HNHacker News
TopNewBestAskShowJobs

AndrewGaspar

913 karma · joined May 2, 2013

Software Engineer @ Meta Reality Labs. Formerly Windows Core Engineering at Microsoft and Applied Computer Science at Los Alamos National Laboratory.

[ my public key: https://keybase.io/angasp; my proof: https://keybase.io/angasp/sigs/JGo37h7yq1cqScBqR8GSHg9KLsd1lUbJA5Zskx87l-E ]

submissionscomments
AndrewGaspar··on Cons of Polyamory
That you're just now discovering a possible argument for traditional gender roles is a failure on your part to engage with any contrary points of view (on this topic we've been hashing it over for centuries) not that those arguments are actually two word "but muh"s.

You might even find that there are billions of people still living in societies with some form of traditional gender roles, and it turns out they're not all superstitious rubes!

AndrewGaspar··on Rust Is Hard, Or: The Misery of Mainstream Programming
> This is why I think we need a high-level, no-BS version of Rust.

F# is probably what you're looking for if you want functional-lite programming with a strong type system, but don't want to deal with the pains caused by static resource management.

AndrewGaspar··on Mortgages are a manufactured product
A non-trivial percentage of Hacker News was 8 years old in 2008. :)
AndrewGaspar··on Employees given three months to return to Facebook office
The OP seems to only source its claims to the Daily Mail[0], which itself does not contain any new claims except for this:

> The need to request formal permission also applies to employees offered pay cuts in return for staying at home in cases where the local cost of living is lower than around the office where they are usually based.

To start, this claim is unsourced.

I'm having a hard time interpreting this claim, but it seems like _maybe_ it's saying that FB is requesting individuals who were already approved for remote work to RTO. But it also seems like they're just describing the existing process, which is that if you want to remain remote beyond January 2022, you must be explicitly approved for remote work. Anyway, I can't speak for everyone, but as a remote approved FB employee, I've received no such request to RTO.

[0] https://www.dailymail.co.uk/news/article-10064865/Facebook-g...

AndrewGaspar··on Employees given three months to return to Facebook office
There is no policy change - RTO in January 2022 was already the policy.

https://www.cnbc.com/2021/08/12/facebook-delays-return-to-of...

AndrewGaspar··on This Can’t Go On
World population in 1700 was 600 million, which is a single order of magnitude.
AndrewGaspar··on It’s possible to pay $150 in taxes on an income of $150K
Capital losses? In this market?
AndrewGaspar··on U.S. adds just 266,000 jobs in April, far below expectations
Most millionaires don't get that way and stay that way by YOLO'ing it all on meme stocks.
AndrewGaspar··on SpaceX wins contract to develop spacecraft to land astronauts on the moon
Why in the title (not reflected in HN title) is it "Elon Musk's SpaceX" rather than just "SpaceX"? I could imagine the distinction being relevant if he had just taken the helm and this was the company taking some new direction because of it (e.g. "Patrick Gelsinger's Intel"), but it doesn't seem to make a lot of sense given Elon Musk has been the CEO since its founding.

Even the subhead betrays how weird this journalistic formulation is: "The company beat out Jeff Bezos’ Blue Origin and Dynetics, a defense contractor". Like, ok, who is the CEO of Dynetics? Is it just some "public person" standard - because readers will say "Oh, that guy!"? Or is it because they're the majority shareholders in their respective companies and they do this for any company with a singular majority stake?

Just seems like a weird tic to me.

AndrewGaspar··on Rust for Windows
Since Windows ships a stable ABI, why does this project need to generate the bindings at build time? Couldn't all of the bindings be pre-generated, eliminating the build-dependencies?
AndrewGaspar··on Bjarne Stroustrup had his Reddit account suspended
They ruled that elected politicians cannot block individuals from participating in the replies to their tweets because they deemed it was the equivalent of a public space. They didn't rule that Twitter couldn't themselves moderate who posts on the platform.
AndrewGaspar··on Peter Thiel on Progress and Stagnation
It would be nice if there was an .epub format version of this.
AndrewGaspar··on New inline assembly syntax available in Rust nightly
I used inline assembly once when playing around with Rust on GPUs to get Cuda's implicit thread ID and grid ID: https://docs.nvidia.com/cuda/parallel-thread-execution/index...
AndrewGaspar··on Xbox and Windows NT 3.5 source code leaks online
Believe it or not, Microsoft engineers already have access to old NT source code.
AndrewGaspar··on Windows package manager does not permit opting out of telemetry
That's probably a function of WSL2 currently only being available in Insider, no? If you want to use pre-release builds, I think it's fair to expect that you need to enable telemetry.
AndrewGaspar··on U.S. Birthrates Fall to Record Low
Respectfully, I think reducing childcare costs is in direct tension with returning to an expectation that grandparents share some of the child rearing responsibilities.

Hopefully an expansion of remote work can improve this situation.

AndrewGaspar··on U.S. Birthrates Fall to Record Low
If you push out the whole distribution of child births, some of those would-be child births will not happen. Female fertility falls rapidly after 35[1]. It's not something you can simply delay across the population and expect to get the same total reproduction rate. People will have fewer kids than they otherwise would have.

[1] https://en.wikipedia.org/wiki/Age_and_female_fertility#Quant...

