| Author |
Message |
|
bossier330
Newbie
Joined: 28 Jun 2006
Posts: 28
|
 Posted: 15 Mar 2007 05:57:10 pm Post subject: |
|
|
i'm trying to create a number between 2 bounds and store it somewhere using as few b_calls as possible. here's my code. i keep gettin an ERR:INVALID DIM
basically, there is a list or prices that are in word format (.dw). the lower bound is the first number and the upper is the next. the temp location is so i can save the lower bound (to add to the random number times the difference of the upper and lower bounds) without using a b_call.
Code:
ld b,8;
ld ix,prices;
ld de,randBounds;
randomPrices:
push bc;
ld a,(de);
ld l,a;
inc de;
ld a,(de);
ld h,a;
inc de;
ld (temp),hl;
ld b,h;
ld c,l;
ld a,(de);
ld l,a;
inc de;
ld a,(de);
ld h,a;
inc de;
scf
ccf
sbc hl,bc
push de;
B_CALL(_SetXXXXOP2);
B_CALL(_OP2ToOP5);
B_CALL(_Random);
B_CALL(_OP5ToOP2);
B_CALL(_FPMult);
ld d,0;
B_CALL(_Round);
B_CALL(_ConvOP1);
ld hl,(temp);
add hl,de;
ld a,l;
ld (ix),a;
ld a,h;
ld (ix + 1),a;
inc ix;
inc ix;
pop de;
pop bc;
djnz randomPrices;
if anyone can help that would be great.
Last edited by Guest on 16 Mar 2007 10:11:16 am; edited 1 time in total |
|
| Back to top |
|
|
Recursive Acronym
Advanced Member
Joined: 11 Dec 2006
Posts: 499
|
| Posted: 16 Mar 2007 10:19:28 am Post subject: |
|
|
The best way to get random integers is to use Ion and its random number routine. You could also download the ion.zip file, open the source to Ion, and copy the code for the random routine into your program.
Last edited by Guest on 16 Mar 2007 10:21:37 am; edited 1 time in total |
|
| Back to top |
|
|
WikiGuru
ADOS (Attention deficit... Oh! Shiny!)
Elite
Joined: 15 Sep 2005
Posts: 923
|
| Posted: 16 Mar 2007 04:22:38 pm Post subject: |
|
|
You should be able to create a random number generator without any bcall's. Here's a simple method:
- Use hl to point to a place in memory stored in the seed, or start
- load that value into the seed, and then find that place in memory again
- re-load value into seed
- repeat this however many times you want
Code:
;Example code
;a holds random number
;
randInt:
ld hl,seed ;get address in seed
ld b,(hl)
inc hl
ld c,(hl)
ld a,(bc) ;get new seed
ld (seed),a
inc bc
ld (seed+1),a
jr randInt ;add own conditions for exiting loop
;after done, get value pointed to by seed
ld hl,seed
ld a,(hl)
seed:
.dw 0
Note that this code is not complete, you'll have to implement limits yourself and other tid-bits you want to add. Also note that I haven't tested it, so their might be some errors.
Also note that if this is for a program to be run solo, then you'll have to implement some way that will make it so the seed doesn't always start at 0 (or what ever value you set it to).
Last edited by Guest on 16 Mar 2007 04:26:04 pm; edited 1 time in total |
|
| Back to top |
|
|
bossier330
Newbie
Joined: 28 Jun 2006
Posts: 28
|
| Posted: 16 Mar 2007 07:01:38 pm Post subject: |
|
|
thanks, but i'm trying to create random 2byte integers between 2 2byte bounds. like i want to have random*(2byteHi-2byteLow)+2byteLow. i already have a "seed" table and the upper bounds. for example:
Code:
seeds:
.dw 500,1000;
.dw 450,2000;
|
|
| Back to top |
|
|
WikiGuru
ADOS (Attention deficit... Oh! Shiny!)
