I thought I'd give an update on my BitBlt reimplementation in case
anyone was interested for SoC or otherwise. Since I last mentioned it here, I've moved it to Slang, so it is developed and built similarly to the regular BitBlt. Large portions are auto-generated from templates (namely many of the combination rules and pixel depth conversion), cutting down the amount of real code to be developed. While it can't yet run a Squeak image (well, not for long...), it can run some basic benchmarks and tests on itself. There are bugs, but it will at least run most of the internal benchmarks. The performance is promising, with most supported operations being faster than stock BitBlt, and a few being slower (these need to be special-cased). Some operations will not be supported soon, especially the WarpBlt functionality. I'm no longer working heavily on this, but do occasionally open it up to fix bugs, add docs, etc. The latest dump is available at: http://home.earthlink.net/~cottongim/NuBltMar21-2007.st.gz Eddie |
Why don't you finish what you have done so that everybody can us it
and be happy? 2007/3/21, Eddie Cottongim <[hidden email]>: > I thought I'd give an update on my BitBlt reimplementation in case > anyone was interested for SoC or otherwise. > > Since I last mentioned it here, I've moved it to Slang, so it is > developed and built similarly to the regular BitBlt. Large portions are > auto-generated from templates (namely many of the combination rules and > pixel depth conversion), cutting down the amount of real code to be > developed. While it can't yet run a Squeak image (well, not for > long...), it can run some basic benchmarks and tests on itself. There > are bugs, but it will at least run most of the internal benchmarks. The > performance is promising, with most supported operations being faster > than stock BitBlt, and a few being slower (these need to be > special-cased). Some operations will not be supported soon, especially > the WarpBlt functionality. > > I'm no longer working heavily on this, but do occasionally open it up to > fix bugs, add docs, etc. > > The latest dump is available at: > http://home.earthlink.net/~cottongim/NuBltMar21-2007.st.gz > > Eddie > > -- Damien Cassou |
Because I'm sick of it.
Damien Cassou wrote: > Why don't you finish what you have done so that everybody can us it > and be happy? > > 2007/3/21, Eddie Cottongim <[hidden email]>: >> I thought I'd give an update on my BitBlt reimplementation in case >> anyone was interested for SoC or otherwise. >> >> Since I last mentioned it here, I've moved it to Slang, so it is >> developed and built similarly to the regular BitBlt. Large portions are >> auto-generated from templates (namely many of the combination rules and >> pixel depth conversion), cutting down the amount of real code to be >> developed. While it can't yet run a Squeak image (well, not for >> long...), it can run some basic benchmarks and tests on itself. There >> are bugs, but it will at least run most of the internal benchmarks. The >> performance is promising, with most supported operations being faster >> than stock BitBlt, and a few being slower (these need to be >> special-cased). Some operations will not be supported soon, especially >> the WarpBlt functionality. >> >> I'm no longer working heavily on this, but do occasionally open it up to >> fix bugs, add docs, etc. >> >> The latest dump is available at: >> http://home.earthlink.net/~cottongim/NuBltMar21-2007.st.gz >> >> Eddie >> >> > > |
Free forum by Nabble | Edit this page |