Re: WT-FPGA? - Sinclair

This is a discussion on Re: WT-FPGA? - Sinclair ; On Aug 7, 1:34 pm, Dylan Smith wrote: > FPGA - From mid thousands of gates on upwards. Generally, use static RAM > for configuration, so need an external ROM to work, otherwise, like a > bigger version of a ...

+ Reply to Thread
Results 1 to 2 of 2

Thread: Re: WT-FPGA?

  1. Re: WT-FPGA?

    On Aug 7, 1:34 pm, Dylan Smith wrote:
    > FPGA - From mid thousands of gates on upwards. Generally, use static RAM
    > for configuration, so need an external ROM to work, otherwise, like a
    > bigger version of a CPLD. Again, Xilinx and Altera are dominant in the
    > FPGA market.


    Do you think someday there'll be a compiler than goes from C to FPGA?
    Then all those emulators out there could be used as the basis of new
    chips to replace the rapidly depleting stock of old silicon out there.

  2. Re: WT-FPGA?

    On 2008-08-07, DanSolo wrote:
    > Do you think someday there'll be a compiler than goes from C to FPGA?
    > Then all those emulators out there could be used as the basis of new
    > chips to replace the rapidly depleting stock of old silicon out there.


    What Matt Westcott says. To amplify a bit, programmable logic isn't
    anything like a procedural, step-by-step program. Programmable logic is
    exactly like designing any other electronic circuit - it's an exercise
    in electronics design, rather than procedural programming.

    Indeed, you can simply draw an electronic schematic using Xilinx's ISE
    (an IDE for CPLDs and FPGAa) and have it turn it into a gate
    configuration for the target chip. The schematic drawing comprises of
    gates, flip flops, half adders, latches, counters etc. just* as if you
    were drawing a schematic for a circuit using a pile of 74-series logic
    ICs that you were planning to solder together on a PCB.

    Chris Smith has implemented the Spectrum's ULA in an Xilinx XC95144
    (nearly using all the resources of said chip) for a project to make a
    drop in replacement of the ULA with a modern and widely available
    equivalent. My Spectranet project uses an Xilinx XC9572 (half the size
    of the '144) for its glue logic.

    The 'old silicon' situation isn't nearly as bad as you might expect -
    apart from the ULA. The Z80 is still manufactured in volume, 74-series
    logic (forming the 'glue' logic in the Spectrum) is still manufactured
    in volume, and with a little finagling, it would be possible to make an
    adaptor such that a currently available static RAM IC could plug into
    where the Spectrum's dynamic RAM sits... but there are still heaps of
    DRAM floating around that'll work with a Spectrum so it's not a pressing
    problem at present.

    * not quite as low level - you don't start drawing out transistors,
    diodes, resistors etc. when doing a CPLD schematic.
    --
    From the sunny Isle of Man.
    Yes, the Reply-To email address is valid.

+ Reply to Thread