Hacker News Reader

gcpon Intel to Develop Discrete GPUs, Hires AMD's Raja Koduri as Chief Architect

Also the fact that OpenCL lost to CUDA for being stuck in C for so long, shows what most GPU devs actually prefer.

Aw come on, you're sure it has nothing to do with the largest GPGPU vendor pushing CUDA very heavily and intentionally gutting their OpenCL tools? Or putting out a ton of very high performance libraries with no OpenCL equivalent? Putting out a shitton of marketing and tutorial videos for CUDA only?

Yeah, that surely was totally unrelated.

Also pushing a proprietary standard goes faster than a standardized one. No surprise there.

gcpon Intel to Develop Discrete GPUs, Hires AMD's Raja Koduri as Chief Architect

What GPU world though?

People still write regular shaders in languages that are much more C like.

People writing GPGPU code are few. Most of the DL GPU use is in Python through several layers, and in the end you are running hand-written SASS assembly sitting in an NVIDIA DLL or whatever.

I guess some people must think it's handy to have C++ support in GPU kernels, or they wouldn't have added the feature. But for it to drive technology, hard to believe.

gcpon Don’t use C++ auto? Restricting auto is not the best decision

But why?

Duck typing works fine.

gcpon Don’t use C++ auto? Restricting auto is not the best decision

This doesn't sound like a case where auto is the one making things harder...

gcpon I bought a Switch from Nintendo and they threatened me with legal action

Some of the vendors allow you to select the delivery company. I like this.

It's pretty hard to get back to a seller and try to explain him the delivery company he is using is potentially scamming, especially as it only affects me (the receiver), not the shipper.

gcpon I bought a Switch from Nintendo and they threatened me with legal action

Believe me, even if you are home 100% of the time, packages will still not get delivered because "nobody was home".

IIRC DHL is especially prone to this. I think the drivers try to catch up time or whatever.

gcpon I bought a Switch from Nintendo and they threatened me with legal action

I don't doubt you're right about the "anything but basic refund/reship is a nightmare", but I think the reason people are praising Amazon (and I would too) is that most other companies (big OR small) don't even do that, even though it resolves like 90+% of the customer support issues and leaves happy customers behind.

As for sellers, yes, it's obvious they're getting screwed.

gcpon Intel to Develop Discrete GPUs, Hires AMD's Raja Koduri as Chief Architect

I'll continue using it as long as the users are still split between NVIDIA and AMD cards.

If AMD GPUs die out, CUDA it is.

Or maybe give up and use a wrapper library for whatever 10 alternatives-only-supported-by-one-marginal-vendor there are. (like Apple)

gcpon Intel to Develop Discrete GPUs, Hires AMD's Raja Koduri as Chief Architect

The OpenCL kernel language is very similar to the CUDA one, before they started adding more C++ support. It's pretty easy to translate between the two if you don't use hw intrinsics.

And segmenting the codebase is a MAJOR feature. With CUDA you are stuck on an old compiler until NVIDIA issues an update. How anyone can think this is a good idea...

gcpon Intel to Develop Discrete GPUs, Hires AMD's Raja Koduri as Chief Architect

Same here. What else can you do to ship GPU acceleration to AMD and NVIDIA people?

The alternatives recommended here aren't even serious IMHO. I'd rather switch to CUDA and wait till Intel/AMD sort out a REAL compatibility layer than deal with those.

Unless I'm mistaken, HIP still requires a separate compile for either platform and what runtime do they expect end users to have exactly?! At least CUDA and OpenCL are integrated in the vendor drivers.

Vulkan compute with SPIR-V seems to be the only real solution, but even that is still very early. Sill waiting for proper OpenCL 2.0 support in NVIDIA drivers :P

gcpon Attack of the week: DUHK

So we're hoping CloudFlare and Amazon don't have such bugs? (Ignoring the fact that it was already demonstrated at least once that they in fact do...)

gcpon Attack of the week: DUHK

Random (no pun intended) thought after reading this: how much of the supposedly secure HTTPS/TLS traffic is terminated by hardware load balancers and such that aren't really open and might have similar vulnerabilities?

gcpon Creator of cURL wins Polhem Prize 2017

Straight from the GNU page:

"The original author of GNU Wget is Hrvoje Nikšić."

gcpon Creator of cURL wins Polhem Prize 2017

I wonder if the media attention due to him getting banned from traveling to the USA contributed to him being known to a wider audience, and thus on the radar of whoever picks the award winner.

