Commit graph

178 commits

Author SHA1 Message Date
Fernando Sahmkow
7fcd0fee6d Buffer Cache: Use vAddr instead of physical memory. 2020-04-06 09:23:06 -04:00
Fernando Sahmkow
9c0f40a1f5 GPU: Setup Flush/Invalidate to use VAddr instead of CacheAddr 2020-04-06 09:21:46 -04:00
Markus Wick
56672b8c98 core/memory: Create a special MapMemoryRegion for physical memory.
This allows us to create a fastmem arena within the memory.cpp helpers.
2020-01-18 08:38:47 +01:00
Markus Wick
0986caa8d8 core/memory + arm/dynarmic: Use a global offset within our arm page table.
This saves us two x64 instructions per load/store instruction.

TODO: Clean up our memory code. We can use this optimization here as well.
2020-01-01 12:24:54 +01:00
Lioncash
e7e939104b core/memory; Migrate over SetCurrentPageTable() to the Memory class
Now that literally every other API function is converted over to the
Memory class, we can just move the file-local page table into the Memory
implementation class, finally getting rid of global state within the
memory code.
2019-11-26 21:55:39 -05:00
Lioncash
50a518be69 core/memory: Migrate over GetPointerFromVMA() to the Memory class
Now that everything else is migrated over, this is essentially just code
relocation and conversion of a global accessor to the class member
variable.

All that remains is to migrate over the page table.
2019-11-26 21:55:39 -05:00
Lioncash
e4c381b885 core/memory: Migrate over Write{8, 16, 32, 64, Block} to the Memory class
The Write functions are used slightly less than the Read functions,
which make these a bit nicer to move over.

The only adjustments we really need to make here are to Dynarmic's
exclusive monitor instance. We need to keep a reference to the currently
active memory instance to perform exclusive read/write operations.
2019-11-26 21:55:39 -05:00
Lioncash
b05bfc6036 core/memory: Migrate over Read{8, 16, 32, 64, Block} to the Memory class
With all of the trivial parts of the memory interface moved over, we can
get right into moving over the bits that are used.

Note that this does require the use of GetInstance from the global
system instance to be used within hle_ipc.cpp and the gdbstub. This is
fine for the time being, as they both already rely on the global system
instance in other functions. These will be removed in a change directed
at both of these respectively.

For now, it's sufficient, as it still accomplishes the goal of
de-globalizing the memory code.
2019-11-26 21:55:39 -05:00
Lioncash
89ef3ef575 core/memory: Migrate over ZeroBlock() and CopyBlock() to the Memory class
These currently aren't used anywhere in the codebase, so these are very
trivial to move over to the Memory class.
2019-11-26 21:55:38 -05:00
Lioncash
849581075a core/memory: Migrate over RasterizerMarkRegionCached() to the Memory class
This is only used within the accelerated rasterizer in two places, so
this is also a very trivial migration.
2019-11-26 21:55:38 -05:00
Lioncash
b2165c6b35 core/memory: Migrate over ReadCString() to the Memory class
This only had one usage spot, so this is fairly straightforward to
convert over.
2019-11-26 21:55:38 -05:00
Lioncash
3f08e8d8d4 core/memory: Migrate over GetPointer()
With all of the interfaces ready for migration, it's trivial to migrate
over GetPointer().
2019-11-26 21:55:38 -05:00
Lioncash
fc7d0a17b6 core/memory: Move memory read/write implementation functions into an anonymous namespace
These will eventually be migrated into the main Memory class, but for
now, we put them in an anonymous namespace, so that the other functions
that use them, can be migrated over separately.
2019-11-26 21:53:35 -05:00
Lioncash
e58748fd80 core/memory: Migrate over address checking functions to the new Memory class
A fairly straightforward migration. These member functions can just be
mostly moved verbatim with minor changes. We already have the necessary
plumbing in places that they're used.

