer resolution.
;;; internal-time-units-per-second is so coarse -- on a Symbolics it is
;;; 977, in MACL it is 60.
;;;
;;;
;;; ****************************************************************
;;; Interface ******************************************************
;;; ****************************************************************
;;;
;;; WITH-MONITORING (&rest functions) [Macro]
;;; (&optional (nested :exclusive)
;;; (threshold 0.01)
;;; (key :percent-time))
;;; &body body
;;; The named functions will be set up for monitoring, the body forms executed,
;;; a table of results printed, and the functions unmonitored. The nested,
;;; threshold, and key arguments are passed to report-monitoring below.
;;;
;;; MONITOR-FORM form [Macro]
;;; &optional (nested :exclusive)
;;; (threshold 0.01)
;;; (key :percent-time)
;;; All functions in the current package are set up for monitoring while
;;; the form is executed, and automatically unmonitored after a table of
;;; results has been printed. The nested, threshold, and key arguments
;;; are passed to report-monitoring below.
;;;
;;; *MONITORED-FUNCTIONS* [Variable]
;;; This holds a list of all functions that are currently being monitored.
;;;
;;; MONITOR &rest names [Macro]
;;; The named functions will be set up for monitoring by augmenting
;;; their function definitions with code that gathers statistical information
;;; about code performance. As with the TRACE macro, the function names are
;;; not evaluated. Calls the function SWANK-MONITOR::MONITORING-ENCAPSULATE on each
;;; function name. If no names are specified, returns a list of all
;;; monitored functions.
;;;
;;; If name is not a symbol, it is evaled to return the appropriate
;;; closure. This allows you to monitor closures stored anywhere like
;;; in a variable, array or structure. Most other monitoring packages
;;; can't handle this.
;;;
;;; MONITOR-ALL &optional (package *package*) [Function]
;;; Monitors all functions in the specified package, which defaults to
;;; the current package.
;;;
;;; UNMONITOR &rest names [Macro]
;;; Removes monitoring code from the named functions. If no names are
;;; specified, all currently monitored functions are unmonitored.
;;;
;;; RESET-MONITORING-INFO name [Function]
;;; Resets the monitoring statistics for the specified function.
;;;
;;; RESET-ALL-MONITORING [Function]
;;; Resets the monitoring statistics for all monitored functions.
;;;
;;; MONITORED name [Function]
;;; Predicate to test whether a function is monitored.
;;;
;;; REPORT-MONITORING &optional names [Function]
;;; (nested :exclusive)
;;; (threshold 0.01)
;;; (key :percent-time)
;;; Creates a table of monitoring information for the specified list
;;; of names, and displays the table using display-monitoring-results.
;;; If names is :all or nil, uses all currently monitored functions.
;;; Takes the following arguments:
;;; - NESTED specifies whether nested calls of monitored functions
;;; are included in the times for monitored functions.
;;; o If :inclusive, the per-function information is for the entire
;;; duration of the monitored function, including any calls to
;;; other monitored functions. If functions A and B are monitored,
;;; and A calls B, then the accumulated time and consing for A will
;;; include the time and consing of B. Note: if a function calls
;;; itself recursively, the time spent in the inner call(s) may
;;; be counted several times.
;;; o If :exclusive, the information excludes time attributed to
;;; calls to other monitored functions. This is the default.
;;; - THRESHOLD specifies that only functions which have been executed
;;; more than threshold percent of the time will be reported. Defaults
;;; to 1%. If a threshold of 0 is specified, all functions are listed,
;;; even those with 0 or negative running times (see note on overhead).
;;; - KEY specifies that the table be sorted by one of the following
;;; sort keys:
;;; :function alphabetically by function name
;;; :percent-time by percent of total execution time
;;; :percent-cons by percent of total consing
;;; :calls by number of times the function was called
;;; :time-per-call by average execution time per function
;;; :cons-per-call by average consing per function
;;; :time same as :percent-time
;;; :cons same as :percent-cons
;;;
;;; REPORT &key (names :all) [Function]
;;; (nested :exclusive)
;;; (threshold 0.01)
;;; (sort-key :percent-time)
;;; (ignore-no-calls nil)
;;;
;;; Same as REPORT-MONITORING but we use a nicer keyword interface.
;;;
;;; DISPLAY-MONITORING-RESULTS &optional (threshold 0.01) [Function]
;;; (key :percent-time)
;;; Prints a table showing for each named function:
;;; - the total CPU time used in that function for all calls
;;; - the total number of bytes consed in that function for all calls
;;; - the total number of calls
;;; - the average amount of CPU time per call
;;; - the average amount of consing per call
;;; - the percent of total execution time spent executing that function
;;; - the percent of total consing spent consing in that function
;;; Summary totals of the CPU time, consing, and calls columns are printed.