gcpon Enabling Rust on Embedded Platforms – Linux, RTOS, Bare Metal

they all panic on failure. It’s one thing for Firefox to crash when my computer completely runs out of swap

Actually most large allocations in Firefox are fallible and the browser will only crash when it can't deal with an allocation in security-sensitive code. So that's a terrible example!

gcpon Key Reinstallation Attacks – Breaking WPA2 by Forcing Nonce Reuse

What's the word for pressuring a person until they make a decision they immediately regret?

gcpon The trouble with text-only email

The problem is that they are Mozilla's competitor (Google - Gmail). It's not named explicitly but the dots are pretty damn obvious to connect, I think.

gcpon The trouble with text-only email

Mozilla employs about 1000 people and has revenues of around 300M-400M, so there's been a miscalculation there. 1/400th with 1000 employees doesn't sound very high at all.

The numbers are probably from the Mozilla Foundation, which only employs a few people, so obviously they're going to be remunerated highly percentage-wise.

gcpon The trouble with text-only email

The deal is with Yahoo.

gcpon Key Reinstallation Attacks – Breaking WPA2 by Forcing Nonce Reuse

Somewhat pointless remark as WPA only protects up to the access point. Any vulnerability after that has wider implications regardless of the WPA status.

gcpon Severe flaw in WPA2 protocol leaves Wi-Fi traffic open to eavesdropping

Apparently the relevant specs are behind a paywall, so almost no-one who understood security properly looked at it.

This was rather obvious from WEP, no?

gcpon Firefox 56: Last Stop Before Quantum

File a bug.

gcpon Coffee Lake Review: Initial Numbers on the Core i7-8700K and Core i5-8400

No, no, not at all!

That's exactly what I was pointing at: most reviews are using the most power hungry Ryzen SKUs (95W), so of course it looks bad in comparison. Meanwhile, the price/perf leader and most sold Ryzen SKU is only 65W. This one is ahead of Intel in power efficiency.

gcpon Coffee Lake Review: Initial Numbers on the Core i7-8700K and Core i5-8400

Doesn't Ryzen have better power efficiency compared to these new Coffee Lake's?

The issue is a bit muddled because sites are (for some weird reason?) ignoring the most common Ryzen SKU, which is 65W (R7 1700).

gcpon Coffee Lake Review: Initial Numbers on the Core i7-8700K and Core i5-8400

Does this mean that if you OC a single core to the limit to get the best single threaded performance, you will see drops in performance when your other cores start getting utilized?

This is how all CPUs with some form of "Boost" work, yes. People who overclock generally lock the multiplier to avoid it. So you will tend to have this problem when you do not overclock.

gcpon Coffee Lake Review: Initial Numbers on the Core i7-8700K and Core i5-8400

Computerbase concluded that the R7 1700 and the R5 1600(X) beat (respectively) the 8700K and 8600K in price/performance for non-gaming uses.

Pretty graphs: https://pics.computerbase.de/8/0/0/8/6/diagramme/70-630.1507...

Somewhat surprising to see AMD in this situation given the console design wins and those being 8 core systems, but for now gaming remains Intel's forte, though.

gcpon Coffee Lake Review: Initial Numbers on the Core i7-8700K and Core i5-8400

You can RMA affected chips (I've done so), which were from early batches (apparently). It's inconvenient but it's not the first and won't be the last piece of hardware that has to be RMA-ed. At least you can RMA them! Intel didn't offer that when TSX turned out to be broken.

After compiling things on an 8 thread Haswell Xeon, and now a 16-thread Ryzen, I really want a ThreadRipper...

gcpon Intel Gears Up for FPGA Push

Don't you need the vendor-specific backend for the physical P&R?

gcpon Intel Gears Up for FPGA Push

The applications are the same, but an FPGA allows extremely narrow optimization. Got something that doesn't parallelize all that well, but can be pipelined, or benefits from a wide datapath? Design that. Got something that parallelizes well and only needs a tiny datapath? Make it a 1000 wide vector machine. It needs difficult control flow? No problem, we're flexible as we're not a GPU, etc.

The optimization needs to beat out the inefficiency of the FPGA compared to an ASIC though (not to speak of development time). But sometimes, you can do just that.

gcpon Intel Gears Up for FPGA Push

Sounds about right. I guess the market economics don't work out for a simple design (i.e. with not much IP revealed by open sourcing the spec) that can be programmed with OS tools. I mean, there's these ICE chips, which is a huge step, but realistically I'd want something just a tiny bit bigger and with a hardcore (ARM or MIPS) in it.

Or maybe the market economics work out, but are small relatively compared to the big players, so it moves glacially.