yuzu/FasTC
1
0
Fork 0
mirror of https://github.com/yuzu-emu/FasTC.git synced 2025-01-08 06:25:31 +00:00
Code Issues Packages Projects Releases Wiki Activity
596 commits 1 branch 0 tags 2.4 MiB
Commit graph

596 commits

This branch
This branch
All branches
Author SHA1 Message Date
Pavel Krajcevski 1f45a49102 Caught small bug 2014-09-25 19:10:30 -04:00
Pavel Krajcevski c0d7607998 Merge branch 'master' into DecompressASTC 2014-09-25 16:50:43 -04:00
Pavel Krajcevski 502c15ee29 Respect the new flag in tests. Fixes #12 2014-09-02 16:52:31 -04:00
Pavel Krajcevski e59541e30a Add #define guard to including IO with PVRTC 2014-09-02 16:39:11 -04:00
Pavel Krajcevski 995eacd5b6 Fix small crash bug 2014-08-26 13:44:09 -04:00
Pavel Krajcevski edfc6bde78 Be a little more defensive about calculating SSIM 2014-08-26 13:34:35 -04:00
Pavel Krajcevski 13df3c72c6 Apparently the block layout is different for ASTC... 2014-05-15 22:02:16 -04:00
Pavel Krajcevski 374f5375d5 Fix the way we quantize 16-bit pixel channels 2014-05-15 22:02:03 -04:00
Pavel Krajcevski 2d52f1eabb Actually advance the color values ptr 2014-05-15 22:01:21 -04:00
Pavel Krajcevski 31e01b7b10 Partition index does not include the number of partitons. 2014-05-14 13:29:12 +02:00
Pavel Krajcevski a8514ba9b3 Reverse the bits of the ASTC block in order to read the texel weight data... 2014-05-08 00:45:23 +02:00
Pavel Krajcevski 1269ab1314 Fix a few bugs dealing with color blending 2014-05-07 23:00:42 +02:00
Pavel Krajcevski a58501f56f Fast track zero partition selection 2014-05-07 22:59:49 +02:00
Pavel Krajcevski d13cca3362 The basename default should be the filename itself 2014-05-07 22:58:32 +02:00
Pavel Krajcevski c8e19252e3 Fix another bug that I thought I fixed... I should really make integration tests 2014-04-22 19:52:59 -04:00
Pavel Krajcevski f47c880198 Make sure to use the proper delete 2014-04-22 19:28:49 -04:00
Pavel Krajcevski c40829a3b7 Fix some compiler warnings 2014-04-15 14:18:12 -04:00
Pavel Krajcevski 03b9e7f413 Merge branch 'ModularizeBPTC', remote-tracking branch 'origin' 2014-04-10 13:50:58 -04:00
Pavel Krajcevski 8660b24ffe Fix multithreaded npot bug. 
There was a bug where certain blocks were not being written due to
improper iterating over non power-of-two textures depending on the
thread that got split across a newline of blocks. This should now
be fixed.

The offending command line arguments looked something like:

CLTool/tc -q 0 -t 32 -j 32 /path/to/npot-texture.png
 2014-04-10 13:39:04 -04:00
Pavel Krajcevski 127c825337 Actually clamp the initial value 2014-04-09 13:55:27 -04:00
Pavel Krajcevski ea1c61913a Small patch to fix loading of non-multiple-of-four images 2014-04-02 13:59:08 -04:00
Pavel Krajcevski 640b098af7 Small formatting change 2014-04-02 13:57:17 -04:00
Pavel Krajcevski ed2bcc3838 Fix some small bugs to hook into NVTT 2014-03-28 11:40:52 -04:00
Pavel Krajcevski fabf6830ee Fix bug so that we at least generate *some* endpoints 2014-03-28 11:40:38 -04:00
Pavel Krajcevski 2141d71dd0 More perf fixes 2014-03-26 18:11:12 -04:00
Pavel Krajcevski 1bfda210c0 Merge from origin 2014-03-25 18:28:05 -04:00
Pavel Krajcevski 7fa561f67f Formatting. 2014-03-25 18:27:26 -04:00
Pavel Krajcevski 6505944f57 Cache the pixel computation for the cluster upon creation 2014-03-25 16:24:30 -04:00
Pavel Krajcevski 4128bcf073 Template the number of buckets during error calc so the compiler can unroll... 2014-03-25 16:24:08 -04:00
Pavel Krajcevski 6012b53cdb Formatting 2014-03-25 16:23:04 -04:00
Pavel Krajcevski 7af6cca4d5 Get rid of float conversion lookup table 2014-03-25 16:21:17 -04:00
Pavel Krajcevski 663caada50 Generalize BPTC compression. 
1. Split compression parameter generation and compression parameter
packing. This gives a good performance boost, since we don't pack every
single time we compress. The error is computed each time, and only the
best parameters are packed.

