48k Timing Tests...

edited April 2010 in Development
Here's v1 of the timing tests for the 48k Spectrum, with a little background

timing_test_v1.jpg

http://www.zxspectrum4.net/op_timing.php

Please use the form to submit the results and we can add them to the table.

Thanks Richard

:smile:
Post edited by Richard_Butler on

Comments

  • edited April 2010
    Are these completely new tests? (just trying to catch up)
  • edited April 2010
    Yes, as much as they were the original tests when we started looking at the timing albeit slightly adjusted, the test for "type" is the same as the previously posted but thats just the start of it the other 35 should give more detail and possibilities for variations
  • edited April 2010
    I know that only results from real machines counts, but i cant resist to run this test on couple of emus..
    Results are quite good.
    SpecEmu showed the expected accuracy, but still failed just one test (35).
    If Woody fix this, we can put a SpecEmu in the hardware section, without problem. :)
    Also great new java emulator JSpec failed just two tests (36,37), and once again confirms their high quality.
    However, the biggest positive surprise is the Sinclairean (latest B5 build), which has passed all tests with no errors, but also failed some other tests big time, like whole MEMPTR section in z80test.
    Spin (latest beta 0.7s) failed 6 tests.
    All results are, for early timing.
    I didnt test Fuse and Spectaculator, so someone who use them, can add the results.
  • edited April 2010
    Pegaz wrote: »
    SpecEmu showed the expected accuracy, but still failed just one test (35).

    No. SpecEmu matches a real 48K machine when all extra hardware is disabled on test 35. I know. I've tested ;)
  • edited April 2010
    A .tzx file would make it easier to test in real machines.
  • edited April 2010
    Woody wrote: »
    No. SpecEmu matches a real 48K machine when all extra hardware is disabled on test 35. I know. I've tested ;)

    Thanks Woody, i didnt know that. :)
  • edited April 2010
    Pegaz wrote: »
    However, the biggest positive surprise is the Sinclairean (latest B5 build), which has passed all tests with no errors, but also failed some other tests big time, like whole MEMPTR section in z80test.

    :smile: :smile: :smile: :smile:

    I was currently working on the 50 Hz interframe blender but, intrigued by this topic, I think I'll implement proper MEMPTR emulation before that ;)

    BTW, where can I find that pesky z80test? I've googled it without luck :-?
  • edited April 2010
    :smile: :smile: :smile: :smile:

    I was currently working on the 50 Hz interframe blender but, intrigued by this topic, I think I'll implement proper MEMPTR emulation before that ;)

    BTW, where can I find that pesky z80test? I've googled it without luck :-?

    Heres the link:
    http://www.sendspace.com/file/5ku7eu
  • edited April 2010
    Thanks Pegaz :smile:
  • edited April 2010
    Richard, could you please provide a .TAP and a .TZX of this new test suite? Thanks.
  • edited April 2010
    BTW, where can I find that pesky z80test? I've googled it without luck :-?

    Its home is the z80 page on the tech wiki site :)

    Obviously you've already got it now!
  • edited April 2010
    Pegaz: SpecEmu, fuse and ZXSpectrum4.net pass all the tests based on type1.

    If anyone really feels the need to test someone elses emulator please keep it to yourself :wink:



    zxbruno:
    please provide a .TAP and a .TZX of this new test suite

    Will do.
    Done. Check download folder
    http://www.zxspectrum4.net/downloads/timing_tests/
  • I have a question... in a real Speccy all the tests should pass? Because I tested it last weekend on my two 48K and I got some fails...

    Even more strange is that if I load the SNA using k7zx I got that one of my machines has late timings, but if I convert the SNA to WAV using Woody's Tape2WAV... that same machine appear to have early timings.

    So, I didn't upload the results because I thougth they weren't very reliable...
  • edited April 2010
    Hern?n wrote: »
    I have a question... in a real Speccy all the tests should pass? Because I tested it last weekend on my two 48K and I got some fails...

    You may well find a spectrum failing tests this was the purpose of the exercise, just means some timing is out/different. Remember to leave the spectrum to warm up 30mins should do it at room temp ~21C.
    Hern?n wrote: »
    Even more strange is that if I load the SNA using k7zx I got that one of my machines has late timings, but if I convert the SNA to WAV using Woody's Tape2WAV... that same machine appear to have early timings.

    The only thing that can change the "type" test result is multiple times the loop is passed, so it could be the different programs are starting the code in a different place, theres a .wav on there use that. But again youll find more than likely it was warming up, run it lots of times and unplug any interfaces.

    http://www.zxspectrum4.net/downloads/timing_tests/
  • edited April 2010
    Woody wrote: »
    Its home is the z80 page on the tech wiki site :)

    Obviously you've already got it now!

    Yup. Thanks Woody :smile:
  • edited April 2010
    My Spectrum 48k+ just failed two tests (12 contended and 27 contended). It had warmed only for around 10 minutes, so I'll try again later and see if it still fails.

    PD:
    After 40 more minutes, the computer is now detected as a Type 1 (Early Timings) instead of Type 2, and fails on these tests:
    Test 3: R=49 SP=23315
    Test 12: R=80 SP=56830
    Test 19: R=39 SP=56818
  • edited April 2010
    Pegaz wrote: »
    Also great new java emulator JSpec failed just two tests (36,37), and once again confirms their high quality.

    Thanks by your fair comment about JSpeccy. I've corrected the problem in JSpeccy now. The tests #36 & #37 fails though, why JSpeccy don't have any method to disable the Kempston emulation. But if you disables the Kempston support, the tests shows OK. :)

    The new JSpeccy version is at the project webpage.
Sign In or Register to comment.