Tcl Source Code

Check-in [e41cbd042a]
Login
Bounty program for improvements to Tcl and certain Tcl packages.

Many hyperlinks are disabled.
Use anonymous login to enable hyperlinks.

Overview
Comment:merge 8.7 (TIP#527, New measurement facilities in TCL: New command timerate, performance test suite)
Downloads: Tarball | ZIP archive | SQL archive
Timelines: family | ancestors | descendants | both | trunk
Files: files | file ages | folders
SHA3-256: e41cbd042af0e394e4931de8c3c4bd99cc5c4ae9017526cd18516befdffaf981
User & Date: sebres 2019-03-05 18:23:26
Context
2019-03-05
19:46
integrate/close sebres-trunk-timerate branch (no functional changes, simple merge-point) check-in: 1cbf574661 user: sebres tags: trunk
18:23
merge 8.7 (TIP#527, New measurement facilities in TCL: New command timerate, performance test suite) check-in: e41cbd042a user: sebres tags: trunk
18:02
integrate dgp-sebres-timerate-review: no functional changes (already backported) check-in: 61fe0670ed user: sebres tags: core-8-branch
2019-03-02
16:53
Merge 8.7 check-in: e766d23655 user: jan.nijtmans tags: trunk
Changes
Hide Diffs Unified Diffs Ignore Whitespace Patch

Changes to .travis.yml.

65
66
67
68
69
70
71












72
73
74
75
76
77
78
        apt:
          sources:
            - ubuntu-toolchain-r-test
          packages:
            - g++-7
      env:
        - BUILD_DIR=unix












    - os: osx
      osx_image: xcode8
      env:
        - BUILD_DIR=unix
    - os: osx
      osx_image: xcode8
      env:






>
>
>
>
>
>
>
>
>
>
>
>







65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
        apt:
          sources:
            - ubuntu-toolchain-r-test
          packages:
            - g++-7
      env:
        - BUILD_DIR=unix
    - os: linux
      dist: xenial
      compiler: gcc-7
      addons:
        apt:
          sources:
            - ubuntu-toolchain-r-test
          packages:
            - g++-7
      env:
        - BUILD_DIR=unix
        - CFGOPT=CFLAGS="-DTCL_UTF_MAX=6"
    - os: osx
      osx_image: xcode8
      env:
        - BUILD_DIR=unix
    - os: osx
      osx_image: xcode8
      env:

Added doc/timerate.n.


































































































































































































































































>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
'\"
'\" Copyright (c) 2005 Sergey Brester aka sebres.
'\"
'\" See the file "license.terms" for information on usage and redistribution
'\" of this file, and for a DISCLAIMER OF ALL WARRANTIES.
'\"
.TH timerate n "" Tcl "Tcl Built-In Commands"
.so man.macros
.BS
'\" Note:  do not modify the .SH NAME line immediately below!
.SH NAME
timerate \- Time-related execution resp. performance measurement of a script
.SH SYNOPSIS
\fBtimerate \fIscript\fR \fI?time ?max-count??\fR
.sp
\fBtimerate \fI?-direct?\fR \fI?-overhead double?\fR \fIscript\fR \fI?time ?max-count??\fR
.sp
\fBtimerate \fI?-calibrate?\fR \fI?-direct?\fR \fIscript\fR \fI?time ?max-count??\fR
.BE
.SH DESCRIPTION
.PP
The first and second form will evaluate \fIscript\fR until the interval
\fItime\fR given in milliseconds elapses, or for 1000 milliseconds (1 second)
if \fItime\fR is not specified.
.sp
The parameter \fImax-count\fR could additionally impose a further restriction
by the maximal number of iterations to evaluate the script.
If \fImax-count\fR is specified, the evalution will stop either this count of
iterations is reached or the time is exceeded.
.sp
It will then return a canonical tcl-list of the form
.PP
.CS
\fB0.095977 \(mcs/# 52095836 # 10419167 #/sec 5000.000 nett-ms\fR
.CE
.PP
which indicates:
.IP \(bu
the average amount of time required per iteration, in microseconds ([\fBlindex\fR $result 0])
.IP \(bu
the count how many times it was executed ([\fBlindex\fR $result 2])
.IP \(bu
the estimated rate per second ([\fBlindex\fR $result 4])
.IP \(bu
the estimated real execution time without measurement overhead ([\fBlindex\fR $result 6])
.PP
Time is measured in elapsed time using the finest timer resolution as possible,
not CPU time.
This command may be used to provide information as to how well the script or a
tcl-command is performing and can help determine bottlenecks and fine-tune
application performance.
.TP
\fI-calibrate\fR
.
To measure very fast scripts as exact as posible the calibration process
may be required.

The \fI-calibrate\fR option is used to calibrate timerate, calculating the
estimated overhead of the given script as the default overhead for future 
invocations of the \fBtimerate\fR command. If the \fItime\fR parameter is not 
specified, the calibrate procedure runs for up to 10 seconds.
.TP
\fI-overhead double\fR
.
The \fI-overhead\fR parameter supplies an estimate (in microseconds) of the
measurement overhead of each iteration of the tested script. This quantity
will be subtracted from the measured time prior to reporting results.
.TP
\fI-direct\fR
.
The \fI-direct\fR option causes direct execution of the supplied script,
without compilation, in a manner similar to the \fBtime\fR command. It can be
used to measure the cost of \fBTcl_EvalObjEx\fR, of the invocation of canonical
lists, and of the uncompiled versions of bytecoded commands.
.PP
As opposed to the \fBtime\fR commmand, which runs the tested script for a fixed
number of iterations, the timerate command runs it for a fixed time.
Additionally, the compiled variant of the script will be used during the entire
measurement, as if the script were part of a compiled procedure, if the \fI-direct\fR
option is not specified. The fixed time period and possibility of compilation allow
for more precise results and prevent very long execution times by slow scripts, making
it practical for measuring scripts with highly uncertain execution times.

.SH EXAMPLE
Estimate how fast it takes for a simple Tcl \fBfor\fR loop (including
operations on variable \fIi\fR) to count to a ten:
.PP
.CS
# calibrate:
timerate -calibrate {}
# measure:
timerate { for {set i 0} {$i<10} {incr i} {} } 5000
.CE
.PP
Estimate how fast it takes for a simple Tcl \fBfor\fR loop, ignoring the
overhead for to perform ten iterations, ignoring the overhead of the management
of the variable that controls the loop:
.PP
.CS
# calibrate for overhead of variable operations:
set i 0; timerate -calibrate {expr {$i<10}; incr i} 1000
# measure:
timerate { for {set i 0} {$i<10} {incr i} {} } 5000
.CE
.PP
Estimate the speed of calculating the hour of the day using \fBclock format\fR only,
ignoring overhead of the portion of the script that prepares the time for it to
calculate:
.PP
.CS
# calibrate:
timerate -calibrate {}
# estimate overhead:
set tm 0
set ovh [lindex [timerate { incr tm [expr {24*60*60}] }] 0]
# measure using esimated overhead:
set tm 0
timerate -overhead $ovh {
    clock format $tm -format %H
    incr tm [expr {24*60*60}]; # overhead for this is ignored
} 5000
.CE
.SH "SEE ALSO"
time(n)
.SH KEYWORDS
script, timerate, time
.\" Local Variables:
.\" mode: nroff
.\" End:

Changes to generic/tclBasic.c.

296
297
298
299
300
301
302

303
304
305
306
307
308
309
...
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
...
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
....
6817
6818
6819
6820
6821
6822
6823
6824
6825
6826
6827
6828
6829
6830
6831
    {"pwd",		Tcl_PwdObjCmd,		NULL,			NULL,	0},
    {"read",		Tcl_ReadObjCmd,		NULL,			NULL,	CMD_IS_SAFE},
    {"seek",		Tcl_SeekObjCmd,		NULL,			NULL,	CMD_IS_SAFE},
    {"socket",		Tcl_SocketObjCmd,	NULL,			NULL,	0},
    {"source",		Tcl_SourceObjCmd,	NULL,			TclNRSourceObjCmd,	0},
    {"tell",		Tcl_TellObjCmd,		NULL,			NULL,	CMD_IS_SAFE},
    {"time",		Tcl_TimeObjCmd,		NULL,			NULL,	CMD_IS_SAFE},

    {"unload",		Tcl_UnloadObjCmd,	NULL,			NULL,	0},
    {"update",		Tcl_UpdateObjCmd,	NULL,			NULL,	CMD_IS_SAFE},
    {"vwait",		Tcl_VwaitObjCmd,	NULL,			NULL,	CMD_IS_SAFE},
    {NULL,		NULL,			NULL,			NULL,	0}
};

/*
................................................................................
{
    Interp *iPtr;
    Tcl_Interp *interp;
    Command *cmdPtr;
    const BuiltinFuncDef *builtinFuncPtr;
    const OpCmdInfo *opcmdInfoPtr;
    const CmdInfo *cmdInfoPtr;
    Tcl_Namespace *mathfuncNSPtr, *mathopNSPtr;
    Tcl_HashEntry *hPtr;
    int isNew;
    CancelInfo *cancelInfo;
    union {
	char c[sizeof(short)];
	short s;
    } order;
................................................................................
    Tcl_CreateObjCommand(interp, "::tcl::dtrace", DTraceObjCmd, NULL, NULL);
#endif /* USE_DTRACE */

    /*
     * Register the builtin math functions.
     */

    mathfuncNSPtr = Tcl_CreateNamespace(interp, "::tcl::mathfunc", NULL,NULL);
    if (mathfuncNSPtr == NULL) {
	Tcl_Panic("Can't create math function namespace");
    }
#define MATH_FUNC_PREFIX_LEN 17 /* == strlen("::tcl::mathfunc::") */
    memcpy(mathFuncName, "::tcl::mathfunc::", MATH_FUNC_PREFIX_LEN);
    for (builtinFuncPtr = BuiltinFuncTable; builtinFuncPtr->name != NULL;
	    builtinFuncPtr++) {
	strcpy(mathFuncName+MATH_FUNC_PREFIX_LEN, builtinFuncPtr->name);
	Tcl_CreateObjCommand(interp, mathFuncName,
		builtinFuncPtr->objCmdProc, builtinFuncPtr->clientData, NULL);
	Tcl_Export(interp, mathfuncNSPtr, builtinFuncPtr->name, 0);
    }

    /*
     * Register the mathematical "operator" commands. [TIP #174]
     */

    mathopNSPtr = Tcl_CreateNamespace(interp, "::tcl::mathop", NULL, NULL);
    if (mathopNSPtr == NULL) {
	Tcl_Panic("can't create math operator namespace");
    }
    Tcl_Export(interp, mathopNSPtr, "*", 1);
#define MATH_OP_PREFIX_LEN 15 /* == strlen("::tcl::mathop::") */
    memcpy(mathFuncName, "::tcl::mathop::", MATH_OP_PREFIX_LEN);
    for (opcmdInfoPtr=mathOpCmds ; opcmdInfoPtr->name!=NULL ; opcmdInfoPtr++){
	TclOpCmdClientData *occdPtr = Tcl_Alloc(sizeof(TclOpCmdClientData));

	occdPtr->op = opcmdInfoPtr->name;
	occdPtr->i.numArgs = opcmdInfoPtr->i.numArgs;
................................................................................
	    }
	}
	break;
    case TCL_NUMBER_BIG:
	if (Tcl_GetBignumFromObj(interp, objv[1], &big) != TCL_OK) {
	    return TCL_ERROR;
	}
	if (SIGN(&big) == MP_NEG) {
	    mp_clear(&big);
	    goto negarg;
	}
	break;
    default:
	if (TclGetWideIntFromObj(interp, objv[1], &w) != TCL_OK) {
	    return TCL_ERROR;






>







 







|







 







|
|









|






|
|


|







 







|







296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
...
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
...
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
....
6818
6819
6820
6821
6822
6823
6824
6825
6826
6827
6828
6829
6830
6831
6832
    {"pwd",		Tcl_PwdObjCmd,		NULL,			NULL,	0},
    {"read",		Tcl_ReadObjCmd,		NULL,			NULL,	CMD_IS_SAFE},
    {"seek",		Tcl_SeekObjCmd,		NULL,			NULL,	CMD_IS_SAFE},
    {"socket",		Tcl_SocketObjCmd,	NULL,			NULL,	0},
    {"source",		Tcl_SourceObjCmd,	NULL,			TclNRSourceObjCmd,	0},
    {"tell",		Tcl_TellObjCmd,		NULL,			NULL,	CMD_IS_SAFE},
    {"time",		Tcl_TimeObjCmd,		NULL,			NULL,	CMD_IS_SAFE},
    {"timerate",	Tcl_TimeRateObjCmd,	NULL,			NULL,	CMD_IS_SAFE},
    {"unload",		Tcl_UnloadObjCmd,	NULL,			NULL,	0},
    {"update",		Tcl_UpdateObjCmd,	NULL,			NULL,	CMD_IS_SAFE},
    {"vwait",		Tcl_VwaitObjCmd,	NULL,			NULL,	CMD_IS_SAFE},
    {NULL,		NULL,			NULL,			NULL,	0}
};

/*
................................................................................
{
    Interp *iPtr;
    Tcl_Interp *interp;
    Command *cmdPtr;
    const BuiltinFuncDef *builtinFuncPtr;
    const OpCmdInfo *opcmdInfoPtr;
    const CmdInfo *cmdInfoPtr;
    Tcl_Namespace *nsPtr;
    Tcl_HashEntry *hPtr;
    int isNew;
    CancelInfo *cancelInfo;
    union {
	char c[sizeof(short)];
	short s;
    } order;
................................................................................
    Tcl_CreateObjCommand(interp, "::tcl::dtrace", DTraceObjCmd, NULL, NULL);
#endif /* USE_DTRACE */

    /*
     * Register the builtin math functions.
     */

    nsPtr = Tcl_CreateNamespace(interp, "::tcl::mathfunc", NULL,NULL);
    if (nsPtr == NULL) {
	Tcl_Panic("Can't create math function namespace");
    }
#define MATH_FUNC_PREFIX_LEN 17 /* == strlen("::tcl::mathfunc::") */
    memcpy(mathFuncName, "::tcl::mathfunc::", MATH_FUNC_PREFIX_LEN);
    for (builtinFuncPtr = BuiltinFuncTable; builtinFuncPtr->name != NULL;
	    builtinFuncPtr++) {
	strcpy(mathFuncName+MATH_FUNC_PREFIX_LEN, builtinFuncPtr->name);
	Tcl_CreateObjCommand(interp, mathFuncName,
		builtinFuncPtr->objCmdProc, builtinFuncPtr->clientData, NULL);
	Tcl_Export(interp, nsPtr, builtinFuncPtr->name, 0);
    }

    /*
     * Register the mathematical "operator" commands. [TIP #174]
     */

    nsPtr = Tcl_CreateNamespace(interp, "::tcl::mathop", NULL, NULL);
    if (nsPtr == NULL) {
	Tcl_Panic("can't create math operator namespace");
    }
    Tcl_Export(interp, nsPtr, "*", 1);
#define MATH_OP_PREFIX_LEN 15 /* == strlen("::tcl::mathop::") */
    memcpy(mathFuncName, "::tcl::mathop::", MATH_OP_PREFIX_LEN);
    for (opcmdInfoPtr=mathOpCmds ; opcmdInfoPtr->name!=NULL ; opcmdInfoPtr++){
	TclOpCmdClientData *occdPtr = Tcl_Alloc(sizeof(TclOpCmdClientData));

	occdPtr->op = opcmdInfoPtr->name;
	occdPtr->i.numArgs = opcmdInfoPtr->i.numArgs;
................................................................................
	    }
	}
	break;
    case TCL_NUMBER_BIG:
	if (Tcl_GetBignumFromObj(interp, objv[1], &big) != TCL_OK) {
	    return TCL_ERROR;
	}
	if (mp_isneg(&big)) {
	    mp_clear(&big);
	    goto negarg;
	}
	break;
    default:
	if (TclGetWideIntFromObj(interp, objv[1], &w) != TCL_OK) {
	    return TCL_ERROR;

Changes to generic/tclClock.c.

1768
1769
1770
1771
1772
1773
1774
1775
1776
1777
1778
1779
1780
1781
1782
1783
....
1839
1840
1841
1842
1843
1844
1845
1846
1847
1848
1849
1850
1851
1852
1853
1854
1855
1856
1857
1858
1859
1860
1861
#ifdef TCL_WIDE_CLICKS
	clicks = TclpGetWideClicks();
#else
	clicks = (Tcl_WideInt) TclpGetClicks();
#endif
	break;
    case CLICKS_MICROS:
	Tcl_GetTime(&now);
	clicks = ((Tcl_WideInt) now.sec * 1000000) + now.usec;
	break;
    }

    Tcl_SetObjResult(interp, Tcl_NewWideIntObj(clicks));
    return TCL_OK;
}
 
................................................................................
int
ClockMicrosecondsObjCmd(
    ClientData clientData,	/* Client data is unused */
    Tcl_Interp *interp,		/* Tcl interpreter */
    int objc,			/* Parameter count */
    Tcl_Obj *const *objv)	/* Parameter values */
{
    Tcl_Time now;

    if (objc != 1) {
	Tcl_WrongNumArgs(interp, 1, objv, NULL);
	return TCL_ERROR;
    }
    Tcl_GetTime(&now);
    Tcl_SetObjResult(interp, Tcl_NewWideIntObj(
	    ((Tcl_WideInt) now.sec * 1000000) + now.usec));
    return TCL_OK;
}
 
/*
 *-----------------------------------------------------------------------------
 *
 * ClockParseformatargsObjCmd --






|
<







 







<
<




<
|
<







1768
1769
1770
1771
1772
1773
1774
1775

1776
1777
1778
1779
1780
1781
1782
....
1838
1839
1840
1841
1842
1843
1844


1845
1846
1847
1848

1849

1850
1851
1852
1853
1854
1855
1856
#ifdef TCL_WIDE_CLICKS
	clicks = TclpGetWideClicks();
#else
	clicks = (Tcl_WideInt) TclpGetClicks();
#endif
	break;
    case CLICKS_MICROS:
	clicks = TclpGetMicroseconds();

	break;
    }

    Tcl_SetObjResult(interp, Tcl_NewWideIntObj(clicks));
    return TCL_OK;
}
 
................................................................................
int
ClockMicrosecondsObjCmd(
    ClientData clientData,	/* Client data is unused */
    Tcl_Interp *interp,		/* Tcl interpreter */
    int objc,			/* Parameter count */
    Tcl_Obj *const *objv)	/* Parameter values */
{


    if (objc != 1) {
	Tcl_WrongNumArgs(interp, 1, objv, NULL);
	return TCL_ERROR;
    }

    Tcl_SetObjResult(interp, Tcl_NewWideIntObj(TclpGetMicroseconds()));

    return TCL_OK;
}
 
/*
 *-----------------------------------------------------------------------------
 *
 * ClockParseformatargsObjCmd --

Changes to generic/tclCmdMZ.c.

13
14
15
16
17
18
19

20
21
22
23
24
25
26
....
4038
4039
4040
4041
4042
4043
4044
4045
4046
4047
4048
4049
4050
4051
4052
....
4074
4075
4076
4077
4078
4079
4080
































































































































































































































































































































































































4081
4082
4083
4084
4085
4086
4087
 * Copyright (c) 2003-2009 Donal K. Fellows.
 *
 * See the file "license.terms" for information on usage and redistribution of
 * this file, and for a DISCLAIMER OF ALL WARRANTIES.
 */

#include "tclInt.h"

#include "tclRegexp.h"
#include "tclStringTrim.h"

static inline Tcl_Obj *	During(Tcl_Interp *interp, int resultCode,
			    Tcl_Obj *oldOptions, Tcl_Obj *errorInfo);
static Tcl_NRPostProc	SwitchPostProc;
static Tcl_NRPostProc	TryPostBody;
................................................................................
    i = count;
#ifndef TCL_WIDE_CLICKS
    Tcl_GetTime(&start);
#else
    start = TclpGetWideClicks();
#endif
    while (i-- > 0) {
	result = Tcl_EvalObjEx(interp, objPtr, 0);
	if (result != TCL_OK) {
	    return result;
	}
    }
#ifndef TCL_WIDE_CLICKS
    Tcl_GetTime(&stop);
    totalMicroSec = ((double) (stop.sec - start.sec)) * 1.0e6
................................................................................
    TclNewLiteralStringObj(objs[1], "microseconds");
    TclNewLiteralStringObj(objs[2], "per");
    TclNewLiteralStringObj(objs[3], "iteration");
    Tcl_SetObjResult(interp, Tcl_NewListObj(4, objs));

    return TCL_OK;
}
































































































































































































































































































































































































 
/*
 *----------------------------------------------------------------------
 *
 * Tcl_TryObjCmd, TclNRTryObjCmd --
 *
 *	This procedure is invoked to process the "try" Tcl command. See the






>







 







|







 







>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>







13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
....
4039
4040
4041
4042
4043
4044
4045
4046
4047
4048
4049
4050
4051
4052
4053
....
4075
4076
4077
4078
4079
4080
4081
4082
4083
4084
4085
4086
4087
4088
4089
4090
4091
4092
4093
4094
4095
4096
4097
4098
4099
4100
4101
4102
4103
4104
4105
4106
4107
4108
4109
4110
4111
4112
4113
4114
4115
4116
4117
4118
4119
4120
4121
4122
4123
4124
4125
4126
4127
4128
4129
4130
4131
4132
4133
4134
4135
4136
4137
4138
4139
4140
4141
4142
4143
4144
4145
4146
4147
4148
4149
4150
4151
4152
4153
4154
4155
4156
4157
4158
4159
4160
4161
4162
4163
4164
4165
4166
4167
4168
4169
4170
4171
4172
4173
4174
4175
4176
4177
4178
4179
4180
4181
4182
4183
4184
4185
4186
4187
4188
4189
4190
4191
4192
4193
4194
4195
4196
4197
4198
4199
4200
4201
4202
4203
4204
4205
4206
4207
4208
4209
4210
4211
4212
4213
4214
4215
4216
4217
4218
4219
4220
4221
4222
4223
4224
4225
4226
4227
4228
4229
4230
4231
4232
4233
4234
4235
4236
4237
4238
4239
4240
4241
4242
4243
4244
4245
4246
4247
4248
4249
4250
4251
4252
4253
4254
4255
4256
4257
4258
4259
4260
4261
4262
4263
4264
4265
4266
4267
4268
4269
4270
4271
4272
4273
4274
4275
4276
4277
4278
4279
4280
4281
4282
4283
4284
4285
4286
4287
4288
4289
4290
4291
4292
4293
4294
4295
4296
4297
4298
4299
4300
4301
4302
4303
4304
4305
4306
4307
4308
4309
4310
4311
4312
4313
4314
4315
4316
4317
4318
4319
4320
4321
4322
4323
4324
4325
4326
4327
4328
4329
4330
4331
4332
4333
4334
4335
4336
4337
4338
4339
4340
4341
4342
4343
4344
4345
4346
4347
4348
4349
4350
4351
4352
4353
4354
4355
4356
4357
4358
4359
4360
4361
4362
4363
4364
4365
4366
4367
4368
4369
4370
4371
4372
4373
4374
4375
4376
4377
4378
4379
4380
4381
4382
4383
4384
4385
4386
4387
4388
4389
4390
4391
4392
4393
4394
4395
4396
4397
4398
4399
4400
4401
4402
4403
4404
4405
4406
4407
4408
4409
4410
4411
4412
4413
4414
4415
4416
4417
4418
4419
4420
4421
4422
4423
4424
4425
4426
4427
4428
4429
4430
4431
4432
4433
4434
4435
4436
4437
4438
4439
4440
4441
4442
4443
4444
4445
4446
4447
4448
4449
4450
4451
4452
4453
4454
4455
4456
4457
4458
4459
4460
4461
4462
4463
4464
4465
4466
4467
4468
4469
4470
4471
4472
 * Copyright (c) 2003-2009 Donal K. Fellows.
 *
 * See the file "license.terms" for information on usage and redistribution of
 * this file, and for a DISCLAIMER OF ALL WARRANTIES.
 */