IsKernelVirtualAddress() can remain a non-member function, since it
doesn't rely on class state in any form.
2019-11-26 21:53:34 -05:00
Lioncash
323680e5ad core/memory: Migrate over memory mapping functions to the new Memory class
Migrates all of the direct mapping facilities over to the new memory
class. In the process, this also obsoletes the need for memory_setup.h,
so we can remove it entirely from the project.
2019-11-26 21:53:34 -05:00
Lioncash
4c2ed2706e core/memory: Introduce skeleton of Memory class
Currently, the main memory management code is one of the remaining
places where we have global state. The next series of changes will aim
to rectify this.

This change simply introduces the main skeleton of the class that will
contain all the necessary state.
2019-11-26 21:53:34 -05:00
Lioncash
f1382cf0e7 core: Remove Core::CurrentProcess()
This only encourages the use of the global system instance (which will
be phased out long-term). Instead, we use the direct system function
call directly to remove the appealing but discouraged short-hand.
2019-10-06 13:53:12 -04:00
Fernando Sahmkow
50259d7bdc Core/Memory: Only FlushAndInvalidate GPU if the page is marked as RasterizerCachedMemory
This commit avoids Invalidating and Flushing the GPU if the page is not
marked as a RasterizerCache Page.
2019-09-18 22:47:22 -04:00
Lioncash
65c748fbd3 memory: Remove unused includes
These aren't used within the central memory management code, so they can
be removed.
2019-07-06 02:24:34 -04:00
Lioncash
f2331a804a core/cpu_core_manager: Create threads separately from initialization.
Our initialization process is a little wonky than one would expect when
it comes to code flow. We initialize the CPU last, as opposed to
hardware, where the CPU obviously needs to be first, otherwise nothing
else would work, and we have code that adds checks to get around this.

For example, in the page table setting code, we check to see if the
system is turned on before we even notify the CPU instances of a page
table switch. This results in dead code (at the moment), because the
only time a page table switch will occur is when the system is *not*
running, preventing the emulated CPU instances from being notified of a
page table switch in a convenient manner (technically the code path
could be taken, but we don't emulate the process creation svc handlers
yet).

This moves the threads creation into its own member function of the core
manager and restores a little order (and predictability) to our
initialization process.

Previously, in the multi-threaded cases, we'd kick off several threads
before even the main kernel process was created and ready to execute (gross!).
Now the initialization process is like so:

