Titanfall
08.10.2025 15:42 β π 1 π 0 π¬ 1 π 0@parkbot.bsky.social
Burrito scientist, AMD Infinity Fabric performance/architecture. @philparkbot in the Bad Place. Personal account. He/Him. πBritish Columbia https://linktr.ee/philpark
Titanfall
08.10.2025 15:42 β π 1 π 0 π¬ 1 π 0Conversion therapy is another term for child abuse.
07.10.2025 19:15 β π 4 π 0 π¬ 0 π 0Itβs 4:15 am here and after four attempts over the span of a month I finally completed my language test. It is now time for dog snuggles.
04.10.2025 11:17 β π 4 π 0 π¬ 0 π 01) yes, agreed. I do also recall around this time Google was growing like crazy, and they used commodity x86 parts to build their search cluster, which it seems like contributed to this shift
2) what (i had never heard this)
When AMD64 was proposed, Athlon was already in the market and was scaling beyond what Intel believed was possible less than 10 years prior. But by then, Intel had already committed to IA64 with HP
22.09.2025 21:06 β π 1 π 0 π¬ 0 π 0Folks may not realize that when Intel announced IA64, their flagship part was the OG Pentium - superscalar but in-order. While some in the company didnβt believe CISC could continue improving, Pentium Pro was still in active development and showed that superscalar and out-of-order was possible.
22.09.2025 21:06 β π 1 π 0 π¬ 1 π 0Quote from Linus Torvalds: "[Intel] threw out all the good parts of the x86 because people thought those parts were ugly. They aren't ugly, they're the 'charming oddity' that makes it do well."
A lot of people rightfully complain that x86 is complex and ugly, but this quote from Linus Torvalds on x86-64 vs IA64 was also revealing
22.09.2025 20:29 β π 2 π 0 π¬ 1 π 0βWhen the IA64 project was started it was perceived that the performance of RISC machines would outstrip that of [Complex Instruction Set Computing (CISC)] machines. ... They felt that they could get higher performance with a [Very Large Instruction Word (VLIW)] architecture than they could with a CISC architecture.β - Dave Cutler, "A History of Modern 64-bit Computing"
One fact I had memory holed goes back to the classic CISC vs RISC debate. My recollection for IA64βs justification was that Intel believed CPUs were getting too complicated and that they wanted a new ISA. In reality it was because they believed CISC could not scale or be competitive with RISC!
22.09.2025 20:25 β π 1 π 0 π¬ 1 π 0The Colwell post Iβm referring to btw:
x.com/philparkbot/...
I wrote my first and maybe only Substack post. This is a follow up to my Colwell post because I found a 2007 technical report by two MS employees who interviewed folks from MS, AMD, and Intel about the history of the transition, and I learned a lot!
computerparkitecture.substack.com/p/the-long-m...
Protest banner at Incheon airport of a person in a Trump mask wearing an ICE shirt, with a bag of cash and a rifle. The banner reads, βWeβre friends! Arenβt we?β
Love this photo
www.theguardian.com/global-devel...
I was wondering how you exceeded the 500 tab limit on Safari and I see itβs not Safari. Incredible job.
09.09.2025 19:16 β π 0 π 0 π¬ 0 π 0A picture of the hazy skies in metro Vancouver due to wildfires in northern BC
Hazy skies from wildfires in the Pacific NW
05.09.2025 03:39 β π 1 π 0 π¬ 0 π 0Post by Todd Austin on LinkedIn The TECH NEWS doesn't want to talk about it, but today EVERY computing system-every last oneβis LITTLE ENDIAN. In the dark ages of computer architecture, there were big-endian and little-endian architectures, and it was TOTAL CHAOS. Through the leadership of myself and other HIGHLY TALENTED computer architects, it became clear that BIG ENDIAN was a TOTAL DISASTER, confusing, slow, and frankly a disgrace to the world of computing. Eventually, our sage leadership FIXED it, unified it, and made memory systems BEAUTIFUL again. Last year at HotChips, a very famous computer architect came up to me (with tears in their eyes) and said, "Thank you, professor, for bringing peace and harmony to our memory systems." Today everyone knows it: computer architects LOVE little endian. But now l've just learned something SHOCKING-maybe the biggest cover-up in the HISTORY of technology. The INTERNET-yes, the whole thingβis COMPLETELY BIG-ENDIAN! After all the tremendous work we did, shadowy standards committees and so-called networking "experts" quietly locked in BIG-ENDIAN as "network byte order." It's a GLOBAL DARK NET conspiracy designed to undermine our BEAUTIFUL memory systems and to bring back confusion, chaos, and misaligned data. We must now come together ONCE AGAIN TO BRING ORDER TO THE CHAOS. We must END big endian once and for all. We must make the Internet little endian and make it STRONGER than ever before. BIG-ENDIAN is for LOSERS-LITTLE-ENDIAN is for WINNERS. Thank you for your attention to this matter.
Is Todd Austin running for office? π#TeamLittleEndian
02.09.2025 17:52 β π 3 π 0 π¬ 0 π 0Reading other sources and it seems like UChicago lost $20 million on crypto? π¬π¬π¬
stanfordreview.org/uchicago-los...
