TheBackShed.com

Forum Index : Microcontroller and PC projects : CMM2: V5.05.06RC1

Page 3 of 4

Author

Message

LeoNicolas

Guru

Joined: 07/10/2020
Location: Canada
Posts: 445

Posted: 06:27pm 18 Oct 2020

Copy link to clipboard

Print this post

matherp said Try OPTION KEYBOARD REPEAT 600,25 and see if that fixes the issue for you

Peter

It's much better with this keyboard config.

Thank you

vegipete

Guru

Joined: 29/01/2013
Location: Canada
Posts: 1083

Posted: 11:07pm 19 Oct 2020

Copy link to clipboard

Print this post

Something slightly odd appears to happen when a program ends if the font has been changed.

mode 1,8
cls
box 300,0,400,100
font 5
print "123456"
pause 1000
end

Watch closely. The image on screen scrolls downwards when the test program ends. If you comment out "font 5", nothing moves. The size of the scroll depends on the size of the particular font. If you change back to font 1 before ending, nothing moves.

Visit Vegipete's *Mite Library for cool programs.

thwill

Guru

Joined: 16/09/2019
Location: United Kingdom
Posts: 3848

Posted: 10:48am 20 Oct 2020

Copy link to clipboard

Print this post

May be a bug, may be a cheap USB keyboard, run the following:

Do
If keydown(0) > 0 Then
Print keydown(0);
For i = 1 To keydown(0)
Print keydown(i);
Next i
Print
EndIf
Loop

If I hold down Q,W,E,R then I get the output:

4, 113, 119, 101, 114

But if I hold down R,T,Y then is starts registering 0 keysdown and reports nothing.

T,Y works, but T,Y,U doesn't.

Also isn't there a potential race condition with keydown() if keys are pressed and released quickly enough between calls ?

Best wishes,

Tom

Game*Mite, CMM2 Welcome Tape, Creaky old text adventures

TassyJim

Guru

Joined: 07/08/2011
Location: Australia
Posts: 5913

Posted: 09:46pm 21 Oct 2020

Copy link to clipboard

Print this post

GUI CURSOR colour.
Running the test program, if you end the program while the red cursor is showing, the next run of the program will start with a red cursor.

It looks like, if you just run "GUI CURSOR ON" instead of "GUI CURSOR ON 0,x,y" the last colour is used, even surviving a restart.

Perhaps program start should default to white.

'mouse and cursor test
CLS
MODE 1
GUI CURSOR ON
CONTROLLER MOUSE OPEN 2, leftclick, rightclick

SETTICK 20, myint
DO : LOOP UNTIL INKEY$ <> "" '
GUI CURSOR OFF
CONTROLLER MOUSE CLOSE 2
END

SUB myint
GUI CURSOR MOUSE(x,2),MOUSE(y,2)
END SUB
'
SUB leftclick
STATIC INTEGER n
LOCAL INTEGER x=MOUSE(x,2),y=MOUSE(y,2)
GUI CURSOR OFF
GUI CURSOR ON 1,x,y, RGB(RED)
n=n+1
PRINT @(100,100)"left",n
END SUB
'
SUB rightclick
STATIC INTEGER n=0
LOCAL INTEGER x=MOUSE(x,2),y=MOUSE(y,2)
GUI CURSOR OFF
GUI CURSOR ON 0,x,y
n=n+1
PRINT @(200,100)"Right",n
END SUB

Jim

VK7JH
MMedit MMBasic Help

twofingers
Guru

Joined: 02/06/2014
Location: Germany
Posts: 1133

Posted: 12:55am 22 Oct 2020

Copy link to clipboard

Print this post

Bug report regarding SORT() for integers (MMBASIC v5.05.06RC2)

Hi Peter,

I think there is a less important issue with SORT() for integers (maybe other sorting functions too?). If the number of elements exceeds 2 ^ 16 - 1, the program crashes.
Maybe the manual should mention the limit?

FYI:
My new quicksort CSUB is working correctly now and sorts an array with up to ~260000 (262132=4*&HFFFF-8) elements in 435ms.
So far I haven't been able to figure out why it crashes too over this number (recursion?)

