What is Digg?53 Comments
- geminitojanus, on 10/12/2007, -1/+9It's actually kinda funny. I just digested what I could of the Octopiler's "documentation" they had online (thanks Ars for pointing me in that direction), to be seriously honest, if I were Sony, I'd be sitting with a really terrible judgement call on my hands. Our cheif chip provider who swore up and down about this chip's capabilities and its availability, has completely and entirely fumbled the ball on us. They've got the chips and they're being made and everything's well and dandy there (though late), but now they're having problems making the chip work for developers.
The fact is it's just damned hard to fill up something so vectorized like the Cell processor. I didn't think it would be at first, but then I started to do some vector work on my own (research mostly, I was going to develop an application to attempt to use the Cell processor to do some artificial intellegence research), and I've come to find that it's just really damned hard work to maximally vectorize your code without writing it out in assembly; the C compilers out there just aren't that good at auto-vectorizing, and Hannibal goes into it a little in the article about the struggles they are having with it.
Sony just doesn't have a year or two to wait on this chip coming to maturity and the applications on it coming to maturity. It has got to roll out the doors in a stunning blaze of glory, or it's lost to the XBox 360 with its eternally easier to use processor (and to be honest, most likely superior operating system; there are so many developers out there who are used to working with the NT kernel and DirectX that it's going to feel just like home to them). That was an all around smart move by Microsoft, and if I were Sony, I'd feel a bit shafted by IBM and Toshiba on this.
The hard decision for Sony now is, look, do we really still want to go with this processor? It's clear now that it's simply late, and that nothing IBM or anyone else can do right now can speed it up. Sure, the game developers are going to be at a loss changing chips this late in the game, but it's very likely they've been having troubles vectorizing their code too (especially if IBM is having troubles). Maybe now's the time to conceed defeat on the Cell front. It would be a crushing blow as it'd put them at least a year behind, but honestly it might be better at this point to consider something like the Power970M and redesigning your boards in a flash than shipping a half crippled product to your consumers. (Hell, if they had the foresight, they might have even had this second board in development in deep secret, just in case something like this happened with the Cell, especially after their collossal fallthrough for Apple and delivering a mobile G5).
Yeah, I know Sony's too entrenched at this point to make that kind of choice, but it's definitely got to be something on their minds. This was a bad move, and strategically it's killing them in this gaming generation. They're going to be playing catchup to the XBox 360 (and in a lot of ways the Revolution, but the Rev just won't be at the same level of competition [no hd, radical new controller design, much slimmer hardware specs.. budget box gaming system], so I'm not quite going to go there).
It's a sad predicament to be in really. - altodarknight, on 10/12/2007, -0/+4Clearly not why the PS3 is running late, but highlights why the PS3 will be late, which is the too much hardware. Not only will it be a bitch to get a decent chip yield, the costs will be going through the roof, all for numbers such as 2 teraflops, etc. The sad fact is that these numbers have very little meaning in gaming and happen to make it a bitch to program for. Apart from some obvious advantages from blu ray (not including cost), the hardware does little more than the 360 in places and less in some places, such as bus speed, disc drive speed, GPU. Sony has been weak in the past when it comes to developer support, something which microsoft it good at, and this shows a tool that may be useful to developers, at least in the theory of the tool. Will it be enough, because in the end, it's the games that are important (even if the add on stuff is really cool such as live and media centre extender) and developer's make games. Sony has got to support developers heavily, because the tools supplied my Micosoft are fantastic (I've used them) and make it cost effective to develop for PC/xbox/360 very easily. Sony has got to open cost effective methods to combat these 3 income streams that Microsoft has opened for developers. Already games such as the next wolfenstin will be PC and 360 only.
- Ultim8Fury, on 10/12/2007, -0/+2kudos to arstechnica for being able to get past the first page on that IBM report and to transcribe the important stuff into English.
- geminitojanus, on 10/12/2007, -0/+1"No Digg. I know a person who works in the IBM chip factory. That person told me that they have already shipped Sony every chip that they've ordered and more. The holdup is definitely elsewhere."
