Slashdot Mirror
Bochs x86 IA-32 Emulator 2.1 Released
Asmodeus writes "Just noticed that the 2.1 release of the Bochs IA-32 emulator is out at the Bochs home page
For those not in the know, Bochs is an open source implementation of the x86 instruction set(s) and a virtual PC (al la VMWare) which is capable of booting FreeDOS and Linux under the host control of another OS."
Bochs emulates the IA-32 instruction set and enables you to run IA-32 software on any sort of hardware that you can compile Bochs on. (eg. I once ran it on a MicroVAX at an incredibly slow speed)
VMWare requires IA-32 hardware. Most of the instructions are executed natively and only some of the priviledged operations are emulated so that whatever is run under VMWare can work as if it has full control over the CPU while in fact being an un-priviledged task.
Bochs isn't just capable of running DOS clones:
Operating Systems inside the emulation including Linux, Windows(R) 95, DOS, and Windows(R) NT 4
It can also run Windows 2000 - and probably XP as well if product activation works.
Wow.... ummmm.... slashdot?
;-)).
Could we not post "news" about things that came out an eon ago? Seriously... ROFL,,,,
----->
Bochs is kind-of OK. I use it regularly when I work on my exokernel project and it really IS A GREAT developing/debugging tool (especially if compiled with the GDB stubs
However, however, however... I wouldn't consider Bochs useful for anything other than hacking around with kernel/os stuff. Bochs needs a re-write from scratch and emulate a real standard PC motherboard - not an 80386 with i486, pentium, athlon, mmx, PCI, USB, ATA etc... hacks around it. PCI support is non-existent. Video is flakey - well you can get VESA-compliant > 800x600 if you physically change the source (easy). All emulated devices are ISA "bus"-based. Over the years stuff just kind-of gotten piled on, and on and on - with no sensible strucure. I am not talking out of my ass either - at some point in my life I felt that Bochs would be a great project to hack.
Many others have already posted this, but VMware != Bochs, because VMware uses virtulization to run a guest OS with minor overhead on a host system. Bochs, on the other hand, emulates everything, even if the host system is IA32, causing massive performance degredation. I see that your applications are rather large scale(3DSMAX and Adobe applications) - and probably would rely heavily on graphics adaptor and memory. Bochs is definately not your answer, as if you could even get it to work, it would be so incredibly slow that you'll forget why you were doing it in the first place before the program even loads (trust me, it has happened before).
Look to VMware to do things like this - it may have a fee attached, but its fast and capable, but not open source.
Perhaps the person who submitted the article did so from inside Bochs.
/. article.
Gotta love that blazing static instruction translation speed! Only a month or so to fire up a web browser and post a
"The worst tyrannies were the ones where a governance required its own logic on every embedded node." - Vernor Vinge
But if it's retro DOS games you're after check out dosbox which runs pretty fast and runs on many platforms.
Doesn't it make you feel good to know that our freedoms are protected by politicans, lawyers and journalists.
I know that in the past, the number of registers of the PPC was far in excess of the capabillities of the x86. Example: No PPC emulator yet exists, no matter what vaporware merchants have said in the past.
Finally, my one experience with Bochs was on BeOS. I couldn't figure it out. On the other hand, Virtual PC was easy as pie. Why doesn't Bochs copy the usabillity of Virtual PC --- the gui is neat and clean, plenty of options; throw open source in the mix and we could have a weiner. (And a real alternative to MS owned, newly activation-coded Virtual PC.)
I stopped upgrading when MS bought it. It was only fair.
Xen has already been covered on slashdot
great for OS Systems development, but not much more. If you are programming an operating system, there is nothing better then Bochs running with gdb debugging stubs. Peroid. All this talk "Bochs can't run photoshop at a good speed" and "Bochs takes for ever to load windows" is bulljive. Of course it does, because that's not the point! Why do I own a 386 and use it with DOS 6.22? Because I want to do assembly programming and test out algorithms written in IA32 assembly. If I tried to run PS 8 and WinXP on it and subsequently complained about the speeds, I'd be flamed to death. The same goes for bochs. Kudos to the developers! A lot of great improvements were put into this release, everything from 3D assembly instructions to a whole new disassembler. Bochs is every OS Developer's dream come true. And it's just gettin' better... (Also, the best "bug fix" imho is that you don't need an extra font installed in X-Windows now). And if you want to emulate windows and have it run fast, go buy a $400 PC from Walmart. They play quake fine while waiting for the latest kernel to compile. :->
- Simrook
'Truth' is linked in a circular relation with systems of power which produce and sustain it...