' Usage: Quicksort ArrayOfIntegers%(), NumberOfIntegers
' Sort starts at element 0
' Based on Quicksort @ Rosetta code
' Limit: 262132 elements
'File quicksort.bas written 22-10-2020 01:06:04
CSUB quicksort INTEGER,INTEGER
00000000
'quicksort
4FF0E92D 7800E9D1 2F02B089 F1789000 E9CD0100 DB5B7802 24000879 32FFF117
01C1EB00 F14846A2 46A333FF 46A63808 94059404 6700E9D1 4502E9F0 EB7542B4
DB260107 EB019900 E9DC0CC2 45468900 0109EB77 F112DA11 99003CFF EB019001
F11201CC 468C32FF 8902E871 33FFF143 EB774546 DBF40009 45929801 0103EB7B
F112DA0E F14332FF E9C033FF E9CC8900 F11A4500 F04F0A01 F14B0E01 E7CB0B00
0F00F1BE E9CDD001 E9DDAB04 A9044502 020AEBB4 EB659C00 4620030B 2306E9CD
FF9EF7FF A9069804 00C0EB04 FF98F7FF E8BDB009 BF008FF0
End CSUB

Regards
Michael

vegipete

Guru

Joined: 29/01/2013
Location: Canada
Posts: 1083

Posted: 06:22am 22 Oct 2020

Copy link to clipboard

Print this post

Could the built-in mouse pointers be changed slightly to improve visibility?

The center of the pointer could be black instead of transparent.

The cross hair could have a black perimeter, and the very center pixel could be transparent.

========
Hopefully more than 1 user defined pointer is possible too. For example, two players might each have a mouse, requiring two different pointers.

Visit Vegipete's *Mite Library for cool programs.

matherp
Guru

Joined: 11/12/2012
Location: United Kingdom
Posts: 8592

Posted: 07:08am 22 Oct 2020

Copy link to clipboard

Print this post

Quote the next run of the program will start with a red cursor.

Will fix

Quote I think there is a less important issue with SORT() for integers (maybe other sorting functions too?). If the number of elements exceeds 2 ^ 16 - 1, the program crashes.

Thanks - will fix

Quote Could the built-in mouse pointers be changed slightly to improve visibility?

Pixels can't be black as that is the transparent colour
Edited 2020-10-22 17:09 by matherp

twofingers
Guru

Joined: 02/06/2014
Location: Germany
Posts: 1133

Posted: 02:15pm 22 Oct 2020

Copy link to clipboard

Print this post

twofingers said Hi Peter,

... maybe other sorting functions too?

Hi Peter,

I have to report that the MMBasic SORT for string arrays also has a problem.
This test program occasionally crashes. The larger the array or the MAX_LENGTH of the string, the more frequently it occurs.

'Stringsort for CMM2 (5.05.06RC2/400MHz) - test program

option base 0
option explicit

CONST LF$ =chr$(10)

DIM INTEGER MAX_ARRAY =1000
DIM INTEGER MAX_LENGTH= 30
DIM S$(MAX_ARRAY) LENGTH MAX_LENGTH
DIM temp$
DIM INTEGER N(MAX_ARRAY),i,j,ts

PRINT "Creating"MAX_ARRAY" random strings ..."
FOR j = 1 TO 100

FOR i = 3 TO MAX_ARRAY-1
S$(i) = RNDSTR$(MAX_LENGTH)
NEXT i

TIMER = 0
Sort S$(), N()
ts=TIMER

temp$=S$(0)

FOR i = 0 TO MAX_ARRAY-1
PRINT S$(i),temp$,N(i),j

IF S$(i)<temp$ THEN PRINT "Sorting Error at element " i:END
temp$=S$(i)
NEXT i

PRINT LF$ "Array successfully sorted!" ts "ms for " MAX_ARRAY " elements"
NEXT J

END '*************************************************************

' create a random string
Function Rndstr$(s_len as integer) as string
LOCAL i
FOR i = 1 TO s_len
Rndstr$=Rndstr$+CHR$(65+(RND*25))
NEXT i
End Function

The SORT for floats works also only below &HFFFF elements.

System: CMM2/400MHz/MMBASIC 5.05.06RC2

Kind regards
Michael

matherp
Guru

Joined: 11/12/2012
Location: United Kingdom
Posts: 8592

Posted: 06:09pm 23 Oct 2020

Copy link to clipboard

Print this post

The latest experimental version has fixes for the SORT command and the BOX command (bug when the border and fill colours are the same)

twofingers
Guru

Joined: 02/06/2014
Location: Germany
Posts: 1133

Posted: 12:42am 24 Oct 2020

Copy link to clipboard

Print this post

Thanks, Peter!