;;; An estimate of the monitoring overhead is also printed. May be run
;;; even after unmonitoring all the functions, to play with the data.
;;;
;;; SAMPLE TABLE:
#|
Cons
% % Per Total Total
Function Time Cons Calls Sec/Call Call Time Cons
----------------------------------------------------------------------
FIND-ROLE: 0.58 0.00 136 0.003521 0 0.478863 0
GROUP-ROLE: 0.35 0.00 365 0.000802 0 0.292760 0
GROUP-PROJECTOR: 0.05 0.00 102 0.000408 0 0.041648 0
FEATURE-P: 0.02 0.00 570 0.000028 0 0.015680 0
----------------------------------------------------------------------
TOTAL: 1173 0.828950 0
Estimated total monitoring overhead: 0.88 seconds
|#
;;; ****************************************************************
;;; METERING *******************************************************
;;; ****************************************************************
;;; ********************************
;;; Warn people using the wrong Lisp
;;; ********************************
#-(or clisp openmcl)
(warn "metering.lisp does not support your Lisp implementation!")
;;; ********************************
;;; Packages ***********************
;;; ********************************
;;; For CLtL2 compatible lisps
(defpackage "SWANK-MONITOR" (:use "COMMON-LISP")
(:export "*MONITORED-FUNCTIONS*"
"MONITOR" "MONITOR-ALL" "UNMONITOR" "MONITOR-FORM"
"WITH-MONITORING"
"RESET-MONITORING-INFO" "RESET-ALL-MONITORING"
"MONITORED"
"REPORT-MONITORING"
"DISPLAY-MONITORING-RESULTS"
"MONITORING-ENCAPSULATE" "MONITORING-UNENCAPSULATE"
"REPORT"))
(in-package "SWANK-MONITOR")
;;; Warn user if they're loading the source instead of compiling it first.
(eval-when (eval)
(warn "This file should be compiled before loading for best results."))
;;; ********************************
;;; Version ************************
;;; ********************************
(defparameter *metering-version* "v2.1 25-JAN-94"
"Current version number/date for Metering.")
;;; ****************************************************************
;;; Implementation Dependent Definitions ***************************
;;; ****************************************************************
;;; ********************************
;;; Timing Functions ***************
;;; ********************************
;;; The get-time function is called to find the total number of ticks since
;;; the beginning of time. time-units-per-second allows us to convert units
;;; to seconds.
#-(or clisp openmcl)
(eval-when (compile eval)
(warn
"You may want to supply implementation-specific get-time functions."))
(defconstant time-units-per-second internal-time-units-per-second)
#+openmcl
(progn
(deftype time-type () 'unsigned-byte)
(deftype consing-type () 'unsigned-byte))
(defmacro get-time ()
`(the time-type (get-internal-run-time)))
;;; NOTE: In Macintosh Common Lisp, CCL::GCTIME returns the number of
;;; milliseconds spent during GC. We could subtract this from
;;; the value returned by get-internal-run-time to eliminate
;;; the effect of GC on the timing values, but we prefer to let
;;; the user run without GC on. If the application is so big that
;;; it requires GC to complete, then the GC times are part of the
;;; cost of doing business, and will average out in the long run.
;;; If it seems really important to a user that GC times not be
;;; counted, then uncomment the following three lines and read-time
;;; conditionalize the definition of get-time above with #-:openmcl.
;#+openmcl
;(defmacro get-time ()
; `(the time-type (- (get-internal-run-time) (ccl:gctime))))
;;; ********************************
;;; Consing Functions **************
;;; ********************************
;;; The get-cons macro is called to find the total number of bytes
;;; consed since the beginning of time.
#+clisp
(defun get-cons ()
(multiple-value-bind (real1 real2 run1 run2 gc1 gc2 space1 space2 gccount)
(sys::%%time)
(declare (ignore real1 real2 run1 run2 gc1 gc2 gccount))
(dpb space1 (byte 24 24) space2)))
;;; Macintosh Common Lisp 2.0
;;; Note that this includes bytes that were allocated during GC.
;;; We could subtract this out by advising GC like we did under
;;; MCL 1.3.2, but I'd rather users ran without GC. If they can't
;;; run without GC, then the bytes consed during GC are a cost of
;;; running their program. Metering the code a few times will
;;; avoid the consing values being too lopsided. If a user really really
;;; wants to subtract out the consing during GC, replace the following
;;; two lines with the commented out code.