Initialization:
  1. Timers

  2. CPU

  3. Kernel

  4. Filesystem stuff (kind of gross, but can be amended trivially)

  5. Applet stuff (ditto in terms of being kind of gross)

  6. Main process (will be moved into the loading step in a following
                   change)

  7. Telemetry (this should be initialized last in the future).

  8. Services (4 and 5 should ideally be alongside this).

  9. GDB (gross. Uses namespace scope state. Needs to be refactored into a
          class or booted altogether).

  10. Renderer

  11. GPU (will also have its threads created in a separate step in a
           following change).

Which... isn't *ideal* per-se, however getting rid of the wonky
intertwining of CPU state initialization out of this mix gets rid of
most of the footguns when it comes to our initialization process.
2019-04-11 22:11:40 -04:00
Lioncash
abae7577d2 core/memory: Remove GetCurrentPageTable()
Now that nothing actually touches the internal page table aside from the
memory subsystem itself, we can remove the accessor to it.
2019-04-07 02:47:37 -04:00
bunnei
3ae0de9b53 memory: Check that core is powered on before attempting to use GPU.
- GPU will be released on shutdown, before pages are unmapped.
- On subsequent runs, current_page_table will be not nullptr, but GPU might not be valid yet.
2019-03-20 22:36:03 -04:00
bunnei
93da8e0abf core: Move PageTable struct into Common. 2019-03-16 22:05:40 -04:00
bunnei
10118c71e0 memory: Simplify rasterizer cache operations. 2019-03-16 00:41:08 -04:00
bunnei
2eaf6c41a4 gpu: Use host address for caching instead of guest address. 2019-03-14 22:34:42 -04:00
bunnei
7b574f406b gpu: Move command processing to another thread. 2019-03-06 21:48:57 -05:00
Weiyi Wang
5159f4eee8 Memory: don't lock hle mutex in memory read/write
The comment already invalidates itself: neither MMIO nor rasterizer cache belongsHLE kernel state. This mutex has a too large scope if MMIO or cache is included, which is prone to dead lock when multiple thread acquires these resource at the same time. If necessary, each MMIO component or rasterizer should have their own lock.
2019-03-02 15:20:05 +01:00
Annomatg
ef84c70d22 Speed up memory page mapping (#2141)
- Memory::MapPages total samplecount was reduced from 4.6% to 1.06%.
- From main menu into the game from 1.03% to 0.35%
2019-02-27 17:22:47 -05:00
David Marcec
fdd649e2ef Fixed uninitialized memory due to missing returns in canary
Functions which are suppose to crash on non canary builds usually don't return anything which lead to uninitialized memory being used.
2018-12-19 12:52:32 +11:00
Lioncash
15e3d4f357 memory: Convert ASSERT into a DEBUG_ASSERT within GetPointerFromVMA()
Given memory should always be expected to be valid during normal
execution, this should be a debug assertion, rather than a check in
regular builds.
2018-12-06 15:02:34 -05:00
Lioncash
d4c1b9d311 vm_manager: Make vma_map private
This was only ever public so that code could check whether or not a
handle was valid or not. Instead of exposing the object directly and
allowing external code to potentially mess with the map contents, we
just provide a member function that allows checking whether or not a
handle is valid.

This makes all member variables of the VMManager class private except
for the page table.
2018-12-06 15:02:17 -05:00
heapo
ca8ab1bc3b Call shrink_to_fit after page-table vector resizing to cause crt to actually lower vector capacity. For 36-bit titles saves 800MB of commit. 2018-12-05 14:40:56 -08:00
Frederic L
7a5eda5914 global: Use std::optional instead of boost::optional (#1578)
* get rid of boost::optional

* Remove optional references

* Use std::reference_wrapper for optional references

* Fix clang format

* Fix clang format part 2

* Adressed feedback

* Fix clang format and MacOS build
2018-10-30 00:03:25 -04:00
Lioncash
cf9d6c6f52 kernel/process: Make data member variables private
Makes the public interface consistent in terms of how accesses are done
on a process object. It also makes it slightly nicer to reason about the
logic of the process class, as we don't want to expose everything to
external code.
2018-09-30 02:30:01 -04:00
Lioncash
83377113bf memory: Dehardcode the use of fixed memory range constants
The locations of these can actually vary depending on the address space
layout, so we shouldn't be using these when determining where to map
memory or be using them as offsets for calculations. This keeps all the
memory ranges flexible and malleable based off of the virtual memory
manager instance state.
2018-09-24 22:16:03 -04:00
Lioncash
7fd598636e memory: Dehardcode the use of a 36-bit address space
Given games can also request a 32-bit or 39-bit address space, we
shouldn't be hardcoding the address space range as 36-bit.
2018-09-24 22:15:53 -04:00
fearlessTobi
63c2e32e20 Port #4182 from Citra: "Prefix all size_t with std::" 2018-09-15 15:21:06 +02:00
bunnei
d647d9550c gl_renderer: Cache textures, framebuffers, and shaders based on CPU address. 2018-08-31 13:07:27 -04:00
Lioncash
45fb74d262 gpu: Make memory_manager private
Makes the class interface consistent and provides accessors for
obtaining a reference to the memory manager instance.

