yuzu/FasTC
1
0
Fork 0
mirror of https://github.com/yuzu-emu/FasTC.git synced 2025-01-08 22:05:37 +00:00
Code Issues Packages Projects Releases Wiki Activity
658 commits 1 branch 0 tags 2.4 MiB
Commit graph

157 commits

Author SHA1 Message Date
Pavel Krajcevski 623f792904 Refactor 2013-10-02 11:19:36 -04:00
Pavel Krajcevski 0112f2cca7 Move the intensity calculation to the Pixel class... 2013-10-01 18:03:59 -04:00
Pavel Krajcevski 03f693f104 Add a hook into the PVRTexLib compressor if it's been found via cmake. 2013-09-30 12:38:09 -04:00
Pavel Krajcevski a31de8f797 Wrong indices for modulation values 2013-09-30 11:20:08 -04:00
Pavel Krajcevski dbc7798655 Add modulation image debug output. 2013-09-30 11:08:51 -04:00
Pavel Krajcevski baab69dc99 Fix some MSVC compiler snafus 2013-09-28 22:21:31 -04:00
Pavel Krajcevski 157295efa3 Squelch some compiler warnings. 2013-09-27 17:38:36 -04:00
Pavel Krajcevski dcf389d346 Merge PVRTC compressor into split library. 2013-09-27 17:30:16 -04:00
Pavel Krajcevski 8ea39890ac Update compressor to do a simple bounding box algorithm... results are still bad but better than what we've been getting. 2013-09-26 20:19:19 -04:00
Pavel Krajcevski 264e447e80 Deal with this bug once and for all. If we have an image in block stream order, then explicitly reorder it before doing any work. Then keep it that way. I probably could have fixed this in the amount of time I've wasted on it. -____- 2013-09-26 20:17:07 -04:00
Pavel Krajcevski 4de5f90edf Some small bug fixes to our still awful compressor. 2013-09-25 20:10:18 -04:00
Pavel Krajcevski a57c40005c Small bug: don't allocate more memory than we have to. 2013-09-25 20:07:44 -04:00
Pavel Krajcevski 2c8254d6c3 Small optimization: reserve vector data before we populate blocks vector. 2013-09-25 16:38:10 -04:00
Pavel Krajcevski 33a1dab0a9 The modulation value is the index. 2013-09-25 16:37:34 -04:00
Pavel Krajcevski 088481afe4 Meant resize instead of reserve. 2013-09-25 16:37:17 -04:00
Pavel Krajcevski a587899048 Fix some compiler errors with MSVC (and honestly, other compilers should catch them, too.) 2013-09-25 15:06:45 -04:00
Pavel Krajcevski ed12d53cd4 Merge in a few changes from master 2013-09-25 15:05:56 -04:00
Pavel Krajcevski f19b324d28 Make sure to use the proper directory when running RUN_TESTS from MSVC. Note, this doesn't work in Release. 2013-09-25 14:42:35 -04:00
Pavel Krajcevski 79d2ad79ac Fix some snafus on windows 2013-09-25 14:30:48 -04:00
Pavel Krajcevski c6d7bdc670 Very preliminary compressor 2013-09-24 20:35:36 -04:00
Pavel Krajcevski 8f4dcca4d7 Add some utility functions for manipulating the block data, including packing colors back into the 64-bit word. 2013-09-24 20:33:48 -04:00
Pavel Krajcevski 3de9abc908 Make PVRTC block size a constant accessible from the global header 2013-09-24 19:30:31 -04:00
