b2 - new emulator

[tom_seddon]

I was inspired by Thomas Harte's Electron emulator to write another BBC emulator, that works internally along vaguely similar principles. Sadly I clearly still haven't quite got things right, because mine doesn't look anything as nice as Thomas's, even ignoring the composite encoding stuff he's done (mine just outputs boxy pixels)... but I think 50Hz stuff still feels quite good, even though it can look rather scrappy.

Anyway, I'll be continuing to chip away at this over time, along with all the other issues it has.

You can get it here: https://github.com/tom-seddon/b2/releases - binary builds for OS X (the dmg) and Windows (the zip). Builds are ordered most recent first, so just grab the one at the top!

Linux users need to build it themselves, I'm afraid, but the dependencies are not terribly complicated...

This is a work in progress, so there will be bugs. If you have any problems, please post here or file a GitHub issue (https://github.com/tom-seddon/b2/issues).

Features:

* Supports Windows, OS X and Linux
* Emulates BBC B (multiple disc interface options), BBC B+, Master 128
* UI for editing ROMs, keyboard mappings and emulator options
* Quick state save/reload
* Timeline view for visualizing state history
* Windows/Linux: generate video from timeline
* Experimental turbo disc access
* Debug version with instruction logging functionality (this is not documented at all yet, though, I'm afraid...)

--Tom

[ctr]

This makes such a realistic - and horrible - disc drive noise I thought something had gone wrong my laptop. Then I remembered I don't have a floppy drive!

The timeline and clone features are really nice. I had thought of adding a state recorder to beebem but I needn't bother now. Is it possible to save the timeline? If the timeline were integrated with a debugger it would be awesome.

I had a quick play of ChrisB's Castle Defender and it worked flawlessly.

The text in the overlay graphics is very "bitty". Picture attached. I suspect it's targeted at a higher resolution screen. Mine is only 1366x768, running Windows 10 64-bit.

I'll have a go at building it later.

[tom_seddon]

Thanks for trying it out! I'm sorry it looks so hideous :( - it's definitely never meant to look that way! The text rendering is supposed to take window size into account, so it's supposed to render correctly whatever the resolution, but clearly there's something wrong. I'll try it this evening on my laptop at 1280x800 and see what I get.

Please could you go to Tools|Messages, click Copy, and paste the output here? This might help narrow things down, though maybe not (the output doesn't go into much detail). Also what sort of GPU does your laptop have?

Regarding save timeline: at the moment you can't, but this is on my to do list (which I'll be adding to the GitHub issues list at some point so that future features are visible). Would be nice to be able to use it for doing something like a full, perfect Repton 2/Exile/Citadel/etc. playthrough, which you could then produce a video of, or just as a more comprehensive alternative to saving stuff to disc.

(I made a video of me playing through the first few levels of Thrust (don't know why YouTube made a mess of the colours...). That probably took me about half an hour... being able to undo was an absolute godsend. But I doubt it'd be much fun to use if you're creating something a lot longer than that.)

A debugger is also on the cards - my plan is that the debugger will be a non-modal popup, like the timeline window, so when you load a state then the debugger will just instantly reflect the new state like the rest of the UI does.

--Tom

P.S. Let me know if you have any problems building it... it ought to work but it's only ever been built on my laptop.

[Rich Talbot-Watkins]

Looks like a great start Tom! A couple of comments from my first try:

- Exile isn't detecting Sideways RAM in any model
- The disc drive noises in the default configuration are odd, as if the buffer is underflowing and it's repeating the same little bit of sample over and over.

What's the "50Hz stuff" you allude to in your original post?

By the way, I've also had those grey lines when video encoding - seems to be an artefact of the encoding process when you have striped horizontal lines like that (I saw it when capturing an interlaced screen where every other line was blank).

[tom_seddon]

Thanks! I take it the text output was OK for you?

The disc noises do need looking at - I basically spent about 10 minutes copying what MAME does, wasn't really happy with the results, decided to revisit it later, and then never did... or at least, didn't just yet.

The sideways RAM thing is fixed in the github repo, thanks (will do another release after I've poked at the text rendering stuff a bit). For now there's not much in the way of automated testing for anything but the lowest levels of the code unfortunately...