My system: MMBasic 5.05.06EXP (23.10.2020)/400 MHz/Option RAM

Sort for integers now works stable for up to 229,374 elements (with=2950ms/without indexarray=1150ms).
For arrays >229,374 (&H37FFE) elements (with/without indexarray) I get occasionally crashes. The sorting time without indexarray increases up to 3316ms (from 1150ms for one element less).

Sort for floats now works stable for up to ~240,000 elements (with=2850ms/without indexarray=1320ms).
With 500,000 elements (without index array), crashes will almost certainly occur.

My Quicksort for integers (without index array!) works perfectly now after I included the RoutineChecks-routine (580000 integers in 1200ms). Thanks for the hint!

BTW. Can you please describe the conditions when the "RoutineChecks"-routine should/must be used? Are there guidelines?

#include "ARMCFunctions.h"
// quicksort from rosetta stone
void quicksort(long long A[], unsigned long long *nElements) {
�unsigned long long len=*nElements; // or uint64_t
�if (len < 2) return;

�unsigned long long pivot = A[len / 2];
�unsigned long long i, j, h;
�unsigned long long temp;

�for (i = 0, j = len - 1; ; i++, j--) {
� �while (A[i] < pivot) i++;
� �while (A[j] > pivot) j--;

� �if (i >= j) break;

� �temp = A[i];
� �A[i] = A[j];
� �A[j] = temp;
�}
�h=len-i;
�RoutineChecks();
�quicksort(A, &i);
�quicksort(A + i, &h);
}

'File quicksort_2.bas written 24-10-2020 02:15:12
CSUB quicksort_2 INTEGER, INTEGER
00000000
'quicksort
4FF0E92D 4500E9D1 2D00B089 BF089001 E9CD2C02 D35C4502 F1140861 F04F32FF
EB000400 F14501C1 46A233FF 380846A3 940446A6 E9D19405 E9F06700 42BD4502
42B4BF08 9901D323 0CC2EB01 8900E9DC BF08454F D20E4546 3CFFF112 01CCEB01
32FFF112 E871468C F1438902 454F33FF 4546BF08 459BD3F4 4592BF08 F112D20E
F14332FF E9C033FF E9CC8900 F11A4500 F04F0A01 F14B0E01 E7CE0B00 0F00F1BE
E9CDD001 E9DDAB04 4B0B4502 010AEBB4 EB65681B 9106020B 47989207 A9049C01
F7FF4620 9804FF9D EB04A906 F7FF00C0 B009FF97 8FF0E8BD 0800033C
End CSUB

I haven't finished testing the string sort yet. It is noticeable that it appears extremely slow for large amounts of data. Much more than expected.

Kind regards
Michael
Edited 2020-10-24 10:58 by twofingers

vegipete

Guru

Joined: 29/01/2013
Location: Canada
Posts: 1083

Posted: 05:21am 24 Oct 2020

Copy link to clipboard

Print this post

matherp said

I said Could the built-in mouse pointers be changed slightly to improve visibility?

Pixels can't be black as that is the transparent colour

Does that mean the 2 built-in cursors are mono-chrome? Only transparent and white pixels where white is changeable with the GUI CURSOR COLOUR command?

Visit Vegipete's *Mite Library for cool programs.

twofingers
Guru

Joined: 02/06/2014
Location: Germany
Posts: 1133

Posted: 02:06am 25 Oct 2020

Copy link to clipboard

Print this post

Hi Peter,

unfortunately I have to inform you that the CMM2 string sort is technically unusable. A minor bug eats the biggest string and sets the first value to zero length.
EDIT: The same bug bothers "float" and "integer".