Pavel Krajcevski 340f4f3141 Add pixel packing routines 2013-09-24 19:29:03 -04:00
Pavel Krajcevski 75e570ed16 Fix some compiler snafus 2013-09-24 17:46:09 -04:00
Pavel Krajcevski 1636232717 Make sure that when changing from higher to lower bit depth (i.e. perform a division) that we round to the nearest value 2013-09-24 17:45:19 -04:00
Pavel Krajcevski ea68f3a138 PVRTC image quality of life improvements. 2013-09-24 14:00:14 -04:00
Pavel Krajcevski e9a9988a56 Fix small bug in downsampling that causes artifacts 2013-09-20 12:48:56 -04:00
Pavel Krajcevski 1093447055 Plug in the infrastructure to start passing images to be compressed. 2013-09-20 12:43:59 -04:00
Pavel Krajcevski 995c237e5e Add downscale test and fix a few bugs and formatting. 2013-09-19 17:47:36 -04:00
Pavel Krajcevski 1d58ea2385 Add RGBA pixel constructor 2013-09-19 15:11:27 -04:00
Pavel Krajcevski 9f4fa671d9 Add a first pass at content aware downscaling. 2013-09-18 18:03:44 -04:00
Pavel Krajcevski e609075d04 Split apart the index calculation and pixel lookup functions 2013-09-18 18:03:08 -04:00
Pavel Krajcevski 16cc7f4a93 Use GetHeight/GetWidth() where appropriate instead of m_Width/m_Height 2013-09-18 18:02:35 -04:00
Pavel Krajcevski 4135e38f22 Set the default wrap behavior to wrap for image upscale, too 2013-09-18 18:00:59 -04:00
Pavel Krajcevski d4ca50f8ed Fix the directories in the PVRTC compressor 2013-09-17 18:35:56 -04:00
Pavel Krajcevski 29bd1368e6 Fix a few compiler warnings and add the BPTCEncoder license. 2013-09-15 14:56:09 -04:00
Pavel Krajcevski 28cf254fe5 Initial decoupling of base library from core library. Includes a few formatting changes as well. 2013-09-13 19:36:37 -04:00
Pavel Krajcevski 093576c9e7 Add initial code for our compressor. 2013-09-13 13:10:22 -04:00
Pavel Krajcevski 0afa1281a8 Change default wrap mode to wrap 2013-09-13 13:10:07 -04:00
Pavel Krajcevski 571bf1ebd0 Remove dead code 2013-09-12 15:01:30 -04:00
Pavel Krajcevski f7404c44b9 Refactor debug output for modulation info. 2013-09-12 15:01:19 -04:00
Pavel Krajcevski 1b073eb4b7 Add 2BPP decompression 2013-09-12 14:45:56 -04:00
Pavel Krajcevski b3de3f2c01 If our compile flag is set to output debug images with tests, then actually do it. 2013-09-12 14:45:42 -04:00
Pavel Krajcevski b3f8fcd454 Decouple upscale directions so that we can non-uniformly scale images. 2013-09-12 14:44:54 -04:00
Pavel Krajcevski 7184d49ccd Add unpack RGBA for our pixels 2013-09-12 14:41:00 -04:00
Pavel Krajcevski 9f6e6e7233 Fix bug with debug images 2013-09-12 14:32:33 -04:00
Pavel Krajcevski 1115c2f9e4 Add 2BPP helper functions for our blocks. 
Namely, there are two things that we need to do:
1. Figure out the sub-mode based on the mode bit and the structure of the modulation data. The comments in Block.h describe how we do this.
