6f1a63fe89Cocoa Port: Support Automatic Graphics Switching, allowing the use of the integrated GPU to run video display windows. (Requires a dual-GPU MacBook Pro with Metal-capable GPUs.) - While this results in a 20% performance loss for video display window functions (like video filters), this also dramatically extends battery life. To return to the old way of using the discrete GPU for video display windows, the user must disable Automatic Graphics Switching in their System Preferences. - The Troubleshooting Window now correctly reports the emulated 3D renderer that is currently active, rather than the one that is selected in the GUI. - Also fix a bug where creating a OpenGL 3D renderer's context would immediately fall back to Apple Software Renderer if context creation failed. Now, context creation falls back to Apple Software Renderer as the last resort, only after all other Core Profile contexts have failed.
rogerman
2025-03-26 16:55:25 -0700
25356b70f1GPU: Add some more optimizations hints in CopyLineExpand() for SSSE3.
rogerman
2025-03-26 13:26:42 -0700
75766495f9Cocoa Port: Add some missing file references to the Xcode 3 project.
rogerman
2025-03-20 22:55:40 -0700
1091e69726Cocoa Port: Fix a bunch of bugs related to moving windows between Retina and non-Retina displays. (Related to commit 2bc5b0d.)
rogerman
2025-03-19 03:11:16 -0700
f6e0feb13eCocoa Port: Report which OpenGL renderer is actually selected for 3D rendering. (Debug mode only.)
rogerman
2025-03-18 15:55:01 -0700
33754b706eCocoa Port: Add support for performing OpenGL 3D rendering on the 2013 Mac Pro's second GPU. - For OpenGL 3D rendering, this yields a performance improvement of 2% - 4%.
rogerman
2025-03-17 17:03:35 -0700
50e39989f7GPU: Do some small optimizations to SSE2 and AVX2.
rogerman
2025-03-16 23:15:46 -0700
c73a7ffe53GPU: Add NEON-accelerated functions for 2D layer compositing. (For 64-bit ARM CPUs only.) - This improves GPU performance by up to 20% on the Raspberry Pi 5, and up to 50% on Apple Silicon CPUs.
rogerman
2025-03-16 16:23:21 -0700
e2379a66d6matrix.h: Fix const-correctness for some *_fast function parameters. - In practice, this only affected compiling for NEON on certain compilers. Other SIMD ISAs should remain unaffected.
rogerman
2025-03-16 16:13:12 -0700
Merge pull request #886 from Anonym271/master
zeromus
2025-02-11 16:55:35 -0500
06bef8b560Bugfix: MemROMReaderWrite was not working for 1 byte length because it would copy the current mem's byte into the source buffer instead of the opposite way.
Anonym271
2025-02-11 22:45:25 +0100
9f30151675types.h: let mingw use regparm(3) calling convention for jit
rofl0r
2024-12-06 18:17:46 +0000
18f166bc97rename FASTCALL to DESMUME_FASTCALL
rofl0r
2024-12-06 18:03:16 +0000
thesource