Creating 10 random strings ...
0>00000000<
1>11111111<
2>22222222<
3>33333333<
4>44444444<
5>55555555<
6>66666666<
7>77777777<
8>88888888<
9>99999999<
Sorting ... (output)
0>< 0
1>00000000< 8
2>11111111< 8
3>22222222< 8
4>33333333< 8
5>44444444< 8
6>55555555< 8
7>66666666< 8
8>77777777< 8
9>88888888< 8

It should more look like this, obviously:

Creating 10 random strings ...
0>00000000<
1>11111111<
2>22222222<
3>33333333<
4>44444444<
5>55555555<
6>66666666<
7>77777777<
8>88888888<
9>99999999<
Sorting ...
0>00000000< 8
1>11111111< 8
2>22222222< 8
3>33333333< 8
4>44444444< 8
5>55555555< 8
6>66666666< 8
7>77777777< 8
8>88888888< 8
9>99999999< 8

My system: CMM2 MMBasic 5.05.06EXP (23.10.2020)/400 MHz/Option RAM
Test code:

'Stringsort for CMM2 (5.05.06RC2/400MHz) - test program

option base 0
option explicit

CONST LF$ � � � =chr$(10)

DIM INTEGER MAX_ARRAY =10
DIM INTEGER MAX_LENGTH= 8
DIM S$(MAX_ARRAY) LENGTH MAX_LENGTH
DIM temp$
DIM INTEGER N(MAX_ARRAY),i,j,ts

�PRINT "Creating"MAX_ARRAY" random strings ..."
�FOR i = 0 TO MAX_ARRAY-1
� �S$(i) = RNDSTR$(MAX_LENGTH)
� �PRINT i">"s$(i)"<":' if i >10 then end
�NEXT i

�PRINT "Sorting ..."

�TIMER = 0
' Sort S$() �', N()' � � � � � � � � � � 1233s/66000 elements/random strings
� � � � � � � � � � � � � � � � � � � �'1430s/66000 elements/sorted strings
�StringSort S$(), MAX_ARRAY,MAX_LENGTH �'870s/66000 elements/random strings
� � � � � � � � � � � � � � � � � � � '0.052s/66000 elements/sorted strings
�ts=TIMER
�temp$=S$(0)

�FOR i = 0 TO MAX_ARRAY-1
� PRINT i">"S$(i)"<"len(s$(i))

� IF S$(i)<temp$ THEN PRINT "Sorting Error at element " i:END
� temp$=S$(i)
�NEXT i

�PRINT LF$ "Array successfully sorted!" ts "ms for " MAX_ARRAY " elements"

END '*************************************************************

' create a random string
Function Rndstr$(s_len as integer) as string
LOCAL integer ii
�'FOR ii = 1 TO s_len:Rndstr$=Rndstr$+CHR$(65+(RND*25)):NEXT ii 'random string
�FOR ii = 1 TO s_len:Rndstr$=Rndstr$+CHR$(48+i):NEXT ii 'Presorted string
End Function

'File StringSort.bas written 25-10-2020 00:37:15
CSUB StringSort STRING, INTEGER, INTEGER
�00000000
�'stringsort
�4FF0E92D B0912300 26014683 27009308 33016813 E9D19305 E9CD2300 E9DD2302
�429F2302 4296BF08 80B9F080 F11646B0 9B0530FF FB033601 F147F808 FB030700
�F81BB000 23009008 A000F890 0C08EB0B 45D14649 4651BF28 F112B2CA F14334FF
�290035FF 80B2F000 46614686 F81E9000 F8112F01 429A3F01 EA54D3D1 F0000005
�F11480A1 F14534FF 429A35FF 9800D0EF 0204F108 F10045CA 445A0304 46CABF38
�0900F04F BF38459C FA5F4290 F0C0F88A F1BA8091 F2400F04 F118808D 46E60301
�930A4601 0300F149 F1B8930B F1490303 089B32FF 7382EA43 0A92EA4F 33012200
�F14A9306 93070300 E9CD2300 E9CD6700 E9CD4508 4665890E 460746D9 AB06E9DD
�68083201 C000F8DE 0300F143 CB04F841 F84E455B BF080B04 D3F14552 46CB9907
�46AC4638 9906008B 890EE9DD EA43008A EBB87391 44100102 910C4494 0103EB69
�890AE9DD 6700E9DD 910D4599 4590BF08 7803D01C 2000F89C 890CE9DD F88C7002
�EA583000 D0110309 F89C7843 70422001 F88C4642 464B3001 BF082B00 D0052A01
�F89C7883 70822002 3002F88C 93082301 2302E9DD BF08429F F4FF4296 4B1BAF47
�4798681B B3739B08 2302E9DD BF082B00 D3222A02 F8DD2300 46588014 27002602
�E73E9308 9800429A AF62F47F F4BF45D1 E75DAF26 F1183801 F81038FF F89C3F01
�F1492000 700239FF 3FFFF1B9 3B01F80C F1B8BF08 D1EE3FFF 4B04E7C8 B011681B
�4FF0E8BD B0114718 8FF0E8BD 0800033C
End CSUB

------------------
About speed:
� � � � � �
The CMM2 Sort (without indexarray) needs 1233s/66000 elements for random strings and
1430s/66000 elements for (pre-)sorted strings.
The CSUB StringSort needs 870s/66000 elements for random strings and 0.052s/66000 elements for (pre-)sorted strings.