#include "tclInt.h"
#include "tclCompile.h"
#include "tclRegexp.h"
#include "tclStringTrim.h"

static inline Tcl_Obj *	During(Tcl_Interp *interp, int resultCode,
			    Tcl_Obj *oldOptions, Tcl_Obj *errorInfo);
static Tcl_NRPostProc	SwitchPostProc;
static Tcl_NRPostProc	TryPostBody;
................................................................................
    i = count;
#ifndef TCL_WIDE_CLICKS
    Tcl_GetTime(&start);
#else
    start = TclpGetWideClicks();
#endif
    while (i-- > 0) {
	result = TclEvalObjEx(interp, objPtr, 0, NULL, 0);
	if (result != TCL_OK) {
	    return result;
	}
    }
#ifndef TCL_WIDE_CLICKS
    Tcl_GetTime(&stop);
    totalMicroSec = ((double) (stop.sec - start.sec)) * 1.0e6
................................................................................
    TclNewLiteralStringObj(objs[1], "microseconds");
    TclNewLiteralStringObj(objs[2], "per");
    TclNewLiteralStringObj(objs[3], "iteration");
    Tcl_SetObjResult(interp, Tcl_NewListObj(4, objs));

    return TCL_OK;
}
 
/*
 *----------------------------------------------------------------------
 *
 * Tcl_TimeRateObjCmd --
 *
 *	This object-based procedure is invoked to process the "timerate" Tcl
 *	command. 
 *	This is similar to command "time", except the execution limited by 
 *	given time (in milliseconds) instead of repetition count.
 *
 * Example:
 *	timerate {after 5} 1000 ; # equivalent for `time {after 5} [expr 1000/5]`
 *
 * Results:
 *	A standard Tcl object result.
 *
 * Side effects:
 *	See the user documentation.
 *
 *----------------------------------------------------------------------
 */

int
Tcl_TimeRateObjCmd(
    ClientData dummy,		/* Not used. */
    Tcl_Interp *interp,		/* Current interpreter. */
    int objc,			/* Number of arguments. */
    Tcl_Obj *const objv[])	/* Argument objects. */
{
    static 
    double measureOverhead = 0; /* global measure-overhead */
    double overhead = -1;	/* given measure-overhead */
    register Tcl_Obj *objPtr;
    register int result, i;
    Tcl_Obj *calibrate = NULL, *direct = NULL;
    Tcl_WideUInt count = 0;	/* Holds repetition count */
    Tcl_WideInt  maxms  = WIDE_MIN;
				/* Maximal running time (in milliseconds) */
    Tcl_WideUInt maxcnt = WIDE_MAX;
				/* Maximal count of iterations. */
    Tcl_WideUInt threshold = 1;	/* Current threshold for check time (faster
				 * repeat count without time check) */
    Tcl_WideUInt maxIterTm = 1;	/* Max time of some iteration as max threshold
				 * additionally avoid divide to zero (never < 1) */
    unsigned short factor = 50;	/* Factor (4..50) limiting threshold to avoid
				 * growth of execution time. */
    register Tcl_WideInt start, middle, stop;
#ifndef TCL_WIDE_CLICKS
    Tcl_Time now;
#endif

    static const char *const options[] = {
	"-direct",	"-overhead",	"-calibrate",	"--",	NULL
    };
    enum options {
	TMRT_EV_DIRECT,	TMRT_OVERHEAD,	TMRT_CALIBRATE,	TMRT_LAST
    };

    NRE_callback *rootPtr;
    ByteCode	 *codePtr = NULL;

    for (i = 1; i < objc - 1; i++) {
    	int index;
	if (Tcl_GetIndexFromObj(NULL, objv[i], options, "option", TCL_EXACT,
		&index) != TCL_OK) {
	    break;
	}
	if (index == TMRT_LAST) {
	    i++;
	    break;
	}
	switch (index) {
	case TMRT_EV_DIRECT:
	    direct = objv[i];
	    break;
	case TMRT_OVERHEAD:
	    if (++i >= objc - 1) {
		goto usage;
	    }
	    if (Tcl_GetDoubleFromObj(interp, objv[i], &overhead) != TCL_OK) {
		return TCL_ERROR;
	    }
	    break;
	case TMRT_CALIBRATE:
	    calibrate = objv[i];
	    break;
	}
    }

    if (i >= objc || i < objc-3) {
usage:
	Tcl_WrongNumArgs(interp, 1, objv, "?-direct? ?-calibrate? ?-overhead double? command ?time ?max-count??");
	return TCL_ERROR;
    }
    objPtr = objv[i++];
    if (i < objc) {	/* max-time */
	result = Tcl_GetWideIntFromObj(interp, objv[i++], &maxms);
	if (result != TCL_OK) {
	    return result;
	}
	if (i < objc) {	/* max-count*/
	    Tcl_WideInt v;
	    result = Tcl_GetWideIntFromObj(interp, objv[i], &v);
	    if (result != TCL_OK) {
		return result;
	    }
	    maxcnt = (v > 0) ? v : 0;
	}
    }

    /* if calibrate */
    if (calibrate) {

	/* if no time specified for the calibration */
	if (maxms == WIDE_MIN) {
	    Tcl_Obj *clobjv[6];
	    Tcl_WideInt maxCalTime = 5000;
	    double lastMeasureOverhead = measureOverhead;
	    
	    clobjv[0] = objv[0]; 
	    i = 1;
	    if (direct) {
	    	clobjv[i++] = direct;
	    }
	    clobjv[i++] = objPtr; 

	    /* reset last measurement overhead */
	    measureOverhead = (double)0;

	    /* self-call with 100 milliseconds to warm-up,
	     * before entering the calibration cycle */
	    TclNewIntObj(clobjv[i], 100);
	    Tcl_IncrRefCount(clobjv[i]);
	    result = Tcl_TimeRateObjCmd(dummy, interp, i+1, clobjv);
	    Tcl_DecrRefCount(clobjv[i]);
	    if (result != TCL_OK) {
		return result;
	    }

	    i--;
	    clobjv[i++] = calibrate;
	    clobjv[i++] = objPtr; 

	    /* set last measurement overhead to max */
	    measureOverhead = (double)UWIDE_MAX;

	    /* calibration cycle until it'll be preciser */
	    maxms = -1000;
	    do {
		lastMeasureOverhead = measureOverhead;
		TclNewIntObj(clobjv[i], (int)maxms);
		Tcl_IncrRefCount(clobjv[i]);
		result = Tcl_TimeRateObjCmd(dummy, interp, i+1, clobjv);
		Tcl_DecrRefCount(clobjv[i]);
		if (result != TCL_OK) {
		    return result;
		}
		maxCalTime += maxms;
		/* increase maxms for preciser calibration */
		maxms -= (-maxms / 4);
		/* as long as new value more as 0.05% better */
	    } while ( (measureOverhead >= lastMeasureOverhead
		    || measureOverhead / lastMeasureOverhead <= 0.9995)
		    && maxCalTime > 0
	    );

	    return result;
	}
	if (maxms == 0) {
	    /* reset last measurement overhead */
	    measureOverhead = 0;
	    Tcl_SetObjResult(interp, Tcl_NewLongObj(0));
	    return TCL_OK;
	}

	/* if time is negative - make current overhead more precise */
	if (maxms > 0) {
	    /* set last measurement overhead to max */
	    measureOverhead = (double)UWIDE_MAX;
	} else {
	    maxms = -maxms;
	}

    }

    if (maxms == WIDE_MIN) {
    	maxms = 1000;
    }
    if (overhead == -1) {
	overhead = measureOverhead;
    }

    /* be sure that resetting of result will not smudge the further measurement */
    Tcl_ResetResult(interp);

    /* compile object */
    if (!direct) {
	if (TclInterpReady(interp) != TCL_OK) {
	    return TCL_ERROR;
	}
	codePtr = TclCompileObj(interp, objPtr, NULL, 0);
	TclPreserveByteCode(codePtr);
    }

    /* get start and stop time */
#ifdef TCL_WIDE_CLICKS
    start = middle = TclpGetWideClicks();
    /* time to stop execution (in wide clicks) */
    stop = start + (maxms * 1000 / TclpWideClickInMicrosec());
#else
    Tcl_GetTime(&now);
    start = now.sec; start *= 1000000; start += now.usec;
    middle = start;
    /* time to stop execution (in microsecs) */
    stop = start + maxms * 1000;
#endif

    /* start measurement */
    if (maxcnt > 0)
    while (1) {
    	/* eval single iteration */
    	count++;

	if (!direct) {
	    /* precompiled */
	    rootPtr = TOP_CB(interp);
	    result = TclNRExecuteByteCode(interp, codePtr);
	    result = TclNRRunCallbacks(interp, result, rootPtr);
	} else {
	    /* eval */
	    result = TclEvalObjEx(interp, objPtr, 0, NULL, 0);
	}
	if (result != TCL_OK) {
	    /* allow break from measurement cycle (used for conditional stop) */
	    if (result != TCL_BREAK) {
		goto done;
	    }
	    /* force stop immediately */
	    threshold = 1;
	    maxcnt = 0;
	    result = TCL_OK;
	}
	
	/* don't check time up to threshold */
	if (--threshold > 0) continue;

	/* check stop time reached, estimate new threshold */
    #ifdef TCL_WIDE_CLICKS
	middle = TclpGetWideClicks();
    #else
	Tcl_GetTime(&now);
	middle = now.sec; middle *= 1000000; middle += now.usec;
    #endif
	if (middle >= stop || count >= maxcnt) {
	    break;
	}

	/* don't calculate threshold by few iterations, because sometimes first
	 * iteration(s) can be too fast or slow (cached, delayed clean up, etc) */
	if (count < 10) {
	   threshold = 1; continue;
	}

	/* average iteration time in microsecs */
	threshold = (middle - start) / count;
	if (threshold > maxIterTm) {
	    maxIterTm = threshold;
	    /* interations seems to be longer */
	    if (threshold > (maxIterTm * 2)) {
		if ((factor *= 2) > 50) factor = 50;
	    } else {
		if (factor < 50) factor++;
	    }
	} else if (factor > 4) {
	    /* interations seems to be shorter */
	    if (threshold < (maxIterTm / 2)) {
		if ((factor /= 2) < 4) factor = 4;
	    } else {
		factor--;
	    }
	}
	/* as relation between remaining time and time since last check,
	 * maximal some % of time (by factor), so avoid growing of the execution time
	 * if iterations are not consistent, e. g. wax continuously on time) */
	threshold = ((stop - middle) / maxIterTm) / factor + 1;
	if (threshold > 100000) {	    /* fix for too large threshold */
	    threshold = 100000;
	}
	/* consider max-count */
	if (threshold > maxcnt - count) {
	    threshold = maxcnt - count;
	}
    }

    {
	Tcl_Obj *objarr[8], **objs = objarr;
	Tcl_WideInt val;
	const char *fmt;

	middle -= start;		     /* execution time in microsecs */

    #ifdef TCL_WIDE_CLICKS
	/* convert execution time in wide clicks to microsecs */
	middle *= TclpWideClickInMicrosec();
    #endif

	/* if not calibrate */
	if (!calibrate) {
	    /* minimize influence of measurement overhead */
	    if (overhead > 0) {
		/* estimate the time of overhead (microsecs) */
		Tcl_WideUInt curOverhead = overhead * count;
		if (middle > (Tcl_WideInt)curOverhead) {
		    middle -= curOverhead;
		} else {
		    middle = 0;
		}
	    }
	} else {
	    /* calibration - obtaining new measurement overhead */
	    if (measureOverhead > (double)middle / count) {
		measureOverhead = (double)middle / count;
	    }
	    objs[0] = Tcl_NewDoubleObj(measureOverhead);
	    TclNewLiteralStringObj(objs[1], "\xC2\xB5s/#-overhead"); /* mics */
	    objs += 2;
	}

	val = middle / count;		     /* microsecs per iteration */
	if (val >= 1000000) {
	    objs[0] = Tcl_NewWideIntObj(val);
	} else {
	    if (val < 10)    { fmt = "%.6f"; } else
	    if (val < 100)   { fmt = "%.4f"; } else
	    if (val < 1000)  { fmt = "%.3f"; } else
	    if (val < 10000) { fmt = "%.2f"; } else
			     { fmt = "%.1f"; };
	    objs[0] = Tcl_ObjPrintf(fmt, ((double)middle)/count);
	}

	objs[2] = Tcl_NewWideIntObj(count); /* iterations */
	
	/* calculate speed as rate (count) per sec */
	if (!middle) middle++; /* +1 ms, just to avoid divide by zero */
	if (count < (WIDE_MAX / 1000000)) {
	    val = (count * 1000000) / middle;
	    if (val < 100000) {
		if (val < 100)	{ fmt = "%.3f"; } else
		if (val < 1000) { fmt = "%.2f"; } else
				{ fmt = "%.1f"; };
		objs[4] = Tcl_ObjPrintf(fmt, ((double)(count * 1000000)) / middle);
	    } else {
		objs[4] = Tcl_NewWideIntObj(val);
	    }
	} else {
	    objs[4] = Tcl_NewWideIntObj((count / middle) * 1000000);
	}

	/* estimated net execution time (in millisecs) */
	if (!calibrate) {
	    objs[6] = Tcl_ObjPrintf("%.3f", (double)middle / 1000);
	    TclNewLiteralStringObj(objs[7], "nett-ms");
	}

	/*
	* Construct the result as a list because many programs have always parsed
	* as such (extracting the first element, typically).
	*/

	TclNewLiteralStringObj(objs[1], "\xC2\xB5s/#"); /* mics/# */
	TclNewLiteralStringObj(objs[3], "#");
	TclNewLiteralStringObj(objs[5], "#/sec");
	Tcl_SetObjResult(interp, Tcl_NewListObj(8, objarr));
    }

done:

    if (codePtr != NULL) {
	TclReleaseByteCode(codePtr);
    }

    return result;
}
 
/*
 *----------------------------------------------------------------------
 *
 * Tcl_TryObjCmd, TclNRTryObjCmd --
 *
 *	This procedure is invoked to process the "try" Tcl command. See the

Changes to generic/tclInt.h.

2803
2804
2805
2806
2807
2808
2809
2810
2811
2812
2813
2814
2815
2816
2817
2818
....
2820
2821
2822
2823
2824
2825
2826
2827
2828
2829
2830
2831
2832
2833
2834
2835
2836
2837
....
3190
3191
3192
3193
3194
3195
3196

3197
3198
3199









3200


3201
3202
3203
3204
3205
3206
3207
....
3463
3464
3465
3466
3467
3468
3469



3470
3471
3472
3473
3474
3475
3476
				/* Procedure that unloads a loaded module */
};

/* Flags for conversion of doubles to digit strings */

#define TCL_DD_SHORTEST 		0x4
				/* Use the shortest possible string */
#define TCL_DD_STEELE   		0x5
				/* Use the original Steele&White algorithm */
#define TCL_DD_E_FORMAT 		0x2
				/* Use a fixed-length string of digits,
				 * suitable for E format*/
#define TCL_DD_F_FORMAT 		0x3
				/* Use a fixed number of digits after the
				 * decimal point, suitable for F format */

................................................................................
				/* Allow return of a shorter digit string
				 * if it converts losslessly */
#define TCL_DD_NO_QUICK 		0x8
				/* Debug flag: forbid quick FP conversion */

#define TCL_DD_CONVERSION_TYPE_MASK	0x3
				/* Mask to isolate the conversion type */
#define TCL_DD_STEELE0 			0x1
				/* 'Steele&White' after masking */
#define TCL_DD_SHORTEST0		0x0
				/* 'Shortest possible' after masking */

/*
 *----------------------------------------------------------------
 * Procedures shared among Tcl modules but not used by the outside world:
 *----------------------------------------------------------------
 */

................................................................................
MODULE_SCOPE int	TclpLoadMemory(Tcl_Interp *interp, void *buffer,
			    int size, int codeSize, Tcl_LoadHandle *loadHandle,
			    Tcl_FSUnloadFileProc **unloadProcPtr, int flags);
#endif
MODULE_SCOPE void	TclInitThreadStorage(void);
MODULE_SCOPE void	TclFinalizeThreadDataThread(void);
MODULE_SCOPE void	TclFinalizeThreadStorage(void);

#ifdef TCL_WIDE_CLICKS
MODULE_SCOPE Tcl_WideInt TclpGetWideClicks(void);
MODULE_SCOPE double	TclpWideClicksToNanoseconds(Tcl_WideInt clicks);









#endif


MODULE_SCOPE int	TclZlibInit(Tcl_Interp *interp);
MODULE_SCOPE void *	TclpThreadCreateKey(void);
MODULE_SCOPE void	TclpThreadDeleteKey(void *keyPtr);
MODULE_SCOPE void	TclpThreadSetMasterTSD(void *tsdKeyPtr, void *ptr);
MODULE_SCOPE void *	TclpThreadGetMasterTSD(void *tsdKeyPtr);
MODULE_SCOPE void	TclErrorStackResetIf(Tcl_Interp *interp,
			    const char *msg, size_t length);
................................................................................
MODULE_SCOPE int	Tcl_TellObjCmd(void *clientData,
			    Tcl_Interp *interp, int objc,
			    Tcl_Obj *const objv[]);
MODULE_SCOPE int	Tcl_ThrowObjCmd(void *dummy, Tcl_Interp *interp,
			    int objc, Tcl_Obj *const objv[]);
MODULE_SCOPE int	Tcl_TimeObjCmd(void *clientData,
			    Tcl_Interp *interp, int objc,



			    Tcl_Obj *const objv[]);
MODULE_SCOPE int	Tcl_TraceObjCmd(void *clientData,
			    Tcl_Interp *interp, int objc,
			    Tcl_Obj *const objv[]);
MODULE_SCOPE int	Tcl_TryObjCmd(void *clientData,
			    Tcl_Interp *interp, int objc,
			    Tcl_Obj *const objv[]);






<
<







 







<
<
<
<







 







>



>
>
>
>
>
>
>
>
>

>
>







 







>
>
>







2803
2804
2805
2806
2807
2808
2809


2810
2811
2812
2813
2814
2815
2816
....
2818
2819
2820
2821
2822
2823
2824




2825
2826
2827
2828
2829
2830
2831
....
3184
3185
3186
3187
3188
3189
3190
3191
3192
3193
3194
3195
3196
3197
3198
3199
3200
3201
3202
3203
3204
3205
3206
3207
3208
3209
3210
3211
3212
3213
....
3469
3470
3471
3472
3473
3474
3475
3476
3477
3478
3479
3480
3481
3482
3483
3484
3485
				/* Procedure that unloads a loaded module */
};

/* Flags for conversion of doubles to digit strings */

#define TCL_DD_SHORTEST 		0x4
				/* Use the shortest possible string */


#define TCL_DD_E_FORMAT 		0x2
				/* Use a fixed-length string of digits,
				 * suitable for E format*/
#define TCL_DD_F_FORMAT 		0x3
				/* Use a fixed number of digits after the
				 * decimal point, suitable for F format */

................................................................................
				/* Allow return of a shorter digit string
				 * if it converts losslessly */
#define TCL_DD_NO_QUICK 		0x8
				/* Debug flag: forbid quick FP conversion */

#define TCL_DD_CONVERSION_TYPE_MASK	0x3
				/* Mask to isolate the conversion type */





/*
 *----------------------------------------------------------------
 * Procedures shared among Tcl modules but not used by the outside world:
 *----------------------------------------------------------------
 */

................................................................................
MODULE_SCOPE int	TclpLoadMemory(Tcl_Interp *interp, void *buffer,
			    int size, int codeSize, Tcl_LoadHandle *loadHandle,
			    Tcl_FSUnloadFileProc **unloadProcPtr, int flags);
#endif
MODULE_SCOPE void	TclInitThreadStorage(void);
MODULE_SCOPE void	TclFinalizeThreadDataThread(void);
MODULE_SCOPE void	TclFinalizeThreadStorage(void);

#ifdef TCL_WIDE_CLICKS
MODULE_SCOPE Tcl_WideInt TclpGetWideClicks(void);
MODULE_SCOPE double	TclpWideClicksToNanoseconds(Tcl_WideInt clicks);
MODULE_SCOPE double	TclpWideClickInMicrosec(void);
#else
#   ifdef _WIN32
#	define TCL_WIDE_CLICKS 1
MODULE_SCOPE Tcl_WideInt TclpGetWideClicks(void);
MODULE_SCOPE double	TclpWideClickInMicrosec(void);
#	define		TclpWideClicksToNanoseconds(clicks) \
				((double)(clicks) * TclpWideClickInMicrosec() * 1000)
#   endif
#endif
MODULE_SCOPE Tcl_WideInt TclpGetMicroseconds(void);

MODULE_SCOPE int	TclZlibInit(Tcl_Interp *interp);
MODULE_SCOPE void *	TclpThreadCreateKey(void);
MODULE_SCOPE void	TclpThreadDeleteKey(void *keyPtr);
MODULE_SCOPE void	TclpThreadSetMasterTSD(void *tsdKeyPtr, void *ptr);
MODULE_SCOPE void *	TclpThreadGetMasterTSD(void *tsdKeyPtr);
MODULE_SCOPE void	TclErrorStackResetIf(Tcl_Interp *interp,
			    const char *msg, size_t length);
................................................................................
MODULE_SCOPE int	Tcl_TellObjCmd(void *clientData,
			    Tcl_Interp *interp, int objc,
			    Tcl_Obj *const objv[]);
MODULE_SCOPE int	Tcl_ThrowObjCmd(void *dummy, Tcl_Interp *interp,
			    int objc, Tcl_Obj *const objv[]);
MODULE_SCOPE int	Tcl_TimeObjCmd(void *clientData,
			    Tcl_Interp *interp, int objc,
			    Tcl_Obj *const objv[]);
MODULE_SCOPE int	Tcl_TimeRateObjCmd(void *clientData,
			    Tcl_Interp *interp, int objc,
			    Tcl_Obj *const objv[]);
MODULE_SCOPE int	Tcl_TraceObjCmd(void *clientData,
			    Tcl_Interp *interp, int objc,
			    Tcl_Obj *const objv[]);
MODULE_SCOPE int	Tcl_TryObjCmd(void *clientData,
			    Tcl_Interp *interp, int objc,
			    Tcl_Obj *const objv[]);

Changes to generic/tclObj.c.

