Commit Graph

6193 Commits

Author SHA1 Message Date
rogerman 37afaefa2f OpenGL Renderer: Replace the accuracy/performance tradeoff "Enable Depth Equals Test Tolerance" with "Enable NDS-Style Depth Calculation", where disabling this option allows the host GPU to natively calculate depth which significantly improves performance in many games.
- New Behavior: In addition to emulating the existing Depth Equals Test Tolerance, NDS-Style Depth Calculation accounts for all NDS depth calculations within the fragment shader. Most notably, disabling this option forgoes the W-depth / Z-depth differentiation that the NDS uses, instead preferring the GPU's native Z-depth calculation. Using the GPU's native depth calculation significantly improves performance, but many games use W-depth calculations or are sensitive to subtleties in the Z-depth calculation, and so this option must remain ON by default for compatibility's sake.
- Also fixes a shader initialization issue on the Windows port. (Regression from commit 7080e21.)
2018-12-18 10:50:41 -08:00
rogerman 7080e2156b OpenGL Renderer: Rework the rendering shaders so that the shader program code is more dynamically generated. This may yield some performance improvement for certain 3D rendering cases, especially when running on lesser GPUs with fewer and/or slower shader execution units. 2018-12-17 16:16:50 -08:00
rogerman ae8fb2c3bb GPU: Fix a bug where using VRAM as a display capture source would sometimes cause graphical glitches under certain conditions. (Regression from commit 2c6a5f9.) 2018-12-17 15:33:16 -08:00
zeromus 88d930ce82 winport - fix loading files named things like Splookékrong from commandline, fixes #238 2018-12-15 17:26:28 -05:00
rogerman 7ff5c5eece Render 3D: Improve the overall rendering accuracy of Edge Mark. Most notably, Edge Mark now properly renders at screen edges. As of now, the current algorithm is as accurate as its ever going to get under our current 3D rendering engine. 2018-12-14 17:08:16 -08:00
rogerman e6d6f2e10d Cocoa Port: Fix a bug where clipboard copies and screenshots taken from Metal display views will cause the image to be Y-flipped. 2018-12-14 15:41:05 -08:00
rogerman 2c6a5f9868 GPU: Do some code cleanup of the display capture code. 2018-12-12 18:34:47 -08:00
zeromus e604631413 winport - fix things named like Blorkénflarge in the recent roms menu (#238) 2018-12-12 15:50:52 -05:00
rogerman 8c2379f6f8 Firmware: Fix various endianness issues. Most importantly, this fixes a bug with touch input not working correctly on big-endian systems. (Regression from commit bb38022.) 2018-12-12 02:49:44 -08:00
rogerman 471f53e506 Cocoa Port: Fix various issues on the PPC build.
- Fix compiling issues for big-endian systems.
- Fix bug where the Recent ROMs menu and also launching the app while loading a ROM file would fail to load the ROM on macOS v10.5 Leopard.
- Fix bug where GPU main memory display mode would show incorrect pixels on big-endian systems when running at 15-bit color depth.
- As an unintended collateral improvement, GPUEngineA::_HandleDisplayModeMainMemory() now has SSE2-accelerated versions for 18-bit and 24-bit color depths. This was done less for its performance benefit (main memory display mode is an extremely rare feature) and more for better code consistency and code completeness.
2018-12-11 17:45:36 -08:00
zeromus e56059872f winport - fix loading games named things like "Yokémorp". It was probably only open through drag and drop.
Probably broke japanese. If I did, write a bug so I can fix japanese and break latin characters again.
2018-12-11 18:20:39 -05:00
zeromus b5477b608b
Merge pull request #236 from NetwideRogue/master
don't clobber existing screenshots
2018-12-06 08:08:40 -06:00
Declan Hoare a3eebbac21
don't clobber existing screenshots 2018-12-06 23:59:18 +11:00
rogerman 35e834ff2c GFX3D: Revert the polygon sorting code back to its original state, which should result in a minor performance improvement for high polygon-count scenes.
- After years of testing, no one has reported running into the assert in gfx3d_ysort_compare() so I think we should be safe in reverting std::stable_sort() back to std::sort().
- For the sorting function, use gfx3d_ysort_compare_orig() since this function compiles down to fewer instructions than gfx3d_ysort_compare_kalven() does, resulting in better sorting performance.
- Of note, I'm pretty sure that SF commit r5132 is what fixed the original bug (see SF#1461 for more details) by getting rid of the NaN comparisons that were tripping up std::sort(). In the future, we should research why we're dividing by 0 in the first place, since r5132 is clearly a hack of a fix.