Kind regards
Michael
Edited 2020-10-26 00:51 by twofingers

matherp
Guru

Joined: 11/12/2012
Location: United Kingdom
Posts: 8592

Posted: 04:29pm 25 Oct 2020

Copy link to clipboard

Print this post

Quote unfortunately I have to inform you that the CMM2 string sort is technically unusable. A minor bug eats the biggest string and sets the first value to zero length.

Unfortunately you have fallen into the array dimension trap. How many elements are there in an array declared?

OPTION BASE 0
DIM array(10)

twofingers
Guru

Joined: 02/06/2014
Location: Germany
Posts: 1133

Posted: 05:55pm 25 Oct 2020

Copy link to clipboard

Print this post

Thanks Peter. I'm relieved!

matherp said How many elements are there in an array declared?

OPTION BASE 0
DIM array(10)

My best guess? 11.

You are correct, my fault! Sorry.
I know that you are currently under heavy strain!

matherp said The latest experimental �version has fixes for the SORT command ...

Not completely (for MMBasic 5.05.06EXP (23.10.2020))! Crashes still occur with arrays larger than 4 * & Hffff. For the CSUB Integersort and stringsort, the inclusion of the RoutineChecks() helped.

As I asked above:
Can you please describe the conditions when the "RoutineChecks"-routine should/must be used? Are there guidelines?

Many thanks!
Michael

matherp
Guru

Joined: 11/12/2012
Location: United Kingdom
Posts: 8592

Posted: 06:03pm 25 Oct 2020

Copy link to clipboard

Print this post

Can you please describe the conditions when the "RoutineChecks"-routine should/must be used? Are there guidelines?

Every millisecond of processing is a good starting point. For CSUBs that take less than 1 mSec it can be omitted.
It does four things:
Pokes the watchdog if required
Checks for SDcard presence
polls the USB port
Updates sound buffers if any music playing

In all cases it will return immediately if nothing to be done so the overhead is very small

Quote Crashes still occur with arrays larger than 4 * & Hffff

Try the latest posted. Also if you are doing serial output as in your earlier example you are filling the output TX buffer which is then blocking and serial I/O is still happening when you start the next sort - try waiting for output to stop

I've just sorted 600,001 random floats on a 400MHz processor which took �4.57 seconds

dim a(600000)
for i=0 to 600000
a(i)=rnd
next i
timer=0
sort a()
? timer

Edited 2020-10-26 04:15 by matherp

twofingers
Guru

Joined: 02/06/2014
Location: Germany
Posts: 1133

Posted: 06:35pm 25 Oct 2020

Copy link to clipboard

Print this post

matherp said Can you please describe the conditions when the "RoutineChecks"-routine should/must be used? Are there guidelines?

Every millisecond of processing is a good starting point. For CSUBs that take less than 1 mSec it can be omitted.
It does four things:
Pokes the watchdog if required
Checks for SDcard presence
polls the USB port
Updates sound buffers if any music playing

In all cases it will return immediately if nothing to be done so the overhead is very small
Thanks Peter, that helped a bit, makes it clearer. But it doesn't explain why a routine crashes that doesn't use any of the four things you're describing.

matherp said

Quote Crashes still occur with arrays larger than 4 * & Hffff

Try the latest posted.

Will do.

