Don't automatically assume that Thread::Create will only be called when the parent process is currently scheduled. This assumption will be broken when applets or system modules are loaded.

There is still an overload of IsValidVirtualAddress that only takes the VAddr and will default to the current process.

Audio: Use std::deque instead of std::vector for the audio buffer type (StereoBuffer16)

The current code inserts and deletes elements from the beginning of the audio buffer, which is very inefficient in an std::vector.

Profiling was done using VisualStudio2017's Performance Analyzer in Super Mario 3D Land.

Before this change: AudioInterp::Linear had 14.14% of the runtime (inclusive) and most of that time was spent in std::vector's insert implementation.
After this change: AudioInterp::Linear has 0.36% of the runtime (inclusive)

HLE/Archives: Allow multiple loaded applications to access their SelfNCCH archive independently.

The loaders now register each loaded ROM with the SelfNCCH factory, which keeps the data around for the duration of the emulation session.

When opening the SelfNCCH archive, the factory queries the current program's programid and uses that as a key to the map that contains the NCCHData structure (RomFS, Icon, Banner, etc).

3dsx files do not have a programid and will use a default of 0 for this value, thus, only 1 3dsx file with RomFS is loadable at the same time.

Switchable Page Tables

Optimized Morton

citra-qt: fix some untranslated strings

HLE/SRV: Implemented RegisterService.

* loader/ncch: split NCCH parsing into its own file

* loader/ncch: add support for loading update NCCHs from the SD card

* loader/ncch: fix formatting

* file_sys/ncch_container: Return a value for OpenFile

* loader/ncch: cleanup, always instantiate overlay_ncch to base_ncch

* file_sys/ncch_container: better encryption checks, allow non-app NCCHs to load properly and for the existence of NCCH structures to be checked

* file_sys/ncch_container: pass filepath as a const reference

* Services/UDS: Added a function to generate the EAPoL-Start packet body.

* Services/UDS: Added filter for beacons.

* Services/UDS: Lock a mutex when accessing connection_status from both the emulation and network thread.

* Services/UDS: Handle the Association Response frame and respond with the EAPoL-Start frame.

* fixup: make use of current_node, changed received_beacons into a list, mutex and assert corrections

* fixup: fix damn clang-format

Optimized Float<M,E> multiplication

Before:

ucomiss xmm1, xmm1
jp      .L9
pxor    xmm2, xmm2
mov     edx, 1
ucomiss xmm0, xmm2
setp    al
cmovne  eax, edx
test    al, al
jne     .L9
.L3:
movaps  xmm0, xmm2
ret
.L9:
ucomiss xmm0, xmm0
jp      .L10
pxor    xmm2, xmm2
mov     edx, 1
ucomiss xmm1, xmm2
setp    al
cmovne  eax, edx
test    al, al
je      .L3

After:

movaps  xmm2, xmm1
mulss   xmm2, xmm0
ucomiss xmm2, xmm2
jnp     .L3
ucomiss xmm1, xmm0
jnp     .L11
.L3:
movaps  xmm0, xmm2
ret
.L11:
pxor    xmm2, xmm2
jmp     .L3

Don't expose Memory::current_page_table as a global.

GPU: Add draw for immediate and batch modes

Now system modules can do more than just crash immediately on startup.

Fixed framebuffer warning

Improved performance of FromAttributeBuffer

WebService: Set USE_SYSTEM_CURL for travis linux builds

* WebService: Verify username and token; Log errors in PostJson

* Fixup: added docstrings to the functions

* Webservice: Added Icons to the verification, imrpved error detection in cpr, fixup nits

* fixup: fmt warning

Ternary operator is optimized by the compiler
whereas std::min() is meant to return a value.

I've noticed a 5%-10% emulation speed increase.

Services/NS: Port ns:s to the new service framework.

PICA: implement custom clip plane

Kernel/Memory: Give each process its own page table and allow switching the current page table upon reschedule

CPU/Dynarmic: Disable the fast page-table access in dynarmic until it supports switching page tables at runtime.

This fixes building the tests

Kernel/Memory: Make IsValidPhysicalAddress not go through the current process' virtual memory mapping.

Kernel/Threads: Don't clear the CPU instruction cache when performing a context switch from an idle thread into a thread in the same process.

We were unnecessarily clearing the cache when going from Process A -> Idle -> Process A, this caused extreme performance regressions.

Kernel/Memory: Changed GetPhysicalPointer so that it doesn't go through the current process' page table to obtain a pointer.