Welcome, Guest. Please login or register.

Author Topic: Vampire 600: FPGA A600 Accelerator Preorder  (Read 8527 times)

Description:

0 Members and 1 Guest are viewing this topic.

Offline Mrs Beanbag

  • Sr. Member
  • ****
  • Join Date: Sep 2011
  • Posts: 455
    • Show all replies
Re: Vampire 600: FPGA A600 Accelerator Preorder
« on: August 10, 2013, 06:00:23 PM »
If there are any plans to make an A1200 version I would put money down for that.
Signature intentionally left blank
 

Offline Mrs Beanbag

  • Sr. Member
  • ****
  • Join Date: Sep 2011
  • Posts: 455
    • Show all replies
Re: Vampire 600: FPGA A600 Accelerator Preorder
« Reply #1 on: August 11, 2013, 09:16:01 PM »
Majsta is it possible for the user to update the FPGA code? We do actually have an A600 that we don't use much, if it were possible, I could use it to try developing a faster CPU softcore, which I would like to try.
Signature intentionally left blank
 

Offline Mrs Beanbag

  • Sr. Member
  • ****
  • Join Date: Sep 2011
  • Posts: 455
    • Show all replies
Re: Vampire 600: FPGA A600 Accelerator Preorder
« Reply #2 on: August 12, 2013, 11:23:25 AM »
I don't have any VHDL or Verilog knowledge, yet...
Signature intentionally left blank
 

Offline Mrs Beanbag

  • Sr. Member
  • ****
  • Join Date: Sep 2011
  • Posts: 455
    • Show all replies
Re: Vampire 600: FPGA A600 Accelerator Preorder
« Reply #3 on: August 12, 2013, 02:38:19 PM »
I gather there are still plenty of spare LUTs on the FPGA, room for a lot of enhancements.
Signature intentionally left blank
 

Offline Mrs Beanbag

  • Sr. Member
  • ****
  • Join Date: Sep 2011
  • Posts: 455
    • Show all replies
Re: Vampire 600: FPGA A600 Accelerator Preorder
« Reply #4 on: August 12, 2013, 05:56:45 PM »
Quote from: bloodline;744458
Perhaps one could simply set a pointer to the chunky buffer in fast ram, then set it off to automatically copy and covert the buffer to chipram as planar data :)
Or simply a command to shuffle the bits around in the data registers, so you can read 32 chunky pixels into D0-D7 and write out 32 planar pixels, or vice versa. That would be very easy to do since it could simply be hardwired into the register file bit by bit.

But I would first of all concentrate on making the CPU core as efficient as possible before adding extra features that no current software will use.
Signature intentionally left blank