As for "50Hz stuff" - you can actually disregard that comment, but it refers to 50Hz games having that somewhat elusive smooth "feel" even when running on the PC with a different refresh rate. Since b2 has decoupled display data production and consumption, you ought not to get the sort of 3:2-pulldown-style effect that I thought I could detect in B-em while working on Rig Attack last week, when I thought it quite noticeable that b2 felt better (though the tearing did make it look a little weird).

But, trying it just now today, I actually think B-em has the edge: no real difference in smoothness, and no tearing to boot. So I take it back. Probably just because I'm not running Twitter in Firefox right now ;)

This side of things does still need some work.

--Tom

[Coeus]

Tom, Thanks for this. Not building on Linux, at least for me, though. I get this:

Code: Select all

CMake Warning at CMakeLists.txt:42 (find_package):
  By not providing "FindSanitizers.cmake" in CMAKE_MODULE_PATH this project
  has asked CMake to find a package configuration file provided by
  "Sanitizers", but CMake did not find one.

  Could not find a package configuration file provided by "Sanitizers" with
  any of the following names:

    SanitizersConfig.cmake
    sanitizers-config.cmake

  Add the installation prefix of "Sanitizers" to CMAKE_PREFIX_PATH or set
  "Sanitizers_DIR" to a directory containing one of the above files.  If
  "Sanitizers" provides a separate development package or SDK, be sure it has
  been installed.


CMake Error at CMakeLists.txt:69 (message):
  SDL not found in folder:
  /home/me/Computing/BBC/b2/b2-0.0.1/submodules/SDL

The SDl directory referred to seems empty. Not sure if it is complaining about the contents of that directory or that SDL is not installed system-wide. It should not be the latter as I have SDL1 and SDL2 installed.

[tom_seddon]

The repo has submodules - have you done a git submodule init/git submodule update? Looks like the README instructions aren't terribly clear, now that I look at them again ;) - I've made a note to update them.

(You can also do the initial clone using git --recursive, which will clone any submodules at the same time.)

The build process uses the submodule's version of SDL2, not the system one - the submodule copy is a customized one with a patch that adds 2D geometry rendering. (See https://bugzilla.libsdl.org/show_bug.cgi?id=1734)

--Tom

[ctr]

If I shrink the window the beeb text also gets all lumpy. There is no anti-aliasing going on.

Please could you go to Tools|Messages, click Copy, and paste the output here?

Renderer: direct3d, 736x576 SDL_PIXELFORMAT_ARGB8888
Sound: directsound, 48000Hz 1-channel (4096 byte buffer)

Also what sort of GPU does your laptop have?

It's an ancient Mobile Intel 4 Series Express Chipset Family GPU, specifically a GMA 4500MHD.

[Coeus]

The repo has submodules - have you done a git submodule init/git submodule update? Looks like the README instructions aren't terribly clear, now that I look at them again - I've made a note to update them.

(You can also do the initial clone using git --recursive, which will clone any submodules at the same time.)

The build process uses the submodule's version of SDL2, not the system one - the submodule copy is a customized one with a patch that adds 2D geometry rendering. (See https://bugzilla.libsdl.org/show_bug.cgi?id=1734)

--Tom

Thanks. I did read that bit but when it referred to using git I assumed it only applied when you had obtained the source for the top level via git clone. I assumed that the tar.gz distribution would have the sub-modules already in. I am not saying you should bundle them, only that the documentation can be read more than one way.

[richmond62]

as "the building instructions" is a dead link,
and as ./compile doesn't, I'm in the dark..

Would be grateful for "the building instructions".