2707
2708
2709
2710
2711
2712
2713
2714
2715
2716
2717
2718
2719
2720

2721
2722
2723
2724
2725
2726
2727
2728
2729
2730
2731
2732
2733
2734
2735
2736
2737
2738
2739
2740
2741
....
2949
2950
2951
2952
2953
2954
2955
2956
2957
2958
2959
2960
2961
2962
2963
2964

2965
2966
2967
2968
2969
2970
2971
2972
2973
2974
2975
2976
2977
2978
2979
2980
2981
2982
2983
2984
2985
....
3408
3409
3410
3411
3412
3413
3414
3415
3416
3417
3418
3419
3420
3421
3422
3423
3424



3425
3426
3427
3428
3429
3430
3431
3432
3433
3434
3435
3436
3437
3438
3439
3440
3441
3442
3443
3444
3445
3446
3447
3448
	     * Must check for those bignum values that can fit in a long, even
	     * when auto-narrowing is enabled. Only those values in the signed
	     * long range get auto-narrowed to tclIntType, while all the
	     * values in the unsigned long range will fit in a long.
	     */

	    mp_int big;

	    UNPACK_BIGNUM(objPtr, big);
	    if ((size_t) big.used <= (CHAR_BIT * sizeof(unsigned long) + DIGIT_BIT - 1)
		    / DIGIT_BIT) {
		unsigned long scratch, value = 0, numBytes = sizeof(unsigned long);
		unsigned char *bytes = (unsigned char *) &scratch;


		if (mp_to_unsigned_bin_n(&big, bytes, &numBytes) == MP_OKAY) {
		    while (numBytes-- > 0) {
			value = (value << CHAR_BIT) | *bytes++;
		    }
		    if (big.sign) {
			if (value <= 1 + (unsigned long)LONG_MAX) {
			    *longPtr = - (long) value;
			    return TCL_OK;
			}
		    } else {
			if (value <= (unsigned long)ULONG_MAX) {
			    *longPtr = (long) value;
			    return TCL_OK;
			}
		    }
		}
	    }
#ifndef TCL_WIDE_INT_IS_LONG
	tooLarge:
#endif
	    if (interp != NULL) {
................................................................................
	if (objPtr->typePtr == &tclBignumType) {
	    /*
	     * Must check for those bignum values that can fit in a
	     * Tcl_WideInt, even when auto-narrowing is enabled.
	     */

	    mp_int big;

	    UNPACK_BIGNUM(objPtr, big);
	    if ((size_t) big.used <= (CHAR_BIT * sizeof(Tcl_WideInt)
		     + DIGIT_BIT - 1) / DIGIT_BIT) {
		Tcl_WideUInt value = 0;
		unsigned long numBytes = sizeof(Tcl_WideInt);
		Tcl_WideInt scratch;
		unsigned char *bytes = (unsigned char *) &scratch;


		if (mp_to_unsigned_bin_n(&big, bytes, &numBytes) == MP_OKAY) {
		    while (numBytes-- > 0) {
			value = (value << CHAR_BIT) | *bytes++;
		    }
		    if (big.sign) {
			if (value <= 1 + ~(Tcl_WideUInt)WIDE_MIN) {
			    *wideIntPtr = - (Tcl_WideInt) value;
			    return TCL_OK;
			}
		    } else {
			if (value <= (Tcl_WideUInt)WIDE_MAX) {
			    *wideIntPtr = (Tcl_WideInt) value;
			    return TCL_OK;
			}
		    }
		}
	    }
	    if (interp != NULL) {
		const char *s = "integer value too large to represent";
		Tcl_Obj *msg = Tcl_NewStringObj(s, -1);

................................................................................
 */

void
Tcl_SetBignumObj(
    Tcl_Obj *objPtr,		/* Object to set */
    mp_int *bignumValue)	/* Value to store */
{
    if (Tcl_IsShared(objPtr)) {
	Tcl_Panic("%s called with shared object", "Tcl_SetBignumObj");
    }
    if ((size_t) bignumValue->used
	    <= (CHAR_BIT * sizeof(Tcl_WideUInt) + DIGIT_BIT - 1) / DIGIT_BIT) {
	Tcl_WideUInt value = 0;
	unsigned long numBytes = sizeof(Tcl_WideUInt);
	Tcl_WideUInt scratch;
	unsigned char *bytes = (unsigned char *) &scratch;




	if (mp_to_unsigned_bin_n(bignumValue, bytes, &numBytes) != MP_OKAY) {
	    goto tooLargeForWide;
	}
	while (numBytes-- > 0) {
	    value = (value << CHAR_BIT) | *bytes++;
	}
	if (value > ((Tcl_WideUInt)WIDE_MAX + bignumValue->sign)) {
	    goto tooLargeForWide;
	}
	if (bignumValue->sign) {
	    TclSetIntObj(objPtr, -(Tcl_WideInt)value);
	} else {
	    TclSetIntObj(objPtr, (Tcl_WideInt)value);
	}
	mp_clear(bignumValue);
	return;
    }
  tooLargeForWide:
    TclInvalidateStringRep(objPtr);
    TclFreeIntRep(objPtr);
    TclSetBignumIntRep(objPtr, bignumValue);
}
 
/*






<
<
<
<
|
|

>
|
|

|
|
|
|
|
|
|
|
|
|
<







 







<
<
<
<
|
|
|
|

>
|
|
|
|
|
|
|
|
|
|
|
|
|
<







 







<
<
<
<
<
|
|
|
|

>
>
>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
<







2707
2708
2709
2710
2711
2712
2713




2714
2715
2716
2717
2718
2719
2720
2721
2722
2723
2724
2725
2726
2727
2728
2729
2730

2731
2732
2733
2734
2735
2736
2737
....
2945
2946
2947
2948
2949
2950
2951




2952
2953
2954
2955
2956
2957
2958
2959
2960
2961
2962
2963
2964
2965
2966
2967
2968
2969
2970

2971
2972
2973
2974
2975
2976
2977
....
3400
3401
3402
3403
3404
3405
3406





3407
3408
3409
3410
3411
3412
3413
3414
3415
3416
3417
3418
3419
3420
3421
3422
3423
3424
3425
3426
3427
3428
3429
3430

3431
3432
3433
3434
3435
3436
3437
	     * Must check for those bignum values that can fit in a long, even
	     * when auto-narrowing is enabled. Only those values in the signed
	     * long range get auto-narrowed to tclIntType, while all the
	     * values in the unsigned long range will fit in a long.
	     */

	    mp_int big;




	    unsigned long scratch, value = 0, numBytes = sizeof(unsigned long);
	    unsigned char *bytes = (unsigned char *) &scratch;

	    UNPACK_BIGNUM(objPtr, big);
	    if (mp_to_unsigned_bin_n(&big, bytes, &numBytes) == MP_OKAY) {
		while (numBytes-- > 0) {
			value = (value << CHAR_BIT) | *bytes++;
		}
		if (big.sign) {
		    if (value <= 1 + (unsigned long)LONG_MAX) {
			*longPtr = - (long) value;
			return TCL_OK;
		    }
		} else {
		    if (value <= (unsigned long)ULONG_MAX) {
			*longPtr = (long) value;
			return TCL_OK;

		    }
		}
	    }
#ifndef TCL_WIDE_INT_IS_LONG
	tooLarge:
#endif
	    if (interp != NULL) {
................................................................................
	if (objPtr->typePtr == &tclBignumType) {
	    /*
	     * Must check for those bignum values that can fit in a
	     * Tcl_WideInt, even when auto-narrowing is enabled.
	     */

	    mp_int big;




	    Tcl_WideUInt value = 0;
	    unsigned long numBytes = sizeof(Tcl_WideInt);
	    Tcl_WideInt scratch;
	    unsigned char *bytes = (unsigned char *) &scratch;

	    UNPACK_BIGNUM(objPtr, big);
	    if (mp_to_unsigned_bin_n(&big, bytes, &numBytes) == MP_OKAY) {
		while (numBytes-- > 0) {
		    value = (value << CHAR_BIT) | *bytes++;
		}
		if (big.sign) {
		    if (value <= 1 + ~(Tcl_WideUInt)WIDE_MIN) {
			*wideIntPtr = - (Tcl_WideInt) value;
			return TCL_OK;
		    }
		} else {
		    if (value <= (Tcl_WideUInt)WIDE_MAX) {
			*wideIntPtr = (Tcl_WideInt) value;
			return TCL_OK;

		    }
		}
	    }
	    if (interp != NULL) {
		const char *s = "integer value too large to represent";
		Tcl_Obj *msg = Tcl_NewStringObj(s, -1);

................................................................................
 */

void
Tcl_SetBignumObj(
    Tcl_Obj *objPtr,		/* Object to set */
    mp_int *bignumValue)	/* Value to store */
{





    Tcl_WideUInt value = 0;
    unsigned long numBytes = sizeof(Tcl_WideUInt);
    Tcl_WideUInt scratch;
    unsigned char *bytes = (unsigned char *) &scratch;

    if (Tcl_IsShared(objPtr)) {
	Tcl_Panic("%s called with shared object", "Tcl_SetBignumObj");
    }
    if (mp_to_unsigned_bin_n(bignumValue, bytes, &numBytes) != MP_OKAY) {
	goto tooLargeForWide;
    }
    while (numBytes-- > 0) {
	value = (value << CHAR_BIT) | *bytes++;
    }
    if (value > ((Tcl_WideUInt)WIDE_MAX + bignumValue->sign)) {
	goto tooLargeForWide;
    }
    if (bignumValue->sign) {
	TclSetIntObj(objPtr, -(Tcl_WideInt)value);
    } else {
	TclSetIntObj(objPtr, (Tcl_WideInt)value);
    }
    mp_clear(bignumValue);
    return;

  tooLargeForWide:
    TclInvalidateStringRep(objPtr);
    TclFreeIntRep(objPtr);
    TclSetBignumIntRep(objPtr, bignumValue);
}
 
/*

Changes to generic/tclPort.h.

21
22
23
24
25
26
27



28
29
30
31
#   include "tclWinPort.h"
#else
#   include "tclUnixPort.h"
#endif
#include "tcl.h"

#define UWIDE_MAX ((Tcl_WideUInt)-1)



#define WIDE_MAX ((Tcl_WideInt)(UWIDE_MAX >> 1))
#define WIDE_MIN ((Tcl_WideInt)((Tcl_WideUInt)WIDE_MAX+1))

#endif /* _TCLPORT */






>
>
>




21
22
23
24
25
26
27
28
29
30
31
32
33
34
#   include "tclWinPort.h"
#else
#   include "tclUnixPort.h"
#endif
#include "tcl.h"

#define UWIDE_MAX ((Tcl_WideUInt)-1)
#define WIDE_MAX ((Tcl_WideInt)(UWIDE_MAX >> 1))
#define WIDE_MIN ((Tcl_WideInt)((Tcl_WideUInt)WIDE_MAX+1))
#define UWIDE_MAX ((Tcl_WideUInt)-1)
#define WIDE_MAX ((Tcl_WideInt)(UWIDE_MAX >> 1))
#define WIDE_MIN ((Tcl_WideInt)((Tcl_WideUInt)WIDE_MAX+1))

#endif /* _TCLPORT */

Changes to generic/tclStrToD.c.

316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
....
2343
2344
2345
2346
2347
2348
2349
2350
2351
2352
2353
2354
2355
2356
2357
2358
2359
2360
2361
2362
2363
2364
2365
2366
2367
2368
2369
2370
2371
2372
2373
2374
2375
....
2377
2378
2379
2380
2381
2382
2383
2384
2385
2386
2387
2388
2389
2390
2391
2392
2393
2394
....
2810
2811
2812
2813
2814
2815
2816
2817
2818
2819
2820
2821
2822
2823
2824
2825
....
2878
2879
2880
2881
2882
2883
2884
2885
2886
2887
2888
2889
2890
2891
2892
....
2907
2908
2909
2910
2911
2912
2913
2914
2915
2916
2917
2918
2919
2920
2921
....
2979
2980
2981
2982
2983
2984
2985
2986
2987
2988
2989
2990
2991
2992
2993
2994
....
3089
3090
3091
3092
3093
3094
3095
3096
3097
3098
3099
3100
3101
3102
3103
....
3125
3126
3127
3128
3129
3130
3131
3132
3133
3134
3135
3136
3137
3138
3139
3140
3141
....
3153
3154
3155
3156
3157
3158
3159
3160
3161
3162
3163
3164
3165
3166
3167
3168
3169
3170
....
3186
3187
3188
3189
3190
3191
3192
3193
3194
3195
3196
3197
3198
3199
3200
3201
....
3273
3274
3275
3276
3277
3278
3279
3280
3281
3282
3283
3284
3285
3286
3287
....
3301
3302
3303
3304
3305
3306
3307
3308
3309
3310
3311
3312
3313
3314
3315
....
3379
3380
3381
3382
3383
3384
3385
3386
3387
3388
3389
3390
3391
3392
3393
3394
....
3401
3402
3403
3404
3405
3406
3407
3408
3409
3410
3411
3412
3413
3414
3415
....
3420
3421
3422
3423
3424
3425
3426
3427
3428
3429
3430
3431
3432
3433
3434
....
3469
3470
3471
3472
3473
3474
3475
3476
3477
3478
3479
3480
3481
3482
3483
....
3533
3534
3535
3536
3537
3538
3539
3540
3541
3542
3543
3544
3545
3546

3547
3548
3549
3550
3551

3552
3553

3554
3555
3556
3557
3558
3559
3560
3561
3562
3563
3564
3565
3566
3567
3568
3569
....
3584
3585
3586
3587
3588
3589
3590
3591
3592
3593
3594
3595
3596
3597
3598
....
3605
3606
3607
3608
3609
3610
3611
3612
3613
3614
3615
3616
3617
3618
3619
....
3641
3642
3643
3644
3645
3646
3647
3648
3649
3650
3651
3652
3653
3654
3655
....
3662
3663
3664
3665
3666
3667
3668
3669
3670
3671
3672
3673
3674
3675
3676
3677
....
3682
3683
3684
3685
3686
3687
3688
3689
3690
3691
3692
3693
3694
3695
3696
3697
....
3770
3771
3772
3773
3774
3775
3776
3777
3778
3779
3780
3781
3782
3783
3784
....
3800
3801
3802
3803
3804
3805
3806
3807
3808
3809
3810
3811
3812
3813
3814
....
3818
3819
3820
3821
3822
3823
3824
3825
3826
3827
3828
3829
3830
3831
3832
3833
3834
3835
3836
3837
3838
3839
3840
3841
....
3934
3935
3936
3937
3938
3939
3940
3941
3942
3943
3944
3945
3946
3947
3948
....
3970
3971
3972
3973
3974
3975
3976
3977
3978
3979
3980
3981
3982
3983
3984
3985
3986
3987
3988
3989
3990
3991
3992
....
4028
4029
4030
4031
4032
4033
4034
4035
4036
4037
4038
4039
4040
4041
4042
4043
4044
4045
....
4099
4100
4101
4102
4103
4104
4105
4106
4107
4108
4109
4110
4111
4112
4113
4114
4115
4116
4117
4118
4119
4120
4121
4122
4123
4124
....
4183
4184
4185
4186
4187
4188
4189
4190
4191
4192
4193
4194
4195
4196
4197
....
4202
4203
4204
4205
4206
4207
4208
4209
4210
4211
4212
4213
4214
4215
4216
4217
4218
4219
4220
4221
4222
4223
4224
4225
....
4239
4240
4241
4242
4243
4244
4245
4246
4247
4248
4249
4250
4251
4252
4253
....
4256
4257
4258
4259
4260
4261
4262
4263
4264
4265
4266
4267
4268
4269
4270
4271
4272
4273
4274
4275
4276
4277
4278
4279
4280
static char *		ShorteningQuickFormat(double, int, int, double,
			    char *, int *);
static char *		StrictQuickFormat(double, int, int, double,
			    char *, int *);
static char *		QuickConversion(double, int, int, int, int, int, int,
			    int *, char **);
static void		CastOutPowersOf2(int *, int *, int *);
static char *		ShorteningInt64Conversion(Double *, int, Tcl_WideUInt,
			    int, int, int, int, int, int, int, int, int,
			    int, int, int *, char **);
static char *		StrictInt64Conversion(Double *, int, Tcl_WideUInt,
			    int, int, int, int, int, int,
			    int, int, int *, char **);
static int		ShouldBankerRoundUpPowD(mp_int *, int, int);
static int		ShouldBankerRoundUpToNextPowD(mp_int *, mp_int *,
			    int, int, int, mp_int *);
static char *		ShorteningBignumConversionPowD(Double *dPtr,
			    int convType, Tcl_WideUInt bw, int b2, int b5,
			    int m2plus, int m2minus, int m5,
			    int sd, int k, int len,
			    int ilim, int ilim1, int *decpt,
			    char **endPtr);
static char *		StrictBignumConversionPowD(Double *dPtr, int convType,
			    Tcl_WideUInt bw, int b2, int b5,
			    int sd, int k, int len,
			    int ilim, int ilim1, int *decpt,
			    char **endPtr);
static int		ShouldBankerRoundUp(mp_int *, mp_int *, int);
static int		ShouldBankerRoundUpToNext(mp_int *, mp_int *,
			    mp_int *, int, int, mp_int *);
static char *		ShorteningBignumConversion(Double *dPtr, int convType,
			    Tcl_WideUInt bw, int b2,
			    int m2plus, int m2minus,
			    int s2, int s5, int k, int len,
			    int ilim, int ilim1, int *decpt,
			    char **endPtr);
static char *		StrictBignumConversion(Double *dPtr, int convType,
			    Tcl_WideUInt bw, int b2,
			    int s2, int s5, int k, int len,
			    int ilim, int ilim1, int *decpt,
			    char **endPtr);
static double		BignumToBiasedFrExp(const mp_int *big, int *machexp);
static double		Pow10TimesFrExp(int exponent, double fraction,
			    int *machexp);
................................................................................
 *	one too high.
 *
 *----------------------------------------------------------------------
 */

static inline void
SetPrecisionLimits(
    int convType,		/* Type of conversion: TCL_DD_SHORTEST,
				 * TCL_DD_STEELE0, TCL_DD_E_FMT,
				 * TCL_DD_F_FMT. */
    int k,			/* Floor(log10(number to convert)) */
    int *ndigitsPtr,		/* IN/OUT: Number of digits requested (will be
				 *         adjusted if needed). */
    int *iPtr,			/* OUT: Maximum number of digits to return. */
    int *iLimPtr,		/* OUT: Number of digits of significance if
				 *      the bignum method is used.*/
    int *iLim1Ptr)		/* OUT: Number of digits of significance if
				 *      the quick method is used. */
{
    switch (convType) {
    case TCL_DD_SHORTEST0:
    case TCL_DD_STEELE0:
	*iLimPtr = *iLim1Ptr = -1;
	*iPtr = 18;
	*ndigitsPtr = 0;
	break;
    case TCL_DD_E_FORMAT:
	if (*ndigitsPtr <= 0) {
	    *ndigitsPtr = 1;
	}
	*iLimPtr = *iLim1Ptr = *iPtr = *ndigitsPtr;
	break;
    case TCL_DD_F_FORMAT:
................................................................................
	*iLimPtr = *iPtr;
	*iLim1Ptr = *iPtr - 1;
	if (*iPtr <= 0) {
	    *iPtr = 1;
	}
	break;
    default:
	*iPtr = -1;
	*iLimPtr = -1;
	*iLim1Ptr = -1;
	Tcl_Panic("impossible conversion type in TclDoubleDigits");
    }
}
 
/*
 *----------------------------------------------------------------------
 *
 * BumpUp --
................................................................................
 *
 *----------------------------------------------------------------------
 */

static inline char *
ShorteningInt64Conversion(
    Double *dPtr,		/* Original number to convert. */
    int convType,		/* Type of conversion (shortest, Steele,
				 * E format, F format). */
    Tcl_WideUInt bw,		/* Integer significand. */
    int b2, int b5,		/* Scale factor for the significand in the
				 * numerator. */
    int m2plus, int m2minus, int m5,
				/* Scale factors for 1/2 ulp in the numerator
				 * (will be different if bw == 1. */
    int s2, int s5,		/* Scale factors for the denominator. */
................................................................................

	/*
	 * Does the current digit put us on the low side of the exact value
	 * but within within roundoff of being exact?
	 */

	if (b < mplus || (b == mplus
		&& convType != TCL_DD_STEELE0 && (dPtr->w.word1 & 1) == 0)) {
	    /*
	     * Make sure we shouldn't be rounding *up* instead, in case the
	     * next number above is closer.
	     */

	    if (2 * b > S || (2 * b == S && (digit & 1) != 0)) {
		++digit;
................................................................................

	/*
	 * Does one plus the current digit put us within roundoff of the
	 * number?
	 */

	if (b > S - mminus || (b == S - mminus
		&& convType != TCL_DD_STEELE0 && (dPtr->w.word1 & 1) == 0)) {
	    if (digit == 9) {
		*s++ = '9';
		s = BumpUp(s, retval, &k);
		break;
	    }
	    ++digit;
	    *s++ = '0' + digit;
................................................................................
 *
 *----------------------------------------------------------------------
 */

static inline char *
StrictInt64Conversion(
    Double *dPtr,		/* Original number to convert. */
    int convType,		/* Type of conversion (shortest, Steele,
				 * E format, F format). */
    Tcl_WideUInt bw,		/* Integer significand. */
    int b2, int b5,		/* Scale factor for the significand in the
				 * numerator. */
    int s2, int s5,		/* Scale factors for the denominator. */
    int k,			/* Number of output digits before the decimal
				 * point. */
    int len,			/* Number of digits to allocate. */
................................................................................
static inline int
ShouldBankerRoundUpPowD(
    mp_int *b,			/* Numerator of the fraction. */
    int sd,			/* Denominator is 2**(sd*DIGIT_BIT). */
    int isodd)			/* 1 if the digit is odd, 0 if even. */
{
    int i;
    static const mp_digit topbit = 1 << (DIGIT_BIT - 1);

    if (b->used < sd || (b->dp[sd-1] & topbit) == 0) {
	return 0;
    }
    if (b->dp[sd-1] != topbit) {
	return 1;
    }
................................................................................
 */

static inline int
ShouldBankerRoundUpToNextPowD(
    mp_int *b,			/* Numerator of the fraction. */
    mp_int *m,			/* Numerator of the rounding tolerance. */
    int sd,			/* Common denominator is 2**(sd*DIGIT_BIT). */
    int convType,		/* Conversion type: STEELE defeats
				 * round-to-even (not sure why one wants to do
				 * this; I copied it from Gay). FIXME */
    int isodd,			/* 1 if the integer significand is odd. */
    mp_int *temp)		/* Work area for the calculation. */
{
    int i;

    /*
     * Compare B and S-m - which is the same as comparing B+m and S - which we
................................................................................
    }
    for (i = sd-1; i >= 0; --i) {
				/* Check for ==s */
	if (temp->dp[i] != 0) {	/* > s */
	    return 1;
	}
    }
    if (convType == TCL_DD_STEELE0) {
				/* Biased rounding. */
	return 0;
    }
    return isodd;
}
 
/*
 *----------------------------------------------------------------------
 *
 * ShorteningBignumConversionPowD --
................................................................................
 *
 *----------------------------------------------------------------------
 */

static inline char *
ShorteningBignumConversionPowD(
    Double *dPtr,		/* Original number to convert. */
    int convType,		/* Type of conversion (shortest, Steele,
				 * E format, F format). */
    Tcl_WideUInt bw,		/* Integer significand. */
    int b2, int b5,		/* Scale factor for the significand in the
				 * numerator. */
    int m2plus, int m2minus, int m5,
				/* Scale factors for 1/2 ulp in the numerator
				 * (will be different if bw == 1). */
    int sd,			/* Scale factor for the denominator. */
................................................................................
	/*
	 * Does the current digit put us on the low side of the exact value
	 * but within within roundoff of being exact?
	 */

	r1 = mp_cmp_mag(&b, (m2plus > m2minus)? &mplus : &mminus);
	if (r1 == MP_LT || (r1 == MP_EQ
		&& convType != TCL_DD_STEELE0 && (dPtr->w.word1 & 1) == 0)) {
	    /*
	     * Make sure we shouldn't be rounding *up* instead, in case the
	     * next number above is closer.
	     */

	    if (ShouldBankerRoundUpPowD(&b, sd, digit&1)) {
		++digit;
................................................................................
	}

	/*
	 * Does one plus the current digit put us within roundoff of the
	 * number?
	 */

	if (ShouldBankerRoundUpToNextPowD(&b, &mminus, sd, convType,
		dPtr->w.word1 & 1, &temp)) {
	    if (digit == 9) {
		*s++ = '9';
		s = BumpUp(s, retval, &k);
		break;
	    }
	    ++digit;
................................................................................
 *
 *----------------------------------------------------------------------
 */

static inline char *
StrictBignumConversionPowD(
    Double *dPtr,		/* Original number to convert. */
    int convType,		/* Type of conversion (shortest, Steele,
				 * E format, F format). */
    Tcl_WideUInt bw,		/* Integer significand. */
    int b2, int b5,		/* Scale factor for the significand in the
				 * numerator. */
    int sd,			/* Scale factor for the denominator. */
    int k,			/* Number of output digits before the decimal
				 * point. */
    int len,			/* Number of digits to allocate. */
................................................................................
    char *retval = Tcl_Alloc(len + 1);
				/* Output buffer. */
    mp_int b;			/* Numerator of the fraction being
				 * converted. */
    mp_digit digit;		/* Current output digit. */
    char *s = retval;		/* Cursor in the output buffer. */
    int i;			/* Index in the output buffer. */
    mp_int temp;

    /*
     * b = bw * 2**b2 * 5**b5
     */

    TclInitBignumFromWideUInt(&b, bw);
    MulPow5(&b, b5, &b);
................................................................................
     */

    if (b.used <= sd) {
	mp_mul_d(&b, 10, &b);
	ilim = ilim1;
	--k;
    }
    mp_init(&temp);

    /*
     * Loop through the digits. Do division and mod by s == 2**(sd*DIGIT_BIT)
     * by mp_digit extraction.
     */

    i = 1;
................................................................................
    }

    /*
     * Endgame - store the location of the decimal point and the end of the
     * string.
     */

    mp_clear_multi(&b, &temp, NULL);
    *s = '\0';
    *decpt = k;
    if (endPtr) {
	*endPtr = s;
    }
    return retval;
}
................................................................................

static inline int
ShouldBankerRoundUpToNext(
    mp_int *b,			/* Remainder from the division that produced
				 * the last digit. */
    mp_int *m,			/* Numerator of the rounding tolerance. */
    mp_int *S,			/* Denominator. */
    int convType,		/* Conversion type: STEELE0 defeats
				 * round-to-even. (Not sure why one would want
				 * this; I coped it from Gay). FIXME */
    int isodd,			/* 1 if the integer significand is odd. */
    mp_int *temp)		/* Work area needed for the calculation. */
{
    int r;


    /*
     * Compare b and S-m: this is the same as comparing B+m and S.
     */


    mp_add(b, m, temp);
    r = mp_cmp_mag(temp, S);

    switch(r) {
    case MP_LT:
	return 0;
    case MP_EQ:
	if (convType == TCL_DD_STEELE0) {
	    return 0;
	} else {
	    return isodd;
	}
    case MP_GT:
	return 1;
    }
    Tcl_Panic("in ShouldBankerRoundUpToNext, trichotomy fails!");
    return 0;
}
 
................................................................................
 *
 *----------------------------------------------------------------------
 */

static inline char *
ShorteningBignumConversion(
    Double *dPtr,		/* Original number being converted. */
    int convType,		/* Conversion type. */
    Tcl_WideUInt bw,		/* Integer significand and exponent. */
    int b2,			/* Scale factor for the significand. */
    int m2plus, int m2minus,	/* Scale factors for 1/2 ulp in numerator. */
    int s2, int s5,		/* Scale factors for denominator. */
    int k,			/* Guessed position of the decimal point. */
    int len,			/* Size of the digit buffer to allocate. */
    int ilim,			/* Number of digits to convert if b >= s */
................................................................................
    char *s = retval;		/* Cursor in the return value. */
    mp_int b;			/* Numerator of the result. */
    mp_int mminus;		/* 1/2 ulp below the result. */
    mp_int mplus;		/* 1/2 ulp above the result. */
    mp_int S;			/* Denominator of the result. */
    mp_int dig;			/* Current digit of the result. */
    int digit;			/* Current digit of the result. */
    mp_int temp;		/* Work area. */
    int minit = 1;		/* Fudge factor for when we misguess k. */
    int i;
    int r1;

    /*
     * b = bw * 2**b2 * 5**b5
     * S = 2**s2 * 5*s5
................................................................................

    mp_init_set_int(&mminus, minit);
    mp_mul_2d(&mminus, m2minus, &mminus);
    if (m2plus > m2minus) {
	mp_init_copy(&mplus, &mminus);
	mp_mul_2d(&mplus, m2plus-m2minus, &mplus);
    }
    mp_init(&temp);

    /*
     * Loop through the digits.
     */

    mp_init(&dig);
    i = 1;
................................................................................

