chiark / gitweb /
Richard Kettlewell [Sun, 2 Dec 2012 20:13:36 +0000]
Generate a movie from the GUI.
A bit rough and ready but works.
The default is to use libx264, which seems to be less available,
but produces much better results than mpeg4.
Richard Kettlewell [Sun, 2 Dec 2012 20:10:11 +0000]
Control panel split.
The control panel itself is now logically separated from the
generic control widget infrastructure. A string control
is added.
Richard Kettlewell [Sun, 2 Dec 2012 20:09:03 +0000]
draw: shell quoting and automatic selection of encoder.
avconv is used if it is on the path, otherwise ffmpeg. If you have
neither then it won't work.
Richard Kettlewell [Sun, 2 Dec 2012 20:01:18 +0000]
Shell-related utilities.
This provides shell quoting and path searching.
Richard Kettlewell [Sun, 2 Dec 2012 19:59:36 +0000]
job: more caution about where things are dequeued.
If polling for a particular completion_data, only dequeue jobs that
match. See the expanded comments in Job.h for the intent.
Richard Kettlewell [Sun, 2 Dec 2012 19:57:50 +0000]
draw: don't use a pixbuf when writing a PPM.
Really we want to avoid GTK for other file types too but that's
a lot harder.
Richard Kettlewell [Sun, 2 Dec 2012 17:28:36 +0000]
Default Job does nothing.
Richard Kettlewell [Sun, 2 Dec 2012 17:18:00 +0000]
Poll/cancel jobs by completion data.
This should be more consistent and reliable than the previous
rather wonky infrastructure.
Richard Kettlewell [Sun, 2 Dec 2012 16:38:50 +0000]
Move work cancellation call.
FractalJob::recompute() now just schedules the work, it doesn't cancel
old work as well.
Richard Kettlewell [Sun, 2 Dec 2012 11:59:55 +0000]
Pass an argument to newly created threads
Richard Kettlewell [Sun, 2 Dec 2012 14:38:24 +0000]
controls: valgrind-clean
Overenthusiastic wcslen can still produce some false +ves from
around this area.
Richard Kettlewell [Sun, 25 Nov 2012 19:27:02 +0000]
gitignore
Richard Kettlewell [Sun, 25 Nov 2012 19:24:57 +0000]
Adjust colorization
Richard Kettlewell [Sun, 25 Nov 2012 19:22:32 +0000]
--dive option to make a video
Richard Kettlewell [Sun, 25 Nov 2012 16:58:21 +0000]
In-progress jobs are pending too
Richard Kettlewell [Sun, 25 Nov 2012 15:11:51 +0000]
Drop a log() from iteration count
Richard Kettlewell [Sun, 25 Nov 2012 15:11:35 +0000]
Show pointer position and iteration count
Richard Kettlewell [Sun, 25 Nov 2012 11:46:34 +0000]
Split pixel generation from job completion
Richard Kettlewell [Mon, 9 Apr 2012 10:07:02 +0000]
autogen: can use /bin/sh, therefore do so
Richard Kettlewell [Sat, 17 Dec 2011 19:20:44 +0000]
Add missing function attributes
Richard Kettlewell [Sat, 17 Dec 2011 19:20:31 +0000]
Correct test.
Richard Kettlewell [Mon, 22 Nov 2010 09:25:50 +0000]
Add missing dependency rule
Richard Kettlewell [Sun, 21 Nov 2010 20:39:23 +0000]
Dist visual studio build files.
Richard Kettlewell [Sun, 21 Nov 2010 20:31:07 +0000]
distcheck-clean
Richard Kettlewell [Sun, 21 Nov 2010 20:20:03 +0000]
Merge branch 'master' into win32
Richard Kettlewell [Sun, 21 Nov 2010 20:18:47 +0000]
Dist getopt sources
Richard Kettlewell [Sun, 21 Nov 2010 20:15:50 +0000]
Restore option parsing under Windows.
Swiped the getopt sources from GCC 4.5.0 and modified them slightly to
build on Windows. This commit doesn't attempt to support Unix
platforms without getopt_long() but that wouldn't be especially hard
(although all the mainstream ones have it anyway now).
Richard Kettlewell [Sun, 21 Nov 2010 19:45:25 +0000]
Use Microsoft-friendly logo.h
Richard Kettlewell [Sun, 21 Nov 2010 19:37:13 +0000]
distcheck-clean
Richard Kettlewell [Sun, 21 Nov 2010 19:35:45 +0000]
Translate embedded image to a format that Microsoft's compiler can
digest.
