[hatari-devel] LMC1992 and Microwire emulation

Laurent Sallafranque laurent.sallafranque at free.fr
Sun Mar 21 13:26:48 CET 2010


I'll have a look at the patches tonight.

> The code after:
>     case 2:
>            /* DMA sound only */
> multiplies DMA sound by 256, whereas, in other cases by 128...
> This will make the DMA sound louder when it is not mixed.


I've done it on purpose :

In case 1, I mix (dma + Yamaha sound) / 2 to have a value from 0 -> 65536

Regards

Laurent


Eero Tamminen a écrit :
> Hi,
>
> On Sunday 21 March 2010, David Savinkoff wrote:
>   
>> I have a 3 part explanation in the hataribug.txt attachment.
>>     
>
>   
>> As for converting floats to ints; Maybe floats are ok. They
>> can simplify code to the point of where it is more efficient
>> on computers since the 486 era.
>>
>> What platforms will have problems?
>>     
>
> There are still CPUs without FPU being used in devices, but those kind of
> CPUs are anyway a bit too slow to run Hatari, so I think they can be
> ignored.
>
> I think desktop computers deal faster with doubles, whereas with latest ARM
> CPUs (currently used in latest phones and soon also in netbooks) doubles are
> much slower than floats (besides taking double the memory), despite them
> having an FPU.
>
> Whereas integer operations are always faster than *same* operations in
> floating point, if CPU includes FPU, floats are faster than emulating
> floating point calculations with e.g. fixed point integers arithmetics.  And
> if one needs to do conversions between floating point and integers, that's
> a heavy floating point operation too...
>
> So, it makes sense to use integers, if the work can be done "naturally" with
> them, using floats loses accuracy, or there would be a need for a lot of
> conversions between floating point and integers (calculated by amount
> of data on which it's done, not by lines of code :-)).
>
>
> 	- Eero
> _______________________________________________
> hatari-devel mailing list
> hatari-devel at lists.berlios.de
> https://lists.berlios.de/mailman/listinfo/hatari-devel
>
>
>   




More information about the hatari-devel mailing list