	/*
	 * Does the current digit leave us with a remainder small enough to
	 * round to it?
	 */

	r1 = mp_cmp_mag(&b, (m2plus > m2minus)? &mplus : &mminus);
	if (r1 == MP_LT || (r1 == MP_EQ
		&& convType != TCL_DD_STEELE0 && (dPtr->w.word1 & 1) == 0)) {
	    mp_mul_2d(&b, 1, &b);
	    if (ShouldBankerRoundUp(&b, &S, digit&1)) {
		++digit;
		if (digit == 10) {
		    *s++ = '9';
		    s = BumpUp(s, retval, &k);
		    break;
................................................................................
	}

	/*
	 * Does the current digit leave us with a remainder large enough to
	 * commit to rounding up to the next higher digit?
	 */

	if (ShouldBankerRoundUpToNext(&b, &mminus, &S, convType,
		dPtr->w.word1 & 1, &temp)) {
	    ++digit;
	    if (digit == 10) {
		*s++ = '9';
		s = BumpUp(s, retval, &k);
		break;
	    }
	    *s++ = '0' + digit;
................................................................................
     * Endgame - store the location of the decimal point and the end of the
     * string.
     */

    if (m2plus > m2minus) {
	mp_clear(&mplus);
    }
    mp_clear_multi(&b, &mminus, &temp, &dig, &S, NULL);
    *s = '\0';
    *decpt = k;
    if (endPtr) {
	*endPtr = s;
    }
    return retval;
}
................................................................................
 *
 *----------------------------------------------------------------------
 */

static inline char *
StrictBignumConversion(
    Double *dPtr,		/* Original number being converted. */
    int convType,		/* Conversion type. */
    Tcl_WideUInt bw,		/* Integer significand and exponent. */
    int b2,			/* Scale factor for the significand. */
    int s2, int s5,		/* Scale factors for denominator. */
    int k,			/* Guessed position of the decimal point. */
    int len,			/* Size of the digit buffer to allocate. */
    int ilim,			/* Number of digits to convert if b >= s */
    int ilim1,			/* Number of digits to convert if b < s */
................................................................................
    char *retval = Tcl_Alloc(len+1);
				/* Buffer of digits to return. */
    char *s = retval;		/* Cursor in the return value. */
    mp_int b;			/* Numerator of the result. */
    mp_int S;			/* Denominator of the result. */
    mp_int dig;			/* Current digit of the result. */
    int digit;			/* Current digit of the result. */
    mp_int temp;		/* Work area. */
    int g;			/* Size of the current digit ground. */
    int i, j;

    /*
     * b = bw * 2**b2 * 5**b5
     * S = 2**s2 * 5*s5
     */

    mp_init_multi(&temp, &dig, NULL);
    TclInitBignumFromWideUInt(&b, bw);
    mp_mul_2d(&b, b2, &b);
    mp_init_set_int(&S, 1);
    MulPow5(&S, s5, &S); mp_mul_2d(&S, s2, &S);

    /*
     * Handle the case where we guess the position of the decimal point wrong.
................................................................................
    ++s;

    /*
     * Endgame - store the location of the decimal point and the end of the
     * string.
     */

    mp_clear_multi(&b, &S, &temp, &dig, NULL);
    *s = '\0';
    *decpt = k;
    if (endPtr) {
	*endPtr = s;
    }
    return retval;
}
................................................................................
 * according to the 'flags' argument. Valid values for 'flags' include:
 *	TCL_DD_SHORTEST - This is the default for floating point conversion.
 *		It constructs the shortest string of
 *		digits that will reconvert to the given number when scanned.
 *		For floating point numbers that are exactly between two
 *		decimal numbers, it resolves using the 'round to even' rule.
 *		With this value, the 'ndigits' parameter is ignored.
 *	TCL_DD_STEELE - This value is not recommended and may be removed in
 *		the future. It follows the conversion algorithm outlined in
 *		"How to Print Floating-Point Numbers Accurately" by Guy
 *		L. Steele, Jr. and Jon L. White [Proc. ACM SIGPLAN '90,
 *		pp. 112-126]. This rule has the effect of rendering 1e23 as
 *		9.9999999999999999e22 - which is a 'better' approximation in
 *		the sense that it will reconvert correctly even if a
 *		subsequent input conversion is 'round up' or 'round down'
 *		rather than 'round to nearest', but is surprising otherwise.
 *	TCL_DD_E_FORMAT - This value is used to prepare numbers for %e format
 *		conversion. It constructs a string of at most 'ndigits' digits,
 *		choosing the one that is closest to the given number (and
 *		resolving ties with 'round to even').  It is allowed to return
 *		fewer than 'ndigits' if the number converts exactly; if the
 *		TCL_DD_E_FORMAT|TCL_DD_SHORTEN_FLAG is supplied instead, it
 *		also returns fewer digits if the shorter string will still
................................................................................
    int flags,			/* Conversion flags. */
    int *decpt,			/* OUTPUT: Position of the decimal point. */
    int *sign,			/* OUTPUT: 1 if the result is negative. */
    char **endPtr)		/* OUTPUT: If not NULL, receives a pointer to
				 *	   one character beyond the end of the
				 *	   returned string. */
{
    int convType = (flags & TCL_DD_CONVERSION_TYPE_MASK);
				/* Type of conversion being performed:
				 * TCL_DD_SHORTEST0, TCL_DD_STEELE0,
				 * TCL_DD_E_FORMAT, or TCL_DD_F_FORMAT. */
    Double d;			/* Union for deconstructing doubles. */
    Tcl_WideUInt bw;		/* Integer significand. */
    int be;			/* Power of 2 by which b must be multiplied */
    int bbits;			/* Number of bits needed to represent b. */
    int denorm;			/* Flag == 1 iff the input number was
				 * denormalized. */
    int k;			/* Estimate of floor(log10(d)). */
................................................................................

    ComputeScale(be, k, &b2, &b5, &s2, &s5);

    /*
     * Correct an incorrect caller-supplied 'ndigits'.  Also determine:
     *	i = The maximum number of decimal digits that will be returned in the
     *      formatted string.  This is k + 1 + ndigits for F format, 18 for
     *      shortest and Steele, and ndigits for E format.
     *  ilim = The number of significant digits to convert if k has been
     *         guessed correctly. This is -1 for shortest and Steele (which
     *         stop when all significance has been lost), 'ndigits' for E
     *         format, and 'k + 1 + ndigits' for F format.
     *  ilim1 = The minimum number of significant digits to convert if k has
     *	        been guessed 1 too high. This, too, is -1 for shortest and
     *	        Steele, and 'ndigits' for E format, but it's 'ndigits-1' for F
     *	        format.
     */

    SetPrecisionLimits(convType, k, &ndigits, &i, &ilim, &ilim1);

    /*
     * Try to do low-precision conversion in floating point rather than
     * resorting to expensive multiprecision arithmetic.
     */

    if (ilim >= 0 && ilim <= QUICK_MAX && !(flags & TCL_DD_NO_QUICK)) {
................................................................................
	     * If 10*2**s2*5**s5 == 2**(s2+1)+5**(s5+1) fits in a 64-bit word,
	     * then all our intermediate calculations can be done using exact
	     * 64-bit arithmetic with no need for expensive multiprecision
	     * operations. (This will be true for all numbers in the range
	     * [1.0e-3 .. 1.0e+24]).
	     */

	    return ShorteningInt64Conversion(&d, convType, bw, b2, b5, m2plus,
		    m2minus, m5, s2, s5, k, len, ilim, ilim1, decpt, endPtr);
	} else if (s5 == 0) {
	    /*
	     * The denominator is a power of 2, so we can replace division by
	     * digit shifts. First we round up s2 to a multiple of DIGIT_BIT,
	     * and adjust m2 and b2 accordingly. Then we launch into a version
	     * of the comparison that's specialized for the 'power of mp_digit
................................................................................
		int delta = DIGIT_BIT - (s2 % DIGIT_BIT);

		b2 += delta;
		m2plus += delta;
		m2minus += delta;
		s2 += delta;
	    }
	    return ShorteningBignumConversionPowD(&d, convType, bw, b2, b5,
		    m2plus, m2minus, m5, s2/DIGIT_BIT, k, len, ilim, ilim1,
		    decpt, endPtr);
	} else {
	    /*
	     * Alas, there's no helpful special case; use full-up bignum
	     * arithmetic for the conversion.
	     */

	    return ShorteningBignumConversion(&d, convType, bw, b2, m2plus,
		    m2minus, s2, s5, k, len, ilim, ilim1, decpt, endPtr);
	}
    } else {
	/*
	 * Non-shortening conversion.
	 */

................................................................................
	    /*
	     * If 10*2**s2*5**s5 == 2**(s2+1)+5**(s5+1) fits in a 64-bit word,
	     * then all our intermediate calculations can be done using exact
	     * 64-bit arithmetic with no need for expensive multiprecision
	     * operations.
	     */

	    return StrictInt64Conversion(&d, convType, bw, b2, b5, s2, s5, k,
		    len, ilim, ilim1, decpt, endPtr);
	} else if (s5 == 0) {
	    /*
	     * The denominator is a power of 2, so we can replace division by
	     * digit shifts. First we round up s2 to a multiple of DIGIT_BIT,
	     * and adjust m2 and b2 accordingly. Then we launch into a version
	     * of the comparison that's specialized for the 'power of mp_digit
................................................................................

	    if (s2 % DIGIT_BIT != 0) {
		int delta = DIGIT_BIT - (s2 % DIGIT_BIT);

		b2 += delta;
		s2 += delta;
	    }
	    return StrictBignumConversionPowD(&d, convType, bw, b2, b5,
		    s2/DIGIT_BIT, k, len, ilim, ilim1, decpt, endPtr);
	} else {
	    /*
	     * There are no helpful special cases, but at least we know in
	     * advance how many digits we will convert. We can run the
	     * conversion in steps of DIGIT_GROUP digits, so as to have many
	     * fewer mp_int divisions.
	     */

	    return StrictBignumConversion(&d, convType, bw, b2, s2, s5, k,
		    len, ilim, ilim1, decpt, endPtr);
	}
    }
}
 
/*
 *----------------------------------------------------------------------






|


|




|

|




|






|
|





|







 







|
|
<









|
<
<
<
<
<
<







 







|
|
|
|







 







<
<







 







|







 







|







 







<
<







 







|







 







<
<
<







 







<
<
<
<







 







<
<







 







|







 







|







 







<
<







 







<







 







<







 







|







 







<
<
<
|
<


>





>
|
|
>




<
<
<
|
<







 







<







 







<







 







<







 







|
<







 







|
|







 







|







 







<







 







<








|







 







|







 







<
<
<
<
<
<
<
<
<







 







<
<
<
<







 







|

|



|
|



|







 







|







 







|








|







 







|







 







|









|







316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
....
2343
2344
2345
2346
2347
2348
2349
2350
2351

2352
2353
2354
2355
2356
2357
2358
2359
2360
2361






2362
2363
2364
2365
2366
2367
2368
....
2370
2371
2372
2373
2374
2375
2376
2377
2378
2379
2380
2381
2382
2383
2384
2385
2386
2387
....
2803
2804
2805
2806
2807
2808
2809


2810
2811
2812
2813
2814
2815
2816
....
2869
2870
2871
2872
2873
2874
2875
2876
2877
2878
2879
2880
2881
2882
2883
....
2898
2899
2900
2901
2902
2903
2904
2905
2906
2907
2908
2909
2910
2911
2912
....
2970
2971
2972
2973
2974
2975
2976


2977
2978
2979
2980
2981
2982
2983
....
3078
3079
3080
3081
3082
3083
3084
3085
3086
3087
3088
3089
3090
3091
3092
....
3114
3115
3116
3117
3118
3119
3120



3121
3122
3123
3124
3125
3126
3127
....
3139
3140
3141
3142
3143
3144
3145




3146
3147
3148
3149
3150
3151
3152
....
3168
3169
3170
3171
3172
3173
3174


3175
3176
3177
3178
3179
3180
3181
....
3253
3254
3255
3256
3257
3258
3259
3260
3261
3262
3263
3264
3265
3266
3267
....
3281
3282
3283
3284
3285
3286
3287
3288
3289
3290
3291
3292
3293
3294
3295
....
3359
3360
3361
3362
3363
3364
3365


3366
3367
3368
3369
3370
3371
3372
....
3379
3380
3381
3382
3383
3384
3385

3386
3387
3388
3389
3390
3391
3392
....
3397
3398
3399
3400
3401
3402
3403

3404
3405
3406
3407
3408
3409
3410
....
3445
3446
3447
3448
3449
3450
3451
3452
3453
3454
3455
3456
3457
3458
3459
....
3509
3510
3511
3512
3513
3514
3515



3516

3517
3518
3519
3520
3521
3522
3523
3524
3525
3526
3527
3528
3529
3530
3531
3532



3533

3534
3535
3536
3537
3538
3539
3540
....
3555
3556
3557
3558
3559
3560
3561

3562
3563
3564
3565
3566
3567
3568
....
3575
3576
3577
3578
3579
3580
3581

3582
3583
3584
3585
3586
3587
3588
....
3610
3611
3612
3613
3614
3615
3616

3617
3618
3619
3620
3621
3622
3623
....
3630
3631
3632
3633
3634
3635
3636
3637

3638
3639
3640
3641
3642
3643
3644
....
3649
3650
3651
3652
3653
3654
3655
3656
3657
3658
3659
3660
3661
3662
3663
3664
....
3737
3738
3739
3740
3741
3742
3743
3744
3745
3746
3747
3748
3749
3750
3751
....
3767
3768
3769
3770
3771
3772
3773

3774
3775
3776
3777
3778
3779
3780
....
3784
3785
3786
3787
3788
3789
3790

3791
3792
3793
3794
3795
3796
3797
3798
3799
3800
3801
3802
3803
3804
3805
3806
....
3899
3900
3901
3902
3903
3904
3905
3906
3907
3908
3909
3910
3911
3912
3913
....
3935
3936
3937
3938
3939
3940
3941









3942
3943
3944
3945
3946
3947
3948
....
3984
3985
3986
3987
3988
3989
3990




3991
3992
3993
3994
3995
3996
3997
....
4051
4052
4053
4054
4055
4056
4057
4058
4059
4060
4061
4062
4063
4064
4065
4066
4067
4068
4069
4070
4071
4072
4073
4074
4075
4076
....
4135
4136
4137
4138
4139
4140
4141
4142
4143
4144
4145
4146
4147
4148
4149
....
4154
4155
4156
4157
4158
4159
4160
4161
4162
4163
4164
4165
4166
4167
4168
4169
4170
4171
4172
4173
4174
4175
4176
4177
....
4191
4192
4193
4194
4195
4196
4197
4198
4199
4200
4201
4202
4203
4204
4205
....
4208
4209
4210
4211
4212
4213
4214
4215
4216
4217
4218
4219
4220
4221
4222
4223
4224
4225
4226
4227
4228
4229
4230
4231
4232
static char *		ShorteningQuickFormat(double, int, int, double,
			    char *, int *);
static char *		StrictQuickFormat(double, int, int, double,
			    char *, int *);
static char *		QuickConversion(double, int, int, int, int, int, int,
			    int *, char **);
static void		CastOutPowersOf2(int *, int *, int *);
static char *		ShorteningInt64Conversion(Double *, Tcl_WideUInt,
			    int, int, int, int, int, int, int, int, int,
			    int, int, int *, char **);
static char *		StrictInt64Conversion(Double *, Tcl_WideUInt,
			    int, int, int, int, int, int,
			    int, int, int *, char **);
static int		ShouldBankerRoundUpPowD(mp_int *, int, int);
static int		ShouldBankerRoundUpToNextPowD(mp_int *, mp_int *,
			    int, int, mp_int *);
static char *		ShorteningBignumConversionPowD(Double *dPtr,
			    Tcl_WideUInt bw, int b2, int b5,
			    int m2plus, int m2minus, int m5,
			    int sd, int k, int len,
			    int ilim, int ilim1, int *decpt,
			    char **endPtr);
static char *		StrictBignumConversionPowD(Double *dPtr,
			    Tcl_WideUInt bw, int b2, int b5,
			    int sd, int k, int len,
			    int ilim, int ilim1, int *decpt,
			    char **endPtr);
static int		ShouldBankerRoundUp(mp_int *, mp_int *, int);
static int		ShouldBankerRoundUpToNext(mp_int *, mp_int *,
			    mp_int *, int);
static char *		ShorteningBignumConversion(Double *dPtr,
			    Tcl_WideUInt bw, int b2,
			    int m2plus, int m2minus,
			    int s2, int s5, int k, int len,
			    int ilim, int ilim1, int *decpt,
			    char **endPtr);
static char *		StrictBignumConversion(Double *dPtr,
			    Tcl_WideUInt bw, int b2,
			    int s2, int s5, int k, int len,
			    int ilim, int ilim1, int *decpt,
			    char **endPtr);
static double		BignumToBiasedFrExp(const mp_int *big, int *machexp);
static double		Pow10TimesFrExp(int exponent, double fraction,
			    int *machexp);
................................................................................
 *	one too high.
 *
 *----------------------------------------------------------------------
 */

static inline void
SetPrecisionLimits(
    int flags,		/* Type of conversion: TCL_DD_SHORTEST,
				 * TCL_DD_E_FMT, TCL_DD_F_FMT. */

    int k,			/* Floor(log10(number to convert)) */
    int *ndigitsPtr,		/* IN/OUT: Number of digits requested (will be
				 *         adjusted if needed). */
    int *iPtr,			/* OUT: Maximum number of digits to return. */
    int *iLimPtr,		/* OUT: Number of digits of significance if
				 *      the bignum method is used.*/
    int *iLim1Ptr)		/* OUT: Number of digits of significance if
				 *      the quick method is used. */
{
    switch (flags & TCL_DD_CONVERSION_TYPE_MASK) {






    case TCL_DD_E_FORMAT:
	if (*ndigitsPtr <= 0) {
	    *ndigitsPtr = 1;
	}
	*iLimPtr = *iLim1Ptr = *iPtr = *ndigitsPtr;
	break;
    case TCL_DD_F_FORMAT:
................................................................................
	*iLimPtr = *iPtr;
	*iLim1Ptr = *iPtr - 1;
	if (*iPtr <= 0) {
	    *iPtr = 1;
	}
	break;
    default:
	*iLimPtr = *iLim1Ptr = -1;
	*iPtr = 18;
	*ndigitsPtr = 0;
	break;
    }
}
 
/*
 *----------------------------------------------------------------------
 *
 * BumpUp --
................................................................................
 *
 *----------------------------------------------------------------------
 */

static inline char *
ShorteningInt64Conversion(
    Double *dPtr,		/* Original number to convert. */


    Tcl_WideUInt bw,		/* Integer significand. */
    int b2, int b5,		/* Scale factor for the significand in the
				 * numerator. */
    int m2plus, int m2minus, int m5,
				/* Scale factors for 1/2 ulp in the numerator
				 * (will be different if bw == 1. */
    int s2, int s5,		/* Scale factors for the denominator. */
................................................................................

	/*
	 * Does the current digit put us on the low side of the exact value
	 * but within within roundoff of being exact?
	 */

	if (b < mplus || (b == mplus
		&& (dPtr->w.word1 & 1) == 0)) {
	    /*
	     * Make sure we shouldn't be rounding *up* instead, in case the
	     * next number above is closer.
	     */

	    if (2 * b > S || (2 * b == S && (digit & 1) != 0)) {
		++digit;
................................................................................

	/*
	 * Does one plus the current digit put us within roundoff of the
	 * number?
	 */

	if (b > S - mminus || (b == S - mminus
		&& (dPtr->w.word1 & 1) == 0)) {
	    if (digit == 9) {
		*s++ = '9';
		s = BumpUp(s, retval, &k);
		break;
	    }
	    ++digit;
	    *s++ = '0' + digit;
................................................................................
 *
 *----------------------------------------------------------------------
 */

static inline char *
StrictInt64Conversion(
    Double *dPtr,		/* Original number to convert. */


    Tcl_WideUInt bw,		/* Integer significand. */
    int b2, int b5,		/* Scale factor for the significand in the
				 * numerator. */
    int s2, int s5,		/* Scale factors for the denominator. */
    int k,			/* Number of output digits before the decimal
				 * point. */
    int len,			/* Number of digits to allocate. */
................................................................................
static inline int
ShouldBankerRoundUpPowD(
    mp_int *b,			/* Numerator of the fraction. */
    int sd,			/* Denominator is 2**(sd*DIGIT_BIT). */
    int isodd)			/* 1 if the digit is odd, 0 if even. */
{
    int i;
    static const mp_digit topbit = ((mp_digit)1) << (DIGIT_BIT - 1);

    if (b->used < sd || (b->dp[sd-1] & topbit) == 0) {
	return 0;
    }
    if (b->dp[sd-1] != topbit) {
	return 1;
    }
................................................................................
 */

static inline int
ShouldBankerRoundUpToNextPowD(
    mp_int *b,			/* Numerator of the fraction. */
    mp_int *m,			/* Numerator of the rounding tolerance. */
    int sd,			/* Common denominator is 2**(sd*DIGIT_BIT). */



    int isodd,			/* 1 if the integer significand is odd. */
    mp_int *temp)		/* Work area for the calculation. */
{
    int i;

    /*
     * Compare B and S-m - which is the same as comparing B+m and S - which we
................................................................................
    }
    for (i = sd-1; i >= 0; --i) {
				/* Check for ==s */
	if (temp->dp[i] != 0) {	/* > s */
	    return 1;
	}
    }




    return isodd;
}
 
/*
 *----------------------------------------------------------------------
 *
 * ShorteningBignumConversionPowD --
................................................................................
 *
 *----------------------------------------------------------------------
 */

static inline char *
ShorteningBignumConversionPowD(
    Double *dPtr,		/* Original number to convert. */


    Tcl_WideUInt bw,		/* Integer significand. */
    int b2, int b5,		/* Scale factor for the significand in the
				 * numerator. */
    int m2plus, int m2minus, int m5,
				/* Scale factors for 1/2 ulp in the numerator
				 * (will be different if bw == 1). */
    int sd,			/* Scale factor for the denominator. */
................................................................................
	/*
	 * Does the current digit put us on the low side of the exact value
	 * but within within roundoff of being exact?
	 */

	r1 = mp_cmp_mag(&b, (m2plus > m2minus)? &mplus : &mminus);
	if (r1 == MP_LT || (r1 == MP_EQ
		&& (dPtr->w.word1 & 1) == 0)) {
	    /*
	     * Make sure we shouldn't be rounding *up* instead, in case the
	     * next number above is closer.
	     */

	    if (ShouldBankerRoundUpPowD(&b, sd, digit&1)) {
		++digit;
................................................................................
	}

	/*
	 * Does one plus the current digit put us within roundoff of the
	 * number?
	 */

	if (ShouldBankerRoundUpToNextPowD(&b, &mminus, sd,
		dPtr->w.word1 & 1, &temp)) {
	    if (digit == 9) {
		*s++ = '9';
		s = BumpUp(s, retval, &k);
		break;
	    }
	    ++digit;
................................................................................
 *
 *----------------------------------------------------------------------
 */

static inline char *
StrictBignumConversionPowD(
    Double *dPtr,		/* Original number to convert. */


    Tcl_WideUInt bw,		/* Integer significand. */
    int b2, int b5,		/* Scale factor for the significand in the
				 * numerator. */
    int sd,			/* Scale factor for the denominator. */
    int k,			/* Number of output digits before the decimal
				 * point. */
    int len,			/* Number of digits to allocate. */
................................................................................
    char *retval = Tcl_Alloc(len + 1);
				/* Output buffer. */
    mp_int b;			/* Numerator of the fraction being
				 * converted. */
    mp_digit digit;		/* Current output digit. */
    char *s = retval;		/* Cursor in the output buffer. */
    int i;			/* Index in the output buffer. */


    /*
     * b = bw * 2**b2 * 5**b5
     */

    TclInitBignumFromWideUInt(&b, bw);
    MulPow5(&b, b5, &b);
................................................................................
     */

    if (b.used <= sd) {
	mp_mul_d(&b, 10, &b);
	ilim = ilim1;
	--k;
    }


    /*
     * Loop through the digits. Do division and mod by s == 2**(sd*DIGIT_BIT)
     * by mp_digit extraction.
     */

    i = 1;
................................................................................
    }

    /*
     * Endgame - store the location of the decimal point and the end of the
     * string.
     */

    mp_clear(&b);
    *s = '\0';
    *decpt = k;
    if (endPtr) {
	*endPtr = s;
    }
    return retval;
}
................................................................................

static inline int
ShouldBankerRoundUpToNext(
    mp_int *b,			/* Remainder from the division that produced
				 * the last digit. */
    mp_int *m,			/* Numerator of the rounding tolerance. */
    mp_int *S,			/* Denominator. */



    int isodd)			/* 1 if the integer significand is odd. */

{
    int r;
    mp_int temp;

    /*
     * Compare b and S-m: this is the same as comparing B+m and S.
     */

    mp_init(&temp);
    mp_add(b, m, &temp);
    r = mp_cmp_mag(&temp, S);
    mp_clear(&temp);
    switch(r) {
    case MP_LT:
	return 0;
    case MP_EQ:



	return isodd;

    case MP_GT:
	return 1;
    }
    Tcl_Panic("in ShouldBankerRoundUpToNext, trichotomy fails!");
    return 0;
}
 
................................................................................
 *
 *----------------------------------------------------------------------
 */

static inline char *
ShorteningBignumConversion(
    Double *dPtr,		/* Original number being converted. */

    Tcl_WideUInt bw,		/* Integer significand and exponent. */
    int b2,			/* Scale factor for the significand. */
    int m2plus, int m2minus,	/* Scale factors for 1/2 ulp in numerator. */
    int s2, int s5,		/* Scale factors for denominator. */
    int k,			/* Guessed position of the decimal point. */
    int len,			/* Size of the digit buffer to allocate. */
    int ilim,			/* Number of digits to convert if b >= s */
................................................................................
    char *s = retval;		/* Cursor in the return value. */
    mp_int b;			/* Numerator of the result. */
    mp_int mminus;		/* 1/2 ulp below the result. */
    mp_int mplus;		/* 1/2 ulp above the result. */
    mp_int S;			/* Denominator of the result. */
    mp_int dig;			/* Current digit of the result. */
    int digit;			/* Current digit of the result. */

    int minit = 1;		/* Fudge factor for when we misguess k. */
    int i;
    int r1;

    /*
     * b = bw * 2**b2 * 5**b5
     * S = 2**s2 * 5*s5
................................................................................

    mp_init_set_int(&mminus, minit);
    mp_mul_2d(&mminus, m2minus, &mminus);
    if (m2plus > m2minus) {
	mp_init_copy(&mplus, &mminus);
	mp_mul_2d(&mplus, m2plus-m2minus, &mplus);
    }


    /*
     * Loop through the digits.
     */

    mp_init(&dig);
    i = 1;
................................................................................