This article has absolutely NOTHING to do with the chip, had you read it you'd know that.
It entirely states that the Compiler development is behind schedule. And Compilers are sort of a "big deal" when it comes to building the actual application that will run on that platform. ;) - norick, on 10/12/2007, -0/+1"the biggest challenge facing programmers who write non-deterministic applications for a highly multithreaded SoC like Cell: debugging."
Yep, we will have tons of bugged games, patched via the "ps3 live" or also called HUB.
Welcome to the PC world.
Nintendo, i am your man. - ThirdPrize, on 10/12/2007, -0/+1Whenever the PS3 does come out, you will have to add 6 months to get any games that actually make use of the architecture.
- WackyT, on 10/12/2007, -0/+1I think if this does progress smoothly, the PS3 will be a kick-ass console in about 1.5 to 2 years, after they iron out all the bugs and optimize the process. Right now it sounds like it's still in an alpha state.
- punisher18, on 10/12/2007, -0/+1@thirdprize : but thats with any new system even the xbox 360. the fact is when a console does come out regardless of who it is some of the consoles have bugs and not all the developers have game that can fully show off their consoles capability being that all dev kits released to developers are fully up to snuff spec wise.
reading into rumors only lead to more rumors and speculation.
no digg though because its nothing new . :/ - eigh, on 10/12/2007, -0/+13rd post :P
- inactive, on 10/12/2007, -0/+1Thus games will be made on the XBox 360 and Ported to PS3.
Is SONY finally biting more than they can chew?
How buggy will this thing be? Add to that the BR Player, oh my, it could be error prone. - smartpatrol, on 10/12/2007, -0/+1When I read stories like this one about Cell, I have to wonder: was it really worth the cost to create Cell? It seems like they would've been able to make a much more developer-friendly console just using an Athlon 64 X2 and perhaps an Ageia physics co-processor (I can't say for sure because I have no clue how much this addition would cost).
This option would create a console that would truly impress gamers RIGHT NOW rather than 2 years from now, would be a dream for developers to create games for (as opposed to the Cell), and would probably be much cheaper considering all the R&D costs for Cell.
Oh well, I am really enjoying the show, seeing Sony bite off more than they can chew. In my opinion, Sony has done absolutely nothing good for games anyway. All they did is that they happened to be in the right place at the right time with PSX, killing off Saturn (which was supposed to be the ultimate 2D graphics system. . . think about how cool that idea is). Then they murdered the Dreamcast with nothing but hype, FUD, and lots of marketing dollars, which I still haven't forgiven them for. It would serve them right for PS3 to become a huge failure, and that's exactly what it's looking like. - drn666, on 10/12/2007, -0/+1Arstechnica - Saturday.
/. - Sunday.
Digg - Today.
Digg sucks at life. - breakneckridge, on 10/12/2007, -0/+1I have to say, from an ex-gamer's pov, the Revolution is the only system that I feel any excitement about.
- zenogais, on 10/12/2007, -0/+1altodarknight: The Xbox/Xbox 360 are both very nice platforms to develop for, because they use a proven, well made API as well as a nice compiler. The people at Sony simply can't compete in those areas.
This is really insane, I mean sure developing for the PS2 was a bit of a bitch, but this makes that look like fisher price stuff. - bigteebo, on 10/12/2007, -0/+1You ask yourself "What if God created a console even he cannot release?", well, now we have the answer.
- inactive, on 10/12/2007, -1/+1those guys might be having a hard time porting pong to it...
leave the good games for atleast 5 years later...
tho you can be sure ea will release its ***** the day the ps3 is released - applepro, on 10/12/2007, -1/+1this is why apple had to go intel. they can't deal with the delays and it left them feeling smucky every time they tried to get some product out
- JohnP, on 10/12/2007, -0/+0Ouch, this is what everyones been saying about the PS3 for a while. Its hardly going to be around in a decade to enjoy a solid knowledge base. Is it? It is another 'before its time' console, and a pain in the arse for devs.