AndrewGaspar··on U.S. Birthrates Fall to Record Low
> it's still weighs heavy the thought of having a child with how expendable engineers are these days.

If you're working in some sort of Information technology, you're one of the least expendable employees in the labor market, at least as of March 2020. Only government and finance have lower layoff rates: https://www.bls.gov/news.release/jolts.t05.htm

AndrewGaspar··on Fast high-level programming languages
Dealing with pointers in C# is not normal.
AndrewGaspar··on Twitter kills off SMS notifications and posting in most countries
I just loaded a link to a tweet in a private browsing window on my iPhone in some sub-second amount of time. So whatever issue you're running into is a bug, not intentional.
AndrewGaspar··on I can't keep up with idiomatic Rust
New Rust versions don't change downstream Rust code either, except in edge cases. And those edge cases are only to fix potential security vulnerabilities.
AndrewGaspar··on Bill Gates calls for nationwide shutdown
1) Bill Gates doesn't have any governmental power - he's stating his opinion and hoping it will be heeded. 2) Bill Gates probably knows better than most medical doctors: https://en.wikipedia.org/wiki/Bill_%26_Melinda_Gates_Foundat...
AndrewGaspar··on [dead]
I'm probably stepping on a landmine here, but I'm not sure who it helps for dioceses' to have to literally sell off parishes to pay out sex abuse claims against people who in many cases are no longer even associated with the diocese.

Forcing dioceses to close and restart from square one doesn't really even hurt any of the people responsible as much as it hurts all the parishioners who now have to bootstrap a brand new diocese. At least when a company files for bankruptcy, the owners' equity is severely damaged. What does a bishop lose? After bankruptcy he probably even retains his position.

AndrewGaspar··on Major union launches campaign to organize video game and tech workers
If we're comparing to "Europe":

- EU Youth Unemployment Rate: August 2019 https://www.statista.com/statistics/266228/youth-unemploymen...

- US Youth Unemployment Rate: November 2018 - December 2019: https://www.statista.com/statistics/217448/seasonally-adjust...

AndrewGaspar··on Typing inside of the default WSL terminal feels amazing (2018)
I've never had a problem with GCC. GCC is, like, my primary use case.
AndrewGaspar··on Edge Chromium Release on January 2020
I believe this is what you're referring to: https://docs.microsoft.com/en-us/microsoft-edge/hosting/webv...
AndrewGaspar··on RLSL: a Rust to SPIR-V Compiler
Thanks, that makes a lot of sense!
AndrewGaspar··on RLSL: a Rust to SPIR-V Compiler
Huh - to make sure I understand correctly, is it your view (ignoring portability concerns) that it should be legal to store and load directly from a non-mut &[f32] when the platform has relaxed-atomic memory consistency by default? Or just that you should use a hypothetical "AtomicF32" with relaxed loads and stores and not worry about it from a performance perspective?
AndrewGaspar··on RLSL: a Rust to SPIR-V Compiler
> I've done enough Hacker News discussions to note that the lay-reader won't understand what this means, and this probably needs more elaboration.

This is why I didn't mention it, but I agree that shared memory and typical uses present a particularly challenging problem for Rust.

AndrewGaspar··on RLSL: a Rust to SPIR-V Compiler
This is a great write-up. I think Rust could be a very nice language for Shader-like applications which are already very functional, and don't involve alot of shared, mutable state across threads.

In HPC, we're very much interested in GPU compute programming, rather than shader programming. In CUDA codes, you're typically doing transformation from input buffers directly into output buffers from your CUDA kernels. This should immediately raise red flags for a Rust developer - you've got shared, mutable state across threads!

Consider this simple CUDA-ish Rust code with threads independently executing over 0..cuda.len() (ignore the bounds bugs at i = 0 and i = in.len()):

  fn stencil(i: usize, in: &[f32], out: &mut [f32]) {
      out[i] = (in[i - 1] + in[i] + in[i + 1]) / 3.0;
  }
(The `i` is a conceit around computing indexing from thread/block IDs, but the input and output arrays are pretty similar to the style CUDA promotes).

It's obvious to me, the programmer, that I don't have any aliasing issue - each thread is only mutating at a single index in the output array. However, Rust is not smart enough to see this. If they allowed the definition of the kernel as is, you could easily write multi-threaded code that has shared mutable access to individual memory locations, violating Rust's memory model. OK, you force the kernel to look more like this, then:

  // `in` is the slice of [i-1,i+1]
  fn stencil(in: &[f32], out: &mut f32) {
      *out = (in[0] + in[1] + in[2]) / 3.0;
  }
And you'd enforce Rust's invariants at the kernel launch site, computing the valid slices at some higher level in the library in some "unsafe" code. But this only solves the simple case where you have some array mapping to another array where the index relationship is obvious, and it's easily provable that there are no aliasing issues. Start layering in things like unique indirect indexing, or perhaps non-unique indexing but with atomic reductions, and it becomes difficult to phrase your correct program in a way to safe(!) Rust that is compatible with the borrow checker, at least without having to build a bunch of abstractions to express each of your parallel patterns. Having to build a bunch of bespoke abstractions may not be scalable to the types of developers building big scientific codes.

Anyway, I'm curious if the folks at "Embark" have spent any time thinking about the issue of shared, mutable state in GPU programming with Rust. It seems like a deal breaker from where I stand.

← PreviousPage 2 of 7Next →