	/*
	 * Does the current digit leave us with a remainder small enough to
	 * round to it?
	 */

	r1 = mp_cmp_mag(&b, (m2plus > m2minus)? &mplus : &mminus);
	if (r1 == MP_LT || (r1 == MP_EQ && (dPtr->w.word1 & 1) == 0)) {

	    mp_mul_2d(&b, 1, &b);
	    if (ShouldBankerRoundUp(&b, &S, digit&1)) {
		++digit;
		if (digit == 10) {
		    *s++ = '9';
		    s = BumpUp(s, retval, &k);
		    break;
................................................................................
	}

	/*
	 * Does the current digit leave us with a remainder large enough to
	 * commit to rounding up to the next higher digit?
	 */

	if (ShouldBankerRoundUpToNext(&b, &mminus, &S,
		dPtr->w.word1 & 1)) {
	    ++digit;
	    if (digit == 10) {
		*s++ = '9';
		s = BumpUp(s, retval, &k);
		break;
	    }
	    *s++ = '0' + digit;
................................................................................
     * Endgame - store the location of the decimal point and the end of the
     * string.
     */

    if (m2plus > m2minus) {
	mp_clear(&mplus);
    }
    mp_clear_multi(&b, &mminus, &dig, &S, NULL);
    *s = '\0';
    *decpt = k;
    if (endPtr) {
	*endPtr = s;
    }
    return retval;
}
................................................................................
 *
 *----------------------------------------------------------------------
 */

static inline char *
StrictBignumConversion(
    Double *dPtr,		/* Original number being converted. */

    Tcl_WideUInt bw,		/* Integer significand and exponent. */
    int b2,			/* Scale factor for the significand. */
    int s2, int s5,		/* Scale factors for denominator. */
    int k,			/* Guessed position of the decimal point. */
    int len,			/* Size of the digit buffer to allocate. */
    int ilim,			/* Number of digits to convert if b >= s */
    int ilim1,			/* Number of digits to convert if b < s */
................................................................................
    char *retval = Tcl_Alloc(len+1);
				/* Buffer of digits to return. */
    char *s = retval;		/* Cursor in the return value. */
    mp_int b;			/* Numerator of the result. */
    mp_int S;			/* Denominator of the result. */
    mp_int dig;			/* Current digit of the result. */
    int digit;			/* Current digit of the result. */

    int g;			/* Size of the current digit ground. */
    int i, j;

    /*
     * b = bw * 2**b2 * 5**b5
     * S = 2**s2 * 5*s5
     */

    mp_init_multi(&dig, NULL);
    TclInitBignumFromWideUInt(&b, bw);
    mp_mul_2d(&b, b2, &b);
    mp_init_set_int(&S, 1);
    MulPow5(&S, s5, &S); mp_mul_2d(&S, s2, &S);

    /*
     * Handle the case where we guess the position of the decimal point wrong.
................................................................................
    ++s;

    /*
     * Endgame - store the location of the decimal point and the end of the
     * string.
     */

    mp_clear_multi(&b, &S, &dig, NULL);
    *s = '\0';
    *decpt = k;
    if (endPtr) {
	*endPtr = s;
    }
    return retval;
}
................................................................................
 * according to the 'flags' argument. Valid values for 'flags' include:
 *	TCL_DD_SHORTEST - This is the default for floating point conversion.
 *		It constructs the shortest string of
 *		digits that will reconvert to the given number when scanned.
 *		For floating point numbers that are exactly between two
 *		decimal numbers, it resolves using the 'round to even' rule.
 *		With this value, the 'ndigits' parameter is ignored.









 *	TCL_DD_E_FORMAT - This value is used to prepare numbers for %e format
 *		conversion. It constructs a string of at most 'ndigits' digits,
 *		choosing the one that is closest to the given number (and
 *		resolving ties with 'round to even').  It is allowed to return
 *		fewer than 'ndigits' if the number converts exactly; if the
 *		TCL_DD_E_FORMAT|TCL_DD_SHORTEN_FLAG is supplied instead, it
 *		also returns fewer digits if the shorter string will still
................................................................................
    int flags,			/* Conversion flags. */
    int *decpt,			/* OUTPUT: Position of the decimal point. */
    int *sign,			/* OUTPUT: 1 if the result is negative. */
    char **endPtr)		/* OUTPUT: If not NULL, receives a pointer to
				 *	   one character beyond the end of the
				 *	   returned string. */
{




    Double d;			/* Union for deconstructing doubles. */
    Tcl_WideUInt bw;		/* Integer significand. */
    int be;			/* Power of 2 by which b must be multiplied */
    int bbits;			/* Number of bits needed to represent b. */
    int denorm;			/* Flag == 1 iff the input number was
				 * denormalized. */
    int k;			/* Estimate of floor(log10(d)). */
................................................................................

    ComputeScale(be, k, &b2, &b5, &s2, &s5);

    /*
     * Correct an incorrect caller-supplied 'ndigits'.  Also determine:
     *	i = The maximum number of decimal digits that will be returned in the
     *      formatted string.  This is k + 1 + ndigits for F format, 18 for
     *      shortest, and ndigits for E format.
     *  ilim = The number of significant digits to convert if k has been
     *         guessed correctly. This is -1 for shortest (which
     *         stop when all significance has been lost), 'ndigits' for E
     *         format, and 'k + 1 + ndigits' for F format.
     *  ilim1 = The minimum number of significant digits to convert if k has
     *	        been guessed 1 too high. This, too, is -1 for shortest,
     *	        and 'ndigits' for E format, but it's 'ndigits-1' for F
     *	        format.
     */

    SetPrecisionLimits(flags, k, &ndigits, &i, &ilim, &ilim1);

    /*
     * Try to do low-precision conversion in floating point rather than
     * resorting to expensive multiprecision arithmetic.
     */

    if (ilim >= 0 && ilim <= QUICK_MAX && !(flags & TCL_DD_NO_QUICK)) {
................................................................................
	     * If 10*2**s2*5**s5 == 2**(s2+1)+5**(s5+1) fits in a 64-bit word,
	     * then all our intermediate calculations can be done using exact
	     * 64-bit arithmetic with no need for expensive multiprecision
	     * operations. (This will be true for all numbers in the range
	     * [1.0e-3 .. 1.0e+24]).
	     */

	    return ShorteningInt64Conversion(&d, bw, b2, b5, m2plus,
		    m2minus, m5, s2, s5, k, len, ilim, ilim1, decpt, endPtr);
	} else if (s5 == 0) {
	    /*
	     * The denominator is a power of 2, so we can replace division by
	     * digit shifts. First we round up s2 to a multiple of DIGIT_BIT,
	     * and adjust m2 and b2 accordingly. Then we launch into a version
	     * of the comparison that's specialized for the 'power of mp_digit
................................................................................
		int delta = DIGIT_BIT - (s2 % DIGIT_BIT);

		b2 += delta;
		m2plus += delta;
		m2minus += delta;
		s2 += delta;
	    }
	    return ShorteningBignumConversionPowD(&d, bw, b2, b5,
		    m2plus, m2minus, m5, s2/DIGIT_BIT, k, len, ilim, ilim1,
		    decpt, endPtr);
	} else {
	    /*
	     * Alas, there's no helpful special case; use full-up bignum
	     * arithmetic for the conversion.
	     */

	    return ShorteningBignumConversion(&d, bw, b2, m2plus,
		    m2minus, s2, s5, k, len, ilim, ilim1, decpt, endPtr);
	}
    } else {
	/*
	 * Non-shortening conversion.
	 */

................................................................................
	    /*
	     * If 10*2**s2*5**s5 == 2**(s2+1)+5**(s5+1) fits in a 64-bit word,
	     * then all our intermediate calculations can be done using exact
	     * 64-bit arithmetic with no need for expensive multiprecision
	     * operations.
	     */

	    return StrictInt64Conversion(&d, bw, b2, b5, s2, s5, k,
		    len, ilim, ilim1, decpt, endPtr);
	} else if (s5 == 0) {
	    /*
	     * The denominator is a power of 2, so we can replace division by
	     * digit shifts. First we round up s2 to a multiple of DIGIT_BIT,
	     * and adjust m2 and b2 accordingly. Then we launch into a version
	     * of the comparison that's specialized for the 'power of mp_digit
................................................................................

	    if (s2 % DIGIT_BIT != 0) {
		int delta = DIGIT_BIT - (s2 % DIGIT_BIT);

		b2 += delta;
		s2 += delta;
	    }
	    return StrictBignumConversionPowD(&d, bw, b2, b5,
		    s2/DIGIT_BIT, k, len, ilim, ilim1, decpt, endPtr);
	} else {
	    /*
	     * There are no helpful special cases, but at least we know in
	     * advance how many digits we will convert. We can run the
	     * conversion in steps of DIGIT_GROUP digits, so as to have many
	     * fewer mp_int divisions.
	     */

	    return StrictBignumConversion(&d, bw, b2, s2, s5, k,
		    len, ilim, ilim1, decpt, endPtr);
	}
    }
}
 
/*
 *----------------------------------------------------------------------

Changes to generic/tclTest.c.

1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
....
1716
1717
1718
1719
1720
1721
1722
1723
1724
1725
1726
1727
1728
1729
1730
1731
1732
1733
1734
1735
1736
1737
 *
 * Usage:
 *	testdoubledigits fpval ndigits type ?shorten"
 *
 * Parameters:
 *	fpval - Floating-point value to format.
 *	ndigits - Digit count to request from Tcl_DoubleDigits
 *	type - One of 'shortest', 'Steele', 'e', 'f'
 *	shorten - Indicates that the 'shorten' flag should be passed in.
 *
 *-----------------------------------------------------------------------------
 */

static int
TestdoubledigitsObjCmd(void *unused,
................................................................................
		       int objc,
				/* Parameter count */
		       Tcl_Obj* const objv[])
				/* Parameter vector */
{
    static const char* options[] = {
	"shortest",
	"Steele",
	"e",
	"f",
	NULL
    };
    static const int types[] = {
	TCL_DD_SHORTEST,
	TCL_DD_STEELE,
	TCL_DD_E_FORMAT,
	TCL_DD_F_FORMAT
    };

    const Tcl_ObjType* doubleType;
    double d;
    int status;






|







 







<






<







1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
....
1716
1717
1718
1719
1720
1721
1722

1723
1724
1725
1726
1727
1728

1729
1730
1731
1732
1733
1734
1735
 *
 * Usage:
 *	testdoubledigits fpval ndigits type ?shorten"
 *
 * Parameters:
 *	fpval - Floating-point value to format.
 *	ndigits - Digit count to request from Tcl_DoubleDigits
 *	type - One of 'shortest', 'e', 'f'
 *	shorten - Indicates that the 'shorten' flag should be passed in.
 *
 *-----------------------------------------------------------------------------
 */

static int
TestdoubledigitsObjCmd(void *unused,
................................................................................
		       int objc,
				/* Parameter count */
		       Tcl_Obj* const objv[])
				/* Parameter vector */
{
    static const char* options[] = {
	"shortest",

	"e",
	"f",
	NULL
    };
    static const int types[] = {
	TCL_DD_SHORTEST,

	TCL_DD_E_FORMAT,
	TCL_DD_F_FORMAT
    };

    const Tcl_ObjType* doubleType;
    double d;
    int status;

Changes to generic/tclUtf.c.

1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
....
2154
2155
2156
2157
2158
2159
2160
2161
2162
2163
2164
2165
2166
2167
2168
		 * quickly if the next char in the pattern isn't a special
		 * character
		 */

		if ((p != '[') && (p != '?') && (p != '\\')) {
		    if (nocase) {
			while (*uniStr && (p != *uniStr)
				&& (p != Tcl_UniCharToLower(*uniStr))) {
			    uniStr++;
			}
		    } else {
			while (*uniStr && (p != *uniStr)) {
			    uniStr++;
			}
		    }
................................................................................
		 * quickly if the next char in the pattern isn't a special
		 * character.
		 */

		if ((p != '[') && (p != '?') && (p != '\\')) {
		    if (nocase) {
			while ((string < stringEnd) && (p != *string)
				&& (p != Tcl_UniCharToLower(*string))) {
			    string++;
			}
		    } else {
			while ((string < stringEnd) && (p != *string)) {
			    string++;
			}
		    }






|







 







|







1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
....
2154
2155
2156
2157
2158
2159
2160
2161
2162
2163
2164
2165
2166
2167
2168
		 * quickly if the next char in the pattern isn't a special
		 * character
		 */

		if ((p != '[') && (p != '?') && (p != '\\')) {
		    if (nocase) {
			while (*uniStr && (p != *uniStr)
				&& (p != (Tcl_UniChar)Tcl_UniCharToLower(*uniStr))) {
			    uniStr++;
			}
		    } else {
			while (*uniStr && (p != *uniStr)) {
			    uniStr++;
			}
		    }
................................................................................
		 * quickly if the next char in the pattern isn't a special
		 * character.
		 */

		if ((p != '[') && (p != '?') && (p != '\\')) {
		    if (nocase) {
			while ((string < stringEnd) && (p != *string)
				&& (p != (Tcl_UniChar)Tcl_UniCharToLower(*string))) {
			    string++;
			}
		    } else {
			while ((string < stringEnd) && (p != *string)) {
			    string++;
			}
		    }

Added tests-perf/clock.perf.tcl.






















































































































































































































































































































































































































































































































































































































































































































































































































































>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
#!/usr/bin/tclsh
# ------------------------------------------------------------------------
#
# test-performance.tcl --
# 
#  This file provides common performance tests for comparison of tcl-speed
#  degradation by switching between branches.
#  (currently for clock ensemble only)
#
# ------------------------------------------------------------------------
# 
# Copyright (c) 2014 Serg G. Brester (aka sebres)
# 
# See the file "license.terms" for information on usage and redistribution
# of this file.
# 

array set in {-time 500}
if {[info exists ::argv0] && [file tail $::argv0] eq [file tail [info script]]} {
  array set in $argv
}

## common test performance framework:
if {![namespace exists ::tclTestPerf]} {
  source [file join [file dirname [info script]] test-performance.tcl]
}

namespace eval ::tclTestPerf-TclClock {

namespace path {::tclTestPerf}

## set testing defaults:
set ::env(TCL_TZ) :CET

# warm-up interpeter compiler env, clock platform-related features:

## warm-up test-related features (load clock.tcl, system zones, locales, etc.):
clock scan "" -gmt 1
clock scan ""
clock scan "" -timezone :CET
clock scan "" -format "" -locale en
clock scan "" -format "" -locale de

## ------------------------------------------

proc test-format {{reptime 1000}} {
  _test_run $reptime {
    # Format : short, week only (in gmt)
    {clock format 1482525936 -format "%u" -gmt 1}
    # Format : short, week only (system zone)
    {clock format 1482525936 -format "%u"}
    # Format : short, week only (CEST)
    {clock format 1482525936 -format "%u" -timezone :CET}
    # Format : date only (in gmt)
    {clock format 1482525936 -format "%Y-%m-%d" -gmt 1}
    # Format : date only (system zone)
    {clock format 1482525936 -format "%Y-%m-%d"}
    # Format : date only (CEST)
    {clock format 1482525936 -format "%Y-%m-%d" -timezone :CET}
    # Format : time only (in gmt)
    {clock format 1482525936 -format "%H:%M" -gmt 1}
    # Format : time only (system zone)
    {clock format 1482525936 -format "%H:%M"}
    # Format : time only (CEST)
    {clock format 1482525936 -format "%H:%M" -timezone :CET}
    # Format : time only (in gmt)
    {clock format 1482525936 -format "%H:%M:%S" -gmt 1}
    # Format : time only (system zone)
    {clock format 1482525936 -format "%H:%M:%S"}
    # Format : time only (CEST)
    {clock format 1482525936 -format "%H:%M:%S" -timezone :CET}
    # Format : default (in gmt)
    {clock format 1482525936 -gmt 1 -locale en}
    # Format : default (system zone)
    {clock format 1482525936 -locale en}
    # Format : default (CEST)
    {clock format 1482525936 -timezone :CET -locale en}
    # Format : ISO date-time (in gmt, numeric zone)
    {clock format 1246379400 -format "%Y-%m-%dT%H:%M:%S %z" -gmt 1}
    # Format : ISO date-time (system zone, CEST, numeric zone)
    {clock format 1246379400 -format "%Y-%m-%dT%H:%M:%S %z"}
    # Format : ISO date-time (CEST, numeric zone)
    {clock format 1246379400 -format "%Y-%m-%dT%H:%M:%S %z" -timezone :CET}
    # Format : ISO date-time (system zone, CEST)
    {clock format 1246379400 -format "%Y-%m-%dT%H:%M:%S %Z"}
    # Format : julian day with time (in gmt):
    {clock format 1246379415 -format "%J %H:%M:%S" -gmt 1}
    # Format : julian day with time (system zone):
    {clock format 1246379415 -format "%J %H:%M:%S"}

    # Format : locale date-time (en):
    {clock format 1246379415 -format "%x %X" -locale en}
    # Format : locale date-time (de):
    {clock format 1246379415 -format "%x %X" -locale de}

    # Format : locale lookup table month:
    {clock format 1246379400 -format "%b" -locale en -gmt 1}
    # Format : locale lookup 2 tables - month and day:
    {clock format 1246379400 -format "%b %Od" -locale en -gmt 1}
    # Format : locale lookup 3 tables - week, month and day:
    {clock format 1246379400 -format "%a %b %Od" -locale en -gmt 1}
    # Format : locale lookup 4 tables - week, month, day and year:
    {clock format 1246379400 -format "%a %b %Od %Oy" -locale en -gmt 1}

    # Format : dynamic clock value (without converter caches):
    setup {set i 0}
    {clock format [incr i] -format "%Y-%m-%dT%H:%M:%S" -locale en -timezone :CET}
    cleanup {puts [clock format $i -format "%Y-%m-%dT%H:%M:%S" -locale en -timezone :CET]}
    # Format : dynamic clock value (without any converter caches, zone range overflow):
    setup {set i 0}
    {clock format [incr i 86400] -format "%Y-%m-%dT%H:%M:%S" -locale en -timezone :CET}
    cleanup {puts [clock format $i -format "%Y-%m-%dT%H:%M:%S" -locale en -timezone :CET]}

    # Format : dynamic format (cacheable)
    {clock format 1246379415 -format [string trim "%d.%m.%Y %H:%M:%S "] -gmt 1}

    # Format : all (in gmt, locale en)
    {clock format 1482525936 -format "%%a = %a | %%A = %A | %%b = %b | %%h = %h | %%B = %B | %%C = %C | %%d = %d | %%e = %e | %%g = %g | %%G = %G | %%H = %H | %%I = %I | %%j = %j | %%J = %J | %%k = %k | %%l = %l | %%m = %m | %%M = %M | %%N = %N | %%p = %p | %%P = %P | %%Q = %Q | %%s = %s | %%S = %S | %%t = %t | %%u = %u | %%U = %U | %%V = %V | %%w = %w | %%W = %W | %%y = %y | %%Y = %Y | %%z = %z | %%Z = %Z | %%n = %n | %%EE = %EE | %%EC = %EC | %%Ey = %Ey | %%n = %n | %%Od = %Od | %%Oe = %Oe | %%OH = %OH | %%Ok = %Ok | %%OI = %OI | %%Ol = %Ol | %%Om = %Om | %%OM = %OM | %%OS = %OS | %%Ou = %Ou | %%Ow = %Ow | %%Oy = %Oy" -gmt 1 -locale en}
    # Format : all (in CET, locale de)
    {clock format 1482525936 -format "%%a = %a | %%A = %A | %%b = %b | %%h = %h | %%B = %B | %%C = %C | %%d = %d | %%e = %e | %%g = %g | %%G = %G | %%H = %H | %%I = %I | %%j = %j | %%J = %J | %%k = %k | %%l = %l | %%m = %m | %%M = %M | %%N = %N | %%p = %p | %%P = %P | %%Q = %Q | %%s = %s | %%S = %S | %%t = %t | %%u = %u | %%U = %U | %%V = %V | %%w = %w | %%W = %W | %%y = %y | %%Y = %Y | %%z = %z | %%Z = %Z | %%n = %n | %%EE = %EE | %%EC = %EC | %%Ey = %Ey | %%n = %n | %%Od = %Od | %%Oe = %Oe | %%OH = %OH | %%Ok = %Ok | %%OI = %OI | %%Ol = %Ol | %%Om = %Om | %%OM = %OM | %%OS = %OS | %%Ou = %Ou | %%Ow = %Ow | %%Oy = %Oy" -timezone :CET -locale de}
  }
}

proc test-scan {{reptime 1000}} {
  _test_run $reptime {
    # Scan : date (in gmt)
    {clock scan "25.11.2015" -format "%d.%m.%Y" -base 0 -gmt 1}
    # Scan : date (system time zone, with base)
    {clock scan "25.11.2015" -format "%d.%m.%Y" -base 0}
    # Scan : date (system time zone, without base)
    {clock scan "25.11.2015" -format "%d.%m.%Y"}
    # Scan : greedy match
    {clock scan "111" -format "%d%m%y" -base 0 -gmt 1}
    {clock scan "1111" -format "%d%m%y" -base 0 -gmt 1}
    {clock scan "11111" -format "%d%m%y" -base 0 -gmt 1}
    {clock scan "111111" -format "%d%m%y" -base 0 -gmt 1}
    # Scan : greedy match (space separated)
    {clock scan "1 1 1" -format "%d%m%y" -base 0 -gmt 1}
    {clock scan "111 1" -format "%d%m%y" -base 0 -gmt 1}
    {clock scan "1 111" -format "%d%m%y" -base 0 -gmt 1}
    {clock scan "1 11 1" -format "%d%m%y" -base 0 -gmt 1}
    {clock scan "1 11 11" -format "%d%m%y" -base 0 -gmt 1}
    {clock scan "11 11 11" -format "%d%m%y" -base 0 -gmt 1}

    # Scan : time (in gmt)
    {clock scan "10:35:55" -format "%H:%M:%S" -base 1000000000 -gmt 1}
    # Scan : time (system time zone, with base)
    {clock scan "10:35:55" -format "%H:%M:%S" -base 1000000000}
    # Scan : time (gmt, without base)
    {clock scan "10:35:55" -format "%H:%M:%S" -gmt 1}
    # Scan : time (system time zone, without base)
    {clock scan "10:35:55" -format "%H:%M:%S"}