- InternetUser, on 10/12/2007, -0/+0Hrm, nothing really to cry about. But it sounds like this is the same problem the PS2 had in the beginning.
To take full advantage of the PS3 hardware (as a developer) you're really going to have to know your Cell assembly language, or, you need IBM's Octopiler. But I suspect most developers will get by knowing only the basics of the hardware for the first few months.
Really, this compiler seems like a bit of after-thought after feedback from the developers ("Help! It's too damn hard to optimise!"). It should have really been available (and stable) 1 year ago.
Good luck PS3 devs! - badnewsblair, on 10/12/2007, -0/+0I've gone cross-eyed.
- SweetsGreen, on 10/12/2007, -0/+0I don't see why they don't just write the games in assembly
- geminitojanus, on 10/12/2007, -0/+0As for if it's really late; it's late in the sense that the XBox 360 has gotten the drop on it, and it's late in the sense that given the current state of the "Octopiler", there's just about no way in hell it's going to ship on time. Go take a look at the IBM specs on it and you can clearly tell the developers don't yet have a complete handle on things; register alignment hacks worse than SSE, block diagrams that would make a detective's head spin.. It's going to take some time for IBM to get a handle on it. And that's time Sony can't afford to be wasting.
- samadam, on 10/12/2007, -0/+0i retract that last statement
- geminitojanus, on 10/12/2007, -0/+0"I don't see why they don't just write the games in assembly"
Because it would take forever given the complexity of current games. It was an option when you were coding games for the NES, but it's just not an option today. That's like asking "Why don't you code Windows in Assembly". Sure, you can code parts of it in asm (and it's smart to), but coding the whole thing? - samadam, on 10/12/2007, -0/+0how did we manage to digg arstechnica?
that is, like, a big site. - BritOverseas, on 10/12/2007, -0/+0Sounds like a bitch to work with, I hate computers.
No really, I wouldn't have a job without them but lets face it, they are a pain in the ass... - geminitojanus, on 10/12/2007, -0/+0"Also, this has absolutely nothing to do with vectorization, compiler have been able to do that. GCC can do that now in 4.1"
Read the article. The "Octopiler", as IBM is calling it, is an auto-vectorizing compiler for the SPEs available on the Cell. And GCC is just now getting auto-vectorization, but it's still not anywhere as good as doing it by hand, and 4.1 is still unavailable to the most of us who don't have time to be hacking our compilers and applications; a great deal of the people I know are still on GCC 3, and don't plan on moving to GCC 4 for at least another 6 months, if then.
Meanwhile, the PS3 is shipping this year. Gives you reason to doubt. - Nelson69, on 10/12/2007, -0/+0Is it actually late?
Also, this has absolutely nothing to do with vectorization, compiler have been able to do that. GCC can do that now in 4.1 - xLiKx, on 10/12/2007, -0/+0probably why the chip costs so damn much to manufacture
- GRIMREAPER187, on 10/12/2007, -0/+0good point chembro
- deepsub, on 10/12/2007, -0/+0"I think if this does progress smoothly, the PS3 will be a kick-ass console in about 1.5 to 2 years, after they iron out all the bugs and optimize the process. Right now it sounds like it's still in an alpha state."
In 1.5 to 2 years, Revolution and 360 will be taking a dump on Sony's market share. - deepsub, on 10/12/2007, -0/+0"My friend who works at IBM on the Cell says that it is insanely hard to program for but insanely powerful"
It can't be insanely powerful if you can't write a high performance application for it.