2018-12-05 14:37:33 -08:00
zeromus d80a84b762
Merge pull request #207 from cosmo-ray/fix-linux-gcc8.2-warnings
Fix gcc8.2 warnings on linux
2018-12-05 14:10:52 -06:00
zeromus 9ea1b5cbda
Merge pull request #223 from intact/gtk-fix-screenshot-path
Gtk+ Port: Use Desktop or Home as fallback directory for screenshots
2018-12-05 14:10:42 -06:00
rogerman 355e4a0fb4 OpenGL Renderer: Remove a now defunct framebuffer texture, significantly reducing VRAM usage at the higher resolutions. 2018-12-04 21:49:09 -08:00
rogerman 3d573e150f OpenGL Renderer: Properly clear the framebuffer during a power-off condition, just like how SoftRasterizer does it. (Related to commit 66b5da1 and commit 759a039. Fixes #234.)
- Also do a minor performance optimization by only doing the framebuffer clear once for each power-off condition, rather than repeatedly and unnecessarily clearing the framebuffer for each and every V-blank.
2018-12-04 21:23:58 -08:00
rogerman df22c6e14d Cocoa Port: Fix some intermittent issues related to launching the app while loading a ROM file (i.e. double-clicking an NDS ROM file to launch DeSmuME.app). 2018-12-04 01:26:27 -08:00
rogerman b9a8bafe8b GPU: Do some code cleanup of the display capture code. 2018-12-03 16:22:12 -08:00
zeromus fe93b70de8
Merge pull request #233 from Jules-A/hideConsole
[Windows] Hide the console window in the current session when disabling in menu.
2018-12-02 11:00:52 -06:00
Jules.A 157717a61f Hide the console window before freeing so if the console is attachted to another process (currently occuring under Win10) it is at least hidden. 2018-12-03 00:52:09 +08:00
rogerman c3614a7e95 GPU: Do some minor code cleanup. 2018-12-01 21:56:32 -08:00
rogerman fb8d937239 GPU: Fix bug where GPUEngineA::RenderLine_Layer3D() was trying to run with uninitialized values. (Regression from commit 37a8ca0.) 2018-12-01 21:38:03 -08:00
rogerman 37a8ca0983 GPU: Do a bunch of minor tweaks and code cleanup to the various pixel compositor methods. Significantly reduces the compiled code size.
- Of note, when running at custom resolutions, we are now being more aggressive in performing early tests for rejecting pixels as soon as possible. This may yield a minor performance improvement in some very specific rendering scenarios that require the window test.
2018-12-01 20:44:45 -08:00
rogerman 6a1d9e4848 GPU: Rendering complete OBJ layer lines is now SSE2-accelerated at the native resolution. This change is less of a performance enhancement and more of improving the code consistency. As of now, ALL complete OBJ layer lines, whether internally generated or from read from VRAM, whether rendering at native resolution or custom resolution, should now be SSE2-accelerated. This commit finalizes this concept. (Related to commit 8e9e7c4 and commit 60c01bd.) 2018-12-01 15:46:23 -08:00
rogerman 60c01bd63a GPU: Do a few more minor optimizations to rendering complete OBJ layer lines. Most notably, all complete OBJ layer lines, not just ones reading directly from custom VRAM, now benefit from the SSE2-accelerated code. 2018-11-30 20:12:58 -08:00
rogerman 9a53e8be69 GPU: Fix bug in GPUEngineBase::_PixelUnknownEffectWithMask16_SSE2() where blending effects for OBJ layers were being handled incorrectly. This bugfix only affects SSE2-enabled systems. (Regression from commit 8e9e7c4. Fixes #232.) 2018-11-30 17:10:56 -08:00
rogerman 2c5c2f6186 GPU: Use the same technique in the commit 6bcd19b GPUEngineBase::_CompositeVRAMLineDeferred() bug fix in order to do a tiny optimization to GPUEngineBase::_CompositeLineDeferred(). Also makes the code more consistent as well. 2018-11-29 22:25:37 -08:00
rogerman d0330fc96e Cocoa Port: Upgrade Interface Builder .xib files to 3.2 format with minimum deployment target of macOS 10.6. This change effectively drops support for building DeSmuME directly from a PowerPC Mac. Fixes #231.
