tamlin

Yes, I used VSS as a solo developer in the 90s. It was a revelation at the time. I met other VCS systems at grad school (RCS, CVS).

I started a job at MSFT in 2004 and I recall someone explaining that VSS was unsafe and prone to corruption. No idea if that was true, or just lore, but it wasn't an option for work anyway.

Source Depot was based on Perforce. Microsoft bought a license for the Perforce source code and made changes to work at Microsoft scale (Windows, Office).

TFS was developed in the Studio team. It was designed to work on Microsoft scale and some teams moved over to it (SQL server). It was also available as a fairly decent product (leagues better than SourceSafe).

Samsung and SK-Hynix have had specs and papers for a few years already for HBM and GDDR. e.g.

* https://www.servethehome.com/sk-hynix-ai-memory-at-hot-chips... * https://www.servethehome.com/samsung-processing-in-memory-te...

Not sure anyone has started using it in production.

Yes, agree. Probably the main thing is the NPU is just a dedicated unit without the generality / complexity of a CPU and so able to crunch matmuls more efficiently.

No worries, and no humble pie required. Peace, good happiness.

Agree on NPU vs CPU memory bandwidth, but not sure about characterizing the GPU that way. GDDR is usually faster than DDR of the same generation, and on higher end graphics cards has a width bus width. A few GPUs have HBM and pretty much all datacenter ML accelerators (NVidia B200 / H100 / A100, Google TPU, etc). The PCIe bus between the host memory and GPU memory is a bottleneck for intensive workloads.

To perform a multiplication on CPU, even SIMD, that values have to fetched and converted to a form the CPU has multipliers for. This means smaller numeric types penalised. For a 128-bit memory bus, an NPU can fetch 32 4-bit values per transfer; the best case for a CPU is 16 8-bit values.

Details are scant on Microsoft's NPU, but it probably has many parallel multipliers; either in the form of tensor cores or a systolic array. The effective number of matmul's per second (or per memory operation) is higher.

Rotor had it's own license:

https://en.wikipedia.org/wiki/Shared_Source_Common_Language_...

I have various snapshots of the Rotor 1 and 2 sources around and they have the SSCLI license. There is a file that contains BSD licensed code (pal\rotor_pal.h).

Microsoft had a research version of the CLR called Rotor (2002) that predated Mono (2004). Rotor built for Windows, FreeBSD, and macOs, albeit with a not-very-open license.

When Mono came along, the internal position at Microsoft was surprisingly positive. There was a dev slide deck that went into Mono in some depth. And a telling slide that said it wasn't a threat because the performance wasn't competitive at the time.

https://en.m.wikipedia.org/wiki/Shared_Source_Common_Languag...

https://en.m.wikipedia.org/wiki/Mono_(software)

A decent chunk of AI computation is the ability to do matrix multiplication fast. Part of that is reducing the amount of data transferred to and from the matrix multiplication hardware on the NPU and GPU; memory bandwidth is a significant bottleneck. The article is highlighting 4-bit format use.

GPUs are an evolving target. New GPUs have tensor cores and support all kinds of interesting numeric formats, older GPUs don't support any of the formats that AI workloads are using today (e.g. BF16, int4, all the various smaller FP types).

NPU will be more efficient because it is much less general an GPU and doesn't have any gates for graphics. However, it is also fairly restricted. Cloud hardware is orders of magnitude faster (due to much higher compute resources I/O bandwidth), e.g. https://cloud.google.com/tpu/docs/v6e.

evroc | Front-End/Developer Experience Engineer | London | Full-time| ONSITE

evroc is building Europe's first sovereign hyperscale cloud. We're building both the physical data centers (flagship on the outskirts of Stockholm in 2024) and the platform that will be running them. We're currently focused on staffing up our Software Engineering teams in Stockholm, London and Sophia Antipolis, France, and are looking for 2 x Front-End/Developer Experience Engineers in London to lay the foundations of our Developer Experience offering.

Experience/skill-set: JavaScript/TypeScript, GCP/AWS/Azure or comparable dev tooling, building UIs for a developer audience.

Apply online at: https://evrocab-1692891239.teamtailor.com/jobs/3166084-devel...

Perhaps someone ex-Intel can comment, but there was a story within Microsoft that Intel proposed a different x86 derived 64-bit architecture after AMD64, but Microsoft didn't want to support two x86-64 ISAs and that put pay to Intel's proposal. There was also talk at Microsoft about having input into the AMD64 design, but no idea what the details were there exactly.

BirdNet is pure joy.

Thanks to everyone who works on it. We've used the app relentlessly for a couple of years in the UK and when you show it to people they are amazed. People thank us for it and all we did was share it with them. Great work!

How many "independent" voices you see depends on your bubble. How many users are there of these platforms? What fraction wrote a rant? Not saying those opinions are wrong, but it's not particularly insightful.

Disappointing article. The title suggested it might have some data rather than being just a non-technical rant.

My manager once presented to Cutler. They said it went remarkably well as Cutler "only interrupted the presentation twice to cuss"...

Typically when building a device like this you get a version of Linux / embedded-OS from the SoC vendor and you are stuck with it because SoC vendor doesn't provide docs that would allow drivers to be maintained. This makes ongoing support harder than it might otherwise be.

Yes. All apps in Android are spawned from the Zygote, but once started an app is free to call as much native code as it likes.

https://medium.com/@voodoomio/what-the-zygote-76f852d887d9

There was a Linux subsystem for NT around early 2000s in Microsoft Research.

There is amazing depth inside parts of Microsoft. For example, a team built Xbox360-on-PC sometime around 2007-08. That was emulating the PowerPC and GPU on Intel PCs. This didn't ship for various reasons as a product, but it was in the draw when Xbox One came along and they wanted game compatibility between Xbox 360 and XBox One.

WSLv1 was built on Drawbridge that originally came from MSR. I heard a rumor that WSL was spun out from a project using Drawbridge to run Android. Obviously a lot of work went into WSL subsequently, but the pieces to get started were in place.

Very cool, that's a decent performance bump.

It'd be nice to have some simple benchmarks that directly compare ignition, sparkplug and turbofan (and full-codegen too). And also time spent compiling.

It looks somewhat similar to context threading as I remember it:

http://www.cs.toronto.edu/~matz/pubs/demkea_context.pdf

but it's been a while...

Ha-ha, the stacks grow downwards button is priceless :-D

MBA and business students are taught the rule of three:

"The Rule of Three applies wherever competitive market forces are allowed to determine market structure with only minor regulatory and technological impediments."

https://iveybusinessjournal.com/publication/competitive-mark... https://en.wikipedia.org/wiki/Rule_of_three_(economics)

The costs here of developing the platform and substitute services / features is high and all investors want some return on the investment. Third place companies are often stuck in the ditch going nowhere. Microsoft has deep pockets and years of experience, but ultimately bailed on Windows Phone because #3 is not a good place to be.