2. For a given texel index, return 2BPP texel modulation bits.
 2013-09-12 14:30:08 -04:00
Pavel Krajcevski 08cad3ba86 Add 2BPP tests 2013-09-12 14:04:20 -04:00
Pavel Krajcevski 0172df4e60 Add final opaque to transparent gradient test. 2013-09-10 13:34:13 -04:00
Pavel Krajcevski dd625fca71 When going from 344(3/4) pixels to 4555, the alpha channel only has a zero appended to it. Hence we must clear that bit for each pixel that was transparent during decoding. 2013-09-10 13:22:53 -04:00
Pavel Krajcevski f5e0aa4f9f Make sure to clear the MSBs of channels that have zero depth that get converted to higher depth values 2013-09-10 13:21:48 -04:00
Pavel Krajcevski 77bb7488cc Fix typo 2013-09-10 11:58:53 -04:00
Pavel Krajcevski eb312ccb0a Small refactor 2013-09-09 14:16:45 -04:00
Pavel Krajcevski 5020f81f00 Add a transparent pvr image test 2013-09-09 10:33:04 -05:00
Pavel Krajcevski 6c8eea5520 Add superfluous code from when I was trying to match the decoder to PVR library. 2013-09-09 10:31:53 -05:00
Pavel Krajcevski 09b5680245 Some small code style fixes 2013-09-06 13:37:50 -04:00
Pavel Krajcevski 5ac6872dc7 Add a test to make sure that after a bilerp the pixels that should remain unaffected do in fact remain unaffected. 2013-09-06 02:02:34 -04:00
Pavel Krajcevski 3806efe6fc Add some more hacky includes to be able to output debug images from our tests. 2013-09-06 02:02:02 -04:00
Pavel Krajcevski 91c79cd717 Fix Bilinear upscale 
There were a couple of fixes.

1. Add another set of pixels to an image that represent the fractional bits that get populated as a result of doing a bilerp.
2. Do not consider individual axes when doing the bilerp. Instead, assign a weight to each of the different texels that we will use for the bilerp. Then, combine them and save the fractional bits.
 2013-09-06 01:58:53 -04:00
Pavel Krajcevski 92827c23fc Add a debug function to our intermediate PVRTC images 2013-09-06 01:57:06 -04:00
Pavel Krajcevski 0d0c65f536 Actually index into our pixels properly... 2013-09-06 01:55:25 -04:00
Pavel Krajcevski 2263080faa Fix the interleaving by doing it when we're actually reading in the blocks 2013-09-06 01:55:06 -04:00
Pavel Krajcevski db914018f1 Two enhancements: 
1. Make sure to use the same lerping code for both modulation modes when decompressing.
2. Accept a flag during decompression that outputs the debug images that were computed during decompression.
 2013-09-06 01:53:10 -04:00
Pavel Krajcevski 7403e77d4d Add some hacky includes and libraries to support writing out debug images 2013-09-06 01:50:46 -04:00
Pavel Krajcevski a6e6138084 Change the lerp function to be more explicit. 2013-09-05 14:49:56 -04:00
Pavel Krajcevski b8e155e11c Switch red and blue channels in our decompressor test. 2013-09-05 14:49:08 -04:00
Pavel Krajcevski 7b72a84d87 Add a working directory for PVR lib tests 
Unfortunately, the WORKING_DIRECTORY directive for ADD_TEST seems to be broken as of CMake 2.8.10, fortunately there was a mail thread about this issue in 2.6 so I was able to implement the workaround.
 2013-09-04 20:43:49 -04:00
Pavel Krajcevski 6149c0af3a Add PVR tests 
If cmake was able to find Imagination's PVR library, then it'd be nice to make sure that our decompressor matches theirs. We incur a bit of overhead by adding .pvr files into git, but hopefully this cost will be beneficial to actually figuring out how things are working...