Given we also return references, this makes our more flimsy uses of
const apparent, given const doesn't propagate through pointers in the
way one would typically expect. This makes our mutable state more
apparent in some places.
2018-08-28 11:11:50 -04:00
Lioncash
2665457f4a renderer_base: Make Rasterizer() return the rasterizer by reference
All calling code assumes that the rasterizer will be in a valid state,
which is a totally fine assumption. The only way the rasterizer wouldn't
be is if initialization is done incorrectly or fails, which is checked
against in System::Init().
2018-08-04 02:36:58 -04:00
Lioncash
6030c5ce41 video_core: Eliminate the g_renderer global variable
We move the initialization of the renderer to the core class, while
keeping the creation of it and any other specifics in video_core. This
way we can ensure that the renderer is initialized and doesn't give
unfettered access to the renderer. This also makes dependencies on types
more explicit.

For example, the GPU class doesn't need to depend on the
existence of a renderer, it only needs to care about whether or not it
has a rasterizer, but since it was accessing the global variable, it was
also making the renderer a part of its dependency chain. By adjusting
the interface, we can get rid of this dependency.
2018-08-04 02:36:57 -04:00
Lioncash
59b04c0df6 memory: Remove unused GetSpecialHandlers() function
This is just unused code, so we may as well get rid of it.
2018-08-03 14:20:50 -04:00
Lioncash
26de4bb521 core/memory: Get rid of 3DS leftovers
Removes leftover code from citra that isn't needed.
2018-08-03 11:22:47 -04:00
bunnei
85421f3406
Merge pull request #690 from lioncash/move
core/memory, core/hle/kernel: Use std::move where applicable
2018-07-18 20:55:55 -07:00
Lioncash
72207577b2 core/memory: Remove unused function GetSpecialHandlers() and an unused variable in ZeroBlock() 2018-07-18 19:55:46 -04:00
Lioncash
46458e7284 core/memory, core/hle/kernel: Use std::move where applicable
Avoids pointless copies
2018-07-18 19:34:31 -04:00
James Rowe
0d46f0df12 Update clang format 2018-07-02 21:45:47 -04:00
James Rowe
638956aa81 Rename logging macro back to LOG_* 2018-07-02 21:45:47 -04:00
Michael Scire
067ac434ba Kernel/Arbiters: Fix casts, cleanup comments/magic numbers 2018-06-22 00:47:59 -06:00
bunnei
a434fdcb10 core: Implement multicore support. 2018-05-10 19:34:46 -04:00
Lioncash
7c9644646f
general: Make formatting of logged hex values more straightforward
This makes the formatting expectations more obvious (e.g. any zero padding specified
is padding that's entirely dedicated to the value being printed, not any pretty-printing
that also gets tacked on).
2018-05-02 09:49:36 -04:00
Lioncash
8475496630
general: Convert assertion macros over to be fmt-compatible 2018-04-27 10:04:02 -04:00
bunnei
42d43ea741
Merge pull request #387 from Subv/maxwell_2d
GPU: Partially implemented the 2D surface copy engine
2018-04-25 20:40:17 -04:00
Subv
5ab597041f Memory: Added a missing shortcut for Memory::CopyBlock for the current process. 2018-04-25 11:55:30 -05:00
Lioncash
6d00780045
core/memory: Amend address widths in asserts
Addresses are 64-bit, these formatting specifiers are simply holdovers from citra. Adjust them to be the correct width.
2018-04-25 12:17:26 -04:00
Lioncash
59dae03dbe
core/memory: Move logging macros over to new fmt-capable ones
While we're at it, correct addresses to print all 64 bits where applicable, which were holdovers from citra.
2018-04-25 12:16:33 -04:00
bunnei
