Computer design: Screen resolution and graphics processor musts
Screens are too small. Of course, with a reasonable 8 x 8 bit font you could make good use of a 1024 x 768 screen, if only you could read the letters! Then again, nobody will produce a big screen with relatively low resolution, only so that programmers are happy that they can see many text lines on it.
Hence the only solution is a big sreen with a normal resolution, meaning a 2048 x 1536 screen.
Why bother with lower resolutions? We can always let the emulator do the downscaling. As far as higher resolutions are concerned, well, we'd have to use several of our computers parallely for that. As long as they are simulated by an emulator, that requires little more than multitasking.
One thing I definitely want from a graphics processor is the ability to add, subtract, multiply and divide a constant value to/from/with an array of n-bit values separated by m bits, allowing me to easily add a green tint to an image etc. This needs to be done in such a way that the maximal and minimal values are not passed. I also want up to four such operations at once. A bicubic rescaler and a bicubic rotator, reading from one address and writing to another, is the next must. Then, in case we run out of sprites, we need blending one image over another, according to the chosen transparency, again reading from two addresses and writing to a third. Finally, you should be able to combine all of these functions in one operation of our Super Blitter.
I also want a chip that does nothing but Fourier transforms.
Beyond that I can't say right now, what else should be included. A vector machine perhaps, some chip that helps sorting sprites and playfields. That would be possibilities.
It's a good question whether you always want transparency as the blending method. I could imagine others, like adding colour values, subtracting or whatever. I think it would be fun to let people choose for every playfield and sprite and also for the blitting what method they want to use. But right now I don't quite know what would be a set of interesting methods. Let's consider this an open question.
Hence the only solution is a big sreen with a normal resolution, meaning a 2048 x 1536 screen.
Why bother with lower resolutions? We can always let the emulator do the downscaling. As far as higher resolutions are concerned, well, we'd have to use several of our computers parallely for that. As long as they are simulated by an emulator, that requires little more than multitasking.
One thing I definitely want from a graphics processor is the ability to add, subtract, multiply and divide a constant value to/from/with an array of n-bit values separated by m bits, allowing me to easily add a green tint to an image etc. This needs to be done in such a way that the maximal and minimal values are not passed. I also want up to four such operations at once. A bicubic rescaler and a bicubic rotator, reading from one address and writing to another, is the next must. Then, in case we run out of sprites, we need blending one image over another, according to the chosen transparency, again reading from two addresses and writing to a third. Finally, you should be able to combine all of these functions in one operation of our Super Blitter.
I also want a chip that does nothing but Fourier transforms.
Beyond that I can't say right now, what else should be included. A vector machine perhaps, some chip that helps sorting sprites and playfields. That would be possibilities.
It's a good question whether you always want transparency as the blending method. I could imagine others, like adding colour values, subtracting or whatever. I think it would be fun to let people choose for every playfield and sprite and also for the blitting what method they want to use. But right now I don't quite know what would be a set of interesting methods. Let's consider this an open question.
Labels: 04, computer design
Eingetragen Am/um
20:19
