Dolphin SVN r6790

Revision 6790:

Worked around the Norwegian Bokm(?/aa)l issue.
Enhanced the gettextize script to update mo files for windows and osx. This script should be run any time that a new language is added (don’t for get to svn add the Data/Languages language subdirectory), and on occasion to update the language files with strings from the source code.
Added the Hebrew mo file for windows and osx (and update the others).

Revision 6789

TextureDecoder: Some misc clean ups. Backport code to SSE2 version. Remove redundancy in RGBA8 (5% speedup).

Revision 6788

Make langs in abc order
Playing a bit with hebrew


r6790 x86 – скачать, зеркало

r6790 x64 – скачать, зеркало

Dolphin SVN r6787

Revision 6787:

Go back to using FindGtk for the cmake build. The module is fixed with later versions of cmake. We may need to check the cmake version and fall back to pkg-config for those using older versions. Please test.

Revision 6786

Added some languages:
Added Norwegian Bokm?l thanks to KHRZ
Added Polish thanks to Baszta
Added Russian thanks to Kein

Also cleaned up the appearance of the GFX config dialog a bit, fixed a minor issue in the cmake build, and fixed the nowx build.

Revision 6785

Fixed language selection by adding English, leaving default as «» language. Removed languages which don’t have translations yet.
As side-effect, changed GameListCtrl sorting to use the configured IPL/GC language instead of the UI language, which is misleading with missing translations, and possibly wrong if we ever get translations that are not part of the GC ones.
In case we want the old behavior back, revert GameListCtrl.cpp and ISOProperties.cpp.

For translators: use either «» if your operating system is the same language as the one you’re translating for, or replace one of the others. Someone will add them to the list when they are committed.


r6787 x86 – скачать, зеркало

r6787 x64 – скачать, зеркало

Dolphin SVN r6784

Revision 6784:

Refactor all the SSSE3 functions in TextureDecoder so that the cpu_info check isn’t looped over. Speeds up most textures dramatically (where it has previously slowed them).

Revision 6783

New SSSE3 implementation for I4 texture decode. 14% speedup over the previous SSE4 implementation (so it was scrapped).

Revision 6782

Add profiling (via oprofile) to the cmake build.

Revision 6781

From my last commit: Fix build on Linux. Use SSSE3 instead of SSE3.

Remove some unused vars from the SSE2 CMPR.

Revision 6780

fix for a error introduced in my last commit.

Revision 6779

New SSSE3 implementation of RGB5A3. About 40% improvement (less cycles) on the plain C version and 17% on the SSE2 version.


r6784 x86 – скачать, зеркало

r6784 x64 – скачать, зеркало

Dolphin SVN r6778

Revision 6778:

Fix alpha reads (hurray for the ZWW Pictograph quest and Killer 7)

Revision 6777

normalize the efb to texture process for color textures to make it work the same in all the plugins and with the same accuracy as real hardware (almost :))
please test for regressions and fixes.
some little changes to make pixel shader more dx9 sm2.0 friendly. the condition is not to use pixel lighting ( sorry no hardware support for the quantity of parameters needed).

Revision 6776

This needs to be in the right place to work for Скачать:

r6778 x86 – скачать, зеркало

r6778 x64 – скачать, зеркало

Dolphin SVN r6767

Revision 6767:

Use CPUDetect.h and bSSSE3 to fix «unknown machine instructions» on DX9 plugin.

Revision 6766

~93% speed improvement (over plain C++) in RGBA -> BGRA conversion for processors that support SSSE3 (Core or newer).
About 40% faster than the SSE2 implementation.

May not be required in future but good for reference.

Revision 6765

45% speed improvement in RGBA -> BGRA conversion over reference C code. This might turn out to be unnecessary once I refactor the CreateTexture2D/ReplaceTexture2D to call the texture decoder instead of vice-versa and have the SSE2 code convert directly to BGRA and write directly to the D3D allocated texture.
Need to investigate source texture alignment and where to introduce that, if it is possible.

Revision 6764

Add anvideo config option to automatically resize the render window to the size of the game’s output resolution. This avoids artifacts that appear when the render target is scaled to fit a window of a different size.


r6767 x86 – скачать, зеркало

r6767 x64 – скачать, зеркало

Dolphin SVN r6763

Revision 6763:

Added a few translatable strings.
Added German language translations. Thanks to JackyCola.
Added Italian language translations. Thanks to RebuMan.

Revision 6762

Missed one MSVC-ism. Should fix build for Linux. Last revision should still work for Windows. No functionality changes this time.