#+openmcl
(defmacro get-cons () `(the consing-type (ccl::total-bytes-allocated)))
#-(or clisp openmcl)
(progn
(eval-when (compile eval)
(warn "No consing will be reported unless a get-cons function is ~
defined."))
(defmacro get-cons () '(the consing-type 0)))
;; actually, neither `get-cons' nor `get-time' are used as is,
;; but only in the following macro `with-time/cons'
#-:clisp
(defmacro with-time/cons ((delta-time delta-cons) form &body post-process)
(let ((start-cons (gensym "START-CONS-"))
(start-time (gensym "START-TIME-")))
`(let ((,start-time (get-time)) (,start-cons (get-cons)))
(declare (type time-type ,start-time)
(type consing-type ,start-cons))
(multiple-value-prog1 ,form
(let ((,delta-time (- (get-time) ,start-time))
(,delta-cons (- (get-cons) ,start-cons)))
,@post-process)))))
#+clisp
(progn
(defmacro delta4 (nv1 nv2 ov1 ov2 by)
`(- (dpb (- ,nv1 ,ov1) (byte ,by ,by) ,nv2) ,ov2))
(let ((del (find-symbol "DELTA4" "SYS")))
(when del (setf (fdefinition 'delta4) (fdefinition del))))
(if (< internal-time-units-per-second 1000000)
;; TIME_1: AMIGA, OS/2, UNIX_TIMES
(defmacro delta4-time (new-time1 new-time2 old-time1 old-time2)
`(delta4 ,new-time1 ,new-time2 ,old-time1 ,old-time2 16))
;; TIME_2: other UNIX, WIN32
(defmacro delta4-time (new-time1 new-time2 old-time1 old-time2)
`(+ (* (- ,new-time1 ,old-time1) internal-time-units-per-second)
(- ,new-time2 ,old-time2))))
(defmacro delta4-cons (new-cons1 new-cons2 old-cons1 old-cons2)
`(delta4 ,new-cons1 ,new-cons2 ,old-cons1 ,old-cons2 24))
;; avoid consing: when the application conses a lot,
;; get-cons may return a bignum, so we really should not use it.
(defmacro with-time/cons ((delta-time delta-cons) form &body post-process)
(let ((beg-cons1 (gensym "BEG-CONS1-")) (end-cons1 (gensym "END-CONS1-"))
(beg-cons2 (gensym "BEG-CONS2-")) (end-cons2 (gensym "END-CONS2-"))
(beg-time1 (gensym "BEG-TIME1-")) (end-time1 (gensym "END-TIME1-"))
(beg-time2 (gensym "BEG-TIME2-")) (end-time2 (gensym "END-TIME2-"))
(re1 (gensym)) (re2 (gensym)) (gc1 (gensym)) (gc2 (gensym)))
`(multiple-value-bind (,re1 ,re2 ,beg-time1 ,beg-time2
,gc1 ,gc2 ,beg-cons1 ,beg-cons2)
(sys::%%time)
(declare (ignore ,re1 ,re2 ,gc1 ,gc2))
(multiple-value-prog1 ,form
(multiple-value-bind (,re1 ,re2 ,end-time1 ,end-time2
,gc1 ,gc2 ,end-cons1 ,end-cons2)
(sys::%%time)
(declare (ignore ,re1 ,re2 ,gc1 ,gc2))
(let ((,delta-time (delta4-time ,end-time1 ,end-time2
,beg-time1 ,beg-time2))
(,delta-cons (delta4-cons ,end-cons1 ,end-cons2
,beg-cons1 ,beg-cons2)))
,@post-process)))))))
;;; ********************************
;;; Required Arguments *************
;;; ********************************
;;;
;;; Required (Fixed) vs Optional Args
;;;
;;; To avoid unnecessary consing in the "encapsulation" code, we find out the
;;; number of required arguments, and use &rest to capture only non-required
;;; arguments. The function Required-Arguments returns two values: the first
;;; is the number of required arguments, and the second is T iff there are any
;;; non-required arguments (e.g. &optional, &rest, &key).
;;; Lucid, Allegro, and Macintosh Common Lisp
#+openmcl
(defun required-arguments (name)
(let* ((function (symbol-function name))
(args (ccl:arglist function))
(pos (position-if #'(lambda (x)
(and (symbolp x)
(let ((name (symbol-name x)))
(and (>= (length name) 1)
(char= (schar name 0)
#\&)))))
args)))
(if pos
(values pos t)
(values (length args) nil))))
#+clisp
(defun required-arguments (name)
(multiple-value-bind (name req-num opt-num rest-p key-p keywords allow-p)
(sys::function-signature name t)
(if name ; no error
(values req-num (or (/= 0 opt-num) rest-p key-p keywords allow-p))
(values 0 t))))
#-(or clisp openmcl)
(progn
(eval-when (compile eval)
(warn
"You may want to add an implementation-specific ~
Required-Arguments function."))
