OpenEuphoria

Re: Out of Memory

• Posted by Pete Eberlein <xseal at HARBORSIDE.COM> Apr 09, 2000
• 849 views

On Sun, 9 Apr 2000 16:12:58 -0700, simulat <simulat at INTERGATE.BC.CA> wrote:

>Thanks Jiri
>
>Yeow - 16 bytes per atom is a lot of overhead! Maybe I'll need to try some
>kind of direct memory access instead of using sequences for this. Darn.

Be prepared to lose a lot of speed peek-poking atom_to_float64s

>But why will the sequence initialize without problems, but then crash later
>as it's being used? As far as I can tell the sequence size should be
staying
>the same with the contents only being modified.

You initialized the sequence to integers which take only 4 bytes apiece plus
sequence overhead:

ps=repeat(repeat(250,1250),1750)

sizeof ps ~= 24 + 1750*(24 + 4*1250) = 8,792,024 bytes

but when you start assigning non-integer values to the sequence, each
integer mutates into a 16-byte atom thus slowing increasing to about:

24 + 1750*(24 + 16*1250) = 35,042,024 bytes like Jiri said


A solution might be to store the return values as fixed-point integers:

        ps[i][j]= floor(
  strange(Degree(Aangle(Field[x], Field[y])), numcolours,
  360) -- evaluates to an atom
      * #10000) -- evaluates back to integer

Then whenever you want to get an element out of ps, you have to remember to
divide by #10000 to get its actual value.  This would have to same memory
requirements as the integer size.

--
Pete Eberlein

>Bye
>Martin
>
>----- Original Message -----
>From: Jiri Babor <J.Babor at GNS.CRI.NZ>
>To: <EUPHORIA at LISTSERV.MUOHIO.EDU>
>Sent: Sunday, April 09, 2000 4:01 PM
>Subject: Re: Out of Memory
>
>
>> Correction, Martin! I wrongly assumed you had a *pair* of coordinates
>> in each case, stupid. In your case you actually need only
>>
>>     ((16 * 1250) + 24) * 1750 = 35,042,000 bytes
>>
>> But the system probably needs to hold *two* copies of the sequence at
>> some stage, not to mention operating system requirements. I hope I am
>> right this time :)  jiri
>>