Yaya48/dolphin
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
29,329 Commits 1 Branch 0 Tags
16f8b7413de41432c9128f1db896149fe8db3e85
Commit Graph
100 Commits
This Branch
This Branch
All Branches
Author SHA1 Message Date
degasus 5333c17cca Jit: Fix branch following. 
The idea of this code was to not unroll loops, but it was completely broken.
So we've unrolled all loops, but only up to the second iteration.
Honestly, a better check would test if we branch to code which is already in the compiling block. But this is out of scope for now.

But testing shows that this unrolling actually improve the performance. So instead of fixing this bug, this check can be dropped.
 2018-07-28 16:35:42 +02:00
degasus ecf86bbf7b JitArm64: Drop the plattform register. 
This register is defined as "optional reserved" within the aarch64 ABI.
Linux doesn't use it, but we must not modify it on ios or windows.
As we have plenty of registers on aarch64, let's just always skip this one.
 2018-06-21 22:39:15 +02:00
degasus 03c88c83ac CachedInterpreter: Implement breakpoints. 
There were missed on the initial implementation of the cached interpreter.
 2018-06-13 08:33:57 +02:00
degasus 7154bfd825 UnitTest: Add StringUtil ToString/TryParse test 
Only a single one, but the main one for ini files:
Check if the written values can be parsed again.
 2018-06-03 23:12:32 +02:00
degasus bde65d8b42 Jit64: Fix MORE_ACCURATE_DOUBLETOSINGLE. 
This is broken since 3d12849967.
 2018-05-29 23:25:30 +02:00
degasus 3307946e2a VertexLoaderArm: Fix 565 color format. 
Copy & paste error for enforcing A=255.
It was copied from the 888X format, which stores the output data in another register.
 2018-05-27 11:27:15 +02:00
degasus fdfd8b19d6 OGL/Streambuffer: Use coherent mapping by default. 
Coherent mappings have a lower overhead and less GL codes.
So enables coherent mapping by default for all drivers.
Both Qualcomm and ARM performs very bad with explicit flushing, so this change helps them as well.

AFAIK there was one GPU generation which was slower on coherent mapping: nvidia tesla
So Geforce 200 and 300 series should be tested with this PR before merging.
As this was last tested many years ago, this issue might have been fixed as well.
Those GPUs are close to 10 years old and not supported any more by nvidia.
 2018-05-24 00:00:02 +02:00
degasus be3ce97560 CMake: Prefer GLVND for OpenGL. 
We have to decide if we want to link to
libGL.so -- legacy
libOpenGL.so -- GLVND

As we use xxxGetProcAddress, we don't care about feature
levels of the linked library, so we're very fine with the
new way.
 2018-05-09 09:08:22 +02:00
degasus a5c0739fab OGL: Scale OSD text on big screens. 2018-03-12 20:38:06 +01:00
degasus d4449971c9 OGL: Warn if buffer_storage isn't available. 
Yes, this commit is only to blame OSX and Mali. Through the former supports unsynchronized mappings, the latter supports *no* way to stream dynamic data at all. Let's try to make bad news, as they ignore friendly feature requests. Maybe we just need to make more noise...
 2018-03-12 20:38:00 +01:00
degasus a5a0599145 CustomTexture: Drop old texture format. 2018-01-20 17:08:47 +01:00
degasus 0b466249e0 CustomTextures: Drop format convertion. 2018-01-20 16:39:04 +01:00
degasus 02dd062518 VideoCommon: Drop now unused efb2tex matrix generation. 2017-12-06 09:30:03 +01:00
degasus e2a9ee384f VideoCommon: Rename TextureConversionShader namespace. 2017-12-02 15:17:39 +01:00
degasus e8febd0cef VideoCommon: Create a namespace for TextureConversionShaderGen. 2017-12-02 15:17:39 +01:00
degasus 6ea3f538b4 JitArm64: Inline GP check in Cleanup. 
We're calling this function up to 2M times per second. Let's inline the pre-check.
 2017-11-18 17:43:38 +01:00
degasus 6c9bb67ca0 JitArm64: Optimize gather pipe writes. 2017-11-18 17:43:38 +01:00
degasus 6ad63c1b35 Jit64: Inline GP fast check in Cleanup. 2017-11-18 17:43:38 +01:00
degasus 36ad887a19 Jit64: Inline GP writes. 