(eval-when (load eval)
(defun required-arguments (name)
(declare (ignore name))
(values 0 t))))
#|
;;;Examples
(defun square (x) (* x x))
(defun square2 (x &optional y) (* x x y))
(defun test (x y &optional (z 3)) 3)
(defun test2 (x y &optional (z 3) &rest fred) 3)
(required-arguments 'square) => 1 nil
(required-arguments 'square2) => 1 t
(required-arguments 'test) => 2 t
(required-arguments 'test2) => 2 t
|#
;;; ****************************************************************
;;; Main METERING Code *********************************************
;;; ****************************************************************
;;; ********************************
;;; Global Variables ***************
;;; ********************************
(defvar *MONITOR-TIME-OVERHEAD* nil
"The amount of time an empty monitored function costs.")
(defvar *MONITOR-CONS-OVERHEAD* nil
"The amount of cons an empty monitored function costs.")
(defvar *TOTAL-TIME* 0
"Total amount of time monitored so far.")
(defvar *TOTAL-CONS* 0
"Total amount of consing monitored so far.")
(defvar *TOTAL-CALLS* 0
"Total number of calls monitored so far.")
(proclaim '(type time-type *total-time*))
(proclaim '(type consing-type *total-cons*))
(proclaim '(fixnum *total-calls*))
;;; ********************************
;;; Accessor Functions *************
;;; ********************************
;;; Perhaps the SYMBOLP should be FBOUNDP? I.e., what about variables
;;; containing closures.
(defmacro PLACE-FUNCTION (function-place)
"Return the function found at FUNCTION-PLACE. Evals FUNCTION-PLACE
if it isn't a symbol, to allow monitoring of closures located in
variables/arrays/structures."
;; Note that (fboundp 'fdefinition) returns T even if fdefinition
;; is a macro, which is what we want.
(if (fboundp 'fdefinition)
`(if (fboundp ,function-place)
(fdefinition ,function-place)
(eval ,function-place))
`(if (symbolp ,function-place)
(symbol-function ,function-place)
(eval ,function-place))))
(defsetf PLACE-FUNCTION (function-place) (function)
"Set the function in FUNCTION-PLACE to FUNCTION."
(if (fboundp 'fdefinition)
;; If we're conforming to CLtL2, use fdefinition here.
`(if (fboundp ,function-place)
(setf (fdefinition ,function-place) ,function)
(eval '(setf ,function-place ',function)))
`(if (symbolp ,function-place)
(setf (symbol-function ,function-place) ,function)
(eval '(setf ,function-place ',function)))))
#|
;;; before using fdefinition
(defun PLACE-FUNCTION (function-place)
"Return the function found at FUNCTION-PLACE. Evals FUNCTION-PLACE
if it isn't a symbol, to allow monitoring of closures located in
variables/arrays/structures."
(if (symbolp function-place)
(symbol-function function-place)
(eval function-place)))
(defsetf PLACE-FUNCTION (function-place) (function)
"Set the function in FUNCTION-PLACE to FUNCTION."
`(if (symbolp ,function-place)
(setf (symbol-function ,function-place) ,function)
(eval '(setf ,function-place ',function))))
|#
(defun PLACE-FBOUNDP (function-place)
"Test to see if FUNCTION-PLACE is a function."
;; probably should be
#|(or (and (symbolp function-place)(fboundp function-place))
(functionp (place-function function-place)))|#
(if (symbolp function-place)
(fboundp function-place)
(functionp (place-function function-place))))
(defun PLACE-MACROP (function-place)
"Test to see if FUNCTION-PLACE is a macro."
(when (symbolp function-place)
(macro-function function-place)))
;;; ********************************
;;; Measurement Tables *************
;;; ********************************
(defvar *monitored-functions* nil
"List of monitored symbols.")
;;; We associate a METERING-FUNCTIONS structure with each monitored function
;;; name or other closure. This holds the functions that we call to manipulate
;;; the closure which implements the encapsulation.