[richmond62]

Refuses to open on Mac OS 10.7.5

Process: launchd [2242]
Path: /Applications/B2/b2 Debug.app/Contents/MacOS/b2
Identifier: com.tom-seddon.b2
Version: ??? (???)
Code Type: X86-64 (Native)
Parent Process: launchd [194]

Date/Time: 2017-05-15 23:01:36.959 +0300
OS Version: Mac OS X 10.7.5 (11G63)
Report Version: 9

Interval Since Last Report: 2281255 sec
Crashes Since Last Report: 3648
Per-App Crashes Since Last Report: 14
Anonymous UUID: 36236C17-6A2F-484F-AFA5-3EB6F9F29C70

Crashed Thread: Unknown

Exception Type: EXC_BAD_ACCESS (SIGSEGV)
Exception Codes: KERN_INVALID_ADDRESS at 0x00007fff5fc01028

Backtrace not available

Unknown thread crashed with X86 Thread State (64-bit):
rax: 0x0000000000000055 rbx: 0x0000000000000000 rcx: 0x0000000000000000 rdx: 0x0000000000000000
rdi: 0x0000000000000000 rsi: 0x0000000000000000 rbp: 0x0000000000000000 rsp: 0x0000000000000000
r8: 0x0000000000000000 r9: 0x0000000000000000 r10: 0x0000000000000000 r11: 0x0000000000000000
r12: 0x0000000000000000 r13: 0x0000000000000000 r14: 0x0000000000000000 r15: 0x0000000000000000
rip: 0x00007fff5fc01028 rfl: 0x0000000000010203 cr2: 0x00007fff5fc01028
Logical CPU: 0

Binary images description not available


External Modification Summary:
Calls made by other processes targeting this process:
task_for_pid: 1
thread_create: 0
thread_set_state: 0
Calls made by this process:
task_for_pid: 0
thread_create: 0
thread_set_state: 0
Calls made by all processes on this machine:
task_for_pid: 1400
thread_create: 2
thread_set_state: 0

Model: iMac5,1, BootROM IM51.0090.B09, 2 processors, Intel Core 2 Duo, 2.16 GHz, 3 GB, SMC 1.9f4
Graphics: ATI Radeon X1600, ATY,RadeonX1600, PCIe, 128 MB
Memory Module: BANK 0/DIMM0, 1 GB, DDR2 SDRAM, 667 MHz, 0xCE00000000000000, 0x4D342037305432393533435A332D43453620
Memory Module: BANK 1/DIMM1, 2 GB, DDR2 SDRAM, 667 MHz, 0x7F7F7F7F7F9B0000, 0x4354323536363441433636372E4B31364600
AirPort: spairport_wireless_card_type_airport_extreme (0x14E4, 0x87), Broadcom BCM43xx 1.0 (5.10.131.36.15)
Bluetooth: Version 4.0.8f17, 2 service, 18 devices, 1 incoming serial ports
Network Service: AirPort, AirPort, en1
Serial ATA Device: WDC WD2500JS-40TGB0, 250.06 GB
Parallel ATA Device: PIONEER DVD-RW DVR-K06
USB Device: Built-in iSight, apple_vendor_id, 0x8501, 0xfd400000 / 3
USB Device: Expansion, 0x0bc2 (Seagate LLC), 0x231a, 0xfd300000 / 2
USB Device: Hub in Apple USB Keyboard, apple_vendor_id, 0x1001, 0x1d100000 / 2
USB Device: Apple Optical USB Mouse, apple_vendor_id, 0x0304, 0x1d120000 / 5
USB Device: composite_device, 0x050d (Belkin Corporation), 0x0805, 0x1d130000 / 4
USB Device: Apple USB Keyboard, apple_vendor_id, 0x0202, 0x1d110000 / 3
USB Device: Bluetooth USB Host Controller, apple_vendor_id, 0x8206, 0x7d100000 / 2
USB Device: IR Receiver, apple_vendor_id, 0x8240, 0x7d200000 / 3

[tom_seddon]

Thanks. I did read that bit but when it referred to using git I assumed it only applied when you had obtained the source for the top level via git clone. I assumed that the tar.gz distribution would have the sub-modules already in. I am not saying you should bundle them, only that the documentation can be read more than one way.

Ahh, yes, thanks - I see what you mean now. Sadly it turns out the .zip and .tar.gz (which are prepared automatically by GitHub) don't contain the submodules, presumably it uses `git archive', which (for reasons unknown) just ignores submodules. What a pain. OK, so those files are completely useless then... maybe there's some easy way I can make equivalents with all the dependencies bundled in?