Todd Austin β’ β’ 1st S. Jack Hu Collegiate Professor of CSE at UofM... I am a COMPUTER ARCHITECT, which means I design computers. I was trained at a great school (UW-Madison), and I worked in industry (Intel) before I became a professor (University of Michigan). Despite these rarefied experiences, I first came to know computer architecture in the KITCHEN, working for 4 years first as a COOK. Let me explain. As a cook, you are the control logic of a complex multi-issue machine, which operates with many of the optimizations found in today's CPUs: IN-ORDER FETCH: The meal orders come in-order from the wait staff, and this order must be preserved. OUT-OF-ORDER ISSUE: With a big menu, the latency to cook items will vary greatly, from fast salads to medium-latency burgers to long-latency steaks. Cooks will scan the upcoming orders and "issue" long latency items first, medium later, and short last, to make sure all items finish at roughly the same time.
IN-ORDER RETIREMENT: Importantly, the meal orders have to finish IN-ORDER. Wait staff will get very annoyed if any out-of-order completion occursβno one wants to see the table that sat down after them get their food first! BTW, Tapas restaurants break this rule, which greatly improves the efficiency of their kitchens. What is the computer architecture dual to the tapas restaurant? PREFETCH: Occasionally, someone from the wait staff will run into the kitchen and yell "BUS!" (typically with the same vigor as someone running into a beach bar and yelling "TSUNAMI!"), which means 50+ additional orders are coming to your line. When this happens, you turn on all the stoves and steak burners and start frying staples, etc.
CACHING and SPECULATION: Many of the "specials" for the night are pre-cooked and kept in an oven (cache). This is optimizing the common case, since these popular meals take seconds to make. Often, if there are specials left over at the end of the night (mispeculation), they are donated to local food shelters. NON-DETERMINISTIC LATENCY: Efficient architectures try to avoid all non-deterministic latencies, since they stymie schedulers and gum up the works. Kitchens operate the same way: everything is prepped to give deterministic cook times. Once, I put a steak on a grill shared with the short-order cook window. Later, when I checked on my steak, it was gone... I confronted the short-order cook, and he said, "Do you want to fight me for it?" A non-deterministic latency made my line a mess that night!
PRECISE INTERRUPTS: When the shift changes, the incoming cook will negotiate which order to transition on. The working cook will finish all their orders, after which the new cook takes over. This is called a precise interrupt, and it is similar to how out-of-order processors effectuate a process switch. POWER GATING: Kitchens are HOT. If stoves are not needed, they are powered down! Besides architecture training, working as a cook is a great way to start a career. All later jobs you will have will be MUCH EASIER and LESS DANGEROUS! #computerarchitecture
I love this post by Todd Austin equating running a kitchen to a modern out of order microprocessor, especially with the inclusion of precise interrupts, prefetch, and power gating.
24.08.2025 19:15 β π 4 π 2 π¬ 0 π 0bsky.app/profile/park...
19.08.2025 02:59 β π 1 π 0 π¬ 0 π 0I do miss 20 episode seasons
15.08.2025 17:44 β π 2 π 0 π¬ 0 π 0The whole series is great. I wish it went on for the full seven seasons. I also enjoyed Prodigy
15.08.2025 03:07 β π 1 π 0 π¬ 1 π 0swear off ATI products for 16 years, and she responded by taking my concern seriously and getting me in contact with the driver team. We fixed the bug in a subsequent release.
In that vein, it feels like she took criticism of ROCm seriously and the software side is making genuine progress
Great interview with my (bossβ bossβ bossβ bossβ) boss.
My Lisa story is that in early 2020 we had a buggy driver launch that caused a game to crash repeatedly. I emailed Lisa telling her that the driver was buggy, my horrible experience with Radeon drivers in 2001 made me
Thereβs also one that uses Grover (korean_teacher_blue)
10.08.2025 16:41 β π 1 π 0 π¬ 0 π 0This is great timing because I just discovered that thereβs a flash mob dance crew called Ajumma EXP
10.08.2025 00:19 β π 0 π 0 π¬ 0 π 0Asking GPT-5 to draw me a map of the United States
GPT-5βs map of the US
Iβve lived in Texas, Colorado, Ionho, Nosava, Ilana, and Ventanr.
10.08.2025 00:12 β π 3 π 0 π¬ 1 π 0This is the consequence of having no state income tax - TX has very high property taxes and sales tax. States have these three pillars of revenue; take away one and the other two go up.
06.08.2025 16:04 β π 1 π 0 π¬ 0 π 0PCIe (bidirectional) bandwidths for various link widths across generations
PCIe 8.0 has been announced, so a reminder that since Gen 3, the link transfer rate is 2^(PCIe gen), and from there you can derive the bandwidth for any link width.
e.g. Gen 8 x16 = 256 GT/s * 16 lanes / 8 bits/B * 2 directions = 1024 GB/s = 1 TB/s
www.businesswire.com/news/home/20...
Not to brag, but I edit text and clean my apartment with vim
01.08.2025 00:46 β π 6 π 0 π¬ 1 π 0Iβm curious who your kidβs favourite groups are
30.07.2025 04:07 β π 0 π 0 π¬ 0 π 0Gaza faces a severe humanitarian crisis as Israeli restrictions on aid lead to widespread malnutrition and starvation, with hospitals overwhelmed by malnourished children.
24.07.2025 11:30 β π 621 π 373 π¬ 58 π 64Yeah, Iβm not surprised it was cut but the abruptness was a little surprising. I noticed that Azure ran Clear so I suppose thatβs going away too
19.07.2025 17:18 β π 1 π 0 π¬ 0 π 0