;;;
(defstruct metering-functions
(name nil)
(old-definition nil :type function)
(new-definition nil :type function)
(read-metering nil :type function)
(reset-metering nil :type function))
;;; In general using hash tables in time-critical programs is a bad idea,
;;; because when one has to grow the table and rehash everything, the
;;; timing becomes grossly inaccurate. In this case it is not an issue
;;; because all inserting of entries in the hash table occurs before the
;;; timing commences. The only circumstance in which this could be a
;;; problem is if the lisp rehashes on the next reference to the table,
;;; instead of when the entry which forces a rehash was inserted.
;;;
;;; Note that a similar kind of problem can occur with GC, which is why
;;; one should turn off GC when monitoring code.
;;;
(defvar *monitor* (make-hash-table :test #'equal)
"Hash table in which METERING-FUNCTIONS structures are stored.")
(defun get-monitor-info (name)
(gethash name *monitor*))
(defsetf get-monitor-info (name) (info)
`(setf (gethash ,name *monitor*) ,info))
(defun MONITORED (function-place)
"Test to see if a FUNCTION-PLACE is monitored."
(and (place-fboundp function-place) ; this line necessary?
(get-monitor-info function-place)))
(defun reset-monitoring-info (name)
"Reset the monitoring info for the specified function."
(let ((finfo (get-monitor-info name)))
(when finfo
(funcall (metering-functions-reset-metering finfo)))))
(defun reset-all-monitoring ()
"Reset monitoring info for all functions."
(setq *total-time* 0
*total-cons* 0
*total-calls* 0)
(dolist (symbol *monitored-functions*)
(when (monitored symbol)
(reset-monitoring-info symbol))))
(defun monitor-info-values (name &optional (nested :exclusive) warn)
"Returns monitoring information values for the named function,
adjusted for overhead."
(let ((finfo (get-monitor-info name)))
(if finfo
(multiple-value-bind (inclusive-time inclusive-cons
exclusive-time exclusive-cons
calls nested-calls)
(funcall (metering-functions-read-metering finfo))
(unless (or (null warn)
(eq (place-function name)
(metering-functions-new-definition finfo)))
(warn "Funtion ~S has been redefined, so times may be inaccurate.~@
MONITOR it again to record calls to the new definition."
name))
(case nested
(:exclusive (values calls
nested-calls
(- exclusive-time
(* calls *monitor-time-overhead*))
(- exclusive-cons
(* calls *monitor-cons-overhead*))))
;; In :inclusive mode, subtract overhead for all the
;; called functions as well. Nested-calls includes the
;; calls of the function as well. [Necessary 'cause of
;; functions which call themselves recursively.]
(:inclusive (values calls
nested-calls
(- inclusive-time
(* nested-calls ;(+ calls)
*monitor-time-overhead*))
(- inclusive-cons
(* nested-calls ;(+ calls)
*monitor-cons-overhead*))))))
(values 0 0 0 0))))
;;; ********************************
;;; Encapsulate ********************
;;; ********************************
(eval-when (compile load eval)
;; Returns a lambda expression for a function that, when called with the
;; function name, will set up that function for metering.
;;
;; A function is monitored by replacing its definition with a closure
;; created by the following function. The closure records the monitoring
;; data, and updates the data with each call of the function.
;;
;; Other closures are used to read and reset the data.
(defun make-monitoring-encapsulation (min-args optionals-p)
(let (required-args)
(dotimes (i min-args) (push (gensym) required-args))
`(lambda (name)
(let ((inclusive-time 0)
(inclusive-cons 0)
(exclusive-time 0)
(exclusive-cons 0)
(calls 0)
(nested-calls 0)
(old-definition (place-function name)))
(declare (type time-type inclusive-time)
(type time-type exclusive-time)
(type consing-type inclusive-cons)
(type consing-type exclusive-cons)
(fixnum calls)
(fixnum nested-calls))
(pushnew name *monitored-functions*)
(setf (place-function name)
#'(lambda (,@required-args
,@(when optionals-p
`(&rest optional-args)))
(let ((prev-total-time *total-time*)
(prev-total-cons *total-cons*)
(prev-total-calls *total-calls*)
;; (old-time inclusive-time)
;; (old-cons inclusive-cons)
;; (old-nested-calls nested-calls)
)
(declare (type time-type prev-total-time)
(type consing-type prev-total-cons)
(fixnum prev-total-calls))
(with-time/cons (delta-time delta-cons)
;; form
,(if optionals-p
`(apply old-definition
,@required-args optional-args)
`(funcall old-definition ,@required-args))
;; post-processing:
;; Calls
(incf calls)
(incf *total-calls*)
;; nested-calls includes this call
(incf nested-calls (the fixnum
(- *total-calls*
prev-total-calls)))
;; (setf nested-calls (+ old-nested-calls
;; (- *total-calls*
;; prev-total-calls)))
;; Time
;; Problem with inclusive time is that it
;; currently doesn't add values from recursive
;; calls to the same function. Change the
;; setf to an incf to fix this?