- It is still possible to create a PowerPC binary, but this now requires some extra steps. From now on, you must use an Intel Mac running Mavericks or earlier to re-save the .xib files with a deployment target of macOS 10.5 in Interface Builder 3.2, and then use Xcode 3 to build a PowerPC binary using the Xcode 3 project file.
2018-11-29 22:02:07 -08:00
zeromus 3e37352bee winport - menu option to control CLI console visibility
fixes #230
2018-11-29 16:25:59 -06:00
rogerman 6bcd19b3cb GPU: Fix a bug in GPUEngineBase::_CompositeVRAMLineDeferred() where compInfo.target.xCustom was overstepping its bounds in X-dimension only custom buffers. This had the effect of causing undefined coloring when running at custom resolutions. (Regression from commit 8e9e7c4. Fixes #228 and fixes #229.) 2018-11-29 13:12:56 -08:00
zeromus f6938dc80a
Merge pull request #217 from SuuperW/paths
Paths
2018-11-29 15:04:31 -06:00
rogerman 8e9e7c4a2a GPU: Enable SSE2-accelerated custom-sized VRAM reads through the OBJ layer. This significantly improves the performance of many games, such as those that make use of dual-screen 3D, when running at the higher resolutions. 2018-11-29 02:00:21 -08:00
rogerman 6fc6ceb294 SoftRasterizer: For SSE2-enabled systems only, fix a rare graphical glitch that can sometimes occur in some games. (Regression from commit 21f04c9.) 2018-11-28 17:31:49 -08:00
rogerman 4f543aa8ca Cocoa Port: Yet another attempt at eliminating microstuttering in Metal display views. While it hasn't been completely eliminated yet, it shouldn't be as bad now. 2018-11-28 13:36:02 -08:00
rogerman 1f9b9e02a4 SoftRasterizer: Fix build issues on Windows. (Regression from commit 21f04c9.) 2018-11-23 15:30:11 -08:00
rogerman 21f04c9ef2 SoftRasterizer: Do some minor improvements to both performance and code size. 2018-11-23 14:59:13 -08:00
Roger Manuel b1abf50a01
Merge pull request #224 from Jules-A/WinOglCompatUI
WIP WinFrontend UI for the new OpenGL compatabilty tradeoffs.
2018-11-23 14:17:00 -08:00
Jules.A a51a6d4b6d Revert pointless changes that only saved a few lines. 2018-11-23 17:18:16 +08:00
Jules.A 20547d38a2 Remove unneeded bool initialisation and now that ini values are sanatized comboboxes can be initialised and set in a loop to save a few lines. 2018-11-23 17:05:40 +08:00
Jules.A 7e4a270644 Ughh... get rid of the extra semicolon... 2018-11-23 14:19:56 +08:00
Jules.A db1ca70f59 Fix label variable naming, spacing and finally get ini value checking working without requiring to check for the value that is getting changed, 2018-11-23 14:12:52 +08:00
Jules.A 156c590ca4 Check if user inputted values are correct on initiation. 2018-11-23 01:47:14 +08:00
Jules.A 54c6205e95 Fix up the UI and delete the unusused function (it's really not worth it until more settings can be checked). 2018-11-22 22:16:36 +08:00
rogerman d410b5c195 SoftRasterizer: Fix Z-depth calculations when compiling on MSVC. Fixes the overworld map in Dragon Quest IV. (Regression from commit d81a75c.)
- Apparently, MSVC has a more strict implementation of IEEE-754 single-precision floats (with 23-bit significands) than Clang and GCC, and so we going to drop 2 LSBs during the calculation so that we're multiplying z by a 22-bit significand. Coincidentally, this now matches what we're doing with the OpenGL renderer, so this tends to better code consistency.
2018-11-21 19:52:30 -08:00
Jules.A 30e9fb9c46 WIP WinFrontend UI for the new OpenGL compatabilty tradeoffs. Should be fully functional except for CheckValid3DSettingInt where I messed up so I left it commented out, honestly it's probably not needed anyways as there's already checks in place, it just doesn't update ini until you save settings. UI is still a mess, will have to work on that later. 2018-11-21 21:48:30 +08:00
intact 2f93fedbf8 Gtk+ Port: Use Desktop or Home as fallback directory for screenshots 2018-11-21 14:44:35 +01:00
rogerman f03a880ef9 WiFi: In Compatibility mode, reset the current RX packet's latency counter for every halfword write like its supposed to be. 2018-11-20 23:14:09 -08:00
rogerman adbbd406b1 Cocoa Port: Add tooltips for each of the OpenGL rendering accuracy options. 2018-11-20 17:56:18 -08:00