4415e00181 gl_rasterizer_cache: Update to be based on GPU addresses, not CPU addresses. 2018-04-24 22:31:45 -04:00
bunnei
d4fb8a887c memory: Fix cast for ReadBlock/WriteBlock/ZeroBlock/CopyBlock. 2018-03-26 21:17:01 -04:00
bunnei
d732142b66 memory: Add RasterizerMarkRegionCached code and cleanup. 2018-03-26 21:16:57 -04:00
bunnei
a10baacf9e
Merge pull request #265 from bunnei/tegra-progress-2
Tegra progress 2
2018-03-23 23:30:48 -04:00
bunnei
b5f3e7951b memory: Fix typo in RasterizerFlushVirtualRegion. 2018-03-23 15:46:21 -04:00
bunnei
1a158dfcd6 memory: RasterizerFlushVirtualRegion should also check process image region. 2018-03-23 15:10:02 -04:00
bunnei
11047d7fd5 rasterizer: Flush and invalidate regions should be 64-bit. 2018-03-23 15:01:45 -04:00
bunnei
63d3924b5b memory: Port RasterizerFlushVirtualRegion from Citra. 2018-03-22 22:56:41 -04:00
N00byKing
8afdbf6a1f Remove more N3DS References 2018-03-22 21:25:06 +01:00
bunnei
7d6653268f core: Move process creation out of global state. 2018-03-14 18:42:19 -04:00
MerryMage
cc368de1a0 memory: LOG_ERROR when falling off end of page table 2018-02-21 21:39:07 +00:00
Lioncash
fe0775d2f4
memory: Silence formatting sepecifier warnings 2018-02-14 01:54:54 -05:00
MerryMage
738f91a57d memory: Replace all memory hooking with Special regions 2018-01-27 15:16:39 +00:00
Rozlette
de7aa3106a memory: Return false for large VAddr in IsValidVirtualAddress 2018-01-20 14:56:15 -06:00
James Rowe
389979018c Remove gpu debugger and get yuzu qt to compile 2018-01-12 19:11:04 -07:00
MerryMage
f10a304ecd fix macos build 2018-01-08 19:20:10 -05:00
bunnei
93480b10ef core/video_core: Fix a bunch of u64 -> u32 warnings. 2018-01-01 15:40:35 -05:00
bunnei
97d52b9f89 memory: Print addresses as 64-bit. 2017-10-18 21:29:23 -04:00
bunnei
b1d5db1cf6 Merge remote-tracking branch 'upstream/master' into nx
# Conflicts:
#	src/core/CMakeLists.txt
#	src/core/arm/dynarmic/arm_dynarmic.cpp
#	src/core/arm/dyncom/arm_dyncom.cpp
#	src/core/hle/kernel/process.cpp
#	src/core/hle/kernel/thread.cpp
#	src/core/hle/kernel/thread.h
#	src/core/hle/kernel/vm_manager.cpp
#	src/core/loader/3dsx.cpp
#	src/core/loader/elf.cpp
#	src/core/loader/ncch.cpp
#	src/core/memory.cpp
#	src/core/memory.h
#	src/core/memory_setup.h
2017-10-09 23:56:20 -04:00
Subv
1f2de7501b Memory: Make WriteBlock take a Process parameter on which to operate 2017-10-01 14:18:36 -05:00
Subv
811c01e5fe Memory: Make ReadBlock take a Process parameter on which to operate 2017-10-01 14:18:36 -05:00
bunnei
46297c9653 memory: Log with 64-bit values. 2017-09-30 14:28:54 -04:00
bunnei
f01472a5ff core: Various changes to support 64-bit addressing. 2017-09-30 14:28:54 -04:00
Huw Pascoe
a13ab958cb Fixed type conversion ambiguity 2017-09-30 09:34:35 +01:00
bunnei
b07af7dda8 Merge pull request #2961 from Subv/load_titles
Loaders: Don't automatically set the current process every time we load an application.
2017-09-29 14:58:42 -04:00
James Rowe
5620327e03 Merge pull request #2954 from Subv/cache_unmapped_mem
Memory/RasterizerCache: Ignore unmapped memory regions when caching physical regions
2017-09-26 16:55:47 -06:00
Subv
35da7f57ef Memory: Allow IsValidVirtualAddress to be called with a specific process parameter.
There is still an overload of IsValidVirtualAddress that only takes the VAddr and will default to the current process.
2017-09-26 17:27:44 -05:00
Subv
41f6c9f87f Memory/RasterizerCache: Ignore unmapped memory regions when caching physical regions.
Not all physical regions need to be mapped into the address space of every process, for example, system modules do not have a VRAM mapping.

