|
Dialectronics - the dialect of electronic communications
|
|
|
About Us Our Mission Minix RISCV PowerPC Lua XCOFF Bootloader Old World Macs Open Firmware Words Journeys Home Repair Top Level |
Words25 Jun 2005 Meet the experts: Arnd Bergmann on the Cell BE processor This was the first Meet The Experts I did. I got most of it right but messed up some parts because I am not familiar with the Linux file system. The interesting aspect to this interview is that later IBM started advocating the position that the best way to maximize Cell performance is to use DMA, instead of file system abstractions. 15 Aug 2005 Shrinking slices: Looking at real time for Linux, PowerPC, and Cell This interview was conducted by McLaren Harris but I participated in it and provided a lot of the questions. The interview inspired me to write Ten as a companion piece which was never published. 11 Oct 2005 Meet the experts: On open source firmware development The SLOF group was an interesting bunch. SLOF was released under BSD license instead of GPL or CPL. You won't find any discussion of licenses in the interview, however. That was cut by IBM Legal as "offering a legal opinion." The fact remains SLOF was released under BSD instead of GPL and there are reasons for this. IBM just won't publically discuss those reasons. 16 Nov 2005 Meet the experts: Dan Brokenshire and Sid Manning on the alphaWorks downloads for the Cell Broadband Engine I didn't enjoy this interview. I didn't get the details of the interview until during the interview. I prepared as best I could, but it was set up to be a "we're great" deal and I didn't buy in. I knew that the interview was going to discuss the Cell BE software downloads being made available, but only for Linux on x86. You'll find that Sid Manning is a strong supporter of the x86 architecture, even if he works for IBM doing Cell/PowerPC core work. I thought this kinda sent the wrong message to PowerPC developers wanting to explore Cell. 16 Nov 2005 Just like being there: Papers from the Fall Processor Forum 2005: Unleashing the power of the Cell Broadband Engine - a programming model approach I did the technical edit on this piece about Alex Chow's Cell BE programming models presentation he gave at the 2005 Fall Processor Forum. It's one of the few times my name actually appears at developerWorks (by my choice). 22 Nov 2005 Meet the experts: Alex Chow on Cell Broadband Engine programming models A great interview all around. It actually took two sessions, and some of the second session ended up comprising a "Five Minutes With" installment on Cell's memory model. Alex Chow discussed at length several different programming models for Cell BE, and reading through this one can start to gras p the complexities programming for a heterogenous processor incurs. 22 Nov 2005 Meet the experts: The Mambo team on the IBM Full-System Simulator for the Cell Broadband Engine processor This one is about a third of the length of the actual interview. It was very difficult to cut anything but in the end I did a slash and burn. I think one undertone is worth illuminating: the project leader runs Mambo on his G5, but IBM only releases Linux/GPL versions (and it took a lot of complaining to get them to release a 970 version as originally it was only x86). 29 Nov 2005 Just like being there: Papers from the Fall Processor Forum 2005: Unleashing the power of the Cell Broadband Engine - the Element Interconnect Bus Another technical edit I did on the FPF papers for Cell BE converted for publication by developerWorks. I did get a credit at the end. 06 Dec 2005 Meet the experts: David Krolak on the Cell Broadband Engine EIB bus The Cell BE Element Interconnect Bus (EIB) is a great piece of engineering that solves 90% of the problems with only 10% of the effort it would take to solve all of them. It's somewhere between a parallel planar and a ring design. Dave Krolak gives a great interview. The only thing that bothered me is that no matter who I asked, no one could ever tell me whose responsibility it was to manage scheduling of EIB elements... 06 Dec 2005 Cell Broadband Engine processor DMA engines, Part 1: The little engines that move data I did the technical edit on this one. It was a pleasure working with the group from India. They had something to prove and I think they did it well. 06 Dec 2005 Just like being there: Papers from the Fall Processor Forum 2005: Application-customized CPU design - the Microsoft Xbox 360 CPU story Another one I got listed in the credits. In all the flurry surrounding Cell, the tri-core PowerPC-based Xenon processor that powers the XBox360 was overshadowed, a rare occurence for a Microsoft event. While the shared L2 cache is small, I really appreciate the 128 Altivec registers (per core). When all was said and done, I found Xenon to be a better design than Cell. It is, after all, shipping... 06 Dec 2005 Just like being there: Papers from the Fall Processor Forum 2005: Introducing the IBM PowerPC 970MP - A new, low-power, high-performance dual-core processor The last time I was listed in the credits. The power specs on this processor are astounding. Flashlights use more power than this thing. Steve Jobs had been complaining that the PowerPC couldn't deliver decent power/performance ratio and that's why they (Apple) switched to Intel. The 970MP showed this was a total fabrication. 