As we're down to 4 instructions now, it is always worth to inline those writes.
 2017-11-18 14:45:09 +01:00
degasus 4feddd7748 PowerPC: Include the gather pipe pointer in the ppc state. 2017-11-18 14:14:45 +01:00
degasus 3a36c964f5 WiimoteEmu: Don't set pack pragma for STL objects. 2017-11-05 13:13:25 +01:00
degasus 47f7e5b454 WiiUtils: Be more explicit about packing. 2017-11-05 10:52:37 +01:00
degasus 992893ba41 Jit64: Use ImmPtr. 2017-09-02 14:33:07 +02:00
degasus da79ddbde7 JitArm64: Rewrite Exit functions. 
The gpr must not be touched in the Exit functions as they are maybe conditional.
So just allocate everything here manually.
 2017-09-02 13:45:24 +02:00
degasus 304e601ad3 JitArm64: Reimplement aarch64 cycle counters. 
CNTVCT_EL0 is force-enabled on all linux plattforms.
Windows is untested, but as this is the best way to get *any* low
overhead performance counters, they likely use it as well.
 2017-09-02 13:24:37 +02:00
degasus 9080192a65 Jit64: Move profiler to the end of th block. 
Within Cleanup(), it is called at *every* end of the block. This generates bigger code,
but it is the only way to handle blocks with multiple exit nodes.
 2017-09-02 13:05:58 +02:00
degasus 958b75b707 JitCommon: Restructure the profiler calls. 2017-09-02 13:05:58 +02:00
degasus 95ce860265 DolphinWX: Enable branch following in the JIT debug interface. 2017-09-02 13:05:58 +02:00
degasus 8ab34b1a3e Jit64: Inline the profiler calls. 2017-09-02 13:05:58 +02:00
degasus b00c60618b JitArm64: Fix rlwinmx. 
Seems like I was wrong that ANDI2R doesn't require a temporary register here.
There is *one* case when the mask won't fit in the ARM AND instruction:
mask = 0xFFFFFFFF
But let's just use MOV instead of AND here for this case...
 2017-08-22 08:47:43 +02:00
degasus cc4b3b180b VideoSW: Drop log level of missing anti aliasing support. 
We don't want to emit such a high priority warning here, else fifoci will die.
 2017-08-05 19:20:19 +02:00
degasus 7389f0f55e Analytics: Set proper thread name. 2017-05-03 23:59:29 +02:00
degasus afb0beb9ab OpenGL: Mute nvidia performance warning about stalling the GPU. 
They are right, our perf query implementation is terrible.
But raising a warning makes it just even slower.
 2017-05-03 23:48:14 +02:00
degasus 90d551e0d1 JitArm64: Drop ps_res. 
The accuracy doesn't match ppc, and worse, it doesn't set the error flags if the input is zero.

Lets stop to ship broken instructions, so right now, the interpreter is the closest one.
 2017-05-03 23:48:14 +02:00
degasus 64ccf803bd Jit64: Use a temporary register for memory references. Part 4. 2017-04-12 20:53:17 +02:00
degasus 0ddd90b539 Jit64: Use a temporary register for memory references. Part 3. 2017-04-12 20:53:17 +02:00
degasus 78b36921d2 Jit64: Use a temporary register for memory references. Part 1. 2017-04-12 20:53:17 +02:00
degasus ca8d9e2215 TextureCache: Don't lock freed rendertargets for one frame. 
New Super Mario Bros on PAL still renders at 60 fps, but skips every 5th XFB copy.
So our detection of "per frame" fails, and we require twice the amound of texture objects.
But our pool frees unused textures after 3 frames, so half of them needs to be reallocated
every few frames.

This commit removes the lock for render targets. It was introduced to not update a texture
while it is still in use. But render targets aren't updated while rendering, so this
lock isn't needed. Non-rendertarget textures however aren't as dynamic, so the lock should
have no performance update.
 2017-03-22 23:28:42 +01:00
degasus c5b9c740c9 JitArm64: Fix gpr.R() usages in conditional code. 2017-03-20 21:18:13 +01:00
degasus f7a52c0c35 JitArm64: Fix gpr.GetReg usages in conditional code. 2017-03-20 21:18:13 +01:00
degasus 5152c997a4 MMU: Create constants for our BAT flags. 
This avoids a few hard-coded constants in several files.
 2017-03-15 09:20:40 +01:00
degasus a3f79f88cd Android: Update android studio + gradle. 