Richard Kettlewell [Sun, 21 Nov 2010 18:07:38 +0000]
Pick up number of CPUs on windows too.
Richard Kettlewell [Sun, 21 Nov 2010 18:07:11 +0000]
...and the release configuration too
Richard Kettlewell [Sun, 21 Nov 2010 17:52:14 +0000]
Add gthread to linker options.
mandy now builds and runs on Windows!
Richard Kettlewell [Sun, 21 Nov 2010 17:47:41 +0000]
WIP compiles but does not link on windows
Why does that manage() call need explicit namespace qualification but
not the adjacent ones? I have NO idea...
Richard Kettlewell [Sun, 21 Nov 2010 17:43:44 +0000]
WIP further Windows support.
Some issues remain.
Richard Kettlewell [Sun, 21 Nov 2010 17:07:07 +0000]
WIP more windows porting
Richard Kettlewell [Sun, 21 Nov 2010 16:54:26 +0000]
WIP Checkpoint windows build of main program.
Richard Kettlewell [Sun, 21 Nov 2010 14:26:48 +0000]
fixed64-test works on Windows too now.
Richard Kettlewell [Sun, 21 Nov 2010 14:19:43 +0000]
inttypes.h is UNIX-only.
Richard Kettlewell [Sun, 21 Nov 2010 14:17:28 +0000]
Add a master .sln file.
Move projects into their own private directory.
Richard Kettlewell [Sun, 21 Nov 2010 14:08:21 +0000]
WIP some initial windows support. fixed-test builds and works,
not tried anything else.
Richard Kettlewell [Sun, 21 Nov 2010 12:18:45 +0000]
declare header
Richard Kettlewell [Sun, 21 Nov 2010 12:14:06 +0000]
Reorganize configure.ac a bit.
Cope in the absence of unistd.h.
Richard Kettlewell [Sun, 21 Nov 2010 12:07:32 +0000]
Request correct version of gthread
Richard Kettlewell [Sun, 21 Nov 2010 12:03:19 +0000]
WIP more threading
Richard Kettlewell [Sun, 21 Nov 2010 11:48:36 +0000]
WIP start of support for alt thread APIs
Richard Kettlewell [Sun, 21 Nov 2010 01:02:44 +0000]
Conflate doubling and rounding, for a performance boost
Richard Kettlewell [Sun, 21 Nov 2010 01:00:13 +0000]
Alternative means of deciding the new sign. Turns out to be faster.
Richard Kettlewell [Sun, 21 Nov 2010 00:57:47 +0000]
comment
Richard Kettlewell [Sun, 21 Nov 2010 00:55:09 +0000]
move rbp<-escape limit way up, helps.
Richard Kettlewell [Sun, 21 Nov 2010 00:54:08 +0000]
move r12/13<-zx^2 down a bit, helps a little
Richard Kettlewell [Sun, 21 Nov 2010 00:52:17 +0000]
align Fixed_iterate to get a bit of consistency
Richard Kettlewell [Sun, 21 Nov 2010 00:50:17 +0000]
speedtest program
Richard Kettlewell [Sun, 21 Nov 2010 00:40:29 +0000]
Furtle with sign handling.
Doing the same in Fixed_mul seems to throw things off, presumably
there is an alignment question to answer.
Richard Kettlewell [Sun, 21 Nov 2010 00:08:50 +0000]
Align square subroutine.
Did NOT align iterloop, it makes it worse! Perhaps branch density too high.
Richard Kettlewell [Sun, 21 Nov 2010 00:05:08 +0000]
We have a register spare for the iteration count
Richard Kettlewell [Sun, 21 Nov 2010 00:00:30 +0000]
As recommended by optimization guide, spill to memory rather than xmm
registers. It really does seem to help - GPR<->XMM transfers are
expensive!
Richard Kettlewell [Sat, 20 Nov 2010 20:10:21 +0000]
Report arithmetic type properly
Richard Kettlewell [Sat, 20 Nov 2010 20:08:57 +0000]
Merge branch 'master' of git+ssh://araminta/~/git/mandy
Richard Kettlewell [Sat, 20 Nov 2010 20:08:36 +0000]
consistent naming
Richard Kettlewell [Sat, 20 Nov 2010 20:07:40 +0000]
update README for --with-short-fixed-point
Richard Kettlewell [Sat, 20 Nov 2010 19:59:48 +0000]
Clear errno before calling strtol
Richard Kettlewell [Sat, 20 Nov 2010 19:52:51 +0000]
move a shrd back a bit (or a mov forward). small benefit.
