Quote from: jacobly on November 27, 2011, 06:17:58 pm

Quote from: Freyaday on November 27, 2011, 12:43:57 am

Perhaps a floating-point-math Axiom, with a function to store a float to an axevar?

How can a float fit in a single axe var? Unless you want truncation, but then you would have to specify how you want it truncated.

There is only one kind of truncation. It's where you chop off the decimal portion. Or, to put it another way, round towards 0

By how you want it truncated, I mean what do you want it truncated to.

I'm pretty sure it's like that, just have the second line of hex under/part of the first one.

...
Ok, I was optimizing last night and wrote down a bunch of Q's as I went.

Do the ?/?? conditionals save space in the executable in relation to conventional If loops?  How about speed?

Can they be nested?  like B?(A++5?-1→A),→A works, nvm that

Which is faster: If (commonly true) / If (sparsely true) ?

So it's optimized to put the constants at the end...what if it's like L2+2?  Would a different order influence anything?

1. ? and ?? are just convenience notation, so they are probably the same as If statements.
2. The speed of an If statement depends on what conditionals you use. However, for If / Else statements, the Else is generally faster due to one fewer jump. This is why I generally put the commonly true branch in the else part.
3. L₂+2 and 2+L₂ both optimize to 35388

Yeah, I know why it's slow, and it's because my algorithm is pretty bad Here's my flood-fill code:

Quote from: TI-BASIC

{~1.01,~.01,.99,~1,1,~.99,.01,1.01->L3
DelVar I{A+sub(B->L2
Repeat I=dim(L2
I+1->I
L2(Ans
F+pxl-Test(600fPart(Ans)+2,6iPart(Ans)+2->F
sum(seq(max(L1=L2(I)+L3(J)),J,1,8
If Ans
Then
" "+sub("12345678",Ans,1)+" 
Else
For(J,1,8)
L2(I)+L3(J
If Ans>=0 and iPart(Ans)<C and |E2fPart(Ans)<D and prod(L2-Ans
Ans->L2(1+dim(L2
End
"     
End
Text(600fPart(L2(I))+1,6iPart(L2(I))+2,Ans
End

The stuff between lines 11 and 15 are what's really slowing the thing down. It goes through the eight surrounding pixels and individually tests them to see if they're valid or not, and if so they get appended to the queue (L2). Because it has to do this for every element on the queue that doesn't have any mines surrounding it, blank squares take a while to show up.

I tried to implement the third algorithm on the Wikipedia article, but since there are eight directions in Minesweeper and three classes of elements (blank, numbered, mine), it turned out to be even slower than the one I have.

Replacing the above code with
:A+sub(B
:sum(seq(max(L₁=Ans+L₃(I)),I,1,8
:If Ans
:Then
:Text(6B+1,6A+2," "+sub("12345678",Ans,1)+"
:Else
:Text(6B+1,6A+2,"    
:DelVar I{A+sub(B➔L₂
:Repeat I=dim(L₂
:I+1➔I
:For(J,1,8
:L₂(I)+L₃(J
:If Ans≥0 and iPart(Ans)<C and ᴇ2fPart(Ans)<D
:Then
:If pxl-Test(600fPart(Ans)+2,6iPart(Ans)+2
:Then
:sum(seq(max(L₁=L₂(I)+L₃(J)+L₃(K)),K,1,8
:If Ans
:Then
:" "+sub("12345678",Ans,1)+"
:Else
:L₂(I)+L₃(J➔L₂(1+dim(L₂
:"    
:End
:Text(600fPart(L₂(I)+L₃(J))+1,6iPart(L₂(I)+L₃(J))+2,Ans
:End
:End
:End
:End
:End
appears faster, although it could probably be optimized some more.

Edit to show how good I am at minesweeper
J/k, actually to see the speed difference (I hope).

For the output command, the column is in the range 0-15 and the row is in the range 0-7.
In your code, I ranges from 1 to 8, so everywhere you have I, you need I-1 (unless you want to use {L₁+0} for something else).
For(I,1,8)
rand^16→{I-1+L₁}
End
For(I,1,8)
Output({I-1+L₁},I-1,"∗")
End
Pause 5000

Edit:
The fact that you have 0→I before For(I,1,8) makes me think that you are trying to make I range from 0 to 7. However, For(I,1,8) starts off by storing 1 to I, causing the 0→I statement to be useless. Instead you may want For(I,0,7).

ASM is not good for writing floating-point math programs

If by floating-point you happen to mean binary floating-point, and not that awkward base 10 format, then coincidentally I happen to be doing that right now.

SAXParser in a very round-about way eventually uses a URLConnection, so you might as well create your own and set its timeout.
public RSS(String uri) {
try {
SAXParser parser = SAXParserFactory.newInstance().newSAXParser();
URL url = new URL(uri);
URLConnection conn = url.openConnection();
conn.setConnectTimeout(0);
conn.setReadTimeout(0);
parser.parse(conn.getInputStream(), new RSSHandler());
} catch (SAXException ex) {
ex.printStackTrace();
} catch (MalformedURLException ex) {
ex.printStackTrace();
} catch (IOException ex) {
ex.printStackTrace();
} catch (ParserConfigurationException ex) {
ex.printStackTrace();
}
}


I think I got it!
 ld hl, (SaveIY)
  push hl
  pop IY
^ That code is really important, right?
but what happens if it is clipped off the bottom... then the routine returns at the clipping code
 bit 7, a
  jp nz, Clip
  ld a, (_ty)
  cp 6
  ret nc // early return, above code is not run!
This results it iy pointing to flash, which is why the os goes crazy... because it is unable to set any of its own flags O.o
Edit: my mistake, it actually points inside saveSScreen, which still causes problems

P.S. ld iy,(nnnn) and ld (nnnn),iy are both documented instructions...

Feel free to change the packages back. I rewrote the RSS Parser like 4 times, and each time I moved it to a new package instead of deleting the old code.
Also, the more info page works fine for me. What version of Android are you testing it on?