    # Scan : date-time (in gmt)
    {clock scan "25.11.2015 10:35:55" -format "%d.%m.%Y %H:%M:%S" -base 0 -gmt 1}
    # Scan : date-time (system time zone with base)
    {clock scan "25.11.2015 10:35:55" -format "%d.%m.%Y %H:%M:%S" -base 0}
    # Scan : date-time (system time zone without base)
    {clock scan "25.11.2015 10:35:55" -format "%d.%m.%Y %H:%M:%S"}

    # Scan : julian day in gmt
    {clock scan 2451545 -format %J -gmt 1}
    # Scan : julian day in system TZ
    {clock scan 2451545 -format %J}
    # Scan : julian day in other TZ
    {clock scan 2451545 -format %J -timezone +0200}
    # Scan : julian day with time:
    {clock scan "2451545 10:20:30" -format "%J %H:%M:%S"}
    # Scan : julian day with time (greedy match):
    {clock scan "2451545 102030" -format "%J%H%M%S"}

    # Scan : century, lookup table month
    {clock scan {1970 Jan 2} -format {%C%y %b %d} -locale en -gmt 1}
    # Scan : century, lookup table month and day (both entries are first)
    {clock scan {1970 Jan 01} -format {%C%y %b %Od} -locale en -gmt 1}
    # Scan : century, lookup table month and day (list scan: entries with position 12 / 31)
    {clock scan {2016 Dec 31} -format {%C%y %b %Od} -locale en -gmt 1}

    # Scan : ISO date-time (CEST)
    {clock scan "2009-06-30T18:30:00+02:00" -format "%Y-%m-%dT%H:%M:%S%z"}
    {clock scan "2009-06-30T18:30:00 CEST" -format "%Y-%m-%dT%H:%M:%S %z"}
    # Scan : ISO date-time (UTC)
    {clock scan "2009-06-30T18:30:00Z" -format "%Y-%m-%dT%H:%M:%S%z"}
    {clock scan "2009-06-30T18:30:00 UTC" -format "%Y-%m-%dT%H:%M:%S %z"}

    # Scan : locale date-time (en):
    {clock scan "06/30/2009 18:30:15" -format "%x %X" -gmt 1 -locale en}
    # Scan : locale date-time (de):
    {clock scan "30.06.2009 18:30:15" -format "%x %X" -gmt 1 -locale de}

    # Scan : dynamic format (cacheable)
    {clock scan "25.11.2015 10:35:55" -format [string trim "%d.%m.%Y %H:%M:%S "] -base 0 -gmt 1}

    break
    # # Scan : long format test (allock chain)
    # {clock scan "25.11.2015" -format "%d.%m.%Y %d.%m.%Y %d.%m.%Y %d.%m.%Y %d.%m.%Y %d.%m.%Y %d.%m.%Y %d.%m.%Y" -base 0 -gmt 1}
    # # Scan : dynamic, very long format test (create obj representation, allock chain, GC, etc):
    # {clock scan "25.11.2015" -format [string repeat "[incr i] %d.%m.%Y %d.%m.%Y" 10] -base 0 -gmt 1}
    # # Scan : again:
    # {clock scan "25.11.2015" -format [string repeat "[incr i -1] %d.%m.%Y %d.%m.%Y" 10] -base 0 -gmt 1}
  } {puts [clock format $_(r) -locale en]}
}

proc test-freescan {{reptime 1000}} {
  _test_run $reptime {
    # FreeScan : relative date
    {clock scan "5 years 18 months 385 days" -base 0 -gmt 1}
    # FreeScan : relative date with relative weekday
    {clock scan "5 years 18 months 385 days Fri" -base 0 -gmt 1}
    # FreeScan : relative date with ordinal month
    {clock scan "5 years 18 months 385 days next 1 January" -base 0 -gmt 1}
    # FreeScan : relative date with ordinal month and relative weekday
    {clock scan "5 years 18 months 385 days next January Fri" -base 0 -gmt 1}
    # FreeScan : ordinal month
    {clock scan "next January" -base 0 -gmt 1}
    # FreeScan : relative week
    {clock scan "next Fri" -base 0 -gmt 1}
    # FreeScan : relative weekday and week offset 
    {clock scan "next January + 2 week" -base 0 -gmt 1}
    # FreeScan : time only with base
    {clock scan "19:18:30" -base 148863600 -gmt 1}
    # FreeScan : time only without base, gmt
    {clock scan "19:18:30" -gmt 1}
    # FreeScan : time only without base, system
    {clock scan "19:18:30"}
    # FreeScan : date, system time zone
    {clock scan "05/08/2016 20:18:30"}
    # FreeScan : date, supplied time zone
    {clock scan "05/08/2016 20:18:30" -timezone :CET}
    # FreeScan : date, supplied gmt (equivalent -timezone :GMT)
    {clock scan "05/08/2016 20:18:30" -gmt 1}
    # FreeScan : date, supplied time zone gmt
    {clock scan "05/08/2016 20:18:30" -timezone :GMT}
    # FreeScan : time only, numeric zone in string, base time gmt (exchange zones between gmt / -0500)
    {clock scan "20:18:30 -0500" -base 148863600 -gmt 1}
    # FreeScan : time only, zone in string (exchange zones between system / gmt)
    {clock scan "19:18:30 GMT" -base 148863600}
    # FreeScan : fast switch of zones in cycle - GMT, MST, CET (system) and EST
    {clock scan "19:18:30 MST" -base 148863600 -gmt 1
     clock scan "19:18:30 EST" -base 148863600
    }
  } {puts [clock format $_(r) -locale en]}
}

proc test-add {{reptime 1000}} {
  set tests {
    # Add : years
    {clock add 1246379415 5 years -gmt 1}
    # Add : months
    {clock add 1246379415 18 months -gmt 1}
    # Add : weeks
    {clock add 1246379415 20 weeks -gmt 1}
    # Add : days
    {clock add 1246379415 385 days -gmt 1}
    # Add : weekdays
    {clock add 1246379415 3 weekdays -gmt 1}

    # Add : hours
    {clock add 1246379415 5 hours -gmt 1}
    # Add : minutes
    {clock add 1246379415 55 minutes -gmt 1}
    # Add : seconds
    {clock add 1246379415 100 seconds -gmt 1}

    # Add : +/- in gmt
    {clock add 1246379415 -5 years +21 months -20 weeks +386 days -19 hours +30 minutes -10 seconds -gmt 1}
    # Add : +/- in system timezone
    {clock add 1246379415 -5 years +21 months -20 weeks +386 days -19 hours +30 minutes -10 seconds -timezone :CET}

    # Add : gmt
    {clock add 1246379415 -5 years 18 months 366 days 5 hours 30 minutes 10 seconds -gmt 1}
    # Add : system timezone
    {clock add 1246379415 -5 years 18 months 366 days 5 hours 30 minutes 10 seconds -timezone :CET}

    # Add : all in gmt
    {clock add 1246379415 4 years 18 months 50 weeks 378 days 3 weekdays 5 hours 30 minutes 10 seconds -gmt 1}
    # Add : all in system timezone
    {clock add 1246379415 4 years 18 months 50 weeks 378 days 3 weekdays 5 hours 30 minutes 10 seconds -timezone :CET}

  }
  # if does not support add of weekdays:
  if {[catch {clock add 0 3 weekdays -gmt 1}]} {
    regsub -all {\mweekdays\M} $tests "days" tests
  }
  _test_run $reptime $tests {puts [clock format $_(r) -locale en]}
}

proc test-convert {{reptime 1000}} {
  _test_run $reptime {
    # Convert locale (en -> de):
    {clock format [clock scan "Tue May 30 2017" -format "%a %b %d %Y" -gmt 1 -locale en] -format "%a %b %d %Y" -gmt 1 -locale de}
    # Convert locale (de -> en):
    {clock format [clock scan "Di Mai 30 2017" -format "%a %b %d %Y" -gmt 1 -locale de] -format "%a %b %d %Y" -gmt 1 -locale en}

    # Convert TZ: direct
    {clock format [clock scan "19:18:30" -base 148863600 -timezone EST] -timezone MST}
    {clock format [clock scan "19:18:30" -base 148863600 -timezone MST] -timezone EST}
    # Convert TZ: included in scan string & format
    {clock format [clock scan "19:18:30 EST" -base 148863600] -format "%H:%M:%S %z" -timezone MST}
    {clock format [clock scan "19:18:30 EST" -base 148863600] -format "%H:%M:%S %z" -timezone EST}

    # Format locale 1x: comparison values
    {clock format 0 -gmt 1 -locale en} 
    {clock format 0 -gmt 1 -locale de}
    {clock format 0 -gmt 1 -locale fr}
    # Format locale 2x: without switching locale (en, en)
    {clock format 0 -gmt 1 -locale en; clock format 0 -gmt 1 -locale en}
    # Format locale 2x: with switching locale (en, de)
    {clock format 0 -gmt 1 -locale en; clock format 0 -gmt 1 -locale de}
    # Format locale 3x: without switching locale (en, en, en)
    {clock format 0 -gmt 1 -locale en; clock format 0 -gmt 1 -locale en; clock format 0 -gmt 1 -locale en}
    # Format locale 3x: with switching locale (en, de, fr)
    {clock format 0 -gmt 1 -locale en; clock format 0 -gmt 1 -locale de; clock format 0 -gmt 1 -locale fr}

    # Scan locale 2x: without switching locale (en, en) + (de, de)
    {clock scan "Tue May 30 2017" -format "%a %b %d %Y" -gmt 1 -locale en; clock scan "Tue May 30 2017" -format "%a %b %d %Y" -gmt 1 -locale en}
    {clock scan "Di Mai 30 2017" -format "%a %b %d %Y" -gmt 1 -locale de; clock scan "Di Mai 30 2017" -format "%a %b %d %Y" -gmt 1 -locale de}
    # Scan locale 2x: with switching locale (en, de)
    {clock scan "Tue May 30 2017" -format "%a %b %d %Y" -gmt 1 -locale en; clock scan "Di Mai 30 2017" -format "%a %b %d %Y" -gmt 1 -locale de}
    # Scan locale 3x: with switching locale (en, de, fr)
    {clock scan "Tue May 30 2017" -format "%a %b %d %Y" -gmt 1 -locale en; clock scan "Di Mai 30 2017" -format "%a %b %d %Y" -gmt 1 -locale de; clock scan "mar. mai 30 2017" -format "%a %b %d %Y" -gmt 1 -locale fr}

    # Format TZ 2x: comparison values
    {clock format 0 -timezone CET -format "%Y-%m-%d %H:%M:%S %z"}
    {clock format 0 -timezone EST -format "%Y-%m-%d %H:%M:%S %z"}
    # Format TZ 2x: without switching
    {clock format 0 -timezone CET -format "%Y-%m-%d %H:%M:%S %z"; clock format 0 -timezone CET -format "%Y-%m-%d %H:%M:%S %z"}
    {clock format 0 -timezone EST -format "%Y-%m-%d %H:%M:%S %z"; clock format 0 -timezone EST -format "%Y-%m-%d %H:%M:%S %z"}
    # Format TZ 2x: with switching
    {clock format 0 -timezone CET -format "%Y-%m-%d %H:%M:%S %z"; clock format 0 -timezone EST -format "%Y-%m-%d %H:%M:%S %z"}
    # Format TZ 3x: with switching (CET, EST, MST)
    {clock format 0 -timezone CET -format "%Y-%m-%d %H:%M:%S %z"; clock format 0 -timezone EST -format "%Y-%m-%d %H:%M:%S %z"; clock format 0 -timezone MST -format "%Y-%m-%d %H:%M:%S %z"}
    # Format TZ 3x: with switching (GMT, EST, MST)
    {clock format 0 -gmt 1 -format "%Y-%m-%d %H:%M:%S %z"; clock format 0 -timezone EST -format "%Y-%m-%d %H:%M:%S %z"; clock format 0 -timezone MST -format "%Y-%m-%d %H:%M:%S %z"}

    # FreeScan TZ 2x (+1 system-default): without switching TZ
    {clock scan "19:18:30 MST" -base 148863600; clock scan "19:18:30 MST" -base 148863600}
    {clock scan "19:18:30 EST" -base 148863600; clock scan "19:18:30 EST" -base 148863600}
    # FreeScan TZ 2x (+1 system-default): with switching TZ
    {clock scan "19:18:30 MST" -base 148863600; clock scan "19:18:30 EST" -base 148863600}
    # FreeScan TZ 2x (+1 gmt, +1 system-default)
    {clock scan "19:18:30 MST" -base 148863600 -gmt 1; clock scan "19:18:30 EST" -base 148863600}
    
    # Scan TZ: comparison included in scan string vs. given
    {clock scan "2009-06-30T18:30:00 CEST" -format "%Y-%m-%dT%H:%M:%S %z"}
    {clock scan "2009-06-30T18:30:00 CET" -format "%Y-%m-%dT%H:%M:%S %z"}
    {clock scan "2009-06-30T18:30:00" -timezone CET -format "%Y-%m-%dT%H:%M:%S"}
  }
}

proc test-other {{reptime 1000}} {
  _test_run $reptime {
    # Bad zone
    {catch {clock scan "1 day" -timezone BAD_ZONE -locale en}}

    # Scan : julian day (overflow)
    {catch {clock scan 5373485 -format %J}}

    # Scan : test rotate of GC objects (format is dynamic, so tcl-obj removed with last reference)
    {set i 0; time { clock scan "[incr i] - 25.11.2015" -format "$i - %d.%m.%Y" -base 0 -gmt 1 } 50}
    # Scan : test reusability of GC objects (format is dynamic, so tcl-obj removed with last reference)
    {set i 50; time { clock scan "[incr i -1] - 25.11.2015" -format "$i - %d.%m.%Y" -base 0 -gmt 1 } 50}
  }
}

proc test-ensemble-perf {{reptime 1000}} {
  _test_run $reptime {
    # Clock clicks (ensemble)
    {clock clicks}
    # Clock clicks (direct)
    {::tcl::clock::clicks}
    # Clock seconds (ensemble)
    {clock seconds}
    # Clock seconds (direct)
    {::tcl::clock::seconds}
    # Clock microseconds (ensemble)
    {clock microseconds}
    # Clock microseconds (direct)
    {::tcl::clock::microseconds}
    # Clock scan (ensemble)
    {clock scan ""}
    # Clock scan (direct)
    {::tcl::clock::scan ""}
    # Clock format (ensemble)
    {clock format 0 -f %s}
    # Clock format (direct)
    {::tcl::clock::format 0 -f %s}
  }
}

proc test {{reptime 1000}} {
  puts ""
  test-ensemble-perf [expr {$reptime / 2}]; #fast enough
  test-format $reptime
  test-scan $reptime
  test-freescan $reptime
  test-add $reptime
  test-convert [expr {$reptime / 2}]; #fast enough
  test-other $reptime

  puts \n**OK**
}

}; # end of ::tclTestPerf-TclClock

# ------------------------------------------------------------------------

# if calling direct:
if {[info exists ::argv0] && [file tail $::argv0] eq [file tail [info script]]} {
  ::tclTestPerf-TclClock::test $in(-time)
}

Added tests-perf/test-performance.tcl.
































































































































































































































































































>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
# ------------------------------------------------------------------------
#
# test-performance.tcl --
# 
#  This file provides common performance tests for comparison of tcl-speed
#  degradation or regression by switching between branches.
#
#  To execute test case evaluate direct corresponding file "tests-perf\*.perf.tcl".
#
# ------------------------------------------------------------------------
# 
# Copyright (c) 2014 Serg G. Brester (aka sebres)
# 
# See the file "license.terms" for information on usage and redistribution
# of this file.
# 

namespace eval ::tclTestPerf {
# warm-up interpeter compiler env, calibrate timerate measurement functionality:

# if no timerate here - import from unsupported:
if {[namespace which -command timerate] eq {}} {
  namespace inscope ::tcl::unsupported {namespace export timerate}
  namespace import ::tcl::unsupported::timerate
}

# if not yet calibrated:
if {[lindex [timerate {} 10] 6] >= (10-1)} {
  puts -nonewline "Calibration ... "; flush stdout
  puts "done: [lrange \
    [timerate -calibrate {}] \
  0 1]"
}

proc {**STOP**} {args} {
  return -code error -level 4 "**STOP** in [info level [expr {[info level]-2}]] [join $args { }]" 
}

proc _test_get_commands {lst} {
  regsub -all {(?:^|\n)[ \t]*(\#[^\n]*|\msetup\M[^\n]*|\mcleanup\M[^\n]*)(?=\n\s*(?:[\{\#]|setup|cleanup|$))} $lst "\n{\\1}"
}

proc _test_out_total {} {
  upvar _ _

  set tcnt [llength $_(itm)]
  if {!$tcnt} {
    puts ""
    return
  }

  set mintm 0x7fffffff
  set maxtm 0
  set nett 0
  set wtm 0
  set wcnt 0
  set i 0
  foreach tm $_(itm) {
    if {[llength $tm] > 6} {
      set nett [expr {$nett + [lindex $tm 6]}]
    }
    set wtm [expr {$wtm + [lindex $tm 0]}]
    set wcnt [expr {$wcnt + [lindex $tm 2]}]
    set tm [lindex $tm 0]
    if {$tm > $maxtm} {set maxtm $tm; set maxi $i}
    if {$tm < $mintm} {set mintm $tm; set mini $i}
    incr i
  }

  puts [string repeat ** 40]
  set s [format "%d cases in %.2f sec." $tcnt [expr {([clock milliseconds] - $_(starttime)) / 1000.0}]]
  if {$nett > 0} {
    append s [format " (%.2f nett-sec.)" [expr {$nett / 1000.0}]]
  }
  puts "Total $s:"
  lset _(m) 0 [format %.6f $wtm]
  lset _(m) 2 $wcnt
  lset _(m) 4 [format %.3f [expr {$wcnt / (($nett ? $nett : ($tcnt * [lindex $_(reptime) 0])) / 1000.0)}]]
  if {[llength $_(m)] > 6} {
    lset _(m) 6 [format %.3f $nett]
  }
  puts $_(m)
  puts "Average:"
  lset _(m) 0 [format %.6f [expr {[lindex $_(m) 0] / $tcnt}]]
  lset _(m) 2 [expr {[lindex $_(m) 2] / $tcnt}]
  if {[llength $_(m)] > 6} {
    lset _(m) 6 [format %.3f [expr {[lindex $_(m) 6] / $tcnt}]]
    lset _(m) 4 [format %.0f [expr {[lindex $_(m) 2] / [lindex $_(m) 6] * 1000}]]
  }
  puts $_(m)
  puts "Min:"
  puts [lindex $_(itm) $mini]
  puts "Max:"
  puts [lindex $_(itm) $maxi]
  puts [string repeat ** 40]
  puts ""
}

proc _test_run {args} {
  upvar _ _
  # parse args:
  set _(out-result) 1
  if {[lindex $args 0] eq "-no-result"} {
    set _(out-result) 0
    set args [lrange $args 1 end]
  }
  if {[llength $args] < 2 || [llength $args] > 3} {
    return -code error "wrong # args: should be \"[lindex [info level [info level]] 0] ?-no-result? reptime lst ?outcmd?\""
  }
  set outcmd {puts $_(r)}
  set args [lassign $args reptime lst]
  if {[llength $args]} {
    set outcmd [lindex $args 0]
  }
  # avoid output if only once:
  if {[lindex $reptime 0] <= 1 || ([llength $reptime] > 1 && [lindex $reptime 1] == 1)} {
    set _(out-result) 0
  }
  array set _ [list itm {} reptime $reptime starttime [clock milliseconds]]

  # process measurement:
  foreach _(c) [_test_get_commands $lst] {
    puts "% [regsub -all {\n[ \t]*} $_(c) {; }]"
    if {[regexp {^\s*\#} $_(c)]} continue
    if {[regexp {^\s*(?:setup|cleanup)\s+} $_(c)]} {
      puts [if 1 [lindex $_(c) 1]]
      continue
    }
    # if output result (and not once):
    if {$_(out-result)} {
      set _(r) [if 1 $_(c)]
      if {$outcmd ne {}} $outcmd
      if {[llength $_(reptime)] > 1} { # decrement max-count
        lset _(reptime) 1 [expr {[lindex $_(reptime) 1] - 1}]
      }
    }
    puts [set _(m) [timerate $_(c) {*}$_(reptime)]]
    lappend _(itm) $_(m)
    puts ""
  }
  _test_out_total
}

}; # end of namespace ::tclTestPerf

Added tests-perf/timer-event.perf.tcl.












































































































































































































































































































































































>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
#!/usr/bin/tclsh

# ------------------------------------------------------------------------
#
# timer-event.perf.tcl --
# 
#  This file provides performance tests for comparison of tcl-speed
#  of timer events (event-driven tcl-handling).
#
# ------------------------------------------------------------------------
# 
# Copyright (c) 2014 Serg G. Brester (aka sebres)
# 
# See the file "license.terms" for information on usage and redistribution
# of this file.
# 


if {![namespace exists ::tclTestPerf]} {
  source [file join [file dirname [info script]] test-performance.tcl]
}


namespace eval ::tclTestPerf-Timer-Event {

namespace path {::tclTestPerf}

proc test-queue {{reptime {1000 10000}}} {

  set howmuch [lindex $reptime 1]

  # because of extremely short measurement times by tests below, wait a little bit (warming-up),
  # to minimize influence of the time-gradation (just for better dispersion resp. result-comparison)
  timerate {after 0} 156

  puts "*** up to $howmuch events ***"
  # single iteration by update, so using -no-result (measure only):
  _test_run -no-result $reptime [string map [list \{*\}\$reptime $reptime \$howmuch $howmuch \\# \#] {
    # generate up to $howmuch idle-events:
    {after idle {set foo bar}}
    # update / after idle:
    {update; if {![llength [after info]]} break}
    
    # generate up to $howmuch idle-events:
    {after idle {set foo bar}}
    # update idletasks / after idle:
    {update idletasks; if {![llength [after info]]} break}

    # generate up to $howmuch immediate events:
    {after 0 {set foo bar}}
    # update / after 0:
    {update; if {![llength [after info]]} break}
    
    # generate up to $howmuch 1-ms events:
    {after 1 {set foo bar}}
    setup {after 1}
    # update / after 1:
    {update; if {![llength [after info]]} break}

    # generate up to $howmuch immediate events (+ 1 event of the second generation):
    {after 0 {after 0 {}}}
    # update / after 0 (double generation):
    {update; if {![llength [after info]]} break}

    # cancel forwards "after idle" / $howmuch idle-events in queue:
    setup {set i 0; timerate {set ev([incr i]) [after idle {set foo bar}]} {*}$reptime}
    setup {set le $i; set i 0; list 1 .. $le; # cancel up to $howmuch events}
    {after cancel $ev([incr i]); if {$i >= $le} break}
    cleanup {update; unset -nocomplain ev}
    # cancel backwards "after idle" / $howmuch idle-events in queue:
    setup {set i 0; timerate {set ev([incr i]) [after idle {set foo bar}]} {*}$reptime}
    setup {set le $i; incr i; list $le .. 1; # cancel up to $howmuch events}
    {after cancel $ev([incr i -1]); if {$i <= 1} break}
    cleanup {update; unset -nocomplain ev}

    # cancel forwards "after 0" / $howmuch timer-events in queue:
    setup {set i 0; timerate {set ev([incr i]) [after 0 {set foo bar}]} {*}$reptime}
    setup {set le $i; set i 0; list 1 .. $le; # cancel up to $howmuch events}
    {after cancel $ev([incr i]); if {$i >= $howmuch} break}
    cleanup {update; unset -nocomplain ev}
    # cancel backwards "after 0" / $howmuch timer-events in queue:
    setup {set i 0; timerate {set ev([incr i]) [after 0 {set foo bar}]} {*}$reptime}
    setup {set le $i; incr i; list $le .. 1; # cancel up to $howmuch events}
    {after cancel $ev([incr i -1]); if {$i <= 1} break}
    cleanup {update; unset -nocomplain ev}
    
    # end $howmuch events.
    cleanup {if [llength [after info]] {error "unexpected: [llength [after info]] events are still there."}}
  }]
}

proc test-access {{reptime {1000 5000}}} {
  set howmuch [lindex $reptime 1]

  _test_run $reptime [string map [list \{*\}\$reptime $reptime \$howmuch $howmuch] {
    # event random access: after idle + after info (by $howmuch events)
    setup {set i -1; timerate {set ev([incr i]) [after idle {}]} {*}$reptime}
    {after info $ev([expr {int(rand()*$i)}])}
    cleanup {update; unset -nocomplain ev}
    # event random access: after 0 + after info (by $howmuch events)
    setup {set i -1; timerate {set ev([incr i]) [after 0 {}]} {*}$reptime}
    {after info $ev([expr {int(rand()*$i)}])}
    cleanup {update; unset -nocomplain ev}

    # end $howmuch events.
    cleanup {if [llength [after info]] {error "unexpected: [llength [after info]] events are still there."}}
  }]
}

proc test-exec {{reptime 1000}} {
  _test_run $reptime {
    # after idle + after cancel
    {after cancel [after idle {set foo bar}]}
    # after 0 + after cancel
    {after cancel [after 0 {set foo bar}]}
    # after idle + update idletasks
    {after idle {set foo bar}; update idletasks}
    # after idle + update
    {after idle {set foo bar}; update}
    # immediate: after 0 + update
    {after 0 {set foo bar}; update}
    # delayed: after 1 + update
    {after 1 {set foo bar}; update}
    # empty update:
    {update}
    # empty update idle tasks:
    {update idletasks}

    # simple shortest sleep:
    {after 0}
  }
}

proc test-nrt-capability {{reptime 1000}} {
  _test_run $reptime {
    # comparison values:
    {after 0 {set a 5}; update}
    {after 0 {set a 5}; vwait a}

    # conditional vwait with very brief wait-time:
    {after 1 {set a timeout}; vwait a; expr {$::a ne "timeout" ? 1 : "0[unset ::a]"}}
    {after 0 {set a timeout}; vwait a; expr {$::a ne "timeout" ? 1 : "0[unset ::a]"}}
  }
}

proc test-long {{reptime 1000}} {
  _test_run $reptime {
    # in-between important event by amount of idle events:
    {time {after idle {after 30}} 10; after 1 {set important 1}; vwait important;}
    cleanup {foreach i [after info] {after cancel $i}}
    # in-between important event (of new generation) by amount of idle events:
    {time {after idle {after 30}} 10; after 1 {after 0 {set important 1}}; vwait important;} 
    cleanup {foreach i [after info] {after cancel $i}}
  }
}

proc test {{reptime 1000}} {
  test-exec $reptime
  foreach howmuch {5000 50000} {
    test-access [list $reptime $howmuch]
  }
  test-nrt-capability $reptime
  test-long $reptime

  puts ""
  foreach howmuch { 10000 20000 40000 60000 } {
    test-queue [list $reptime $howmuch]
  }

  puts \n**OK**
}

}; # end of ::tclTestPerf-Timer-Event

# ------------------------------------------------------------------------

# if calling direct:
if {[info exists ::argv0] && [file tail $::argv0] eq [file tail [info script]]} {
  array set in {-time 500}
  array set in $argv
  ::tclTestPerf-Timer-Event::test $in(-time)
}

Changes to tests/cmdMZ.test.