Elite
Joined: 15 Sep 2005
Posts: 923
|
| Posted: 16 Mar 2007 10:09:55 pm Post subject: |
|
|
Same principle, you just need to load the address after it into the next register:
Code:
;Example code
;bc holds random number
;
randInt:
ld hl,seed ;get address in seed
ld b,(hl)
inc hl
ld c,(hl)
ld a,(bc) ;get new seed
ld (seed),a
inc bc
ld (seed+1),a
jr randInt ;add own conditions for exiting loop
;after done, get value pointed to by seed
ld hl,seed
ld b,(hl)
inc hl
ld c,(hl)
;check if in bounds
ld hl,upper ;upper bounds
ld a,(hl)
cp b
jr c,randInt ;get new number if too big
jr nz,testSmall ;if not equal, no need to check other 8 bits
inc hl
ld a,(hl)
cp c
jr c,randInt
testSmall:
ld hl,lower
ld a,(hl)
cp b
jr z,test2small ;test lower 8 bits if equal
jr nc,randInt ;jump upper 8 bits
ret ;number is okay, return
test2small:
inc hl
ld a,(hl)
cp c
ret z ;return if equal
jr nc,randInt ;check if smaller
ret ;number is okay
lower:
.dw 10
upper:
.dw 500
seed:
.dw 0
again, I haven't tested this code, so there might be some errors. |
|
| Back to top |
|
|
Jim e
Advanced Member
Joined: 05 Sep 2005
Posts: 360
|
| Posted: 16 Mar 2007 11:20:18 pm Post subject: |
|
|
wow...seriously. Not only will that not generate a random number but its possible it may never leave the loop.
@bossier330:
Call ion random twice.
Code:
;in
;BC upper bound
;
;out
; hl = Random number ( 0 <= HL <= BC )
random16:
push bc
inc b
call ionrandom
ld h,a
pop bc
cp b
ld b,0
jr nz,skipsmlr
ld b,c
inc b
skipsmlr:
call ionrandom
ld l,a
ret
Last edited by Guest on 16 Mar 2007 11:47:25 pm; edited 1 time in total |
|
| Back to top |
|
|
bossier330
Newbie
Joined: 28 Jun 2006
Posts: 28
|
| Posted: 17 Mar 2007 07:18:53 am Post subject: |
|
|
| can you pute comments in your code? i'm getting a little confused. |
|
| Back to top |
|
|
bossier330
Newbie
Joined: 28 Jun 2006
Posts: 28
|
| Posted: 17 Mar 2007 07:40:24 am Post subject: |
|
|
| nevermind. thanks for your help. |
|
| Back to top |
|
|
DarkerLine
ceci n'est pas une |
Super Elite (Last Title)
Joined: 04 Nov 2003
Posts: 8328
|
| Posted: 17 Mar 2007 10:29:33 am Post subject: |
|
|
Jim_e: your code is slightly skewed in its probability. The less-than-256 numbers on the high end of the range have the same probability of occurring as any 256 numbers elsewhere in the range. This could have an extreme case of 1 number being 256 times more likely to be chosen than any other.
Granted, I can't think of a better way to do it. You could, of course, round the number up to the nearest 256, then pick numbers until it's in the right range, but that could potentially take any amount of time (though it probably won't). |
|
| Back to top |
|
|
Jim e
Advanced Member
Joined: 05 Sep 2005
Posts: 360
|
| Posted: 17 Mar 2007 11:02:16 am Post subject: |
|
|
DarkerLine wrote: Jim_e: your code is slightly skewed in its probability. The less-than-256 numbers on the high end of the range have the same probability of occurring as any 256 numbers elsewhere in the range. This could have an extreme case of 1 number being 256 times more likely to be chosen than any other. I don't see how.
DarkerLine wrote: Granted, I can't think of a better way to do it. You could, of course, round the number up to the nearest 256, then pick numbers until it's in the right range, but that could potentially take any amount of time (though it probably won't). what? |
|
| Back to top |
|
|
DarkerLine
ceci n'est pas une |
Super Elite (Last Title)
Joined: 04 Nov 2003
Posts: 8328
|
| Posted: 17 Mar 2007 11:30:57 am Post subject: |
|
|
Say you're picking a number between 0 and 256. The routine will get the MSB to be either a 0 or a 1 (with presumably equal probability). If it's a 0, then the routine picks the LSB to be a random number 0-255, yielding a result from 0 to 255. If it's a 1, then the routine picks the LSB to be a random number 0-0, yielding a result from 256 to 256. These two outcomes are equally likely. Making every number 0..255 have a probability of 1/512, and 256 have a probability of 256/512 = 1/2. Of course, this is the most extreme example possible, so it's really not that bad.
Unless I'm misunderstanding something about your code?
Last edited by Guest on 17 Mar 2007 11:32:48 am; edited 1 time in total |
|
| Back to top |
|
|
Jim e
Advanced Member
Joined: 05 Sep 2005
Posts: 360
|
| Posted: 17 Mar 2007 11:46:13 am Post subject: |
|
|
hmmm...yeah that is true. It probably won't be a problem. I suppose another solution could be generating 2 random bytes making a word, and dividing by 2 until it is less than the upper bound.