Sony should just drop Cell while it can and use a PowerPC core or cores. - en3r0, on 10/12/2007, -0/+0i hope to see alot of homebrew for this multi core... i might even buy one then :)
- GRIMREAPER187, on 10/12/2007, -0/+0i still remember authors saying that the ps2 is so hard to program for its the fastest thing out iraq buying up ps2's they didn't think they would ever take advantage of more than half the processing power they were bitching and moaning now god of war looks better than most games on the xbox
so although ps3 won't be an instant money maker when xbox 360 is nearing its end of its 4 year cycle i suspect ps3 will be putting the full circle system out of date yes hype is bad when its false advertised i too am pissed off bout the psp its their first handheld so i'll cut them slack burn me once i'll buy a second burn me twice and im having a yard sale - PhoneGuy, on 10/12/2007, -0/+0Rumours abound that the real hold up is Sony's HDR technology. They have rushed its developement in an effort to include it in the first run of PS3's and sadly have had several problems with safe implementation.
- popper1, on 10/12/2007, -0/+0i know its been a while now, but didnt anyone bother to actually read the ibm site about it.
http://www.alphaworks.ibm.com/tech/cellcompiler
"How does it work?
The IBM XL C/C++ Alpha Edition for Cell BE Processor provides three invocation commands: ppuxlc, ppuxlc++, and spuxlc. The commands ppuxlc and ppuxlc++ are used to generate code for the PPE, and spuxlc is used to generate code for C on the SPE (C++ for SPE is not available in current version).
The compiler invocation commands for the PPE performs all necessary steps for compiling C/C++ source files by ppuxlc or ppuxlc++ into .o files and linking the object files and libraries by ppu-ld into an executable program. Similarly, the compiler invocation command for the SPE performs all necessary steps for compiling C/C++ source files by spuxlc into .s files, assembling .s files into .o files by spu-as, and linking the object files and libraries into an executable program by spu-ld. The Cell BE Software Development Kit also provides the tool ppu-embedspu for linking a PPE executable and a SPE executable into a single executable.
The compiler includes five base optimization levels:
-O0: almost no optimization
-O2: strong, low-level optimization that benefits most programs
-O3: intense, low-level optimization analysis with basic loop optimization
-O4: all of -O3 and detailed loop analysis and good whole-program analysis at link time
-O5: all of -O4 and detailed whole-program analysis at link time.
Auto-SIMDization is enabled at O3 -qhot or O4 and O5 by default for the PPE, and at O3 -qhot or O4 and O5. SIMD has been improved to better handle relatively aligned streams from run time-aligned individual streams. (SIMD stands for Single Instruction and Multiple Data.) "
http://www.alphaworks.ibm.com/tech/cellsw
"What is the IBM Cell Broadband Engine (Cell BE) Software Development Kit (SDK)?
The IBM Cell Broadband Engine™ (Cell BE) SDK, Version 1.1, is a complete Cell BE development environment. The SDK contains binaries and source code that are available for downloading from both alphaWorks and Barcelona Supercomputing Center's Web site. The SDK here on alphaWorks contains IBM-authored material, including Library and Samples Source Code, IBM XL C/C++ Alpha Edition for Cell Broadband Engine Processor (a compiler), and IBM Full-System Simulator for the Cell Broadband Engine Processor. The Barcelona Supercomputing Center's Web site contains open-source projects that have been modified for Cell BE Processor; these include GNU GCC compilers for PPU and SPU, Linux Kernel 2.6.16, SPE Library support, NUMA support, and a system root image for the Full System Simulator. "
http://www.bsc.es/projects/deepcomputing/linuxoncell/gcctoolchain_cbe.html
"GNU Toolchain 4.0.2 and GDB for the Cell BE's SPU
The Cell BE's PPU (Power Processor Unit) uses the same instruction set as the PowerPC 970 processor. To compile code that will run on the Cell BE's PPU, last toolchain release did not include a GCC compiler that optimizes code for it and the default GCC compiler that comes with Fedora was used.
This new toolchain release includes a GCC compiler for PPU to cross-compile on x86 and as a replacement for the native GCC compiler on ppc platforms. The GCC compiler for the PPU is preferred and the make files are configured to use it when building the Libraries and Samples. The GCC compiler also contains a separate SPE cross-compiler that supports the standards defined in the SPU C/C++ Language Extensions V2.1, SPU Application Binary Interface Specification V1.4, and Synergistic Processor Unit (SPU) Instruction Set Architecture V1.0 documents. The associated assembler and linker additionally support the SPU Assembly Language Specification V1.3. The toolchain is based on a port of GCC, GDB and the GNU binutils to the Cell Broadband Engine Architecture (CBEA) provided by Sony Computer Entertainment Inc.