337
338
339
340
341
342
343
































































344
345
346
347
348
349
350
351
352
353
354
355
test cmdMZ-5.7 {Tcl_TimeObjCmd: errors generate right trace} {
    list [catch {time {error foo}} msg] $msg $::errorInfo
} {1 foo {foo
    while executing
"error foo"
    invoked from within
"time {error foo}"}}

































































# The tests for Tcl_WhileObjCmd are in while.test
 
# cleanup
cleanupTests
}
namespace delete ::tcl::test::cmdMZ
return

# Local Variables:
# mode: tcl
# End:






>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>












337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
test cmdMZ-5.7 {Tcl_TimeObjCmd: errors generate right trace} {
    list [catch {time {error foo}} msg] $msg $::errorInfo
} {1 foo {foo
    while executing
"error foo"
    invoked from within
"time {error foo}"}}

test cmdMZ-6.1 {Tcl_TimeRateObjCmd: basic format of command} {
    list [catch {timerate} msg] $msg
} {1 {wrong # args: should be "timerate ?-direct? ?-calibrate? ?-overhead double? command ?time ?max-count??"}}
test cmdMZ-6.2.1 {Tcl_TimeRateObjCmd: basic format of command} {
    list [catch {timerate a b c d} msg] $msg
} {1 {wrong # args: should be "timerate ?-direct? ?-calibrate? ?-overhead double? command ?time ?max-count??"}}
test cmdMZ-6.2.2 {Tcl_TimeRateObjCmd: basic format of command} {
    list [catch {timerate a b c} msg] $msg
} {1 {expected integer but got "b"}}
test cmdMZ-6.2.3 {Tcl_TimeRateObjCmd: basic format of command} {
    list [catch {timerate a b} msg] $msg
} {1 {expected integer but got "b"}}
test cmdMZ-6.3 {Tcl_TimeRateObjCmd: basic format of command} {
    list [catch {timerate -overhead b {} a b} msg] $msg
} {1 {expected floating-point number but got "b"}}
test cmdMZ-6.4 {Tcl_TimeRateObjCmd: compile of script happens even with negative iteration counts} {
    list [catch {timerate "foreach a {c d e} \{" -12456} msg] $msg
} {1 {missing close-brace}}
test cmdMZ-6.5 {Tcl_TimeRateObjCmd: result format and one iteration} {
    regexp {^\d+.\d+ \ws/# 1 # \d+ #/sec \d+.\d+ nett-ms$} [timerate {} 0]
} 1
test cmdMZ-6.6 {Tcl_TimeRateObjCmd: slower commands take longer, but it remains almost the same time of measument} {
    set m1 [timerate {after 0} 20]
    set m2 [timerate {after 1} 20]
    list \
	[expr {[lindex $m1 0] < [lindex $m2 0]}] \
	[expr {[lindex $m1 0] < 100}] \
	[expr {[lindex $m2 0] >= 500}] \
	[expr {[lindex $m1 2] > 1000}] \
	[expr {[lindex $m2 2] <= 50}] \
	[expr {[lindex $m1 4] > 10000}] \
	[expr {[lindex $m2 4] < 10000}] \
	[expr {[lindex $m1 6] > 10 && [lindex $m1 6] < 50}] \
	[expr {[lindex $m2 6] > 10 && [lindex $m2 6] < 50}]
} [lrepeat 9 1]
test cmdMZ-6.7 {Tcl_TimeRateObjCmd: errors generate right trace} {
    list [catch {timerate {error foo} 1} msg] $msg $::errorInfo
} {1 foo {foo
    while executing
"error foo"
    invoked from within
"timerate {error foo} 1"}}
test cmdMZ-6.8 {Tcl_TimeRateObjCmd: allow (conditional) break from timerate} {
    set m1 [timerate {break}]
    list \
	[expr {[lindex $m1 0] < 1000}] \
	[expr {[lindex $m1 2] == 1}] \
	[expr {[lindex $m1 4] > 1000}] \
	[expr {[lindex $m1 6] < 10}]
} {1 1 1 1}
test cmdMZ-6.9 {Tcl_TimeRateObjCmd: max count of iterations} {
    set m1 [timerate {} 1000 5];	# max-count wins
    set m2 [timerate {after 20} 1 5];	# max-time wins
    list [lindex $m1 2] [lindex $m2 2]
} {5 1}
test cmdMZ-6.10 {Tcl_TimeRateObjCmd: huge overhead cause 0us result} {
    set m1 [timerate -overhead 1e6 {after 10} 100 1]
    list \
	[expr {[lindex $m1 0] == 0.0}] \
	[expr {[lindex $m1 2] == 1}] \
	[expr {[lindex $m1 4] == 1000000}] \
	[expr {[lindex $m1 6] <= 0.001}]
} {1 1 1 1}

# The tests for Tcl_WhileObjCmd are in while.test
 
# cleanup
cleanupTests
}
namespace delete ::tcl::test::cmdMZ
return

# Local Variables:
# mode: tcl
# End:

Changes to tools/tcltk-man2html-utils.tcl.

145
146
147
148
149
150
151

152
153
154
155
156
157
158
	    {\%}	{} \
	    "\\\n"	"\n" \
	    {\(+-}	"&#177;" \
	    {\(co}	"&copy;" \
	    {\(em}	"&#8212;" \
	    {\(en}	"&#8211;" \
	    {\(fm}	"&#8242;" \

	    {\(mu}	"&#215;" \
	    {\(mi}	"&#8722;" \
	    {\(->}	"<font size=\"+1\">&#8594;</font>" \
	    {\fP}	{\fR} \
	    {\.}	. \
	    {\(bu}	"&#8226;" \
	    ]






>







145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
	    {\%}	{} \
	    "\\\n"	"\n" \
	    {\(+-}	"&#177;" \
	    {\(co}	"&copy;" \
	    {\(em}	"&#8212;" \
	    {\(en}	"&#8211;" \
	    {\(fm}	"&#8242;" \
	    {\(mc}	"&#181;" \
	    {\(mu}	"&#215;" \
	    {\(mi}	"&#8722;" \
	    {\(->}	"<font size=\"+1\">&#8594;</font>" \
	    {\fP}	{\fR} \
	    {\.}	. \
	    {\(bu}	"&#8226;" \
	    ]

Changes to unix/tclUnixTime.c.

55
56
57
58
59
60
61


























62
63
64
65
66
67
68
...
187
188
189
190
191
192
193













































194
195
196
197
198
199
200
{
    return time(NULL);
}
 
/*
 *----------------------------------------------------------------------
 *


























 * TclpGetClicks --
 *
 *	This procedure returns a value that represents the highest resolution
 *	clock available on the system. There are no garantees on what the
 *	resolution will be. In Tcl we will call this value a "click". The
 *	start time is also system dependant.
 *
................................................................................
#else
#error Wide high-resolution clicks not implemented on this platform
#endif
    }

    return nsec;
}













































#endif /* TCL_WIDE_CLICKS */
 
/*
 *----------------------------------------------------------------------
 *
 * Tcl_GetTime --
 *






>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>







 







>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>







55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
...
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
{
    return time(NULL);
}
 
/*
 *----------------------------------------------------------------------
 *
 * TclpGetMicroseconds --
 *
 *	This procedure returns the number of microseconds from the epoch.
 *	On most Unix systems the epoch is Midnight Jan 1, 1970 GMT.
 *
 * Results:
 *	Number of microseconds from the epoch.
 *
 * Side effects:
 *	None.
 *
 *----------------------------------------------------------------------
 */

Tcl_WideInt
TclpGetMicroseconds(void)
{
    Tcl_Time time;

    tclGetTimeProcPtr(&time, tclTimeClientData);
    return ((Tcl_WideInt)time.sec)*1000000 + time.usec;
}
 
/*
 *----------------------------------------------------------------------
 *
 * TclpGetClicks --
 *
 *	This procedure returns a value that represents the highest resolution
 *	clock available on the system. There are no garantees on what the
 *	resolution will be. In Tcl we will call this value a "click". The
 *	start time is also system dependant.
 *
................................................................................
#else
#error Wide high-resolution clicks not implemented on this platform
#endif
    }

    return nsec;
}
 
/*
 *----------------------------------------------------------------------
 *
 * TclpWideClickInMicrosec --
 *
 *	This procedure return scale to convert click values from the 
 *	TclpGetWideClicks native resolution to microsecond resolution
 *	and back.
 *
 * Results:
 * 	1 click in microseconds as double.
 *
 * Side effects:
 *	None.
 *
 *----------------------------------------------------------------------
 */

double
TclpWideClickInMicrosec(void)
{
    if (tclGetTimeProcPtr != NativeGetTime) {
	return 1.0;
    } else {
#ifdef MAC_OSX_TCL
	static int initialized = 0;
	static double scale = 0.0;

	if (initialized) {
	    return scale;
	} else {
	    mach_timebase_info_data_t tb;

	    mach_timebase_info(&tb);
	    /* value of tb.numer / tb.denom = 1 click in nanoseconds */
	    scale = ((double)tb.numer) / tb.denom / 1000;
	    initialized = 1;
	    return scale;
	}
#else
#error Wide high-resolution clicks not implemented on this platform
#endif
    }
}
#endif /* TCL_WIDE_CLICKS */
 
/*
 *----------------------------------------------------------------------
 *
 * Tcl_GetTime --
 *

Changes to win/tclWinTime.c.

24
25
26
27
28
29
30

31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53


54
55
56
57
58
59
60
..
63
64
65
66
67
68
69

70
71
72
73
74
75
76
77

78
79
80
81
82

83
84
85
86
87











88
89
90
91
92
93
94
..
95
96
97
98
99
100
101

102
103
104
105
106
107
108
...
126
127
128
129
130
131
132








133
134
135
136

137
138
139
140
141
142
143
...
154
155
156
157
158
159
160








161
162
163
164
165
166
167
168
169

170

























































































































171
172
173
174
175
176
177
178
179
180
...
195
196
197
198
199
200
201









202

203
204
205
206
207
208
209
...
228
229
230
231
232
233
234
235
236
237
238
239
240
241

242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257






258
259


260
261
262
263
264
265
266
267
268
269
270




271
272
273
274
275
276
277
...
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428

429
430
431
432
433
434
435





436
437
438


439





















440







441
442
443


444
445
446

447
448
449
450
451
452
453
...
459
460
461
462
463
464
465


466
467
468
469
470
471
472
...
518
519
520
521
522
523
524


525
526
527
528
529
530
531
...
577
578
579
580
581
582
583

584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601








602
603



604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
...
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667

668
669
670


671







672
673
674
675
676





677
678

















679
680












681
682















683
684
685
686
687
688
689
typedef struct {
    CRITICAL_SECTION cs;	/* Mutex guarding this structure. */
    int initialized;		/* Flag == 1 if this structure is
				 * initialized. */
    int perfCounterAvailable;	/* Flag == 1 if the hardware has a performance
				 * counter. */

    HANDLE calibrationThread;	/* Handle to the thread that keeps the virtual
				 * clock calibrated. */
    HANDLE readyEvent;		/* System event used to trigger the requesting
				 * thread when the clock calibration procedure
				 * is initialized for the first time. */
    HANDLE exitEvent; 		/* Event to signal out of an exit handler to
				 * tell the calibration loop to terminate. */
    LARGE_INTEGER nominalFreq;	/* Nominal frequency of the system performance
				 * counter, that is, the value returned from
				 * QueryPerformanceFrequency. */

    /*
     * The following values are used for calculating virtual time. Virtual
     * time is always equal to:
     *    lastFileTime + (current perf counter - lastCounter)
     *				* 10000000 / curCounterFreq
     * and lastFileTime and lastCounter are updated any time that virtual time
     * is returned to a caller.
     */

    ULARGE_INTEGER fileTimeLastCall;
    LARGE_INTEGER perfCounterLastCall;
    LARGE_INTEGER curCounterFreq;



    /*
     * Data used in developing the estimate of performance counter frequency
     */

    Tcl_WideUInt fileTimeSample[SAMPLES];
				/* Last 64 samples of system time. */
................................................................................
    int sampleNo;		/* Current sample number. */
} TimeInfo;

static TimeInfo timeInfo = {
    { NULL, 0, 0, NULL, NULL, 0 },
    0,
    0,

    (HANDLE) NULL,
    (HANDLE) NULL,
    (HANDLE) NULL,
#ifdef HAVE_CAST_TO_UNION
    (LARGE_INTEGER) (Tcl_WideInt) 0,
    (ULARGE_INTEGER) (DWORDLONG) 0,
    (LARGE_INTEGER) (Tcl_WideInt) 0,
    (LARGE_INTEGER) (Tcl_WideInt) 0,

#else
    0,
    0,
    0,
    0,

#endif
    { 0 },
    { 0 },
    0
};












/*
 * Declarations for functions defined later in this file.
 */

static void		StopCalibration(ClientData clientData);
static DWORD WINAPI	CalibrationThread(LPVOID arg);
................................................................................
static void 		UpdateTimeEachSecond(void);
static void		ResetCounterSamples(Tcl_WideUInt fileTime,
			    Tcl_WideInt perfCounter, Tcl_WideInt perfFreq);
static Tcl_WideInt	AccumulateSample(Tcl_WideInt perfCounter,
			    Tcl_WideUInt fileTime);
static void		NativeScaleTime(Tcl_Time* timebuf,
			    ClientData clientData);

static void		NativeGetTime(Tcl_Time* timebuf,
			    ClientData clientData);

/*
 * TIP #233 (Virtualized Time): Data for the time hooks, if any.
 */

................................................................................
 *
 *----------------------------------------------------------------------
 */

Tcl_WideUInt
TclpGetSeconds(void)
{








    Tcl_Time t;

    tclGetTimeProcPtr(&t, tclTimeClientData);	/* Tcl_GetTime inlined. */
    return t.sec;

}
 
/*
 *----------------------------------------------------------------------
 *
 * TclpGetClicks --
 *
................................................................................
 *
 *----------------------------------------------------------------------
 */

Tcl_WideUInt
TclpGetClicks(void)
{








    /*
     * Use the Tcl_GetTime abstraction to get the time in microseconds, as
     * nearly as we can, and return it.
     */

    Tcl_Time now;		/* Current Tcl time */
    unsigned long retval;	/* Value to return */

    tclGetTimeProcPtr(&now, tclTimeClientData);	/* Tcl_GetTime inlined */



























































































































    retval = ((Tcl_WideUInt) now.sec * 1000000) + now.usec;
    return retval;

}
 
/*
 *----------------------------------------------------------------------
 *
 * Tcl_GetTime --
 *
................................................................................
 *----------------------------------------------------------------------
 */

void
Tcl_GetTime(
    Tcl_Time *timePtr)		/* Location to store time information. */
{









    tclGetTimeProcPtr(timePtr, tclTimeClientData);

}
 
/*
 *----------------------------------------------------------------------
 *
 * NativeScaleTime --
 *
................................................................................
     * Native scale is 1:1. Nothing is done.
     */
}
 
/*
 *----------------------------------------------------------------------
 *
 * NativeGetTime --
 *
 *	TIP #233: Gets the current system time in seconds and microseconds
 *	since the beginning of the epoch: 00:00 UCT, January 1, 1970.
 *
 * Results:
 *	Returns the current time in timePtr.

 *
 * Side effects:
 *	On the first call, initializes a set of static variables to keep track
 *	of the base value of the performance counter, the corresponding wall
 *	clock (obtained through ftime) and the frequency of the performance
 *	counter. Also spins a thread whose function is to wake up periodically
 *	and monitor these values, adjusting them as necessary to correct for
 *	drift in the performance counter's oscillator.
 *
 *----------------------------------------------------------------------
 */

static void
NativeGetTime(
    Tcl_Time *timePtr,
    ClientData clientData)






{
    struct _timeb t;



    /*
     * Initialize static storage on the first trip through.
     *
     * Note: Outer check for 'initialized' is a performance win since it
     * avoids an extra mutex lock in the common case.
     */

    if (!timeInfo.initialized) {
	TclpInitLock();
	if (!timeInfo.initialized) {




	    timeInfo.perfCounterAvailable =
		    QueryPerformanceFrequency(&timeInfo.nominalFreq);

	    /*
	     * Some hardware abstraction layers use the CPU clock in place of
	     * the real-time clock as a performance counter reference. This
	     * results in:
................................................................................

    if (timeInfo.perfCounterAvailable && timeInfo.curCounterFreq.QuadPart!=0) {
	/*
	 * Query the performance counter and use it to calculate the current
	 * time.
	 */

	ULARGE_INTEGER fileTimeLastCall;
	LARGE_INTEGER perfCounterLastCall, curCounterFreq;
				/* Copy with current data of calibration cycle */

	LARGE_INTEGER curCounter;
				/* Current performance counter. */
	Tcl_WideInt curFileTime;/* Current estimated time, expressed as 100-ns
				 * ticks since the Windows epoch. */
	static LARGE_INTEGER posixEpoch;
				/* Posix epoch expressed as 100-ns ticks since
				 * the windows epoch. */
	Tcl_WideInt usecSincePosixEpoch;
				/* Current microseconds since Posix epoch. */

	posixEpoch.LowPart = 0xD53E8000;
	posixEpoch.HighPart = 0x019DB1DE;

	QueryPerformanceCounter(&curCounter);

	/*
	 * Hold time section locked as short as possible
	 */
	EnterCriticalSection(&timeInfo.cs);

	fileTimeLastCall.QuadPart = timeInfo.fileTimeLastCall.QuadPart;
	perfCounterLastCall.QuadPart = timeInfo.perfCounterLastCall.QuadPart;
	curCounterFreq.QuadPart = timeInfo.curCounterFreq.QuadPart;

	LeaveCriticalSection(&timeInfo.cs);

	/*
	 * If calibration cycle occurred after we get curCounter
	 */
	if (curCounter.QuadPart <= perfCounterLastCall.QuadPart) {
	    usecSincePosixEpoch =
		(fileTimeLastCall.QuadPart - posixEpoch.QuadPart) / 10;
	    timePtr->sec = (long) (usecSincePosixEpoch / 1000000);
	    timePtr->usec = (unsigned long) (usecSincePosixEpoch % 1000000);
	    return;
	}

	/*
	 * If it appears to be more than 1.1 seconds since the last trip
	 * through the calibration loop, the performance counter may have
	 * jumped forward. (See MSDN Knowledge Base article Q274323 for a
	 * description of the hardware problem that makes this test
	 * necessary.) If the counter jumps, we don't want to use it directly.
	 * Instead, we must return system time. Eventually, the calibration
	 * loop should recover.
	 */

	if (curCounter.QuadPart - perfCounterLastCall.QuadPart <
		11 * curCounterFreq.QuadPart / 10
	) {

	    curFileTime = fileTimeLastCall.QuadPart +
		 ((curCounter.QuadPart - perfCounterLastCall.QuadPart)
		    * 10000000 / curCounterFreq.QuadPart);

	    usecSincePosixEpoch = (curFileTime - posixEpoch.QuadPart) / 10;
	    timePtr->sec = (long) (usecSincePosixEpoch / 1000000);
	    timePtr->usec = (unsigned long) (usecSincePosixEpoch % 1000000);





	    return;
	}
    }
























    /*







     * High resolution timer is not available. Just use ftime.
     */



    _ftime(&t);
    timePtr->sec = (long)t.time;
    timePtr->usec = t.millitm * 1000;

}
 
/*
 *----------------------------------------------------------------------
 *
 * StopCalibration --
 *
................................................................................
 *
 * Side effects:
 *	Sets the 'exitEvent' event in the 'timeInfo' structure to ask the
 *	thread in question to exit, and waits for it to do so.
 *
 *----------------------------------------------------------------------
 */



static void
StopCalibration(
    ClientData unused)		/* Client data is unused */
{
    SetEvent(timeInfo.exitEvent);

................................................................................
     */

    GetSystemTimeAsFileTime(&curFileTime);
    QueryPerformanceCounter(&timeInfo.perfCounterLastCall);
    QueryPerformanceFrequency(&timeInfo.curCounterFreq);
    timeInfo.fileTimeLastCall.LowPart = curFileTime.dwLowDateTime;
    timeInfo.fileTimeLastCall.HighPart = curFileTime.dwHighDateTime;



    ResetCounterSamples(timeInfo.fileTimeLastCall.QuadPart,
	    timeInfo.perfCounterLastCall.QuadPart,
	    timeInfo.curCounterFreq.QuadPart);

    /*
     * Wake up the calling thread. When it wakes up, it will release the
................................................................................
static void
UpdateTimeEachSecond(void)
{
    LARGE_INTEGER curPerfCounter;
				/* Current value returned from
				 * QueryPerformanceCounter. */
    FILETIME curSysTime;	/* Current system time. */

    LARGE_INTEGER curFileTime;	/* File time at the time this callback was
				 * scheduled. */
    Tcl_WideInt estFreq;	/* Estimated perf counter frequency. */
    Tcl_WideInt vt0;		/* Tcl time right now. */
    Tcl_WideInt vt1;		/* Tcl time one second from now. */
    Tcl_WideInt tdiff;		/* Difference between system clock and Tcl
				 * time. */
    Tcl_WideInt driftFreq;	/* Frequency needed to drift virtual time into
				 * step over 1 second. */

    /*
     * Sample performance counter and system time.
     */

    QueryPerformanceCounter(&curPerfCounter);
    GetSystemTimeAsFileTime(&curSysTime);
    curFileTime.LowPart = curSysTime.dwLowDateTime;
    curFileTime.HighPart = curSysTime.dwHighDateTime;









    EnterCriticalSection(&timeInfo.cs);




    /*
     * We devide by timeInfo.curCounterFreq.QuadPart in several places. That
     * value should always be positive on a correctly functioning system. But
     * it is good to be defensive about such matters. So if something goes
     * wrong and the value does goes to zero, we clear the
     * timeInfo.perfCounterAvailable in order to cause the calibration thread
     * to shut itself down, then return without additional processing.
     */

    if (timeInfo.curCounterFreq.QuadPart == 0){
	LeaveCriticalSection(&timeInfo.cs);
	timeInfo.perfCounterAvailable = 0;
	return;
    }

    /*
     * Several things may have gone wrong here that have to be checked for.
     *  (1) The performance counter may have jumped.
................................................................................
     *
     * vt1 = 20000000 + curFileTime
     *
     * The frequency that we need to use to drift the counter back into place
     * is estFreq * 20000000 / (vt1 - vt0)
     */

    vt0 = 10000000 * (curPerfCounter.QuadPart
		- timeInfo.perfCounterLastCall.QuadPart)
	    / timeInfo.curCounterFreq.QuadPart
	    + timeInfo.fileTimeLastCall.QuadPart;
    vt1 = 20000000 + curFileTime.QuadPart;