This fixes a crash when loading applets and system modules.
2017-09-25 19:37:45 -05:00
MerryMage
67a70bd9e1 ARM_Interface: Implement PageTableChanged 2017-09-24 23:08:25 +01:00
MerryMage
4e5eb2044a memory: Remove GetCurrentPageTablePointers 2017-09-24 22:43:28 +01:00
MerryMage
c02bbb7030 memory: Add GetCurrentPageTable/SetCurrentPageTable
Don't expose Memory::current_page_table as a global.
2017-09-24 22:42:42 +01:00
B3n30
813837c5cf Merge pull request #2842 from Subv/switchable_page_table
Kernel/Memory: Give each process its own page table and allow switching the current page table upon reschedule
2017-09-15 22:41:45 +02:00
Subv
f18a176b60 Kernel/Memory: Make IsValidPhysicalAddress not go through the current process' virtual memory mapping. 2017-09-15 14:26:18 -05:00
Subv
214150f00c Kernel/Memory: Changed GetPhysicalPointer so that it doesn't go through the current process' page table to obtain a pointer. 2017-09-15 14:26:13 -05:00
Subv
6d2734a074 Kernel/Memory: Give each Process its own page table.
The loader is in charge of setting the newly created process's page table as the main one during the loading process.
2017-09-10 15:13:41 -05:00
danzel
8266064796 Use recursive_mutex instead of mutex to fix #2902 2017-08-29 20:39:55 +12:00
James Rowe
61442d6afb Merge pull request #2839 from Subv/global_kernel_lock
Kernel/HLE: Use a mutex  to synchronize access to the HLE kernel state between the cpu thread and any other possible threads that might touch the kernel (network thread, etc).
2017-08-23 18:17:44 -06:00
Subv
f484927ed0 Kernel/Memory: Acquire the global HLE lock when a memory read/write operation falls outside of the fast path, for it might perform an MMIO operation. 2017-08-22 09:30:58 -05:00
Weiyi Wang
045d0b5bbd Merge pull request #2799 from yuriks/virtual-cached-range-flush
Add address conversion functions returning optional, Add function to flush virtual region from rasterizer cache
2017-07-22 10:15:52 +03:00
Yuri Kunde Schlesner
d553135748 Memory: Add function to flush a virtual range from the rasterizer cache
This is slightly more ergonomic to use, correctly handles virtual
regions which are disjoint in physical addressing space, and checks only
regions which can be cached by the rasterizer.
2017-06-21 22:57:12 -07:00
Yuri Kunde Schlesner
f2a5a77e27 Memory: Fix crash when unmapping a VMA covering cached surfaces
Unmapping pages tries to flush any cached GPU surfaces touching that
region. When a cached page is invalidated, GetPointerFromVMA() is used
to restore the original pagetable pointer. However, since that VMA has
already been deleted, this hits an UNREACHABLE case in that function.

Now when this happens, just set the page type to Unmapped and continue,
which arrives at the correct end result.
2017-06-21 22:56:31 -07:00
Yuri Kunde Schlesner
6ae0086b39 Memory: Add TryVirtualToPhysicalAddress, returning a boost::optional 2017-06-21 22:55:18 -07:00
Yuri Kunde Schlesner
326e7c7020 Memory: Make PhysicalToVirtualAddress return a boost::optional
And fix a few places in the code to take advantage of that.
2017-06-21 22:55:17 -07:00
Yuri Kunde Schlesner
d3db770cad Memory: Add constants for the n3DS additional RAM
This is 4MB of extra, separate memory that was added on the New 3DS.
2017-05-09 21:43:59 -07:00