Code:
rand16:
ld b,0
call ionrandom
ld d,a
ld b,0
call ionrandom
ld e,a
rand16loop:
ld h,d
ld l,e
or a
sbc hl,bc
ret c
srl d
rr e
jp rand16loop
|
|
| Back to top |
|
|
DarkerLine
ceci n'est pas une |
Super Elite (Last Title)
Joined: 04 Nov 2003
Posts: 8328
|
| Posted: 17 Mar 2007 12:00:13 pm Post subject: |
|
|
Even worse, as it chooses high numbers consistently higher than low ones (sorry about picking your random numbers apart like this). For example, suppose you're looking for a random number 0..511. There's only one way to generate a number 0..255, and that is if it comes up from the start. It can't be generated by shifting another number, since any number that could be shifted into the range 0..255 was already in the correct range, and therefore already returned as a number 256..511.
On the other hand, there are 8 ways to generate a number 256..511. It could come up from the start, or from shifting 1 to 7 times. This would make numbers 256..511 8 times as likely to come up (with a probability of 1/288 as opposed to 1/2304). |
|
| Back to top |
|
|
Jim e
Advanced Member
Joined: 05 Sep 2005
Posts: 360
|
| Posted: 17 Mar 2007 12:17:05 pm Post subject: |
|
|
Theres no satisfying you is there. Especially considering that using ion random in a loop has the randomness of reading bytes from memory.
Well the only other thing is actually multiplying a 16bit seed with the 16bit upper, May as well rewrite the ionrandom. |
|
| Back to top |
|
|
bossier330
Newbie
Joined: 28 Jun 2006
Posts: 28
|
| Posted: 17 Mar 2007 12:59:27 pm Post subject: |
|
|
i really want to do this:
load the lower word to de
load the upper word to hl
subtract the two
set hl to OP2
copy OP2 to OP5 (save it )
make a random number 0<n<1 to OP1
mvoe OP5 back to OP2
multiply OP1 by OP2
set OP2 to de
add lower to de
save new number to RAM |
|
| Back to top |
|
|
DarkerLine
ceci n'est pas une |
Super Elite (Last Title)
Joined: 04 Nov 2003
Posts: 8328
|
| Posted: 17 Mar 2007 01:06:30 pm Post subject: |
|
|
Jim e wrote: Theres no satisfying you is there. Especially considering that using ion random in a loop has the randomness of reading bytes from memory. Not if there's interrupts. Not that I'm recommending doing that, I think it's stupid myself.
Quote: Well the only other thing is actually multiplying a 16bit seed with the 16bit upper, May as well rewrite the ionrandom. And be forced to write a 16bit multiplication routine...
Then again, all this depends on what the random number generator should be used for. More information would tell us what aspects of randomness aren't essential.
bossier330 wrote: i really want to do this:
load the lower word to de
load the upper word to hl
subtract the two
set hl to OP2
copy OP2 to OP5 (save it )
make a random number 0<n<1 to OP1
mvoe OP5 back to OP2
multiply OP1 by OP2
set OP2 to de
add lower to de
save new number to RAM
[post="99030"]<{POST_SNAPBACK}>[/post] This would probably be the best more-or-less-random way in terms of size. In terms of speed, it would be the worst - the TI random routine is so slow BASIC can use it as a delay.
A good compromise in my opinion would be to bcall(_Random) then use the random floating point data directly to get a few bytes of randomness. The annoying aspect of that is converting decimal data to hexadecimal data without loss of randomness.
Last edited by Guest on 17 Mar 2007 01:13:12 pm; edited 1 time in total |
|
| Back to top |
|
|
Jim e
Advanced Member
Joined: 05 Sep 2005
Posts: 360
|
| Posted: 17 Mar 2007 02:38:09 pm Post subject: |
|
|
DarkerLine wrote: Not if there's interrupts. Not that I'm recommending doing that, I think it's stupid myself. Chances of an interrupt happening there can be pretty slim, either way interrupts are fairly uniform after a while.
Quote: Then again, all this depends on what the random number generator should be used for. More information would tell us what aspects of randomness aren't essential. I think its excessive, ion random is good enough.
Quote: bossier330 wrote: i really want to do this:
load the lower word to de
load the upper word to hl
subtract the two
set hl to OP2
copy OP2 to OP5 (save it )
make a random number 0<n<1 to OP1
mvoe OP5 back to OP2
multiply OP1 by OP2
set OP2 to de
add lower to de
save new number to RAM
[post="99030"]<{POST_SNAPBACK}>[/post] This would probably be the best more-or-less-random way in terms of size.
Still just use ion random and a small mod routine would be fairly cheap. |
|
| Back to top |
|
|
|
Z80 & 68k Assembly => z80 & ez80 Assembly