The cross-compiling GNU Toolchain is installed into the /opt/sce/toolchain-3.2 directory while the ppc toolchain will be installed into /usr. "
http://www.cgo.org/cgo2006/html/cell_abs.html
"The complexities of the Cell processor span multiple dimensions. At the elementary level, the Cell system has two distinct processor types, each with its own application level ISA. One ISA (PE) is the familiar 64-bit PowerPC with VMX, the other, (SPE) is a new 128-bit SIMD instruction set for multimedia and general floating point processing. Typical applications on the Cell processor will consist of a combination of codes to exploit both these processors. The pipelines of both processor types must be taken into account, and the SPE presents several challenges not seen in the PX, chief among them the instruction prefetch capabilities and the significant branch miss penalties resulting from the lack of hardware branch prediction. At the next level, the SPE is a short SIMD or multimedia processor with scant support for scalar operations. On the next dimension is the parallelism of the machine when deploying applications across all SPEs.
It has been demonstrated that expert programmers can develop and hand tune applications to exploit the full performance potential of this machine. We believe that sophisticated compiler optimization technology can bridge the gap between usability and performance in this arena. To this end we have developed a research prototype compiler targeting the Cell processor. In this tutorial we discuss a variety of compiler techniques we have investigated/implemented, and their associated performance benefits. These techniques are aimed at automatically generating high quality codes over the broad spectrum of heterogeneous parallelism available on the Cell processor.
The tutorial will begin with a brief overview of the Cell architecture to motivate the discussion of compiling to exploit specific novel features of the architecture. The techniques we describe include compiler supported branch prediction, compiler assisted instruction fetch, and the generation of scalar codes on SIMD units. We will then discuss our techniques for automatic generation of SIMD codes, and automatically parallelizing single programs across the multiple heterogeneous processors. In particular we will describe and discuss the performance of our technique for presenting the user with a single shared memory image through our compiler controlled memory management. We will also report and discuss the results we have achieved to date, which indicate that significant speedup can be achieved on this processor with a high level of support from the compiler.
Intended Audience
This tutorial is intended for those with a background in Computer Architecture and Compiler Writing. Some knowledge of parallelization techniques would also be useful. " - GuyNextDoor, on 10/12/2007, -0/+0They should title this "IBM's Octopiler, or, why the PS3 won't impress us for a while".
I guess we could say the PS3 is "late" if they don't get it out by the end of the "spring", but they still have time to work "a miracle", and considering no hard-dates have been given, then I'm not quite sure we could ever say it is late.
I think the launch titles will look pretty darn unimpressive, even less than the first batch of Xbox360 titles. BUT, whereas on the 360, we are going to see some pretty impressive "next-gen" looking games as early as March (5 months after launch), you won't see this kind of leap with the PS3 for over a year, if not more. The cell is by no means equitable to the difficulty of the emotion engine chip in the PS2, it's in a whole new league of difficulty, throw in the poor software support and dev tools that Sony has given developers, and you have a fairly sad situation.
The plus side is that, the PS3 will be very popular, simply on brand recognition alone. The large installed user base should force devs to grind through issues to make impressive gains in using the cell. The flipside is that only the largert devs will have the financial fortitude to do this, leaving smaller devs out of the loop. - chembro84, on 10/12/2007, -0/+0"i still remember authors saying that the ps2 is so hard to program for its the fastest thing out iraq buying up ps2's they didn't think they would ever take advantage of more than half the processing power they were bitching and moaning now god of war looks better than most games on the xbox"
The reason developers persisted on the PS2 is because it was clearly the winner in that console war and everyone knew it. Think that developers will be so hell bent on supporting the PS3 if it doesn't have the market share that the ps2 had? Probably not, if they can make a game for much cheaper on the 360 or revolution and possibly turn a bigger profit (I have no idea how much more it'd cost to make a ps3 game vs a 360 game, but maybe it's 1/3 more expensive). So if you sell 1/3 less on the console with less market share, you're still making more money because you can devote resources to other games where you would of been spending time on a PS3 game. This may turn the tides in the console wars, 360 might start to look very attractive for developers who don't have quite as much money to spend on good games. - chembro84, on 10/12/2007, -0/+0Interesting article, written well for someone who doesn't know too much about programming like me.