If you want a free, open-source and (fairly) portable x86 emulator that provides better performance than Bochs then you could do far worse than QEMU. It uses a nifty dynamic recompilation techinque for its CPU emulation which gives much better speed than Bochs's interpretive emulation while remaining relatively easy to port.
It's a young project, and it has a long way to go before it'll be a real alternative to VMWare for most people, but it's getting there pretty quickly - the recently released 0.5.2 can already run Windows 98.
Gotta love that blazing static instruction translation speed! Auctually I gotta say I've noticed large speed improvements on i86 hosts between this and 2.02. Now if only they would release 2.1.1 already with the bug fix so you can compile in both x11 anf rfb(VNC) console support Also the bochs people outright admit that it is slow. They refuse to add any kind of trickery like running instructions natively on intel becasue its meant for debugging OSes and the like. Sometimes you need to be able to run through your OS's boot up one instruction at a time to find a bug. This allos you to do that.
--- Justin Dearing http://www.justaprogrammer.net/ We're just programmers.
This post has no point. It just provides some general (hopefully interesting) background info.
As many people pointed out, Bochs is an x86 emulator, rather than a virtualization system like VMWare. Emulation means that you have a representation of an x86 machine in memory, look at each instruction, and change the representation appropriately. Virtualization means the code runs on the actual CPU natively, and uses 386 ninja powers to intercept all I/O calls and reroute them to the base OS.
As a result, Bochs will run on any platform. VMWare will only run on x86. Bochs is slow enough to be useless for most common uses (a bit over a 100x hit in speed). VMWare has almost no hit in speed.
However, the free software community did have a project that attempted to reimplement VMWare. That project was called Plex86 (http://plex86.sourceforge.net/). For reasons that I do not know, Plex86 recently reinvented itself not to do full hardware virtualization -- rather, it does not implement the I/O layer, and instead provides special drivers for Linux to talk to its I/O layer. As a result, it can only run Linux (although it claims to run it reasonably well). They may implement drivers for other platforms, but I would be fairly sceptical of any real Windows support anytime soon. That seems a lot less useful now...
The Plex86 project, however, claims the possibility of using their virtualization technology in conjunction with Bochs to make a useable system: "There is the potential to use plex86 as an accelerator for bochs, as was demonstrated some time ago." (source: Plex86 FAQ). Likewise, it seems that if Bochs was more intelligently implemented, they could use just-in-time recompliation, a la Java or Transmeta, since they are effectively treating the x86 ISA as bytecode. That would be in the very, very distant future, but if either of these is implemented, the Bochs project is not as hopeless for end-user use as it may at first seem... Either or both of these technologies ought to give reasonable performance.
One problem is that VMWare is creating a patent minefield in front of Plex86 and Bochs. I am not familiar with all of the patents, but from what I've heard, they've got a pretty wide field of IP cut out. I'm not sure how hard they'll exploit it, since the people working there seem like nice guys, and understand the whole open/Linux/GNU/free/etc. thing. On the other hand, so did Caldera a few years back, and VMWare is definitely getting those patents for a reason....
One final point -- properly used, emulators like Bochs can provide amazingly powerful debugging tools. You can run a full x86 machine (admittedly at very slow speeds), but grab snapshots of the system memory at different points. You can then roll back, use a capture of all inputs to roll forward, etc.
On their main page in the first paragraph it says ,"Bochs can be compiled to emulate a 386, 486, Pentium, Pentium Pro or AMD64 CPU, including optional MMX, SSE, SSE2 and 3DNow instructions."
Regards,
Steve
Uhh, did you bother to read the link? Ohh wait, this is /.
If you HAD bothered to go to the BOCHS site you would notice that it DOES do 64-bit emulation. More specifically, it emulates the AMD64 instruction set (aka x86-64). This is rather nifty in that it allows developers to test out code for AMD64 without having to purchase the hardware. Obviously not an ideal development platform, but it could be useful for some.
Plex86 (and Xen, VMware, and Connectix, and Ensim, and others) are the things people should look at if they want fast virtualization of x86. The trouble all these technologies run into is that IO has to go through the "host" OS (the one actually running on the metal) - often popping into userspace to do it (read: context & ring switches --> slow!). This is necessary in order to allow multiple virtualized OS's to share the IO devices. This causes stuff that is IO intensive (games, compilers, databases, etc.) take a fairly serious performance hit. Interestingly enough, Intel is working on building this sort of capability in the chips directly - check out Vanderpool for instance. I don't know if AMD is doing anything similar, anybody heard anything?