Android studio 2.3 finally displays the result of the native compilation, so C++ warnings are now shown in the build log.
 2017-03-04 15:53:40 +01:00
degasus ffa61fcf57 JitCache: Also unlink exits of the current block. 
We might still be in the current block. This is fine, but the next one might also be invalidated later on. But we may never also call the next one.
 2017-02-27 23:50:16 +01:00
degasus c1ddc2678e JitCache: Fix removing of blocks. 2017-02-27 23:50:16 +01:00
degasus 674e7430d0 JitArm64: Extract block exit link area generation. 
Always use 3 instructions on linking.
 2017-02-17 23:07:28 +01:00
degasus 8829af62cb JitArm64: Fix for stack push/pop ABI. 2017-02-11 00:59:12 +01:00
degasus 6aa54a029e JitArm64: Optimize GPR register push/pop. 2017-02-11 00:59:12 +01:00
degasus 4d3883a756 JitArm64: Handle stack faults. 2017-02-10 00:32:21 +01:00
degasus 40b7cc9252 JitArm64: Use a custom stack with proper guard pages. 2017-02-10 00:32:17 +01:00
degasus 384efb0cb2 JitArm64: Initial implementation of the BLR optimization. 2017-02-02 09:06:34 +01:00
degasus 2c69e976d0 DSPLLE: Only enable the DSP JIT on x64. 
The x64 JIT is hardcoded right now, and it seems unlikely that we'll support another arch here soon.
So let's just disable the DSP JIT.
 2017-01-30 23:53:58 +01:00
degasus 89cefe3103 JitArm64: Fix immediate versions of cmpl. 2017-01-29 18:38:43 +01:00
degasus c2822ef7d3 PatchEngine: Apply speedhack per PC, not only per block entry. 
This makes speedhacks indepdent of block assembly / inlining / conditional continue.
 2017-01-28 12:36:13 +01:00
degasus ca10cf5afe PPCAnalyst: Update comments 2017-01-28 03:03:04 +01:00
degasus 4855764345 JitArm64: Implement leaf inlining. 
There is no BLR stack, so this is quite trivial.
 2017-01-28 02:48:56 +01:00
degasus f31b25fe39 Jit64: Enable branch following. 2017-01-28 02:48:56 +01:00
degasus 3787af9ee9 CommandProcessor: Limit scope of ugly SCPFifoStruct. 
It's only used as an interface between two classes. So no need to declare
it in the backend export header.
 2017-01-27 19:04:56 +01:00
degasus ba1decf7bc PixelEngine: Use a DirectRead for the token. 
No need to call a std::function, and a few games poll this value very often.
 2017-01-27 07:49:47 +01:00
degasus cbe9c3e040 GameSettings: Don't ship speedhack which are enabled by default. 
This may slow down a few games, but we don't want to support this kind of mess.
 2017-01-27 00:07:19 +01:00
degasus d3aee2de08 JitCache: Split off JIT call from dispatcher. 
This avoid flushing the BLR optimization stack on fast_block_cache misses.
 2017-01-25 01:51:19 +01:00
degasus 7f6b8e3555 JitCache: Extract ErasePhysicalRange as function. 2017-01-23 20:33:44 +01:00
degasus 70caf447b9 JitCache: Get physical addresses from PPCAnalyst. 
So we support all kind of degenerated blocks now, not just range+length based ones.
 2017-01-23 20:33:44 +01:00
degasus 3529af61c4 JitAnalyzer: Remove jit block cache workaround. 2017-01-23 20:33:44 +01:00
degasus f3ed993747 JitCache: Use a map with macro blocks for the occupied memory regions. 
This also allow fast invalidation, without any restritions on the blocks itself.
So we can now implement inlining.
 2017-01-23 20:33:44 +01:00
degasus dc0fbc15f0 JitCache: Drop block_map. 
It is only used for invalidation, and in a bad way. Just scan over all elements,
as it is still in O(n), this shouldn't matter much.
 2017-01-23 20:33:44 +01:00
degasus 819ebfb213 JitCache: Freeing hotfix. 
Sorry, I'm too stupid to test my code.
 2017-01-23 06:58:02 +01:00
degasus a19993c2a0 JitArm64: Fix dcbz. 
This fixes the hidden wall in xenoblade.