Seems like Sony may have bitten off a little more than it could chew (at least for the developers), if IBM can make the "Octopiler" work for a "Tier III" programmer like a "Tier I" programmer, then we'll be in for a treat, but from the looks of it, that's a while off. - HappyScrappy, on 10/12/2007, -0/+0I despite the title. It was bad enough when Arstechnica wrote it. It's worse to repeat it.
There's simply to link offered, let alone proven, that this is why PS3 is late, or even that PS3 is late (although that seems pretty likely to me). - tmcleroy, on 10/12/2007, -0/+0There's more than one reason that the ps3 is delayed out the ass. My friend who works at IBM on the Cell says that it is insanely hard to program for but insanely powerful
- dmoore0100, on 10/12/2007, -1/+0Umm - lets see, it is a straight up and down inorder PowerPC + Vector unit x 8. IBM already have PPC compilers since 1989 or so and I'm guessing the compiler guys have seen the vector architecture before Cell.
As far as I know the PS2 was capable of up to 10 simultaneous FMACs whilst Cell can do (from memory) 32 FMACs. Some of the comments about performance are also out of whack. One of the more realistic figures quoted was an overall x2 speedup which considering the ballyhoo is pretty disappointing especially given the reputations of the companies involved. At launch in 1991 the DEC Alpha was 4 times faster than the best RISC competitors. The story that there is a lot to do I'm thinking refers to the disappointing hype/real world performance index.
The only thing Cell/PS3 will changes is shipping that much hardware to so many gamers. Furthermore if you are going to use Cell in a $500K server and also in a "$500 PS3 connected to the Internet" you are going to need to do some extremely careful market placement. - SweetsGreen, on 10/12/2007, -1/+0geminitojanus.....it was a JOKE.
- rebrad, on 10/12/2007, -1/+0The risk in being on the bleeding edge is that sometimes you get cut. The $ony PS3 is history. The programmers aren't there and won't be for awhile if ever. Why beat you head against a wall when there are better solutions. Maybe this is the scenario of the next Final Fantasy. I guess $ony will simply rename their gamer from PS to FF once they come up with a viable gamer if they don't just quit and buy Nintendo or Apple. Never the less, the PS3 is dead.
- jayf, on 10/12/2007, -1/+0I bet Sony will drop the CELL chip for a later date and slap some other CPU in the PS3 within 2 years. Development houses would certainly be happy if this were true.
- Jadey, on 10/12/2007, -2/+1Kudos to obendega for knowing to use "a lot" instead of "alot" which I see way too often :P
- aggies11, on 10/12/2007, -1/+0To further up the whole "Do it in assembly" idea.
Some studies have shown that a typical programm
er will write, on average, the same amount of "lines of code", regardless if they are working in a high level language like say C, or low level Assembler.
Now given that one line of a high level language can result in several lines of assembler, you can see that this is a dramatic loss in productivity/ efficiency.
Aggies - NekstBestThing, on 10/12/2007, -2/+0The 360 will conquer.
- myskja, on 10/12/2007, -2/+0There is not much news in this story is there...
- gamabunta, on 10/12/2007, -2/+0Late or not, I put 3 pre orders in so I can sell 2 off and break even.
-
Show 51 - 53 of 53 discussions
Browsing Digg on your phone just got easier with our enhancements to the 
© Digg Inc. 2009
— Content created and posted by Digg users is