Anyway, sorry for the bother, but it looks like getting everything from GitHub is the way to go for now. I'll update the README to make this clearer.

--Tom

[tom_seddon]

as "the building instructions" is a dead link,
and as ./compile doesn't, I'm in the dark..

Would be grateful for "the building instructions".

Link fixed, sorry about that.

<<crash report>>

It's very possible the binary won't run on anything pre-El Capitan (which is what I use). I'd be interested to hear how you get on when trying to build it yourself.

This does remind me there's a clang switch you can use to configure the minimum OS X version, so I'll give that a go and see what happens. Perhaps if I set the min version to 10.7 then it'll create a binary that works.

Thanks,

--Tom

[richmond62]

a clang switch you can use to configure the minimum OS X version

That would be very kind as not all of us have up-to-date Macs (my newest is this 2006 iMac).

[Coeus]

Thanks. I did read that bit but when it referred to using git I assumed it only applied when you had obtained the source for the top level via git clone. I assumed that the tar.gz distribution would have the sub-modules already in. I am not saying you should bundle them, only that the documentation can be read more than one way.

Ahh, yes, thanks - I see what you mean now. Sadly it turns out the .zip and .tar.gz (which are prepared automatically by GitHub) don't contain the submodules, presumably it uses `git archive', which (for reasons unknown) just ignores submodules. What a pain. OK, so those files are completely useless then... maybe there's some easy way I can make equivalents with all the dependencies bundled in?

Anyway, sorry for the bother, but it looks like getting everything from GitHub is the way to go for now. I'll update the README to make this clearer.

--Tom

Thanks. Not knowing what "the usual submodule nonsense" was I went for git clone --recursive and it has built fine.

I am interested that you have a GUI with SDL. Thomas was talking about porting B-Em to SDL but that didn't seem to have much in the way of GUI toolkit functionality built in - did you find a library that provides that based on SDL's basic drawing capability?

Also interested to notice a different 6502 implementation and interestingly one which is stepped on externally, i.e. something outide makes calls to it for each clock cycle. That is very different from how B-Em, for example, works where the 6502 runs a burst of instructions in a loop and then makes calls out to the modules that emulate other devices after the appropriate number of cycles. The approach in b2 makes the 6502 much less tightly coupled into the rest of the system. Wearing my purist hat, given that B-Em emulates two 6502s, the main (i/o) one and a tube co-pro I'd ideally want them to be two instances of the same code hooked up to the rest of the system differently whereas in reality they are two copies of what was once the same code adapted in different ways.

Anyway I'll have more of a play but so far it looks very good.

[Rich Talbot-Watkins]

Also interested to notice a different 6502 implementation and interestingly one which is stepped on externally, i.e. something outide makes calls to it for each clock cycle. That is very different from how B-Em, for example, works where the 6502 runs a burst of instructions in a loop and then makes calls out to the modules that emulate other devices after the appropriate number of cycles. The approach in b2 makes the 6502 much less tightly coupled into the rest of the system. Wearing my purist hat, given that B-Em emulates two 6502s, the main (i/o) one and a tube co-pro I'd ideally want them to be two instances of the same code hooked up to the rest of the system differently whereas in reality they are two copies of what was once the same code adapted in different ways.

Yes, it's a much tidier approach, and one I've also played around with recently (including modelling the data bus as a separate piece of architecture which interfaces with the memory 'system'). B-Em's approach is probably a little faster, but it means the CPU emulation is essentially 'running' the entire system, rather than it being a collection of equal components (why is the CPU somehow any 'better' than the CRTC, which is also continually running a state machine of its own? - the answer I guess is the CPU is the most complex component).

It's possible that a compiled language which supported coroutines as a native language feature might offer the best way to approach emulator code like this (as essentially what you have is a whole bunch of components running simultaneously), but I'm not aware of such a language (and coroutine libraries for C or C++ can be a little hacky - it's something which requires true language support, for multiple concurrent stack frames).

[ctr]

It's possible that a compiled language which supported coroutines as a native language feature might offer the best way to approach emulator code like this (as essentially what you have is a whole bunch of components running simultaneously), but I'm not aware of such a language (and coroutine libraries for C or C++ can be a little hacky - it's something which requires true language support, for multiple concurrent stack frames).