Richard Kettlewell [Sat, 20 Nov 2010 19:48:06 +0000]
Shave off some more cycles using lea
Richard Kettlewell [Sat, 20 Nov 2010 18:56:45 +0000]
move shrds up a bit in the hope of parallelizing with the
now-subsequent imul. a small measured improvement.
Richard Kettlewell [Sat, 20 Nov 2010 18:22:11 +0000]
optimization TODO
Richard Kettlewell [Sat, 20 Nov 2010 18:14:37 +0000]
align inner loop. seems to improve performance a tiny amount
Richard Kettlewell [Sat, 20 Nov 2010 18:04:25 +0000]
sacrifice a bit or two at the bottom for ~30% reduction in runtime
Richard Kettlewell [Sat, 20 Nov 2010 17:48:16 +0000]
Restrict Fixed64 to actual iteration; it cannot represent large enough
integers to do all the other work, which we therefore keep in Fixed.
Richard Kettlewell [Sat, 20 Nov 2010 17:37:10 +0000]
fixed64 arith_traits
Richard Kettlewell [Sat, 20 Nov 2010 17:34:41 +0000]
C++ fixed64
Richard Kettlewell [Sat, 20 Nov 2010 17:22:09 +0000]
A couple more assembler fixes
Richard Kettlewell [Sat, 20 Nov 2010 16:30:33 +0000]
better test output
Richard Kettlewell [Sat, 20 Nov 2010 16:13:48 +0000]
Correct register
Richard Kettlewell [Sat, 20 Nov 2010 16:13:39 +0000]
Asserts
Richard Kettlewell [Sat, 20 Nov 2010 16:09:29 +0000]
don't use a temporary to store r^2 return
Richard Kettlewell [Sat, 20 Nov 2010 16:07:16 +0000]
Merge branch 'master' into fixed64
Richard Kettlewell [Sat, 20 Nov 2010 16:06:28 +0000]
insist that test matches
Richard Kettlewell [Sat, 20 Nov 2010 16:04:41 +0000]
Reset z0 for Fixed_iterate test
Richard Kettlewell [Sat, 20 Nov 2010 14:58:21 +0000]
Comments
Richard Kettlewell [Sat, 20 Nov 2010 14:56:06 +0000]
WIP 8.56 mandelbrot iterator
Richard Kettlewell [Sat, 20 Nov 2010 13:53:39 +0000]
WIP 8.56 assembler support
Richard Kettlewell [Sat, 20 Nov 2010 12:59:56 +0000]
fixed multiplication (right this time), square roots
Richard Kettlewell [Sat, 20 Nov 2010 12:39:50 +0000]
division
Richard Kettlewell [Sat, 20 Nov 2010 12:22:04 +0000]
Correct multiplication
Richard Kettlewell [Sat, 20 Nov 2010 12:05:22 +0000]
WIP 8.56 fixed point in C
Richard Kettlewell [Sat, 20 Nov 2010 11:10:31 +0000]
--with-short-fixed-point configure option
Richard Kettlewell [Fri, 19 Nov 2010 21:59:07 +0000]
assembler syntax
Richard Kettlewell [Sun, 14 Nov 2010 19:19:11 +0000]
update docs
Richard Kettlewell [Sun, 14 Nov 2010 19:05:48 +0000]
Save JPEGs and PNGs as well
Richard Kettlewell [Sun, 14 Nov 2010 18:52:18 +0000]
width/height on command line
Richard Kettlewell [Sun, 14 Nov 2010 18:49:11 +0000]
draw a ppm from the command line
Richard Kettlewell [Sun, 14 Nov 2010 18:03:19 +0000]
Smooth coloring for the fixed point version. Cheats by converting
r^2 to double, so might get crufty at high magnifications?
Richard Kettlewell [Sun, 14 Nov 2010 17:57:32 +0000]
Rather less garish
Richard Kettlewell [Sun, 14 Nov 2010 17:53:06 +0000]
Smooth (if rather garish) coloring
Richard Kettlewell [Sun, 14 Nov 2010 17:39:56 +0000]
Color according to n-log2(log(r^2)). Still steppy. Fixed point
version missing.
Richard Kettlewell [Sun, 14 Nov 2010 17:25:23 +0000]
More count_t
Richard Kettlewell [Sun, 14 Nov 2010 14:05:28 +0000]
Preparatory work for smooth coloring.
Richard Kettlewell [Sun, 14 Nov 2010 13:37:06 +0000]
Make color functions templates so we can mess with 'count' type
conveniently.
Richard Kettlewell [Sun, 14 Nov 2010 13:29:25 +0000]
Move color selection into lib/