Revision 6761

Faster SSE2 optimized GX_TF_CMPR texture decoder which gets ~40% speed improvement on x64 and ~50% improvement on x86 as compared to reference C code.
The code now uses direct pointer access from C code to write the colors to the destination texture instead of trying to force them back up into an __m128i and a single write call. This is what produces the major speed-up.

Revision 6760


* Fixed a bug on Texture Converter when Copy EFB to RAM’s cache is disabled, so…

* …accordingly, I revert my latest commit because now it becomes useless.

Revision 6759

— add some more strings for translation, change others for easier translation (in case anyone knows a better way than _(«a») + wxString(wxT(«b»)), feel free to fix it).
— removed strings that shouldnt be translated.
— added gettextize script from glennrics to generate the .pot file; this excludes debugger-related strings which are usually not used by the end user anyways (again, less stuff to translate).


r6763 x86 – скачать, зеркало

r6763 x64 – скачать, зеркало

Dolphin SVN r6758

Revision 6758:

Add an English translation file (with no translations). For some reason OSX seems to need it. Shouldn’t affect other platforms.

Revision 6757

a little cleanup of dx9 depth conversion code that r6751 made unnecessary, and a question about D3DFMT_D24X8 checks, and (most importantly) spelled Stretch right in video config

Revision 6756


Revision 6755

Enabled SSE2 optimization of GX_TF_CMPR decoder only for x86 builds. It can’t compete with the x64 optimized reference C code.

Revision 6754

improve z copy a little further in 4 bits copy,
fix and error introduced by my last commit in dx11 plugin


r6758 x86 – скачать, зеркало

r6758 x64 – скачать, зеркало

Dolphin SVN r6753

Revision 6753:

Add some more translatable strings. Thanks to DavidVag.

Revision 6752

corrected a little error in my last commit

Revision 6751

corrected depth copy shader in dx9/dx11 to improve accuracy when doing efb to texture for depth formats

Revision 6750

Oops, forgot to remove that line.

Revision 6749

Install .mo files on osx as well.

Revision 6748

Disable the Languages target on windows 32 bit builds.


r6753 x86 – скачать, зеркало

r6753 x64 – скачать, зеркало

Dolphin SVN r6747

Revision 6747:

Add internationalization support. There aren’t really any translations yet other than a little French thanks to DavidVag. So now the translators need to go to work. See for directions. The binary .mo files are generated at build time on linux. They are provided in Data/Languages for windows. I don’t know where they need to go on OSX, but they should be able to be generated there at least. I added a target to generate them on windows, but does not build by default as it requires that msgfmt.exe from the gnu gettext tools be installed and in the path.

Revision 6746

Changed order of operations in Profiler.cpp to calculate non-zero values for ‘cost’. Integer division was too early. Credit to my brother, Joe, for figuring this one out.

Revision 6745

Fixed Linux build.
Fixed small undiscovered bug in WII_IPC_HLE_Device_FileIO.cpp when looking at 0-length strings.

Revision 6744

Fixed S3TC DXT1 decoder implementation. I see ~40% speed improvements running on x86 Intel hardware and 0% improvements running on x64 AMD hardware. Strange. More investigation to follow!

Revision 6743

After removing the input queueing in I was still seeing
the occasional HLE wiimote disconnection, although nowhere it was at
near the level before both the recent wiiuse integration and adding
the queue in the first place.

The callback-based bluetooth input method on OS X really requires
that packets be accepted immediately and the 1ms polling done by the
wiimote worker threads isn’t quite good enough.

So just play along with the native OS X model and send the packet up
our stack directly from the callback. With our current API, this is
kind of a layering violation, but it seems less messy than either
putting back internal queueing in or adding the platform
support for higher-frequency polling than what usleep(3) offers.

Perhaps we should just change our own API to a callback model. While
this makes no real difference right now, it would make it a fair bit
cleaner to accept new wiimotes at any time instead of having to click
Refresh, although with the recent realmote changes, that feature can
probably just be hacked in by looping around FindWiimotes().

For other platforms having problems with HLE disconnections caused by
dropped bluetooth packets, as a first stop I would suggest trying to
do more than just one wiimote->Read() per iteration in the thread.

This change also conveniently sidesteps another new bug
where the discovery and I/O code would clash during a refresh because
they use the same default CFRunLoop, so if we switch back to using a
run loop for I/O, must remember to create a separate one.


r6747 x86 – скачать, зеркало

r6747 x64 – скачать, зеркало