Piezoid

(u)int8_t have the same problems, including aliasing because they are just alias of (unsigned) char. Sometimes it's nice to have modular arithmetic mod 256, or compact memory layout for eg. count sketches.

Depression can be seen as protection mechanism that reinforce rigid thought patterns in order to avoid runaway (psychosis, mania). So anything tiping the balance in the other direction allows to deprogram those behaviors.

This is related to the "entropic brain" theory, see https://www.frontiersin.org/articles/10.3389/fnhum.2014.0002...

Thank you for the clear explanation.

Could this find applications in (realtime ?) acoustic room correction on embedded devices ?

Do you think it's planned to bring back noalias in rustc ?

This, along with const generics and simd, would make rust the perfect language for me.

> For instance, the AGC (like many 1960s computers) didn't have a stack, so you had to keep track of the return address for each subroutine call.

> I managed to get everything to fit in one bank by reusing these 16 words for multiple purposes, but I spent a lot of time debugging problems when a variable clobbered a location still in use.

It could be fun to make a slightly higher level ad-hoc assembly language for solving these problems. For example SSA with basic blocks.

Yes, I'm aware of that. Sorry, bad wording. I choose btrfs because it's simpler to setup. It is also lighter on memory for uses on personal machines. I know ZFS can work with less memory but at a performance disadvantage.

Honestly not bad. About 7 years of use on personal machine, single volume. It simpler than ZFS and allows cheap incremental backups with send/receive.

One year ago I built a RAID56 with 5 used 2TB drives from eBay. Risky move, but it went smoothly so far. It's only for home storage and important stuff is backuped off site anyway. One Seagate drive died with lots a bad sectors. The replace command took really long even with the "-r" flag (don't read from replaced drive, in theory), so I ended up unplugging the drive and rebalancing from there.

I have high hopes for bcachefs. We have a real need for a modern FS with tiered caches. I backed the project but I don't have the skills or time to help.

I upgraded a free T410 with two X5675, 32GB ECC RAM, PCIe to NVMe adapter, a RAID array of 2TB drives and an USB3 controller.

I's a quite capable machine. I needed it to learning about NUMA archs and test my software.

However there is no sleep mode. The boot time is not that bad for a server so I start it with IPMI bounced from a SBC.

I made it quiet the hard way, mainly for fun and learning about embedded control loops: water cooling with a passive motorbike radiator and a Arduino to control pumps and monitor temperatures while feeding fake hall sensor data to the original BMC so it doesn't freak out.

Having a very large distributed surface area could be key to drive the peltier cells at low current and get back some efficiency. The thermal contact with the body also helps by reducing the temperature gradient against the charges flow.

> They pushed for Gnome to have a hard dependency on it so that everyone would be _forced_ to use it instead of just letting it be accepted by merits.

We should standardize on the interfaces, avoiding hard dependencies. There is dbus for that.

In my field (bioinformatics), there is a trend of using a single hash function, an over-sized bloom filter which is then compressed with something like entropy optimal RRR[1] or RoaringBitmap bit vectors. The resulting bloom filter ends up taking the same space than a properly sized bloom filter. There is two advantages: knowing the size of the set in advance is not necessary, and a single hash computation and memory access per query. However updates are not well supported.

[1] Raman, R., Raman, V., & Rao, S. S. (2002, January). Succinct indexable dictionaries with applications to encoding k-ary trees and multisets. A blog post: https://alexbowe.com/rrr/

They would probably be buried under sediments before decomposing into micro-plastics.

A good blog post about xor-shift-multiply reversible hash functions: https://nullprogram.com/blog/2018/07/31/ The inverse functions are easily found by computing the modular inverses of the odd multiplicative factors.

Real perfect hashing is more about hashing a non contiguous set of elements with no collisions, unlike the set of all integers that fit in n bits.

A different problem is Minimal Perfect Hashing. In addition to being injective, the function should have a minimal image of the input set: the set of keys is mapped to consecutive integers without vacant slots.

The implementations in optimal space are non-practical but some libraries construct such functions with a bit more memory (2-3bits/key), like emphf or BBhash (both on github). The latter use a series of 1-hash bloom filters where keys having collision on one layer are removed to be handled by deeper layers. The overall rank() of the bit where a key ultimately lands gives its hash.

The mini/micro-brute by Arturia are the new cheap analog bass synths.

They combine a minimal amount of monophonic signal nodes while providing quite a lot of possibilities. The Voltage Control IOs are a bridge to Modular Acquisition Systemic Syndrome :)

Paypal's post login page comes to mind. I want to see the dashboard after login, not a uninformative ad that probably took time and money to put there.

Sadly, 10 years later, they didn't respond to criticism about removing support for non-ext4-yet-xattr-enabled filesystems.

I think F# will be my next learning project. Still keeping an eye on Idris, though.

I will keep my dual X5690 machine for few more years I guess. The increase IPC of modern CPU is only 33% compared to theses old Xeons. Although, power consumption is another factor to consider...

Sometimes the "curse of dimensionality" is a good thing :)

You're right. The main selling point of both turbines is the axial flow: the velocity differences between the water and the blades are the same on every sides of the turbine. The shear forces are low enough to let wildlife pass through unharmed (supposing that it's not heavily torqued by a high electric loads).

Facebook built Haxl[0], a library in Haskell, precisely for this use case: batching, caching and paralyzing requests against external sources.

[0] https://github.com/facebook/Haxl

Domain Specific Languages are good thing for specialized applications. Here we want automatic differentiation, parallelization and vectorization. It gets bad when general purpose language's features are added to allow things that the language wasn't designed for. A good interface to a GP language would be more relevant.

Random projections methods are cheaper alternatives to SVD. For example you can bin contexts with a hash function and count collocations between word and binned contexts the same way this article does. Then apply weighting and SVD if you really want the top n principal components.

What's nice with counting methods is that you can simply add matrices from different collections of documents.