 2013-09-04 20:21:56 -04:00
Pavel Krajcevski 052157ea49 Rename our tests to start with the prefix 'Test' 2013-09-04 20:21:28 -04:00
Pavel Krajcevski abd4ba8b54 Abstract away some utilities that we can use across all of our tests, and add a test for a gray image. The test currently seems to be failing, which is likely due to a bug in our bit depth manipulation code. 2013-09-04 20:20:37 -04:00
Pavel Krajcevski 669b2c0934 Reverse the order for modulating our bits... I'm not convinced that this is right 2013-09-04 20:18:43 -04:00
Pavel Krajcevski 34ab853a62 Make sure to take into account the interleaved nature of the blocks when decompressing 2013-09-04 20:18:03 -04:00
Pavel Krajcevski bb525ae9dc If all of the PVR blocks are the same then we don't need so much code. 2013-09-04 15:28:51 -04:00
Pavel Krajcevski 9273625eeb Add initial decompressor test 2013-09-02 19:14:50 -04:00
Pavel Krajcevski 4bbd80aab2 Make sure to preserve bit depth when bilinearly upscaling 2013-09-02 19:14:31 -04:00
Pavel Krajcevski 3f4ffa61b5 Reverse byte order (and interpret the data as a 64 bit word) for PVRTC blocks 2013-09-02 19:13:53 -04:00
Pavel Krajcevski df81c73456 Default wrap mode to clamp 2013-09-02 19:12:43 -04:00
Pavel Krajcevski 5cde88d3b7 Add Image tests 2013-08-31 18:31:38 -04:00
Pavel Krajcevski 1d25e31d2a Bilinear upscale bug fixes. 2013-08-31 18:30:47 -04:00
Pavel Krajcevski 40b479487c Add accessors for image dimensions 2013-08-31 17:05:26 -04:00
Pavel Krajcevski 36641c2268 Don't set all of the memory to zero when we specify an image without any pixels 2013-08-31 16:43:32 -04:00
Pavel Krajcevski 992c72e4b5 Fix asserts 2013-08-31 16:40:49 -04:00
Pavel Krajcevski fa37aba768 Add comparison operator for pixels 2013-08-31 16:37:58 -04:00
Pavel Krajcevski 8bf682f04e Add comment for RGBA packing 2013-08-31 16:37:40 -04:00
Pavel Krajcevski b3a07e21f7 Add assignment operator 2013-08-31 16:37:07 -04:00
Pavel Krajcevski 26005bfd27 Disambiguate the Image.h from PVRTCEncoder and Core 2013-08-31 16:36:10 -04:00
Pavel Krajcevski 21199bc264 Add Pixel tests 2013-08-31 16:06:00 -04:00
Pavel Krajcevski 8c2127c08a When packing ARGB pixels make sure to convert them to 8-bit depth first 2013-08-31 16:05:50 -04:00
Pavel Krajcevski 1ffbdea2b8 Make sure to handle 0 bit depth a bit more gracefully 2013-08-31 16:05:28 -04:00
Pavel Krajcevski 9a0fbebac3 Fix bug when going from smaller to higher bit depth 2013-08-31 16:04:58 -04:00
Pavel Krajcevski ad17404bf7 Begin counting the bits from the beginning of the byte stream. The way I was doing it before was silly... 2013-08-31 14:33:44 -04:00
Pavel Krajcevski f280c2bb33 Fix some compiler warnings. 2013-08-31 13:30:12 -04:00
Pavel Krajcevski 53a8c8e3cd Fix bug where the defaulted value caused a crash (wtf?) 2013-08-30 19:08:53 -04:00
Pavel Krajcevski 012d9e3ce0 Add preliminary tests for PVRTC pixels 2013-08-30 19:08:20 -04:00
Pavel Krajcevski ab0dbbfa6e Bugfix 
The bug was that bitIdx was not being taken into account when we realized that there was enough pixel depth to stay within the current byte when reading pixel values.
 2013-08-30 18:45:14 -04:00
Pavel Krajcevski 1cc8f57538 Add block tests 2013-08-30 18:44:50 -04:00
Pavel Krajcevski 90662a56cd First pass at PVRTC Decompressor 2013-08-29 18:35:15 -04:00
Pavel Krajcevski c59deae728 Add accessors for the color channels and a method for packing the information into a 32 bit RGBA int 2013-08-29 18:34:53 -04:00
Pavel Krajcevski f822c15868 Pixel depth defaults to 8 2013-08-29 18:34:25 -04:00
Pavel Krajcevski 73f4992081 If we want a color channel with zero depth then it should default to full intensity, but we will remember that there is zero depth there. 2013-08-29 18:33:56 -04:00