    /*
     * If we've gotten more than a second away from system time, then drifting
     * the clock is going to be pretty hopeless. Just let it jump. Otherwise,
     * compute the drift frequency and fill in everything.
     */

    tdiff = vt0 - curFileTime.QuadPart;
    if (tdiff > 10000000 || tdiff < -10000000) {

	timeInfo.fileTimeLastCall.QuadPart = curFileTime.QuadPart;
	timeInfo.curCounterFreq.QuadPart = estFreq;
    } else {


	driftFreq = estFreq * 20000000 / (vt1 - vt0);








	if (driftFreq > 1003*estFreq/1000) {
	    driftFreq = 1003*estFreq/1000;
	} else if (driftFreq < 997*estFreq/1000) {
	    driftFreq = 997*estFreq/1000;





	}


















	timeInfo.fileTimeLastCall.QuadPart = vt0;
	timeInfo.curCounterFreq.QuadPart = driftFreq;












    }
















    timeInfo.perfCounterLastCall.QuadPart = curPerfCounter.QuadPart;

    LeaveCriticalSection(&timeInfo.cs);
}
 
/*
 *----------------------------------------------------------------------






>










<












>
>







 







>








>

|
|
|
|
>





>
>
>
>
>
>
>
>
>
>
>







 







>







 







>
>
>
>
>
>
>
>
|

|
|
>







 







>
>
>
>
>
>
>
>
|
|
|
|

|
<

|
>
|
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
|
<
|







 







>
>
>
>
>
>
>
>
>
|
>







 







|

|
|


|
>












|
|
|
|
>
>
>
>
>
>
|
<
>
>
|










>
>
>
>







 







|
|




<
<
<
<
<
<
<
<
<
<








|
|
|






|
|
|
<
<
<












|
|

>
|
<
|
|
<
<
<
>
>
>
>
>
|
|
|
>
>
|
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>

>
>
>
>
>
>
>
|
|

>
>
|
|
|
>







 







>
>







 







>
>







 







>











|


<



>
>
>
>
>
>
>
>
|
<
>
>
>











<







 







|
|
|
<
<
<








>
|
<

>
>
|
>
>
>
>
>
>
>

<
<
<
<
>
>
>
>
>
|
|
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
|
|
>
>
>
>
>
>
>
>
>
>
>
>
|
|
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>







24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41

42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
..
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
...
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
...
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
...
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200

201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326

327
328
329
330
331
332
333
334
...
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
...
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429

430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
...
544
545
546
547
548
549
550
551
552
553
554
555
556










557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576



577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593

594
595



596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
...
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
...
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
...
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800

801
802
803
804
805
806
807
808
809
810
811
812

813
814
815
816
817
818
819
820
821
822
823
824
825
826

827
828
829
830
831
832
833
...
858
859
860
861
862
863
864
865
866
867



868
869
870
871
872
873
874
875
876
877

878
879
880
881
882
883
884
885
886
887
888
889




890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
typedef struct {
    CRITICAL_SECTION cs;	/* Mutex guarding this structure. */
    int initialized;		/* Flag == 1 if this structure is
				 * initialized. */
    int perfCounterAvailable;	/* Flag == 1 if the hardware has a performance
				 * counter. */
    DWORD calibrationInterv;	/* Calibration interval in seconds (start 1 sec) */
    HANDLE calibrationThread;	/* Handle to the thread that keeps the virtual
				 * clock calibrated. */
    HANDLE readyEvent;		/* System event used to trigger the requesting
				 * thread when the clock calibration procedure
				 * is initialized for the first time. */
    HANDLE exitEvent; 		/* Event to signal out of an exit handler to
				 * tell the calibration loop to terminate. */
    LARGE_INTEGER nominalFreq;	/* Nominal frequency of the system performance
				 * counter, that is, the value returned from
				 * QueryPerformanceFrequency. */

    /*
     * The following values are used for calculating virtual time. Virtual
     * time is always equal to:
     *    lastFileTime + (current perf counter - lastCounter)
     *				* 10000000 / curCounterFreq
     * and lastFileTime and lastCounter are updated any time that virtual time
     * is returned to a caller.
     */

    ULARGE_INTEGER fileTimeLastCall;
    LARGE_INTEGER perfCounterLastCall;
    LARGE_INTEGER curCounterFreq;
    LARGE_INTEGER posixEpoch;	/* Posix epoch expressed as 100-ns ticks since
				 * the windows epoch. */

    /*
     * Data used in developing the estimate of performance counter frequency
     */

    Tcl_WideUInt fileTimeSample[SAMPLES];
				/* Last 64 samples of system time. */
................................................................................
    int sampleNo;		/* Current sample number. */
} TimeInfo;

static TimeInfo timeInfo = {
    { NULL, 0, 0, NULL, NULL, 0 },
    0,
    0,
    1,
    (HANDLE) NULL,
    (HANDLE) NULL,
    (HANDLE) NULL,
#ifdef HAVE_CAST_TO_UNION
    (LARGE_INTEGER) (Tcl_WideInt) 0,
    (ULARGE_INTEGER) (DWORDLONG) 0,
    (LARGE_INTEGER) (Tcl_WideInt) 0,
    (LARGE_INTEGER) (Tcl_WideInt) 0,
    (LARGE_INTEGER) (Tcl_WideInt) 0,
#else
    {0, 0},
    {0, 0},
    {0, 0},
    {0, 0},
    {0, 0},
#endif
    { 0 },
    { 0 },
    0
};

/*
 * Scale to convert wide click values from the TclpGetWideClicks native
 * resolution to microsecond resolution and back.
 */
static struct {
    int initialized;		/* 1 if initialized, 0 otherwise */
    int perfCounter;		/* 1 if performance counter usable for wide clicks */
    double microsecsScale;	/* Denominator scale between clock / microsecs */
} wideClick = {0, 0.0};


/*
 * Declarations for functions defined later in this file.
 */

static void		StopCalibration(ClientData clientData);
static DWORD WINAPI	CalibrationThread(LPVOID arg);
................................................................................
static void 		UpdateTimeEachSecond(void);
static void		ResetCounterSamples(Tcl_WideUInt fileTime,
			    Tcl_WideInt perfCounter, Tcl_WideInt perfFreq);
static Tcl_WideInt	AccumulateSample(Tcl_WideInt perfCounter,
			    Tcl_WideUInt fileTime);
static void		NativeScaleTime(Tcl_Time* timebuf,
			    ClientData clientData);
static Tcl_WideInt	NativeGetMicroseconds(void);
static void		NativeGetTime(Tcl_Time* timebuf,
			    ClientData clientData);

/*
 * TIP #233 (Virtualized Time): Data for the time hooks, if any.
 */

................................................................................
 *
 *----------------------------------------------------------------------
 */

Tcl_WideUInt
TclpGetSeconds(void)
{
    Tcl_WideInt usecSincePosixEpoch;

    /* Try to use high resolution timer */
    if ( tclGetTimeProcPtr == NativeGetTime
      && (usecSincePosixEpoch = NativeGetMicroseconds())
    ) {
	return usecSincePosixEpoch / 1000000;
    } else {
	Tcl_Time t;

	tclGetTimeProcPtr(&t, tclTimeClientData);	/* Tcl_GetTime inlined. */
	return t.sec;
    }
}
 
/*
 *----------------------------------------------------------------------
 *
 * TclpGetClicks --
 *
................................................................................
 *
 *----------------------------------------------------------------------
 */

Tcl_WideUInt
TclpGetClicks(void)
{
    Tcl_WideInt usecSincePosixEpoch;

    /* Try to use high resolution timer */
    if ( tclGetTimeProcPtr == NativeGetTime
      && (usecSincePosixEpoch = NativeGetMicroseconds())
    ) {
	return (Tcl_WideUInt)usecSincePosixEpoch;
    } else {
	/*
	* Use the Tcl_GetTime abstraction to get the time in microseconds, as
	* nearly as we can, and return it.
	*/

	Tcl_Time now;		/* Current Tcl time */


	tclGetTimeProcPtr(&now, tclTimeClientData);	/* Tcl_GetTime inlined */
	return (Tcl_WideUInt)(now.sec * 1000000) + now.usec;
    }
}
 
/*
 *----------------------------------------------------------------------
 *
 * TclpGetWideClicks --
 *
 *	This procedure returns a WideInt value that represents the highest
 *	resolution clock in microseconds available on the system.
 *
 * Results:
 *	Number of microseconds (from some start time).
 *
 * Side effects:
 *	This should be used for time-delta resp. for measurement purposes
 *	only, because on some platforms can return microseconds from some
 *	start time (not from the epoch).
 *
 *----------------------------------------------------------------------
 */

Tcl_WideInt
TclpGetWideClicks(void)
{
    LARGE_INTEGER curCounter;

    if (!wideClick.initialized) {
	LARGE_INTEGER perfCounterFreq;

	/*
	 * The frequency of the performance counter is fixed at system boot and
	 * is consistent across all processors. Therefore, the frequency need 
	 * only be queried upon application initialization.
	 */
	if (QueryPerformanceFrequency(&perfCounterFreq)) {
	    wideClick.perfCounter = 1;
	    wideClick.microsecsScale = 1000000.0 / perfCounterFreq.QuadPart;
	} else {
	    /* fallback using microseconds */
	    wideClick.perfCounter = 0;
	    wideClick.microsecsScale = 1;
	}
	
	wideClick.initialized = 1;
    }
    if (wideClick.perfCounter) {
	if (QueryPerformanceCounter(&curCounter)) {
	    return (Tcl_WideInt)curCounter.QuadPart;
	}
	/* fallback using microseconds */
	wideClick.perfCounter = 0;
	wideClick.microsecsScale = 1;
	return TclpGetMicroseconds();
    } else {
    	return TclpGetMicroseconds();
    }
}
 
/*
 *----------------------------------------------------------------------
 *
 * TclpWideClickInMicrosec --
 *
 *	This procedure return scale to convert wide click values from the 
 *	TclpGetWideClicks native resolution to microsecond resolution
 *	and back.
 *
 * Results:
 * 	1 click in microseconds as double.
 *
 * Side effects:
 *	None.
 *
 *----------------------------------------------------------------------
 */

double
TclpWideClickInMicrosec(void)
{
    if (!wideClick.initialized) {
    	(void)TclpGetWideClicks();	/* initialize */
    }
    return wideClick.microsecsScale;
}
 
/*
 *----------------------------------------------------------------------
 *
 * TclpGetMicroseconds --
 *
 *	This procedure returns a WideInt value that represents the highest
 *	resolution clock in microseconds available on the system.
 *
 * Results:
 *	Number of microseconds (from the epoch).
 *
 * Side effects:
 *	None.
 *
 *----------------------------------------------------------------------
 */

Tcl_WideInt 
TclpGetMicroseconds(void)
{
    Tcl_WideInt usecSincePosixEpoch;

    /* Try to use high resolution timer */
    if ( tclGetTimeProcPtr == NativeGetTime
      && (usecSincePosixEpoch = NativeGetMicroseconds())
    ) {
	return usecSincePosixEpoch;
    } else {
	/*
	* Use the Tcl_GetTime abstraction to get the time in microseconds, as
	* nearly as we can, and return it.
	*/

	Tcl_Time now;

	tclGetTimeProcPtr(&now, tclTimeClientData);	/* Tcl_GetTime inlined */
	return (((Tcl_WideInt)now.sec) * 1000000) + now.usec;

    }
}
 
/*
 *----------------------------------------------------------------------
 *
 * Tcl_GetTime --
 *
................................................................................
 *----------------------------------------------------------------------
 */

void
Tcl_GetTime(
    Tcl_Time *timePtr)		/* Location to store time information. */
{
    Tcl_WideInt usecSincePosixEpoch;

    /* Try to use high resolution timer */
    if ( tclGetTimeProcPtr == NativeGetTime
      && (usecSincePosixEpoch = NativeGetMicroseconds())
    ) {
	timePtr->sec = (long) (usecSincePosixEpoch / 1000000);
	timePtr->usec = (unsigned long) (usecSincePosixEpoch % 1000000);
    } else {
    	tclGetTimeProcPtr(timePtr, tclTimeClientData);
    }
}
 
/*
 *----------------------------------------------------------------------
 *
 * NativeScaleTime --
 *
................................................................................
     * Native scale is 1:1. Nothing is done.
     */
}
 
/*
 *----------------------------------------------------------------------
 *
 * NativeGetMicroseconds --
 *
 *	Gets the current system time in microseconds since the beginning
 *	of the epoch: 00:00 UCT, January 1, 1970.
 *
 * Results:
 *	Returns the wide integer with number of microseconds from the epoch, or
 *	0 if high resolution timer is not available.
 *
 * Side effects:
 *	On the first call, initializes a set of static variables to keep track
 *	of the base value of the performance counter, the corresponding wall
 *	clock (obtained through ftime) and the frequency of the performance
 *	counter. Also spins a thread whose function is to wake up periodically
 *	and monitor these values, adjusting them as necessary to correct for
 *	drift in the performance counter's oscillator.
 *
 *----------------------------------------------------------------------
 */

static inline Tcl_WideInt
NativeCalc100NsTicks(
    ULONGLONG fileTimeLastCall,
    LONGLONG perfCounterLastCall,
    LONGLONG curCounterFreq,
    LONGLONG curCounter
) {
    return fileTimeLastCall + 
	((curCounter - perfCounterLastCall) * 10000000 / curCounterFreq);
}


static Tcl_WideInt
NativeGetMicroseconds(void)
{
    /*
     * Initialize static storage on the first trip through.
     *
     * Note: Outer check for 'initialized' is a performance win since it
     * avoids an extra mutex lock in the common case.
     */

    if (!timeInfo.initialized) {
	TclpInitLock();
	if (!timeInfo.initialized) {

	    timeInfo.posixEpoch.LowPart = 0xD53E8000;
	    timeInfo.posixEpoch.HighPart = 0x019DB1DE;

	    timeInfo.perfCounterAvailable =
		    QueryPerformanceFrequency(&timeInfo.nominalFreq);

	    /*
	     * Some hardware abstraction layers use the CPU clock in place of
	     * the real-time clock as a performance counter reference. This
	     * results in:
................................................................................

    if (timeInfo.perfCounterAvailable && timeInfo.curCounterFreq.QuadPart!=0) {
	/*
	 * Query the performance counter and use it to calculate the current
	 * time.
	 */

	ULONGLONG fileTimeLastCall;
	LONGLONG perfCounterLastCall, curCounterFreq;
				/* Copy with current data of calibration cycle */

	LARGE_INTEGER curCounter;
				/* Current performance counter. */











	QueryPerformanceCounter(&curCounter);

	/*
	 * Hold time section locked as short as possible
	 */
	EnterCriticalSection(&timeInfo.cs);

	fileTimeLastCall = timeInfo.fileTimeLastCall.QuadPart;
	perfCounterLastCall = timeInfo.perfCounterLastCall.QuadPart;
	curCounterFreq = timeInfo.curCounterFreq.QuadPart;

	LeaveCriticalSection(&timeInfo.cs);

	/*
	 * If calibration cycle occurred after we get curCounter
	 */
	if (curCounter.QuadPart <= perfCounterLastCall) {
	    /* Calibrated file-time is saved from posix in 100-ns ticks */
	    return fileTimeLastCall / 10;



	}

	/*
	 * If it appears to be more than 1.1 seconds since the last trip
	 * through the calibration loop, the performance counter may have
	 * jumped forward. (See MSDN Knowledge Base article Q274323 for a
	 * description of the hardware problem that makes this test
	 * necessary.) If the counter jumps, we don't want to use it directly.
	 * Instead, we must return system time. Eventually, the calibration
	 * loop should recover.
	 */

	if (curCounter.QuadPart - perfCounterLastCall <
		11 * curCounterFreq * timeInfo.calibrationInterv / 10
	) {
	    /* Calibrated file-time is saved from posix in 100-ns ticks */
	    return NativeCalc100NsTicks(fileTimeLastCall,

		perfCounterLastCall, curCounterFreq, curCounter.QuadPart) / 10;
	}



    }

    /*
     * High resolution timer is not available.
     */
    return 0;
}
 
/*
 *----------------------------------------------------------------------
 *
 * NativeGetTime --
 *
 *	TIP #233: Gets the current system time in seconds and microseconds
 *	since the beginning of the epoch: 00:00 UCT, January 1, 1970.
 *
 * Results:
 *	Returns the current time in timePtr.
 *
 * Side effects:
 *	See NativeGetMicroseconds for more information.
 *
 *----------------------------------------------------------------------
 */

static void
NativeGetTime(
    Tcl_Time *timePtr,
    ClientData clientData)
{
    Tcl_WideInt usecSincePosixEpoch;

    /*
     * Try to use high resolution timer.
     */
    if ( (usecSincePosixEpoch = NativeGetMicroseconds()) ) {
	timePtr->sec = (long) (usecSincePosixEpoch / 1000000);
	timePtr->usec = (unsigned long) (usecSincePosixEpoch % 1000000);
    } else {
	/*
	* High resolution timer is not available. Just use ftime.
	*/

	struct _timeb t;

	_ftime(&t);
	timePtr->sec = (long)t.time;
	timePtr->usec = t.millitm * 1000;
    }
}
 
/*
 *----------------------------------------------------------------------
 *
 * StopCalibration --
 *
................................................................................
 *
 * Side effects:
 *	Sets the 'exitEvent' event in the 'timeInfo' structure to ask the
 *	thread in question to exit, and waits for it to do so.
 *
 *----------------------------------------------------------------------
 */

void TclWinResetTimerResolution(void);

static void
StopCalibration(
    ClientData unused)		/* Client data is unused */
{
    SetEvent(timeInfo.exitEvent);

................................................................................
     */

    GetSystemTimeAsFileTime(&curFileTime);
    QueryPerformanceCounter(&timeInfo.perfCounterLastCall);
    QueryPerformanceFrequency(&timeInfo.curCounterFreq);
    timeInfo.fileTimeLastCall.LowPart = curFileTime.dwLowDateTime;
    timeInfo.fileTimeLastCall.HighPart = curFileTime.dwHighDateTime;
    /* Calibrated file-time will be saved from posix in 100-ns ticks */
    timeInfo.fileTimeLastCall.QuadPart -= timeInfo.posixEpoch.QuadPart;

    ResetCounterSamples(timeInfo.fileTimeLastCall.QuadPart,
	    timeInfo.perfCounterLastCall.QuadPart,
	    timeInfo.curCounterFreq.QuadPart);

    /*
     * Wake up the calling thread. When it wakes up, it will release the
................................................................................
static void
UpdateTimeEachSecond(void)
{
    LARGE_INTEGER curPerfCounter;
				/* Current value returned from
				 * QueryPerformanceCounter. */
    FILETIME curSysTime;	/* Current system time. */
    static LARGE_INTEGER lastFileTime; /* File time of the previous calibration */
    LARGE_INTEGER curFileTime;	/* File time at the time this callback was
				 * scheduled. */
    Tcl_WideInt estFreq;	/* Estimated perf counter frequency. */
    Tcl_WideInt vt0;		/* Tcl time right now. */
    Tcl_WideInt vt1;		/* Tcl time one second from now. */
    Tcl_WideInt tdiff;		/* Difference between system clock and Tcl
				 * time. */
    Tcl_WideInt driftFreq;	/* Frequency needed to drift virtual time into
				 * step over 1 second. */

    /*
     * Sample performance counter and system time (from posix epoch).
     */


    GetSystemTimeAsFileTime(&curSysTime);
    curFileTime.LowPart = curSysTime.dwLowDateTime;
    curFileTime.HighPart = curSysTime.dwHighDateTime;
    curFileTime.QuadPart -= timeInfo.posixEpoch.QuadPart;
    /* If calibration still not needed (check for possible time switch) */
    if ( curFileTime.QuadPart > lastFileTime.QuadPart
      && curFileTime.QuadPart < lastFileTime.QuadPart +
      				    (timeInfo.calibrationInterv * 10000000)
    ) {
    	/* again in next one second */
	return;
    }

    QueryPerformanceCounter(&curPerfCounter);
    
    lastFileTime.QuadPart = curFileTime.QuadPart;

    /*
     * We devide by timeInfo.curCounterFreq.QuadPart in several places. That
     * value should always be positive on a correctly functioning system. But
     * it is good to be defensive about such matters. So if something goes
     * wrong and the value does goes to zero, we clear the
     * timeInfo.perfCounterAvailable in order to cause the calibration thread
     * to shut itself down, then return without additional processing.
     */

    if (timeInfo.curCounterFreq.QuadPart == 0){

	timeInfo.perfCounterAvailable = 0;
	return;
    }

    /*
     * Several things may have gone wrong here that have to be checked for.
     *  (1) The performance counter may have jumped.
................................................................................
     *
     * vt1 = 20000000 + curFileTime
     *
     * The frequency that we need to use to drift the counter back into place
     * is estFreq * 20000000 / (vt1 - vt0)
     */

    vt0 = NativeCalc100NsTicks(timeInfo.fileTimeLastCall.QuadPart,
	    timeInfo.perfCounterLastCall.QuadPart, timeInfo.curCounterFreq.QuadPart,
	    curPerfCounter.QuadPart);



    /*
     * If we've gotten more than a second away from system time, then drifting
     * the clock is going to be pretty hopeless. Just let it jump. Otherwise,
     * compute the drift frequency and fill in everything.
     */

    tdiff = vt0 - curFileTime.QuadPart;
    if (tdiff > 10000000 || tdiff < -10000000) {
    	/* jump to current system time, use curent estimated frequency */
    	vt0 = curFileTime.QuadPart;

    } else {
    	/* calculate new frequency and estimate drift to the next second */
	vt1 = 20000000 + curFileTime.QuadPart;
	driftFreq = (estFreq * 20000000 / (vt1 - vt0));
	/* 
	 * Avoid too large drifts (only half of the current difference),
	 * that allows also be more accurate (aspire to the smallest tdiff),
	 * so then we can prolong calibration interval by tdiff < 100000
	 */
	driftFreq = timeInfo.curCounterFreq.QuadPart +
		(driftFreq - timeInfo.curCounterFreq.QuadPart) / 2;





	/* 
	 * Average between estimated, 2 current and 5 drifted frequencies,
	 * (do the soft drifting as possible)
	 */
	estFreq = (estFreq + 2 * timeInfo.curCounterFreq.QuadPart + 5 * driftFreq) / 8;
    }
    
    /* Avoid too large discrepancy from nominal frequency */
    if (estFreq > 1003*timeInfo.nominalFreq.QuadPart/1000) {
	estFreq = 1003*timeInfo.nominalFreq.QuadPart/1000;
	vt0 = curFileTime.QuadPart;
    } else if (estFreq < 997*timeInfo.nominalFreq.QuadPart/1000) {
	estFreq = 997*timeInfo.nominalFreq.QuadPart/1000;
	vt0 = curFileTime.QuadPart;
    } else if (vt0 != curFileTime.QuadPart) {
	/* 
	 * Be sure the clock ticks never backwards (avoid it by negative drifting)
	 * just compare native time (in 100-ns) before and hereafter using 
	 * new calibrated values) and do a small adjustment (short time freeze)
	 */
	LARGE_INTEGER newPerfCounter;
	Tcl_WideInt nt0, nt1;

	QueryPerformanceCounter(&newPerfCounter);
	nt0 = NativeCalc100NsTicks(timeInfo.fileTimeLastCall.QuadPart,
		timeInfo.perfCounterLastCall.QuadPart, timeInfo.curCounterFreq.QuadPart,
		newPerfCounter.QuadPart);
	nt1 = NativeCalc100NsTicks(vt0,
		curPerfCounter.QuadPart, estFreq,
		newPerfCounter.QuadPart);
	if (nt0 > nt1) { /* drifted backwards, try to compensate with new base */
	    /* first adjust with a micro jump (short frozen time is acceptable) */
	    vt0 += nt0 - nt1;
	    /* if drift unavoidable (e. g. we had a time switch), then reset it */
	    vt1 = vt0 - curFileTime.QuadPart;
	    if (vt1 > 10000000 || vt1 < -10000000) {
	    	/* larger jump resp. shift relative new file-time */
	    	vt0 = curFileTime.QuadPart;
	    }
	}
    }

    /* In lock commit new values to timeInfo (hold lock as short as possible) */
    EnterCriticalSection(&timeInfo.cs);

    /* grow calibration interval up to 10 seconds (if still precise enough) */
    if (tdiff < -100000 || tdiff > 100000) {
	/* too long drift - reset calibration interval to 1000 second */
	timeInfo.calibrationInterv = 1;
    } else if (timeInfo.calibrationInterv < 10) {
	timeInfo.calibrationInterv++;
    }

    timeInfo.fileTimeLastCall.QuadPart = vt0;
    timeInfo.curCounterFreq.QuadPart = estFreq;
    timeInfo.perfCounterLastCall.QuadPart = curPerfCounter.QuadPart;

    LeaveCriticalSection(&timeInfo.cs);
}
 
/*
 *----------------------------------------------------------------------