(incf inclusive-time (the time-type delta-time))
;; (setf inclusive-time (+ delta-time old-time))
(incf exclusive-time (the time-type
(+ delta-time
(- prev-total-time
*total-time*))))
(setf *total-time* (the time-type
(+ delta-time
prev-total-time)))
;; Consing
(incf inclusive-cons (the consing-type delta-cons))
;; (setf inclusive-cons (+ delta-cons old-cons))
(incf exclusive-cons (the consing-type
(+ delta-cons
(- prev-total-cons
*total-cons*))))
(setf *total-cons*
(the consing-type
(+ delta-cons prev-total-cons)))))))
(setf (get-monitor-info name)
(make-metering-functions
:name name
:old-definition old-definition
:new-definition (place-function name)
:read-metering #'(lambda ()
(values inclusive-time
inclusive-cons
exclusive-time
exclusive-cons
calls
nested-calls))
:reset-metering #'(lambda ()
(setq inclusive-time 0
inclusive-cons 0
exclusive-time 0
exclusive-cons 0
calls 0
nested-calls 0)
t)))))))
);; End of EVAL-WHEN
;;; For efficiency reasons, we precompute the encapsulation functions
;;; for a variety of combinations of argument structures
;;; (min-args . optional-p). These are stored in the following hash table
;;; along with any new ones we encounter. Since we're now precomputing
;;; closure functions for common argument signatures, this eliminates
;;; the former need to call COMPILE for each monitored function.
(eval-when (compile eval)
(defconstant precomputed-encapsulations 8))
(defvar *existing-encapsulations* (make-hash-table :test #'equal))
(defun find-encapsulation (min-args optionals-p)
(or (gethash (cons min-args optionals-p) *existing-encapsulations*)
(setf (gethash (cons min-args optionals-p) *existing-encapsulations*)
(compile nil
(make-monitoring-encapsulation min-args optionals-p)))))
(macrolet ((frob ()
(let ((res ()))
(dotimes (i precomputed-encapsulations)
(push `(setf (gethash '(,i . nil) *existing-encapsulations*)
#',(make-monitoring-encapsulation i nil))
res)
(push `(setf (gethash '(,i . t) *existing-encapsulations*)
#',(make-monitoring-encapsulation i t))
res))
`(progn ,@res))))
(frob))
(defun monitoring-encapsulate (name &optional warn)
"Monitor the function Name. If already monitored, unmonitor first."
;; Saves the current definition of name and inserts a new function which
;; returns the result of evaluating body.
(cond ((not (place-fboundp name)) ; not a function
(when warn
(warn "Ignoring undefined function ~S." name)))
((place-macrop name) ; a macro
(when warn
(warn "Ignoring macro ~S." name)))
(t ; tis a function
(when (get-monitor-info name) ; monitored
(when warn
(warn "~S already monitored, so unmonitoring it first." name))
(monitoring-unencapsulate name))
(multiple-value-bind (min-args optionals-p)
(required-arguments name)
(funcall (find-encapsulation min-args optionals-p) name)))))
(defun monitoring-unencapsulate (name &optional warn)
"Removes monitoring encapsulation code from around Name."
(let ((finfo (get-monitor-info name)))
(when finfo ; monitored
(remprop name 'metering-functions)
(setq *monitored-functions*
(remove name *monitored-functions* :test #'equal))
(if (eq (place-function name)
(metering-functions-new-definition finfo))
(setf (place-function name)
(metering-functions-old-definition finfo))
(when warn
(warn "Preserving current definition of redefined function ~S."
name))))))
;;; ********************************
;;; Main Monitoring Functions ******
;;; ********************************
(defmacro MONITOR (&rest names)
"Monitor the named functions. As in TRACE, the names are not evaluated.
If a function is already monitored, then unmonitor and remonitor (useful
to notice function redefinition). If a name is undefined, give a warning
and ignore it. See also unmonitor, report-monitoring,
display-monitoring-results and reset-time."
`(progn
,@(mapcar #'(lambda (name) `(monitoring-encapsulate ',name)) names)
*monitored-functions*))
(defmacro UNMONITOR (&rest names)
"Remove the monitoring on the named functions.
Names defaults to the list of all currently monitored functions."
`(dolist (name ,(if names `',names '*monitored-functions*) (values))
(monitoring-unencapsulate name)))
(defun MONITOR-ALL (&optional (package *package*))
"Monitor all functions in the specified package."
