1. 16 Sep, 2012 2 commits
  2. 23 Mar, 2001 1 commit
    • Stewart Brodie's avatar
      Updated build structure to use the shared AAsmModule makefile. · 226dfa3c
      Stewart Brodie authored
        Updated to build using objasm instead of aasm.
        Sources changed to be objasm-compatible.
        Removed unused resources.
        Fixed some non 32-bit compatible code.
        Requires Library 0.71 or later.
        Requires BuildSys 3.09 or later.
        Requires Env 0.65 or later.
      Version 0.15. Tagged as 'ITable-0_15'
  3. 07 Dec, 2000 1 commit
  4. 14 Aug, 2000 1 commit
    • Justin Fletcher's avatar
      InverseTable modified to use ColourTrans to generate tables. · 330360ef
      Justin Fletcher authored
        * In the first instance, it is now possible to use the 32k table from
          Resources:$.Resources.Colours.Tables.8desktop instead of the normal
          ...ITable.Tables.8desktop location. This reduces the space used by
          InverseTable by 32k and may free enough room for 2MB ROM builds.
        * Additionally, it is possible to use ColourTrans to generate all
          tables, rather than using the inbuilt table. ColourTrans is much
          faster at generating these tables - presumably at the expense of
          accuracy - and this may help in modes where the default palette is
          in use. Currently, if ColourTrans fails we fall back to the old
          InverseTable method.
        * Both switches can be toggled independantly.
        * Makefile modified to not export the InverseTable in the resources
          phase. If you toggle the UseColourTransForResourceTable switch, you
          must update the makefile to reflect the required export.
        Tested, no noticable degradation of blended font readability observed,
        either with standard palettes or with custom palettes, when displayed
        in the desktop, single tasking, or when redirected to sprites.
        StrongARM timings:
          ColourTrans takes about 1 second to generate translation tables.
          InverseTable takes about 2 seconds.
        Plus, with ColourTrans you're gaining from the fact that it caches more
        than a single table, so a second render after a sprite switch /does/
        remember the cached entry.
        Additional space is required for a 2MB ROM build, and the 32k table
        provided by InverseTable in ResourceFS is really redundant as the
        ColourTrans table will suffice.
        InverseTable provides its own table generation code which is similar
        to that in ColourTrans (colourtrans uses simple weightings, inversetable
        uses sum of squares, as I understand it)
      Version 0.13. Tagged as 'ITable-0_13'
  5. 07 Jul, 2000 1 commit
    • Justin Fletcher's avatar
      Fix for rare bug triggered by multiple nested sprite redirections confusing... · eeb9ef17
      Justin Fletcher authored
      Fix for rare bug triggered by multiple nested sprite redirections confusing the module's palette tracking.
        The problem is that at present InverseTable assumes that when you
        redirect output to a sprite you will be toggling in and out of a sprite
        with a palette it understands. That is that in the sequence Calculate,
        redirect, redirect, Calculate, the final Calculate can be ignored
        because it will have the same table as the first. This is incorrect if
        the second redirect is a redirection to a sprite which was not that
        which the first would have restored to.
        The failure sequence can occur when you generate a bitmap from a font
        for the first time, I believe. It can also be artificially simulated
        (and can happen in real world situations if redirection occurs on a
        The new algorithm is :
          At every redirection, record whether we are redirecting to the screen
          When tables are generated, remember whether this was for a screen
          Mark every redirection as invalid, UNLESS the table we have cached is
            for the screen and we are redirecting to the screen, in which case
            clear the redirection invalid flag.
        The upshot of this is that for any sequence of redirection the
        inversetable is maintained as valid, BUT redirection between multiple
        sprites can cause InverseTable to recache the tables where previously it
        might have used tables it already had. In the latter case, it would be
        likely that redirection back to the screen would have displayed
        'tutti-fruti' coloured text.
        The speed hit of calculating the tables will be noticable if they are
        needed, but the effect of displaying incorrect text is more noticeable
        when it occurs.
        Tested with examples of badly nested sprite redirection, and correctly
        tracked the palettes in use.
      Version 0.12. Tagged as 'ITable-0_12'
  6. 28 Jun, 2000 1 commit
    • Alistair Rodger's avatar
      this is a checkin following a failed checkin. · bfb376d5
      Alistair Rodger authored
        added a usermode gap.
         the central palette calculation loop was taking sufficiently
         long to disrupt the flow of video.  This has been fixed.
         I have tested this on the desktop and the STB.
      Version 0.11. Not tagged
  7. 12 May, 2000 1 commit
  8. 27 Aug, 1999 1 commit
  9. 05 Aug, 1999 2 commits
  10. 21 Jan, 1997 1 commit
  11. 21 Nov, 1996 2 commits
  12. 06 Nov, 1996 1 commit