 2017-01-23 00:26:38 +01:00
degasus 8e00c411a3 JitCache::Rename iCache to fast_block_map. 
iCache sounds too much like emulation.
 2017-01-22 17:10:28 +01:00
degasus 830ae6a2c1 JitCache: Store the JitBlock in the std::map. 2017-01-22 16:50:46 +01:00
degasus 43cdbab515 Jit64: Add a switch for easier fallback to the C++ dispatcher. 2017-01-22 16:50:46 +01:00
degasus 9b77a39767 JitCache: Only call DestroyBlock on valid blocks. 2017-01-22 16:50:46 +01:00
degasus 113d6b3b84 JitCache: Use a multimap for block_map and start_block_map. 
We may have duplicated entries here because of MSR mismatch. Just
store both and validate the matching one on cache access.
 2017-01-22 16:50:46 +01:00
degasus 9d58127dec JitCache: Move JitBlock config variable. 2017-01-22 16:50:46 +01:00
degasus 352909fc4c JitCache: Track the ICache entry of jit blocks. 
This guarantees that no invalidated jit block is still in the icache.
 2017-01-22 16:50:14 +01:00
degasus c9c5437cb8 JitCache: Fix links_to updates on destroying blocks. 2017-01-22 11:37:22 +01:00
degasus 65e3aa796e VideoConfig: Fix boolean default values. 2017-01-21 11:31:44 +01:00
degasus fe41ed7fa1 JitCache: Use pointers in the ICache and dispatchers. 2017-01-12 21:43:28 +01:00
degasus d4a529614a Jit64: Use PPCSTATE reg if possible. 2017-01-12 20:55:17 +01:00
degasus a8b26937f3 JitCache: Clean up GetBlockFromStartAddress. 2017-01-12 20:24:43 +01:00
degasus 214f37ecef JitCache: Rename variables. 2017-01-12 20:23:14 +01:00
degasus 3ee178abe7 JitCache: Drop unused functions. 2017-01-12 20:23:14 +01:00
degasus 7e850361fb JitCache: Add a helper function to iterate over all blocks. 2017-01-12 20:23:14 +01:00
degasus ca026b58ab JitCache: Use a pointer in links_to. 2017-01-12 20:23:14 +01:00
degasus 928ccbef53 JitCache: Use a pointer in block_map. 2017-01-12 20:23:14 +01:00
degasus 74a27d9742 JitCache: Use a pointer in start_block_map. 2017-01-12 20:23:14 +01:00
degasus 2d5288dc10 JitCache: Return a pointer in AllocateBlock. 2017-01-12 20:23:14 +01:00
degasus f14cebf079 JitCache: Use a pointer in FinalizeBlock. 2017-01-12 20:23:14 +01:00
degasus d3aa8c8080 JitCache: Return a pointer in GetBlockFromStartAddress. 2017-01-12 20:23:14 +01:00
degasus f6ec96efbd JitCache: Use a pointer in DestroyBlock. 2017-01-12 20:23:14 +01:00
degasus 68c85d32f7 JitCache: Use a pointer in UnlinkBlock. 2017-01-12 20:23:14 +01:00
degasus eb390f3020 JitCache: Use a pointer in LinkBlock. 2017-01-12 20:23:14 +01:00
degasus ccb8c44a5a JitCache: Use a pointer in LinkBlockExits. 2017-01-12 20:23:14 +01:00
degasus f12460d1f4 VideoSW: Fix GL ES shader. 2017-01-07 12:32:15 +01:00
degasus f6c7a258ed BlendState: Use masking as a fastpath for the logic op NO_OP. 2017-01-06 14:01:39 +01:00
degasus 23b0faeba5 OGL: Use VideoCommon blend state for framebuffer configuration. 2017-01-06 14:01:39 +01:00
degasus 3df828463d VideoCommon: Assembly blending state in a shared state object. 2017-01-06 14:01:36 +01:00
degasus 41b0c74e30 VideoCommon: Make dst_alpha state implicit. 2017-01-04 20:02:31 +01:00
degasus b7d8bd13a6 OGL/Vulkan: Drop dual pass alpha. 2017-01-04 19:59:41 +01:00
degasus d0a73c1d90 TextureCache: Add a dirty bit for partial updates on overlapping EFB copies. 2016-12-27 17:26:04 +01:00
degasus abf9bb170b TextureCache: Drop unused parameter in backend API. 2016-12-26 22:10:32 +01:00