(let ((package (if (packagep package)
package
(find-package package))))
(do-symbols (symbol package)
(when (eq (symbol-package symbol) package)
(monitoring-encapsulate symbol)))))
(defmacro MONITOR-FORM (form
&optional (nested :exclusive) (threshold 0.01)
(key :percent-time))
"Monitor the execution of all functions in the current package
during the execution of FORM. All functions that are executed above
THRESHOLD % will be reported."
`(unwind-protect
(progn
(monitor-all)
(reset-all-monitoring)
(prog1
(time ,form)
(report-monitoring :all ,nested ,threshold ,key :ignore-no-calls)))
(unmonitor)))
(defmacro WITH-MONITORING ((&rest functions)
(&optional (nested :exclusive)
(threshold 0.01)
(key :percent-time))
&body body)
"Monitor the specified functions during the execution of the body."
`(unwind-protect
(progn
(dolist (fun ',functions)
(monitoring-encapsulate fun))
(reset-all-monitoring)
,@body
(report-monitoring :all ,nested ,threshold ,key))
(unmonitor)))
;;; ********************************
;;; Overhead Calculations **********
;;; ********************************
(defconstant overhead-iterations 5000
"Number of iterations over which the timing overhead is averaged.")
;;; Perhaps this should return something to frustrate clever compilers.
(defun STUB-FUNCTION (x)
(declare (ignore x))
nil)
(proclaim '(notinline stub-function))
(defun SET-MONITOR-OVERHEAD ()
"Determines the average overhead of monitoring by monitoring the execution
of an empty function many times."
(setq *monitor-time-overhead* 0
*monitor-cons-overhead* 0)
(stub-function nil)
(monitor stub-function)
(reset-all-monitoring)
(let ((overhead-function (symbol-function 'stub-function)))
(dotimes (x overhead-iterations)
(funcall overhead-function overhead-function)))
; (dotimes (x overhead-iterations)
; (stub-function nil))
(let ((fiter (float overhead-iterations)))
(multiple-value-bind (calls nested-calls time cons)
(monitor-info-values 'stub-function)
(declare (ignore calls nested-calls))
(setq *monitor-time-overhead* (/ time fiter)
*monitor-cons-overhead* (/ cons fiter))))
(unmonitor stub-function))
(set-monitor-overhead)
;;; ********************************
;;; Report Data ********************
;;; ********************************
(defvar *monitor-results* nil
"A table of monitoring statistics is stored here.")
(defvar *no-calls* nil
"A list of monitored functions which weren't called.")
(defvar *estimated-total-overhead* 0)
;; (proclaim '(type time-type *estimated-total-overhead*))
(defstruct (monitoring-info
(:conc-name m-info-)
(:constructor make-monitoring-info
(name calls time cons
percent-time percent-cons
time-per-call cons-per-call)))
name
calls
time
cons
percent-time
percent-cons
time-per-call
cons-per-call)
(defun REPORT (&key (names :all)
(nested :exclusive)
(threshold 0.01)
(sort-key :percent-time)
(ignore-no-calls nil))
"Same as REPORT-MONITORING but with a nicer keyword interface"
(declare (type (member :function :percent-time :time :percent-cons
:cons :calls :time-per-call :cons-per-call)
sort-key)
(type (member :inclusive :exclusive) nested))
(report-monitoring names nested threshold sort-key ignore-no-calls))
(defun REPORT-MONITORING (&optional names
(nested :exclusive)
(threshold 0.01)
(key :percent-time)
ignore-no-calls)
"Report the current monitoring state.
The percentage of the total time spent executing unmonitored code
in each function (:exclusive mode), or total time (:inclusive mode)
will be printed together with the number of calls and
the unmonitored time per call. Functions that have been executed
below THRESHOLD % of the time will not be reported. To report on all
functions set NAMES to be either NIL or :ALL."
(when (or (null names) (eq names :all)) (setq names *monitored-functions*))
(let ((total-time 0)
(total-cons 0)
(total-calls 0))
;; Compute overall time and consing.
(dolist (name names)
(multiple-value-bind (calls nested-calls time cons)
(monitor-info-values name nested :warn)
(declare (ignore nested-calls))
(incf total-calls calls)
(incf total-time time)
(incf total-cons cons)))
;; Total overhead.
(setq *estimated-total-overhead*
(/ (* *monitor-time-overhead* total-calls)
time-units-per-second))
;; Assemble data for only the specified names (all monitored functions)
(if (zerop total-time)
(format *trace-output* "Not enough execution time to monitor.")