2. Allow the shape selection function to specify up to ten shapes to
try for compression. We were already doing this kind of hackily where
we allowed both a three and two partition shape. This makes it a little
cleaner and exposes it to the user.
 2014-03-25 11:40:06 -04:00
Pavel Krajcevski f951dda449 Fix bug caught by clang 2014-03-23 19:52:51 -04:00
Pavel Krajcevski aba347778e Fix a few compiler warnings 2014-03-23 17:48:17 -04:00
Pavel Krajcevski 5725220ba6 Merge branch 'master' into ModularizeBPTC 2014-03-23 17:41:46 -04:00
Pavel Krajcevski d03732fc09 Move BPTC shapes header to include folder 2014-03-22 21:17:46 -04:00
Pavel Krajcevski 671454f57d Some small formatting changes 2014-03-22 20:40:28 -04:00
Pavel Krajcevski 5bd306d09b Actually use the simulated annealing setting. 2014-03-22 20:07:17 -04:00
Pavel Krajcevski 220a736a36 Move the other BPTC settings into the settings struct 2014-03-22 19:52:58 -04:00
Pavel Krajcevski 9144db4de6 Actually pass block coordinates to shape selection function 2014-03-22 19:25:21 -04:00
Pavel Krajcevski 891e2cfee8 Formatting 2014-03-22 19:24:51 -04:00
Pavel Krajcevski 9f259744de Get rid of comment 2014-03-21 20:36:54 -04:00
Pavel Krajcevski e936cce0cb More refactoring. 
Change RGBACluster to be a class that only really persists once per block.
When we switch shapes and do operations on them, then we really only need
to change which points in the block are accessed. We don't need to do this
very often, so just change the mask whenever we need it. This brings us back
closer to our original performance, but we're still not where we were when
we started refactoring.
 2014-03-21 20:27:00 -04:00
Pavel Krajcevski fe69dc9fb5 Fix small bug with large consequences 2014-03-21 20:26:45 -04:00
Pavel Krajcevski dcc12eb28a Intel compiler hates floats 2014-03-21 20:25:59 -04:00
Pavel Krajcevski cf937f2ad3 Refactor shape and mode selection 
We suffered another performance hit. This time it comes from the fact
that we're copying around a lot of data based on what partition we're
choosing. We can get rid of this a tad by only copying the data that we
need once and then using getters/setters that selectively pull from
an array based on our shape index.
 2014-03-21 18:02:02 -04:00
Pavel Krajcevski 26e816b3db Add settings for BPTC compression 2014-03-21 12:45:47 -04:00
Pavel Krajcevski 6954d7b154 Refactor RGBAEndpoints 
Changed the RGBAEndpoints to use the vector/matrix classes in
FasTCBase. This caused a ~20ms performance hit on an 8-core machine
which is likely due to the compiler having difficulty compiling away
some procedure call overheads. Upon profiling, the biggest bottleneck
is still by far the QuantizedError function, so any and all further
optimization should be focused on that.
 2014-03-21 01:21:07 -04:00
Pavel Krajcevski 2b17cf6f97 Add scalar-matrix multiplication 2014-03-21 01:17:19 -04:00
Pavel Krajcevski 63b8744917 Pull out multiplication routines so that they can be specialized if need be 2014-03-21 01:16:45 -04:00