In commit ec325ef14f I did something stupid from
a performance standpoint, which was to select the shading function on a
per-pixel basis even though it can only change on a per-polygon basis. This
commit uses c++ templates and inline functions to so that the compiler's
constant folding can eliminate the switch statement that selects shading
functions (thus making the choice a per-polygon one). I also did the same thing
for the texture and bilinear filtering checks.
This brings back a little less than half of the performance hit on my PC, so
things still aren't as fast as they used to be.
Benchmark (running a full cycle of Crazy Taxi NAOMI's attract) on my PC:
Before ec325ef14f: 30%
After ec325ef14f: 23%
With this commit: 26%
Regarding the other 4% of the drop, that is coming from the pixel shading.
When I hardcode powervr2 to always use shading instruction 0, performance
jumps up to 29.5%, but when I hardcode it to use shading instruction 3
performance falls a bit down to 25%. Shading instruction 3 is the most
complicated of the four and shading instruction 0 is the simplest, so I think
that proves that the shading instructions themselves are a major bottleneck.
I don't have any ideas about how to make them faster yet.
* tlcs870 - takes first interrupt (nw)
still trying to work out exactly how I want this code to be formed, but this keeps pushing the program execution forward at least
* bit of port interface guessing to pass port based shared RAM check (nw)
* (nw)
* add analog ports, fix some tlcs870 exec issues (nw)
The Burning Cavern (31/03/87) [Miguel Bragado, Recreativas.org, ArcadeHacker, David Haywood]
A Day In Space (31/03/87) [Miguel Bragado, Recreativas.org, ArcadeHacker, David Haywood]
new clones marked as WORKING
War Mission (WM 09/04/87) [Miguel Bragado, Recreativas.org, ArcadeHacker, David Haywood]
there was also a disk for Paris Dakar, but the first tracks are damaged beyond repair, and it doesn't work, so haven't added it.
I'm not sure this API is a great idea - see inline comment on 796abaf7f3
I realise it seems intuitive, but it breaks with a use-after-free (not even a
validity error) if you replace/remove the target device. Implementing it in a
way that isn't fragile and doesn't hurt the performance of -romident,
-validate and -listxml seems impossible. If it causes developer confusing and
things start breaking, back to literal tags we go for this case.
* Clean up some corner cases in layouts with repeating blocks
* Make complay.py validate many more elements and attributes
* Make complay.py easier to use for just validating a layout
* Remove redundant view from Sega VMU layout
* Make buttons visually respond to input in whousetc.lay
* Add view with LED displays as well as terminal for aim65_40 and use repeats
* Clean up some outdated "game" terminology in clifront.cpp
* Initiaise a couple of members in tap/tun network module
* Start documenting layout format
* taptun: add support for Windows
Looking for feedback only at this point, because it's the first time I've attempted any OSD stuff - be gentle.
This extends the existing taptun OSD module to support Windows through the TAP-Windows6 driver (https://github.com/OpenVPN/tap-windows6).
* TAP-Windows6 is GPLv2, however only the header file is required in MAME, and the driver itself is entirely optional.
* I've tried to minimise the size of the diff, rather than completely separate the Windows/non-Windows implementations - not sure which is preferable in this case.
* The license file has Mac(?) line-endings - unsure if they should be normalized or used verbatim.
* Uncertain about the non-Windows case (or the Windows SDL case) - existing code uses __linux__, but I'm not clear on how that works with OSX, for example?
* I can't claim this actually "works" yet, because I don't know enough to configure the tap end of it properly to get networking doing something useful, but I will get there and correct anything necessary in a subsequent real PR.
* minor fixes/improvements (nw)
* Use Unicode Windows APIs, and UTF-8 for MAME
* Deal with unterminated registry string values
* Cancel any pending I/O in destructor
-----------------------------------
ibm5170: ABS Self Running Demo, Disney's The Lion King, Disney's The Lion King (Playable Demo), Microsoft Productivity Pack for Windows, Microsoft Works 3.0b, NETCOM NetCruiser [Justin Kerk]
New not working software list additions
-----------------------------------
ibm5170: Mitsumi IDE CD-ROM Driver [Justin Kerk]
- Screen raw parameters (circuitry is surprisingly similar to Galaxian)
- Addressable latch for video CPU
- VBLANK callback instead of MCFG_DEVICE_VBLANK_INT (interrupt also acknowledged now)
- Separate machine configuration for wcup90, which now displays something
This is damn sensitive code, and generates differences all over the
place we don't really explain. The changes should be justified by
themselves and tested in collaboration with Tafoid to ensure the
differences are not a problem.
