A message in the Bottle

[fabrice montupet]

Maybe a completely useless program but I dream about a program that could benchmark the performances of the TI-99/2/4A/5/8 computers family, but also the Geneve 9640 and all the new computers compatible with them. It could measure the mathematics operations, RAM access, internal operations, graphic performances or other else. It could also be interesting to see the speed differences between the real hardware and the emulators.
Just a message in a bottle at the software's sea.... ?

 

Edited May 12, 2021 by fabrice montupet

[+TheBF]

This might best be done like an elephant eating exercise.  One byte   bite at a time.  

Start with one simple benchmark program on one platform and port it to the others one at a time.

 

Then like it says on my shampoo bottle: Lather, rinse and repeat

Eventually you get a set of programs.

[+mizapf]

Not an easy task. Keep in mind that performance on our machines is highly depending on the kind of memory usage. For instance, is it a fair comparison if you manage to put the whole benchmark (or relevant parts) into the on-chip RAM of the 9995 on the Geneve? But if not, and you choose the slowest memory, does that comparison yield any interesting information?

[+TheBF]

 

2 hours ago, mizapf said:

Not an easy task. Keep in mind that performance on our machines is highly depending on the kind of memory usage. For instance, is it a fair comparison if you manage to put the whole benchmark (or relevant parts) into the on-chip RAM of the 9995 on the Geneve? But if not, and you choose the slowest memory, does that comparison yield any interesting information?

I was thinking the same things. Not a simple request.

 

I think the first one would need to be a very small piece of machine code that can run on all machines in their respective fastest memory space.

 

Perhaps just some math operations, logic operations and shifts in registers in the fastest RAM iterating >FFFF times ?

The Tursi sprite benchmark might be another one to test VDP/CPU connection speeds?

 

Getting useful information may depend on M. Montupet.

Fabrice. What do you need these tests to accomplish?

[fabrice montupet]

The idea would be to use a program that could be executed on all machines, as is, without a model detection.  The goal is to see their speed behavior. These tests could be made on each endemic features of the 99/4A only.

The 99/8 features a speed option (99/4A speed/full speed). This speed option is also present on the 99/5 (normal/full speed). The Basic Computer 99/2 was announced by TI as the fast little computer (a speed demo has been done in the CES 1983). Some people have added a 14MHz crystal (selectable by a switch), some other have removed some wait states on their 99/4A (thanks to the excellent Nouspikel's page http://www.unige.ch/medecine/nouspikel/ti99/wait.htm ). Some people or companies , past and present, have made a compatible 99/4A computer or a computer that embeds a compatibility mode. And, finally , some people have made a TI-99/4A emulators and on which we can notice some speed differences with the original computer. For all of them, I would have liked to quantify this speed differences.

 

An example of speed difference between the 99/4A mode and the speed mode of the  99/5:

http://www.ti99.com/ti/data/medias/Parsec_prototype_99_5.mp4

 

Edited May 13, 2021 by fabrice montupet

[oddemann]

99/5 looks... great!