(progn
(setq *monitor-results* nil *no-calls* nil)
(dolist (name names)
(multiple-value-bind (calls nested-calls time cons)
(monitor-info-values name nested)
(declare (ignore nested-calls))
(when (minusp time) (setq time 0.0))
(when (minusp cons) (setq cons 0.0))
(if (zerop calls)
(push (if (symbolp name)
(symbol-name name)
(format nil "~S" name))
*no-calls*)
(push (make-monitoring-info
(format nil "~S" name) ; name
calls ; calls
(/ time (float time-units-per-second)) ; time in secs
(round cons) ; consing
(/ time (float total-time)) ; percent-time
(if (zerop total-cons) 0
(/ cons (float total-cons))) ; percent-cons
(/ (/ time (float calls)) ; time-per-call
time-units-per-second) ; sec/call
(round (/ cons (float calls)))) ; cons-per-call
*monitor-results*))))
(display-monitoring-results threshold key ignore-no-calls)))))
(defun display-monitoring-results (&optional (threshold 0.01)
(key :percent-time)
(ignore-no-calls t))
(let ((max-length 8) ; Function header size
(max-cons-length 8)
(total-time 0.0)
(total-consed 0)
(total-calls 0)
(total-percent-time 0)
(total-percent-cons 0))
(sort-results key)
(dolist (result *monitor-results*)
(when (or (zerop threshold)
(> (m-info-percent-time result) threshold))
(setq max-length
(max max-length
(length (m-info-name result))))
(setq max-cons-length
(max max-cons-length
(m-info-cons-per-call result)))))
(incf max-length 2)
(setf max-cons-length (+ 2 (ceiling (log max-cons-length 10))))
(format *trace-output*
"~%~%~
~VT ~VA~
~% ~VT % % ~VA ~
Total Total~
~%Function~VT Time Cons Calls Sec/Call ~VA ~
Time Cons~
~%~V,,,'-A"
max-length
max-cons-length "Cons"
max-length
max-cons-length "Per"
max-length
max-cons-length "Call"
(+ max-length 62 (max 0 (- max-cons-length 5))) "-")
(dolist (result *monitor-results*)
(when (or (zerop threshold)
(> (m-info-percent-time result) threshold))
(format *trace-output*
"~%~A:~VT~6,2F ~6,2F ~7D ~,6F ~VD ~8,3F ~10D"
(m-info-name result)
max-length
(* 100 (m-info-percent-time result))
(* 100 (m-info-percent-cons result))
(m-info-calls result)
(m-info-time-per-call result)
max-cons-length
(m-info-cons-per-call result)
(m-info-time result)
(m-info-cons result))
(incf total-time (m-info-time result))
(incf total-consed (m-info-cons result))
(incf total-calls (m-info-calls result))
(incf total-percent-time (m-info-percent-time result))
(incf total-percent-cons (m-info-percent-cons result))))
(format *trace-output*
"~%~V,,,'-A~
~%TOTAL:~VT~6,2F ~6,2F ~7D ~9@T ~VA ~8,3F ~10D~
~%Estimated monitoring overhead: ~5,2F seconds~
~%Estimated total monitoring overhead: ~5,2F seconds"
(+ max-length 62 (max 0 (- max-cons-length 5))) "-"
max-length
(* 100 total-percent-time)
(* 100 total-percent-cons)
total-calls
max-cons-length " "
total-time total-consed
(/ (* *monitor-time-overhead* total-calls)
time-units-per-second)
*estimated-total-overhead*)
(when (and (not ignore-no-calls) *no-calls*)
(setq *no-calls* (sort *no-calls* #'string<))
(let ((num-no-calls (length *no-calls*)))
(if (> num-no-calls 20)
(format *trace-output*
"~%~@(~r~) monitored functions were not called. ~
~%See the variable swank-monitor::*no-calls* for a list."
num-no-calls)
(format *trace-output*
"~%The following monitored functions were not called:~
~%~{~<~%~:; ~A~>~}~%"
*no-calls*))))
(values)))
(defun sort-results (&optional (key :percent-time))
(setq *monitor-results*
(case key
(:function (sort *monitor-results* #'string>
:key #'m-info-name))
((:percent-time :time) (sort *monitor-results* #'>
:key #'m-info-time))
((:percent-cons :cons) (sort *monitor-results* #'>
:key #'m-info-cons))
(:calls (sort *monitor-results* #'>
:key #'m-info-calls))
(:time-per-call (sort *monitor-results* #'>
:key #'m-info-time-per-call))
(:cons-per-call (sort *monitor-results* #'>
:key #'m-info-cons-per-call)))))
;;; *END OF FILE*