BITMAP (Allegro 4) to ALLEGRO_BITMAP (Allegro 5) conversion

[MarekRatajczak2024]

Hi everyone,
Does optimised procedure exist converting BITMAP (Allegro 4 style) to ALLEGRO_BITMAP (Allegro 5 style) assuming that both bitmaps are the same dimensions, the same colour depth, so they are "visually" identical, but doing that conversion somehow smart by simply memmove of entire lines (horizontal lines) rather then doing it pixel by pixel, with pixel colour conversion which I realized is platform dependant, and what I have done so far in my program, what is not really optimal, and that I have to do 30 times per second, using Allegro-Legacy library due to nearly entire code I have done in Allegro 4, so all based on BITMAPs. Thanks for your attention and awaiting for any answer.
Marek.

[miquelfire]

First question, are you linking both A4 and A5 to your program? Or are you using some backwards compatibility library that allows you to use A4 style code?

[MarekRatajczak2024]

Hi miquelfire,
Yes, my entire program AlfaCAD (https://alfacad.net) is based on Allegro 4 (for Linux, and for Windows with my own 64bit implementation of Allegro for Windows, because didn't exist), but to do that for MacOS I was enforced to use Allegro 5, due to decommissioned QuickDraw, so eventually I'm using Allegro-Legacy, with some changes - mostly for MacOS platform, however on development stage I do the same on Linux (it's just test version, production version is on Allegro 4 with my extensions), my main programming platform, to replicate all on MacOS x64 then MacOS aa64.
So basically all is nearly done using Allegro-Legacy (https://github.com/NewCreature/Allegro-Legacy), which is the replica of Allegro 4.4 or 4.3.1 it doesn't really matters, but all is drawn on virtual screen as Allegro 4 style BITMAP, than, in separate thread, that screen is converted to ALLEGRO_BITMAP (in fact there is first blit of the "screen" (so BITMAP) to it's copy in the same format, and can be done by blit, or can be done by memmove of entire image on n-lines x m-rows at once, and from my experience both methods are fast, to avoid some flickering in real time, then that copy is copied to ALLEGRO_BITMAP) . But that what takes most time is copying BITMAP to ALLEGRO_BITMAP, and so far I have to do that - as Allegro-Legacy is doing at the very last stage, pixel by pixel. I can provide details how it's done, but maybe this is not important now. So I was asking if maybe somebody invented the procedure avoiding copying pixel-by-pixel, rather doing it line by line or so, what could increase the speed dramatically. There is no problem with M4 processor, it's only important for older platform, x64 especially when we deal with retina screen, because there are many pixels to copy, and this is some 30 FPS or more. As I wrote, I'm happy with M4, I do not see any lagging, all works smooth, because M4 is bloody fast. Not necessary it's like that with i7. As a matter of fact, I do not know ell the structure of ALLEGRO_BITMAP, while I do know very well structure of BITMAP, which I also use to produce printing image, also PDF images etc. I'm not able to say if copying by lines is even possible due to the format of each pixel, what is platform dependant.

[mlt]

Give a try to https://github.com/kazzmir/allegro4-to-5 instead of AllegroLegacy. I somewhat actively commit there to the point I can run Editor On Fire. The main advantage in my view is that 425 allows to slowly migrate to a5 by mixing in a5 code here and there compared to legacy.

[MarekRatajczak2024]

I did try kazzmir's stuff already, as I remember 2 years ago, and I got it working, but due to the nature of my entire program, using vector entities way more than bitmap ones, with entire inventory of vector objects, including writing modes like line types (user defined) filling polygons with seamless bitmaps etc., I realized it won't be that easy to get everything working and I focused on other things, mostly on porting my program to Linux as native Xlib application. On Linux Allegro 4 work like charm, is fast. Then, to touch Mac, I did it on Linux first using Allegro Legacy in parallel. Obviously, it's not that fast due to the nature of the library, but fast enough and eventually I found a way how to get it working using modern Allegro 5 driver with - say - Allegro 4 core, which doesn't require to change anything except of some last-step stuff, one of them is conversion Allegro 4 BITMAP to Allegro 5 ALLEGRO_BITMAP every single frame, some 30 frames per second or more. Then I could back to MacOS, and however I'm not impressed with performance of Intel, especially with high resolution like retina screen, it works more than perfectly fine on M processors. The another problem for me is a difference between BITMAP and ALLEGRO_BITMAP. BITMAP is - say - compatible (not counting bit mixed color bytes) with Windows GDI, which is fast (accelerated), not just for screen, but also for printing, what is a part of my program, and printing sheets formats are actually limitless, where resolution is way more than any screen on the planet. However there is not such big meaning in Linux and MacOS where printing is CUPS and is based on pdf technology, however even in PDFs is easier for me to convert BMP / BITMAP format to other formats also used inside PDFs, because not every vectorial object has representation in PDF technology, and must be substituted with ready to use image made on the side. This is always a matter of tools you use, those existing and those to be done. Maybe one day I will try Kazzmir's again, but I know it won't be painless. Anyway good to know Allegro4-to-5 is alive. Cheers, Marek.