Re: [hatari-devel] DSP performance

[ Thread Index | Date Index | More lists.tuxfamily.org/hatari-devel Archives ]


I was a bit lazy while coding the static table, and I used the same value for all the addressing modes of the FPU cycles.
That's why many FPU cycles were wrong, but not so much.

Example :

/*1821*/ {6, 0, 20,0,1,0, 22,0,1,0}, // FScc.L #<data>.W,(An) /*1822*/ {6, 0, 20,0,1,0, 22,0,1,0}, // FScc.L #<data>.W,(An)+ /*1823*/ {6, 0, 20,0,1,0, 22,0,1,0}, // FScc.L #<data>.W,-(An) /*1824*/ {6, 0, 20,0,1,0, 22,0,1,0}, // FScc.L #<data>.W,(d16,An) /*1825*/ {6, 0, 20,0,1,0, 22,0,1,0}, // FScc.L #<data>.W,(d8,An,Xn) /*1826*/ {6, 0, 20,0,1,0, 22,0,1,0}, // FScc.L #<data>.W,(xxx).W /*1827*/ {6, 0, 20,0,1,0, 22,0,1,0}, // FScc.L #<data>.W,(xxx).L


It would be simple to add the correct values now, but the static table won't be the futur of hatari, so I don't do it. But I can compute the values if one need them to compare with the new core instruction cycles.

That's why old cpu timings were wrong with the static table

Laurent




Le 30/06/2015 22:57, Nicolas Pomarède a écrit :
Le 30/06/2015 22:45, Eero Tamminen a écrit :
Hi,

For references, here are Centurbo bench results:

Today WinUAE CPU core: FPU 812 Mhz, DSP 57 Mhz, CPU 27 Mhz
Dec WinUAE CPU core: FPU 153 Mhz, DSP 57 Mhz, CPU 172 Mhz
Old WinUAE CPU core: FPU 221 Mhz, DSP 32 Mhz, CPU 15 Mhz
Old UAE CPU core: FPU 403 Mhz, DSP 32 Mhz, CPU 78 Mhz
(moving mouse drops FPU and CPU numbers slightly)

I.e. compared to old WinUAE core with Laurent's timings,
current CPU runs 80%, DSP 78% and FPU >3x faster.

DSP timing being derived from CPU would explain why it's
faster by same amount as CPU, but why FPU is now so much
faster?


Hi

haven't look in detail, but AFAIK fpu emulation is not cycle exact in lastest WinUAE and most fpu opcodes take a very small number of cycles.

Maybe it was a bit closer to real thing before, but it was wrong anyway.

Nicolas






Mail converted by MHonArc 2.6.19+ http://listengine.tuxfamily.org/