Gunnar von Boehn (Apollo Team Member) Posts 3431 03 Apr 2018 14:17
Some FPU speedup in upcoming APOLLO Core release.
x11 Core Speed results:
Roger Andre Lassen
Posts 32 03 Apr 2018 14:20
Impressive Gunnar :-)
Posts 347 03 Apr 2018 16:08
Nice 105 MIPS and 51 MFLOPS for what ... A500 V2 with 2.9 CORE X11?
Posts 55 03 Apr 2018 17:31
Posts 68 03 Apr 2018 17:50
Same core revision, but in A600. Even slightly faster :) Kudos, Gunnar!
Posts 131 03 Apr 2018 19:09
Does this next release fix the rounding issue
Gunnar von Boehn (Apollo Team Member) Posts 3431 03 Apr 2018 19:11
Peter, here you can read about the FPU. EXTERNAL LINK The Wiki page should answer your question.
Posts 6 03 Apr 2018 23:26
Looks good. Except fpsp handled instructions, are only a few fastest than MC68882 version, and over 2-3 times slowest than Pentium version. Im not hardware expert, but for me optimised versions for 68080FPU can be much fastest. I know that problem for handling these instructions in FPGA is limited FPGA space. My idea (i dont know if possible to do) is next. Remove FPSP vector, place all highly optimised microcoded instructions inside Apollo core, place all big tables for microcoded instructions in external ROM. I think that it will be much fastest, because only reads from external ROM tables will be slow.
Posts 248 04 Apr 2018 01:26
Wow that is a big boost. Nice work!
Gunnar von Boehn (Apollo Team Member) Posts 3431 07 Apr 2018 10:04
AIBB FPU Beachball 3D Render speed using GOLD 2.9 x11 Core