07 Feb 2006 An introduction to compiling for the Cell Broadband Engine architecture, Part 1: A bird's-eye view This is the first of five pieces I did an in-depth technical review of while they were adapted from presentations for publication at developerWorks. It is an excellent primer on the architecture of the Cell Broadband Engine. 07 Feb 2006 An introduction to compiling for the Cell Broadband Engine architecture, Part 2: Optimizing for the SPE Technical review, adapted from presentations. The Cell Synergistic Processing Elements (SPEs) are completely different than the PowerPC core and have a different programming approach. The instructions aren't even compatible, so code has to be written for either the PowerPC core or for an SPE. This introduces an entire line of questions about whether or not programmers will adopt a technology that has this requirement. Time will tell. 07 Feb 2006 An introduction to compiling for the Cell Broadband Engine architecture, Part 3: Make the most of SIMD Technical review, adapted from presentations. An excellent primer on SIMD for Cell. Truly worth the effort of getting through it. 07 Feb 2006 An introduction to compiling for the Cell Broadband Engine architecture, Part 4: Partitioning large tasks Technical review, adapted from presentations. This one starts diverging significantly from the concrete details of the first three installments. Instead of discussing how programmers can handle difficulties inherent to Cell BE, the article describes how well the internal IBM XL C research compiler does the job. That compiler is not available publically. 07 Feb 2006 An introduction to compiling for the Cell Broadband Engine architecture, Part 5: Managing memory Technical review, adapted from presentations. Continuing with a strong divergence from anything informative, this one ended up being a real deal-breaker. A comment I made about the piece being "fluffy" got back to the authors who basically pulled out of helping at that point. At least, that was the story I was told. 14 Feb 2006 Meet the experts: An interview with the compiler This was a disaster that was badly salvaged. I actually didn't do the final material for this one. I did all of the interviewing and almost all of the research, but some key people dropped the ball and it ended up having to go to someone else. I enjoyed the interview. The participants enjoyed the interview. Everybody else screwed it up. 18 Apr 2006 Five minutes with: Mark Nutter and Max Aguilar on the Cell BE memory model I put this together from cuttings on the Alex Chow interview. Very tight, very concise. 02 May 2006 Cell Broadband Engine processor DMA engines, Part 2: From an SPE point of view The second installment from the India group. Another job well done, with a fraction of the effort it took with groups in the U.S. 10 May 2006 Meet the experts: Peng Wu and Alex Eichenberger on compilers and hardware constraints As I said, the participants in the previous interview (compiler) enjoyed themselves, and to prove my point, here are two of them for a second round. This was a great, great interview. An hour of seriously detailed discussions about a huge range of topics that should be considered required reading for all programmers. I didn't do the final edit on this one, either. All I can say is that I asked straight-up that if a programmer is completely abstracted and virtualized from the hardware, what is the point of having more than one CPU design in the market? (The original interview as I submitted it contained the length of time of the silence that followed that question, and neither person wanted to answer it. As for the length of the silence, let's just say I was starting wonder if the call had been dropped.) 10 May 2006 I did a companion piece on virtualization for the Peng Wu/Alex Eichenberger interview. It wasn't published, either. 31 May 2006 Not once but twice I pitched a new series for developerWorks Power Architecture Zone that focused on PowerPC development. Not once but twice was the series accepted and then later the series collapsed due to disagreements. The first series was called "This Operating System, That Hardware," to play on "This and That." It was based on my work on porting OpenBSD to Old World Macs, but was going to go through the process of porting NetBSD to a JS20. The first two installments can be found in our Old World Macs area. The second series was called "Some Assembly Required" and was going to be an in-depth exploration of writing highly optimized assembly language code for PowerPC. The first three installments of that can be found in our PowerPC area. 18 July 2006 As mentioned above, there were two series pitched, accepted, and buggered by developerWorks. The "Some Assembly Required" series was the second one. The first one was "This Operating System, That Hardware," which was going to begin with a four part series on Open Firmware debugging techniques. We've published Part I. Part II is generally written but needs finishing and images. 17 June 2007 Beginning Open Firmware Debugging Techniques, Part II is posted, with HTML formatting, but not a lot of final editing. It also does not contain the originally proposed images at this time. |