Is lack of demand. Registers aren't relivant. It is a fact of a turing machine that any one can emulate any other. An x86 chip is perfectly capable of emulating a PPC chip. Now it might end up being slow (due to registeres needing to be in memory), but it would work fine. I actually have a feeling you could get it working pretty well. The 32 "general purpose" registers on a PPC actually aren't, many of them have specific tasks, and the number of registers actually on an x86 chip is not related to the number exposed by the ISA.
However, regardless, you can make an emulator. You can make an emulator in 100% C or Perl or Java if you like, and one that is portable to any platform. It needn't be anything low level. It'd be slow, but it'd work just fine.
Basically what it comes down to, is who wants a PPC emulator? I mean if you want a PPC system, get one. There are plenty available from IBM for reasonable prices. If it's Mac emulation you are looking at, well that's a problem. The Mac ROMs are not available outside of Mac hardware, nor is the OS, and without those, it is useless. So to run the emulator, someone would need a legit copy of the ROMs and OS, meaning they'd need to own a Mac. Well if you own the hardware an emulator is worthless.
x86 emulation on the Mac is of much more intrest. First off, it's actually feasable to do. PC BIOS is easy to license from a number of manufacturers, and MS is happy to sell copies of Windows, even for virtual machines. Also there are cases where you have a Mac and 99% of what you do is done natively but there is the ONE app that you need for something that is Windows only. So you get an emulator. Well the only Mac only apps I can think of are things like Final Cut Pro, which would run like shit in an emulator, so you'll have native hardware if you want to use it.
Finally, now I can install windows XP on my xbox!
(XP on Bochs on linux on xbox)
It is a fact of a turing machine that any one can emulate any other
It is a fact that they can. However that does not mean that it will be easy.
It'd be slow, but it'd work just fine.
This is exactly the problem: it would be slow. And up until a certain point, it's slow enough you might as well not do it at all. No, there's no commercial demand for PPC emulation on x86; there
doesn't really need to be. People write emulators just because they can. Do you think there is any "demand" for an emulator for the Amstrad CPC? In the meantime, there's some hobbyist demand from people who are "curious" about OS X; there's the guarantee of instant infamy for anyone who succeeds. People have really tried, put a lot of effort into trying to, emulate the PPC on an x86. I've never seen anyone succeed. As it turns out, though, writing a PPC emulator that runs on the x86 just happens to be unbelievably difficult to do with anything even remotely approaching an acceptable speed of emulation due to the neatly mismatching design philosophies of the two instruction sets. Yeah, if there was a real commercial *NEED* for someone to emulate, an acceptable emulator could probably be created. But the issue is a little more complicated than "oh, no one wants it".
If it's Mac emulation you are looking at, well that's a problem. The Mac ROMs are not available outside of Mac hardware, nor is the OS, and without those, it is useless. So to run the emulator, someone would need a legit copy of the ROMs and OS, meaning they'd need to own a Mac. Well if you own the hardware an emulator is worthless.
Not only is this not the hard part, this is the part that has already been solved. Modern macintoshes no longer have anything significant in ROM. The ROM is just a tiny kickstart thing and the OS is booted entirely using the openly documented Open Firmware protocol. This part is a non-issue.
Since the internals of an apple machine aren't that public, virtualizing the hardware might be a little bit difficult.. but, well, not that difficult, as practically all of the work has already been done for you in the form of the mac-on-linux project, a VMWare-like virtual machine for macintosh hardware that will let you boot OS X within a virtual machine on top of Linux. I am uncertain how much extra work needs to be done on top of that when emulating on the PC platform since I don't know what the internals of mac-on-linux look like. However, at the very least, the hardest and most voodoo-y part, actually getting it to boot, has already been done.
As far as the OS goes, you can buy a copy of the Mac OS without buying an actual mac. As in, you can go to a store and buy a copy of Mac OS X 10.3 in a box. This is not unrealistic; just because someone is emulating doesn't mean they aren't willing to actually buy the OS. Case in point, everyone who emulates Windows on the Mac does in fact actually have to buy a copy of Windows.
BTW, just out of curiousity, where are these PPC systems which you say are "available from IBM for reasonable prices"? I may just be going about it wrong, but I'm looking at IBM's website and the cheapest POWER-based system I can find is nearly $6000.
Irritable, left-wing and possibly humorous bumper stickers and t-shirts
I got bochs v2.1 to run Windows XP without any problems. The trick is to configure bochs with --enable-cpu-level=5 --disable-sse.
Here are some screenshots and a howto