Linux 2.6.20-rc6 Kernel Performance

Michael writes "The Linux 2.6.20 kernel will feature KVM support, Playstation 3 support, and a variety of other improvements. With the Linux 2.6.20-rc6 kernel out the door, Phoronix has written a performance comparison of the Linux 2.6.20-rc6 kernel against the 2.6.19 and 2.6.19.2 kernels in a variety of benchmarks."

What the...

[Jethro]

So the bottom line here is that they're almost exactly the same?

Re:PS-3

[jimstapleton]

I think that means that it is bootable on a PS3. This kind of thing would only be included if it was compiled for a PS3, and as for "is it required", only if you want to run arbitrary distros on a PS3, which there are people who'd want to do that.

Somebody owes me two minutes.

[Kadin2048]

So the bottom line here is that they're almost exactly the same?

Yeah, that was a totally worthwhile read, no?

Let me give everyone else the bottom line, and save you two or three minutes of your life, that you'll otherwise never get back:

Sony Playstation 3 support and Kernel-based Virtual Machine support are among the exciting features in this release. From today's testing in our environment used and set of benchmarks, there were no definitive performance gains or losses seen throughout the set of tests.

Now, back to our regularly scheduled Slashvertising....

Re:Regarding Playstation Support

[Anonymous Coward]

7SPE's are accessible via Linux, as long as you use the Cell BBE SDK..

Re:Regarding Playstation Support

[drinkypoo]

Does anyone know how much of the PS3's hardware is actually supported? When you run Linux on Cell, is it actually using all of the Cell cores, or is it just using the main (PPC-like) one?

Cell as implemented in the PS3 has 8 cells. One is disabled (probably due to poor yields when demanding that all 8 be working.) In Linux, one is devoted to kernel tasks. That leaves you with six Cell SPEs to work with besides the PPC PPE.

Seems like getting software to take advantage of it, would require changes both to the kernel, and also to GCC, in order to produce optimized binaries for it, not to mention various pieces of software themselves (rewriting for greater parallelizability).

Well, yes and no. The real problem is that the SPEs are only good for vector data. Anything else requires that you underutilize them. For instance if you have just two numbers and not a whole matrix to multiply, it takes equally long - you just have one useful result and a bunch of useless results that you didn't want. So certain kinds of tasks will be easier to optimize on the SPEs than others. But in many cases you can probably get good results by just using libraries... for instance if libz and libm were accelerated, that would probably make a big difference. Likewise for widget libraries, sound processing libraries, 3d...

3d brings us to the other point, which is that Linux runs in the PS3 "hypervisor" environment and you do not have unfettered access to the video hardware. I don't know precisely what you're not allowed to do that you can do in the commercial environment though; I've never seen a complete description of that.

Re:Looks like it got worse!?

[Enderandrew]

Did you only skip to the last page? If you looked over every benchmark, the new kernel had improved performance in almost every test, save for two of the last three.

Re:Isn't it just Microsoft-style "bloat"?

[radarsat1]

You misunderstand the Linux kernel ethos. The idea is to include *everything* "in the kernel", but you only have to compile the parts that you want. That way there is a central place to track all changes and maintain compatibility and consistency between all parts of the kernel, without having to set an internal interfaces in stone.
It's not "bloat" if it's only in the source. Simply put, you don't have to include PS3 support in your binary version. In other words, the only way it affects you is a few extra bytes to download when you want to compile it.

Re:You know what?...

[loonicks]

i thought quite the opposite. though, honestly, after seeing a couple of the graphs, i decided not to read the content. comparing application performance on different versions of the kernel seems rather stupid.. like seeing if your car goes faster when you give it nicer seats. consumer application performance is largely dictated by the application code itself and the hardware it runs on. maybe it can handle multi-user I/O and multi-process scheduling better, but I didn't see much of that here.

Re:Regarding Playstation Support

[tlhIngan]

The only thing that is limited is GPU support. Sony doesn't want you running pirated or homebrew games on it, so now graphics hardware acceleration (until someone writes an unofficial driver).

I've played with PS3 linux. I can tell you, the hypervisor is just that. It virtualizes the PS3 hardware. About the only thing Linux has "raw" access to (which could also be virtualized) are the USB ports.

The hard disk must be PS3-formatted before Linux will see it. Otherwise the hypervisor will not see it and make it available.

BTW: /dev/sda - hard disk, if available (else this node is the following devices.) /dev/sdb - flash memory - configuration area storing the bootloader (kboot), and a few configuration flags /dev/sdc - Memory stick, I believe /dev/sdd - SD Card /dev/sde - CompactFlash card /dev/sr0 - blu-ray drive.

The hypervisor is a lot like VMWare/Virtual PC/etc. I suspect the Power Processing Elements aren't even fully accessible and that the hypervisor is trapping everything and passing it on as appropriate, like virtualization software you run.

BTW, the virtualization also causes some issues. When I bought a new hard disk for PS3 Linux, it had bad sectors on it (I returned it in the end), but instead of the usual IDE error messages (DriveError) or SCSI errors (with media sense keys), you get nothing, other than a generic "I/O Error reading sector XXXX", which causes the filesystem in use to suddenly go read-only (not sure if ext3 did that or if the hypervisor just disabled the ability to write to the disk - I never had many bad disks with ext3). Basically, you don't even know it's a bad sector as it isn't reported. I suspected it when I could get dd to consistently put the filesystem into read-only mode 16GB in. Another system helped prove the point.

