c59829ad86
Fix buffer overflow
2011-12-10 16:23:24 -05:00
411d8b6263
Be more conservative in running overflow check
2011-12-10 12:42:40 -05:00
2e4e525d49
Use atomic by default (I'm sold)
2011-12-10 12:32:12 -05:00
6c50e6dadc
New atomic write mode
2011-12-10 12:18:00 -05:00
a0fd3f965f
More template names
2011-12-10 12:16:42 -05:00
a509e58b43
Fix overflow-related bug (hopefully)
2011-12-09 18:46:40 -05:00
c5da1efc74
Lockless lossy shared memory writeback.
...
Barely tested! And yet it's going straight into master. Lucky you!
2011-12-09 16:13:23 -05:00
6bac3b3a95
Use reordered, lossy bit handling
2011-12-09 14:14:36 -05:00
b592cda3db
Fix potential bug in binsearch
2011-12-09 14:14:00 -05:00
d3ee6f36c2
Flat (pre-packed int) palettes in deferred mode.
2011-12-08 20:55:07 -05:00
b76208078f
Deferred works again. Time to break it.
2011-12-08 15:28:10 -05:00
e106524701
Fix code in comment
2011-12-08 13:24:10 -05:00
e006ec1915
Slight tweak to clamp thresholds
2011-12-08 12:13:58 -05:00
b73461132c
Use consts for image size instead of immediates.
...
This saves us from having to recompile if the frame size changes.
2011-12-08 12:07:22 -05:00
084a65c615
Go back to using global accum for now
2011-12-08 11:50:46 -05:00
df8100d1f4
Use new mad.cc instruction in MWC
2011-12-08 11:49:31 -05:00
094df0ae21
Name the variation templates for debugging
2011-12-08 11:48:38 -05:00
e79d9a58fd
Fix cschden, cothden variations
2011-12-07 13:41:45 -05:00
5ce5763da7
changed sense (and name) of values for affine xforms to match smoulder renders.
2011-11-23 07:54:48 -07:00
22fdc98128
Fix point swapping.
2011-11-20 10:08:13 -05:00
efd261bd5b
fixes related to interpolation of palettes; hsv interpolation now goes
...
the 'short way' around the hue circle, and the correct palette is now
chosen when > 2 palettes are present in the knots.
2011-11-14 19:12:41 -07:00
0f848b8bb8
Dither color when packing for deferred write.
2011-11-12 11:06:44 -05:00
6d1c81486c
Don't inline catmull_rom for much faster compiles.
2011-11-12 11:05:44 -05:00
24c0c8ee56
Fix some color foibles (more yet remain)
2011-11-12 10:42:02 -05:00
9ef5363652
Fix dumb overflow bug
2011-11-11 17:54:33 -05:00
eb43b151dc
Deferred writeback.
2011-11-11 17:37:27 -05:00
05e1d08681
Add -1-skipping to sort.
2011-11-11 17:34:43 -05:00
54f411878b
Experiments with multi-pass sort (still has bugs)
2011-11-10 10:49:35 -05:00
13842196ea
Generalize the sort.
2011-11-09 12:00:59 -05:00
3147fd40d2
Support CUDA 4.1. Split filtering into new module.
...
The new toolkit generates code for filtering which uses too many
registers, so this change splits filtering into its own module so that
it can have separate register usage limits during compiling. As a bonus,
this should improve startup time in general, since the filtering code
is now fixed and does not need to be recompiled.
2011-11-08 14:38:45 -05:00
cea91d75bf
A very fast key-only radix sort.
2011-11-07 23:23:20 -05:00
7815c13ba4
Fix camera offset WRT flam3
2011-11-06 10:01:26 -05:00
5179c98254
fixed flawed lazysusan variation. added whorl variation.
2011-11-03 13:31:32 -06:00
3badf0f826
Merge branch 'master' of git://github.com/stevenrobertson/cuburn
...
Conflicts:
cuburn/code/filtering.py
2011-11-03 13:27:31 -06:00
8ea057ff96
fixed highlight_power functionality difference between cuburn and smoulder
2011-11-03 13:18:43 -06:00
d382e0f14a
Fix highlight power
2011-10-31 17:20:13 -04:00
b240fc8459
Use custom "cross" filtering.
...
Sobel was giving too many false positives. This cross seems to detect
the kinds of edges we care about and avoids the rest of the image, and
it does so on pretty much everything I've tried it on. Very satisfying.
2011-10-29 17:36:06 -04:00
0936e34b88
Fix cards stalling out on degenerate flames
2011-10-29 11:14:11 -04:00
6204f36ebc
Fix spline derivative calculation.
2011-10-29 10:51:33 -04:00
bfff915b7e
Two very obvious spline bugs fixed.
2011-10-28 21:34:42 -04:00
28e73d08ee
Add derivative support to SplWrap.
2011-10-28 18:51:33 -04:00
a2c4c90cb2
Upgrade fuse, because, well, why not?
2011-10-28 08:41:20 -04:00
a6177edc0d
Drop the RNG mult shuffle.
...
It's surprisingly time-consuming, and until I have data showing that we
need it, I'm going to leave this bit of extra randomness voodoo out.
2011-10-28 08:36:27 -04:00
185823ad55
Rearrange the main render loop... again.
...
Using one stream with two pagelocked host buffers allows us to keep the
GPU work queue full without pegging the CPU, and also reduces the
incidences where a host buffer will get overwritten before it can be
written. devtid() was flaky, so this patch also introduces a ringbuffer
to handle the 'slots' concept. It also introduces an adaptive number of
temporal samples, which improves efficiency but also killed the
assumption that (ntemporal_samples % 256 == 0), which required some
additional fixes.
2011-10-28 08:30:36 -04:00
15f88383b1
Experimental: real Sobel gradient detection
2011-10-28 08:25:00 -04:00
9b2b3ba011
Fix buffer overrun in filtering
2011-10-28 08:24:16 -04:00
6b2cb024ac
Expand max filter radius to 21 pixels
2011-10-28 08:23:33 -04:00
f3a79b200c
New badvals mechanism.
2011-10-27 12:59:58 -04:00
cac9b691a8
Add a missing semicolin in disc2
2011-10-27 10:37:12 -04:00
77daf5e639
Correct blur radius after Box-Muller
2011-10-27 10:36:44 -04:00
