+- Forum (https://www.boriel.com/forum)
+-- Forum: Compilers and Computer Languages (https://www.boriel.com/forum/forumdisplay.php?fid=12)
+--- Forum: ZX Basic Compiler (https://www.boriel.com/forum/forumdisplay.php?fid=11)
+---- Forum: Bug Reports (https://www.boriel.com/forum/forumdisplay.php?fid=15)
+---- Thread: Odd error (/showthread.php?tid=437)
Odd error - britlion - 02-27-2012
This might be another float bug. I wrote up the distance algorithm na_th_an showed us. It seems to work - but in this test program, it's the floating point values that go wrong...
If you set your emulator to full speed, you see it looking for the biggest error it can find, showing larger errors as it finds them. Eventually it hits distance (250,56) which my function says is > 255 (so returns 255). The value calculated from square rooting in floats is...er 0. So that's an error of 255, apparently. Except the correct answer would be 256.1952 Is the int (256) returning 0 for some reason? On a float?
When I'm happy with this funtion, it will go in the function library as a code snippet.
Code:
FUNCTION fastcall distance (a as ubyte, b as ubyte) as uByte
' returns a fast approximation of SQRT (a^2 + b^2) - the distance formula, generated from taylor series expansion.
asm
POP DE ; return address
; First parameter in A
POP BC ; second parameter -> B
LD C,A ; Put second parameter in C for safekeeping.
LD HL,0
ADD A,B
LD L,A ; Put result in HL
RL H ; Pull in carry bit if necessary.
LD A,B ; get original
CP C ; compare with C
JP NC, distance_Cbigger
LD C,A ; B was smaller, so we replace the value in C with the smaller value in A
distance_Cbigger:
; C was smaller, so we leave it alone.
SRL C ; C=C/2
XOR A
LD B,A
SBC HL,BC ; take half our smallest from HL
SRL C
AND A ; Clear carry flag.
SBC HL,BC
SRL C
SRL C
AND A ; Clear carry flag.
ADD HL,BC
LD A,H
AND A
JP NZ, distance_toobig
LD A,L ; Get result.
EX DE,HL ; return address is in DE.
JP (HL)
distance_toobig:
; If the answer turns out to be > 255, we'll return 255.
LD A,255
EX DE,HL ; return address is in DE.
JP (HL)
END ASM
END FUNCTION
CLS
DIM i,imax as uByte
dim j,jmax as uByte
dim answer,realanswer as uByte
dim errorval,errorvalmax as float
imax=0
jmax=0
errorvalmax=0
for j=1 to 250
for i=1 to 250
answer=distance(i,j)
realanswer=SQR(CAST(float,i)*i+CAST(float,j)*j)
errorval=ABS(CAST(float,realanswer)-answer)
if errorval>errorvalmax then
PRINT i;" ";j;" ";answer;" ";INT realanswer; " ";int errorval
imax=i
jmax=j
errorvalmax=errorval
end if
border j
next i
next jRe: Odd error - LCD - 02-28-2012
If the Answer is defined as uByte, then 256 is 0...
No Error of ZXBC I guess.
Re: Odd error - britlion - 02-29-2012
Float, Jose LCD. Float:
Britlion Wrote:The value calculated from square rooting in floats is...er 0. So that's an error of 255, apparently. Except the correct answer would be 256.1952
The ubyte bit is coming out nicely. I was baffled by the float version of the calculation, here.
I changed how it all worked anyway, so it might not matter. I did wonder if it was pointing to an odd bug, though.
Re: Odd error - boriel - 02-29-2012
britlion Wrote:Float, Jose. Float:Uh:?: I haven't said a word... yet. :roll:
Britlion Wrote:The value calculated from square rooting in floats is...er 0. So that's an error of 255, apparently. Except the correct answer would be 256.1952
Re: Odd error - britlion - 02-29-2012
Ah. Fundamental problem. I can't read. Apart from that, carry on as normal.
Re: Odd error - britlion - 04-06-2012
No comment if this is a bug? Did it go away?