Ada and Modula 2 are compiled languages that have coroutines and no garbage collection. (Then I thought, and surely Modula 3? But it doesn't.) If you're not fussed about garbage collection there are go and Haskell. Iterators in C# also work as a poor man's version.

I guess real threads wouldn't be any good because there's far too much communication needed between the components.

(Fun fact from Wikipedia: Modula-2+ was used by the Acorn Computers Research Centre!)

[ctr]

I built it on Windows with cmake 3.8.1.

gen_6502.py failed because I've got python3. I replaced all the .iteritems() with .items() and that fixed it.

You need to add python2 to the instructions as a prerequisite. I need to read the instructions.

Building open source projects with lots of dependencies is usually a pain on Windows but it basically just worked. Nicely done.

Edit:

In dear_imgui.cpp I set:

Code: Select all

SDL_SetHint(SDL_HINT_RENDER_SCALE_QUALITY,"linear");

where it had previously been "nearest". This improved the text rendering but it's still not very readable.

[tom_seddon]

Great, glad it worked. Previous attempts at building other people's open source projects on Windows were a big influence on how I set this project up ;) And thanks for the note about python 3 - if it's that simple then I ought to be able to add a wrapper function so that gen6502 is version-agnostic.

I've made a new release that hopefully fixes OS X 10.7, fixes that completely stupid sideways RAM thing, and has a filtering toggle that might (feedback sought!) help with broken text rendering: https://github.com/tom-seddon/b2/releases/tag/0.0.2

--Tom

[ctr]

I edited my previous comment while you were posting. The "linear" filter is an improvement.

[Lion]

Does not start on my Mac.

macOS 10.12.5

[Lion]

...but starts up OK if I run Applications/Emulators/b2/b2.app/Contents/MacOS/b2 from a Terminal.

[Lion]

There's something weird about mode 7, the aspect ratio doesn't seem correct. It's too wide, or possibly not tall enough, not sure which.

Other modes seem to look fine.

It's great to have another emulator for the Mac. I was beginning to worry that one day BeebEm would no longer work on some macOS update. Very happy!

[Rich Talbot-Watkins]

Possibly everything is too wide, because Thomas Harte concluded that Beeb pixels are not exactly square, but actually slightly taller than they are wide. That MODE 7 basically looks OK to me though.

I noticed there's no anti-aliasing - does that mean you're actually building a very big bitmap which you can divide by 3, 6 or 12 (for regular modes' pixels) or by 4 (for MODE 7 subpixels)? There seems to be less than a pixel's gap between MODE 7 glyphs so maybe there's something else going on there.

[Rich Talbot-Watkins]

Here's a weird bug (0.0.2) - when I launch, the overlay text looks fine. If I then clone the window, I get something which looks nearest-neighboury and faded out:



Windows 10 64-bit, GTX 1070

[Rich Talbot-Watkins]

Hmm OK, the cloned one is with Filter UI enabled. Basically that doesn't seem to look very good. Unfiltered looks crisp and nice.

[Rich Talbot-Watkins]

More feedback: it looks as if you're not displaying screen data during the vertical total adjust section if the vertical displayed total hasn't yet been reached (see attached disc image; new rows should be scrolling on pixel-by-pixel, not waiting until the entire row is visible).

[Elminster]

Does not start on my Mac.

macOS 10.12.5

That is what happens when you are bleeding edge I am old school 10.12.4, will upgrade tonight.

As to your comment about BeebEm on Mac, you have seen the post with new versions of BeebEm on Mac being pushed out? If so have you tried it and compared to Tom's new version? Will have a look myself as soon as I get around to it.

[richmond62]

run Applications/Emulators/b2/b2.app/Contents/MacOS/b2 from a Terminal

Um: over "here" on 10.7.5 I get:

"Illegal Instruction: 4"

Which I sincerely hope means more to Thee than it doth to Me

[Elminster]

I get say as Lion on MAc OS 10.12.5 i.e. I have to open it from within the package but then I have to do that on BeebEm as well.

I think I must have misunderstood the keyboard remap as I ended up mapping * to <Shift> Key!