Note: It's not 100% perfect, as some of the GPU capablities leak into the
pixel shader UID.
Currently our UIDs don't get exported, so there is no issue. But someone
might want to fix this in the future.
As much as possible, the asserts have been moved out of the GetUID
function. But there are some places where asserts depend on variables
that aren't stored in the shader UID.
Bug Fix: It was theoretically possible for a shader with depth writes
disabled to map to the same UID as a shader with late depth
writes.
No known test cases trigger this.
Bug Fix: The normal stage UIDs were randomly overwriting indirect
stage texture map UID fields. It was possible for multiple
shaders with diffrent indirect texture targets to map to
the same UID.
Once again, it dpesn't look like this bug was ever triggered.
This frees up 21 bits and allows us to shorten the UID struct by an entire
32 bits.
It's not strictly needed (as it's encoded into the length) but I added a
bit for per-pixel lighiting to make my life easier in the following
commits.
The only code which touches xfmem is code which writes directly into
uid_data.
All the rest now read their parameters out of uid_data.
I also simplified the lighting code so it always generated seperate
codepaths for alpha and color channels instead of trying to combine
them on the off-chance that the same equation works for all 4 channels.
As modern (post 2008) GPUs generally don't calcualte all 4 channels
in a single vector, this optimisation is pointless. The shader compiler
will undo it during the GLSL/HLSL to IR step.
Bug Fix: The about optimisation was also broken, applying the color light
equation to the alpha light channel instead of the alpha light
euqation. But doesn't look like anything trigged this bug.
Fixes a major preformance regression in Skies of Arcadia during
battle transisions.
I had plans for a more advanced version of this code after 5.0,
but here is a minimal implemenation for now.
Donkey Kong Country Returns is writing new data to some files in /tmp
when loading each level. But the savestate code was opening the files
a second time and reading some old and stale data out.
As of #3798, dolphin now correctly restores that stale data to /tmp,
which broke DKCR (and probally countless other games).
This PR closes all file handles before saving and loading savestates,
which flushes the data out and pervents this issue. (old savestates
are corrupted and will still cause crashes if loaded)
They all handled it diffrently, so I've just moved it into Advance()
This fixes Pokemon Box booting in JIT/JITIL which shared a bug where
exceptions set in a scheduled event would be ignored untill the next
slice (upto 20,000 cycles).
Because the file handles were open, the recursive delete was
failing. The previous commit stopped the crash but this should
make the restore actually happen has expected.
If the game sent a command to a disconnected controller, the
wii u adapter code would return a diffrent response.
This simply deletes the speclized version of RunBuffer for the
wii-u adapter as the only diffrence was the code which detected
disconnected controllers and returned a error.
During boot of Other M, there is momentarily a period when VICallback's
cycles late is larger than GetTicksPerHalfLine(). Because
GetTicksPerHalfLine() returns a u32 and c++'s weird type promotion rules,
cycleslate gets promoted from a s32 to a u32 and the result of the
substraction is a really large u32.
Before ScheduleEvent accuracy improvements, ScheduleEvent took a s32, so
the result got cast back to the small negitave we expect. But it now takes
a s64 and the u32 to s64 conversion gives us a really large number (around
two seconds) and Other M times out while waiting for something.
Now that the accuracy of ScheduleEvent has changed, 0 cycles will
schedule an event as soon as possible. But this breaks ATV 2.
So we schedule it 100 cycles out (unless it's a really short copy)
The NES games on the Zelda Collecters Edition disk use a XFB which is
only 256 pixels wide, but has a stide of 640 pixels.
This fits our definition of a interlaced xfb, as a second line of data
could fit in the extra space. The solution is to check that we are
actually in a interlaced video mode before activating the force
progressive hack.
Previously the default queryed the controller 4500 times a second,
Wasn't really a problem for most games as they set it to a sane
value. But fifoplayer didn't, and so in my profile dolphin spends
12% of the cpu time reading the controllers.
This new default value (I just took what the gamecube bios set)
drops that to 1.2% of cpu time and increase the framerate of the
silent hill fifo by 10-12%
Events scheduled more than 4.12 seconds in the future (2.96 seconds for
Wii games) would overflow the sign bit and get scheduled in the past
instead, causing them to fire instantly.
Previously GlobalTimer was only updated at the end of each slice
when CoreTiming::Advance() was called, so it could be upto 20,000
cycles off.
This was causing huge problems with games which made heavy use of
the time base register, such as OoT (virtual console) and Pokemon
puzzle.
I've also made it so event scheduling will be accurate to the jit
block level, instead of accurate to the slice.
Just because it was the same game, released in diffrent regions
at around the same time doesn't mean they use the same version
of the Virtual Console emulator.
The previous INIs were based of testing of the US release so I've
move them to 6 letter INIs only effecting the US release. I've
also added a few more and added a few Inis for the few PAL games
that I could confirm also didn't require EFB2Ram.
Since all Virtual console releases use the same emulator (though there
are multiple revisions some emulators) and generally need the same
setting, this commit allows the creation of INI files which cover all
Virtual Console games fow each system.
For example, F.ini can provide settings that all NES games have in common
and C.ini will provide settings for all Commadore 64 games.
If needed, a 3 letter ini can override settings for individual games.
They are now based on signal timings rather than pixels, as it
didn't make a lot of sense to do things with pixels.
Now handles all 240i/240p/480i/480p modes without any special
casing.
Despite the diffrent equaions, this should result in the exact same aspect
ratio as the previous code.
Detected by the previous commit. We had forgotten to make sure the
Vertex Loader's table of normal sizes is initilzied, resulting in
the wrong offsets/sizes being used for vertices with normals.
Scott Mansell