EDIT:
I loaded 2020-10-26_001301_CMM2V1.5.zip (5.05.06RC6) into my CMM2 and ran your little test code. Result:

RUN
Error: System fault - possible causes
- Incorrect use of POKE
- Faulty CSUB
- Reserved words used as variables
- Firmware problem
> RUN
4588.915

> RUN
Error: System fault - possible causes
- Incorrect use of POKE
- Faulty CSUB
- Reserved words used as variables
- Firmware problem
> RUN
Error: System fault - possible causes
- Incorrect use of POKE
- Faulty CSUB
- Reserved words used as variables
- Firmware problem
> RUN
Error: System fault - possible causes
- Incorrect use of POKE
- Faulty CSUB
- Reserved words used as variables
- Firmware problem
> RUN
Error: System fault - possible causes
- Incorrect use of POKE
- Faulty CSUB
- Reserved words used as variables
- Firmware problem
> RUN
4573.616

> RUN
Error: System fault - possible causes
- Incorrect use of POKE
- Faulty CSUB
- Reserved words used as variables
- Firmware problem
> RUN
Error: System fault - possible causes
- Incorrect use of POKE
- Faulty CSUB
- Reserved words used as variables
- Firmware problem
> RUN
4562.95

> RUN
Error: System fault - possible causes
- Incorrect use of POKE
- Faulty CSUB
- Reserved words used as variables
- Firmware problem
> RUN
4495.262

Edited 2020-10-26 06:17 by twofingers

twofingers
Guru

Joined: 02/06/2014
Location: Germany
Posts: 1133

Posted: 09:04pm 25 Oct 2020

Copy link to clipboard

Print this post

Bad news!

This test code for integer sort

dim integer a(600000)
do
for i=0 to 600000
a(i)=rnd*200000
next i
timer=0
sort a()
? timer
loop
end

results in occasional, fairly frequent crashes..
CMM2 MMBasic 5.05.06RC6/400 MHz/Option FLASH
Edited 2020-10-26 08:38 by twofingers

matherp
Guru

Joined: 11/12/2012
Location: United Kingdom
Posts: 8592

Posted: 04:04pm 26 Oct 2020

Copy link to clipboard

Print this post

Quote This test code for integer sort

I'm not quite sure the point of using all the memory for a single array and sorting it however it is easily fixed by moving the routinechecks() call to the inner loop.

I'll be posting a new experimental version later today but I want to get to the bottom of a nasty BLIT issue when both pages are in SDRAM and the move isn't word aligned first
Edited 2020-10-27 02:05 by matherp

twofingers
Guru

Joined: 02/06/2014
Location: Germany
Posts: 1133

Posted: 05:50pm 26 Oct 2020

Copy link to clipboard

Print this post

Thanks, Peter!

Quote ...and sorting it however it is easily fixed by moving the routinechecks() call to the inner loop.

I don't know if that helps (probably not), anyway this is the code for the CSUB StringSort (made by Matherp) that worked for me with no problems and minimal loss of speed. Just for comparing purposes.

#include "ARMCFunctions.h"

// function to sort an array of strings
// StringSort ArrayOfStrings$(), NbrOfStrings, ElementLength
void stringsort(unsigned char *sarray, long long *nstring, long long *maxlength){
unsigned long long i, ii, n=*nstring;
int s=1, k, offset=*maxlength+1;
unsigned char *s1,*s2,*p1,*p2;
unsigned char temp;
while (s){
s=0;
for(i=1;i<n;i++){
s2=i*offset+sarray;
s1=(i-1)*offset+sarray;
ii = *s1 < *s2 ? *s1 : *s2; //get the smaller length
p1 = s1 + 1; p2 = s2 + 1;
k=0; //assume the strings match
while((ii--) && (k==0)) {
if(*p1 > *p2){
k=1; //earlier in the array is bigger
}
if(*p1 < *p2){
k=-1; //later in the array is bigger
}
p1++; p2++;
}
// if up to this point the strings match
// make the decision based on which one is shorter
if(k==0){
if(*s1 > *s2) k=1;
if(*s1 < *s2) k=-1;
}
if (k==1){ // if earlier is bigger swap them round
ii = *s1 > *s2 ? *s1 : *s2; //get the bigger length
ii++;
p1=s1;p2=s2;
while(ii--){
temp=*p1;
*p1=*p2;
*p2=temp;
p1++; p2++;
}
s=1;
}
}
RoutineChecks();
}
}

Michael

matherp
Guru

Joined: 11/12/2012
Location: United Kingdom
Posts: 8592

Posted: 05:53pm 26 Oct 2020

Copy link to clipboard

Print this post

Try the latest code just posted on the performance gains thread - should be fixed in that

Page 3 of 4

Print this page