The Last Word 3.21 Released

[Jacques]

So there will be (at least) 2 reasons to celebrate your 40th birthday

[flashjazzcat]

So there will be (at least) 2 reasons to celebrate your 40th birthday

Hopefully completion of the GUI as well, which makes three.

[w1k]

v3.21b - sometimes doesnt work lowercase letters, only upper..

[flashjazzcat]

v3.21b - sometimes doesnt work lowercase letters, only upper..

OK - can you describe the general scenario where this happens for you? I'll have a look into it.

[w1k]

i overwriting text from notepad to LW, i switching windows (notepad / altirra 2.00-t52).. problem fixed when i reboot LW..

[flashjazzcat]

Sounds like it's outside of LW's hands to me...

[flashjazzcat]

Major new update coming up.

[Jacques]

Great!

[+Stephen]

Major new update coming up.

Great - any details or is that part of the surprise?

[flashjazzcat]

Let's just say that owners of a certain increasingly prolific piece of hardware will be in for a nice surprise.

[Jacques]

Wow, superb! Std + VBXE versions merged into one, I presume?

[flashjazzcat]

Wow, superb! Std + VBXE versions merged into one, I presume?

 

Pretty much: VBXE driver is (optionally) loaded during startup, and you get a 26KB main text buffer.

Edited December 15, 2011 by flashjazzcat

[w1k]

latest LW + q-meg v3 fast sio = slow directory

 

http://www.youtube.com/watch?v=izlRYirIYig

[flashjazzcat]

To elucidate, this also happens occasionally when using SDX 4.4x with SIO2PC, generally when the baud rate is quite high. I advised deactivation of the progress bar in the above case (assuming the rendering of the graphic - which occurs during the stage 1 VBL - was causing problems), but the issue apparently persisted with the Q-Meg OS (stock OS was fine, though).

[Rybags]

I've devised a method of quit / resume VBlank processing that can help fix delayed IRQs if you need a possible fix.

[flashjazzcat]

I've devised a method of quit / resume VBlank processing that can help fix delayed IRQs if you need a possible fix.

Interested!

[flashjazzcat]

Why did it take me thirteen years to figure out that rearranging Atari fonts in ASCII order (instead of - in this case - converting the whole text buffer to and from internal code format) is the oldest trick in the book?

[+Stephen]

LMAO - bet you'll never forget that trick now

[flashjazzcat]

Heh... I thought I'd invented the wheel with that one till Candle told me it was a common technique. The prior technique (conversion of the whole text buffer) can be traced right back to my early investigations into text editors. TextPro and Speedscript before it both convert the whole buffer to internal codes.

 

On a technical note, the VBXE driver performs disappointingly slowly. This is no fault of the hardware - just an underestimation of the overhead of switching the VBXE MEMAC window in and out for every byte written. The idea behind this was to allow text buffer space to occupy the RAM under the MEMAC window. However, the test version is slower than the standard version in 80 columns. The old version 4 beta I wrote a couple of years ago keeps MEMAC open the whole time and blits the line buffer to the display RAM (this of course depends on the line buffer the editor screen refresh routine builds being in the MEMAC region). I'll have to go back to this way of doing things, so much of the major code reorganization I've done over the last week has been a waste of time. Not much chance of this being done before Christmas now. Still - you live and learn, as they say.

[Rybags]

Couldn't you just use a 4K window and map it somewhere "spare" ?

 

Or on the other hand, you can use both MEMACs, the "old" one @ $4000-7FFF and new one wherever you want it.

 

Or maybe some sort of trickery where you could use INC/DEC to switch the window in and out.

[flashjazzcat]

A 4K window would probably do it, actually. This means VBXE users would get an extra 6K on their main text buffer, instead of 10. Regarding $4000-$7FFF: I think one has to be careful here, since some memory expansions override the MEMAC window, and LW leaves extended banks switched in depending on which text bank one has open.

 

Because VBXE can be mapped to $D600 or $D700, we're addressing the hardware registers using an indirect pointer, so at the very least the Y reg needs saving and restoring. The quickest way to switch banks in and out is with bit 7 of MEMAC control, so we need the accumulator as well. I recall (and should have remembered earlier) that without blitting the line buffer to the screen in the version 4 beta, performance was disappointing. I think MEMAC needs to be left open, regardless of the size or position of the window.

 

This might sound unnecessarily complex, but remember we're applying device abstraction to a piece of software which was completely tied to the stock hardware up until now. Changes to the memory map are nightmarish: the program already has dozens of overlaid segments which get jettisioned after initialization, and which are all carefully positioned not to crash into important buffers, etc. At least the stock version is a little faster now, owing to optimisations I discovered while writing the VBXE driver.

[flashjazzcat]

Looks like this might be ready for Christmas after all - at least as a beta version.

[flashjazzcat]

Blitter operations finally working - and this thing is FAST!

[flashjazzcat]

Too buggy and incomplete to release even as a beta yet, but here's a taster before the (short) Christmas break:

 

http://www.youtube.com/watch?v=4QU9mGt_E98

 

Had to film this on an LCD monitor because Altirra's keyboard repeat rate seems slow even with the fast debounce and key buffer turned on (I'm wondering if Altirra has its own keybuffer: the repeat seems slow for a while, then it suddently sprays out buffered keystrokes).

 

Anyway, just progress bar and colour customisation dialogue to do, plus various bug fixes. The VBXE version is FAST, but the stock version is now somewhat quicker too.

 

Merry Christmas!

 

Oh... lovely start to Christmas. Flyback transformer's just died on the new Philips CM8833-II. Wonderful!

Edited December 24, 2011 by flashjazzcat

[flashjazzcat]

DMA wasn't disabled properly in the video: just fixed that, and the editor is faster again. Hoping to get a beta out by the weekend. This thing is smokin'!

Edited December 29, 2011 by flashjazzcat