The video hardware is identical - it's virtualized the same way. It's not a driver issue - it's just that Sony has virtualized the video hardware away, and there's no direct access available. Heck, there aren't any WiFi devices accessible either - not for lack of a driver, but that Sony didn't make the WiFi hardware accessible.

Re:Why not?

[steveha]

There is support for hardware in the kernel that is so obscure that there are probably less than 100 people in the world still using it.

I attended FreedomHEC in Seattle last year. Greg Kroah-Hartman gave a talk, and one point he made was that there are devices supported by the Linux kernel that are literally known to have only one or two users in the whole world; we are talking devices that are so obscure that only one or two people are known to even possess the hardware.

The point he was making is: if you make some hardware, and you are wondering whether your device is too obscure for Linux to accept drivers for it... don't wonder, just submit the drivers.

steveha

Re:Regarding Playstation Support

[drinkypoo]

You didn't answer the parent's question completely, I think. Which cell is the kernel using (I'm assuming the PPC PPE), and more importantly, does the PS3 support in the Linux kernel enable access to the Six Cell SPEs in anyway, or are you basically on your own.

This paragraph proves that you do not know what we are talking about.

The PowerPC in the "Cell" processor in the PS3 is not, repeat NOT a "Cell". It is known as the PPE, or Primary Processing Element (IIRC.) The actual Cell processors in the PS3 are known as the SPEs. I forget what the S stands for, maybe synergistic or something like that. I honestly cannot remember and the information is out there so YOU can look it up :)

The chip in the PS3 has one PPE (a PowerPC with VERY roughly the processing power of a late-model pentium III) and eight SPEs or Cell cores. One of those cores is disabled, probably in order to improve yields. The Linux kernel runs on both the PPE, and ONE of the SPEs. The other six available SPEs are available to the user. Note that all of this information appears in the comment to which you replied, but I am now being more redundant and overusing emphasis in an attempt to get you to actually read it.

3d brings us to the other point, which is that Linux runs in the PS3 "hypervisor" environment and you do not have unfettered access to the video hardware. [...]

Seems like that defeats the whole purpose doesn't it? PS3 is all about the 3D graphics.

Well, I agree that I wouldn't buy one, even if I weren't boycotting Sony. This is hardly the only major flaw; another is that unlike the Xbox 360 which utilizes a unified memory architecture (UMA) the PS3's memory is both physically and logically broken into two pools of 256MB each; one for programs and other resources, and one for graphics. This means that the PS3 has only 256MB (minus whatever hypervisor overhead there is) of user-addressable memory.

There are basically two purposes to using the PS3 for other than its design purpose. One is for scientific computing. For certain types of highly-parallelizable problems, the PS3 is by far the cheapest bang-for-your-buck. Anything that can be broken up into data sets which can reasonably be transferred over 100Mbps ethernet and processed on systems with only 256MB RAM, and which furthermore can be executed as a series of matrix operations, will probably run faster (or for less money, depending on how you look at things) on a cluster of PS3s than any other option available today. Of course, only certain types of data are amenable to this type of manipulation. The other purpose is for use as a media center system. An HD decoder which utilized the available SPEs would certainly be able to handle decoding HD video on the fly (The PS3 does it already, of course, for playing Blu-Ray) and with a little hacking it should even be possible to get the Xbox 360's HD-DVD drive working on it so that you could have one system capable of playing both HD video formats before hybrid players even hit.

Re:Regarding Playstation Support

[SpectreHiro]

This paragraph proves that you do not know what we are talking about.

Well, I think that's self-evident, since he's asking a question about it.

The PowerPC in the "Cell" processor in the PS3 is not, repeat NOT a "Cell". It is known as the PPE, or Primary Processing Element (IIRC.) The actual Cell processors in the PS3 are known as the SPEs. I forget what the S stands for, maybe synergistic or something like that. I honestly cannot remember and the information is out there so YOU can look it up :)

The chip in the PS3 has one PPE (a PowerPC with VERY roughly the processing power of a late-model pentium III) and eight SPEs or Cell cores. One of those cores is disabled, probably in order to improve yields. The Linux kernel runs on both the PPE, and ONE of the SPEs. The other six available SPEs are available to the user. Note that all of this information appears in the comment to which you replied, but I am now being more redundant and overusing emphasis in an attempt to get you to actually read it.

If I could offer a couple clarifications (although I'm surely no expert):

1. The PPE stands for Power Processing Element (IIRC).
2. The S in SPE does in fact stand for Synergistic.
3. I've never heard anyone, in any article or any of the whitepapers I've read, refer to the SPEs as Cell processors, Cell cores or Cells. The Cell Broadband Engine is the entire chip, while the SPEs are precisely that... SPEs. The root of the confusion in this thread stems from A) Referring to processing elements as Cells (etc.) and B) Arbitrarily deciding that the term applies to the SPEs and not the PPE.

I can't really blame the fellow for being confused at your reply. I suppose this is a fine example of why people should do their own research instead of asking around on /.

I usually really enjoy your posts, drinkypoo, but, with all respect, in this case, I think you could stand to fling around a little less attitude.

Re:Same here

[this great guy]

No, it is a password cracker. More info in a couple of week, when I release it as open source.