dist1ll

13 ms·

121.
by dist1ll 3y ago
The difference is that when you're buying a x86 system, the entire CPU bringup (incl. AMEGAS/openSIL on AMD) runs proprietary and poorly documented firmware. You're entirely at the mercy of the vendor. Oxide has put immense e
122.
by dist1ll 3y ago
This server is for cloud operators - i.e. for organizations that are building their own cloud infrastructure. That's why Oxide is drawing comparisons to hyperscalers.
123.
by dist1ll 3y ago
The tabbing feature looks nice. AST-based code navigation has been a huge boon to productivity. My hope is that Helix gets around making /**/-style block comments more ergonomic. It's one of the reasons I'm still using n
124.
by dist1ll 3y ago
I wouldn't call TLBs small: > Consequently, modern processors have extremely large and highly associative two-level TLBs per CPU — for example, Intel’s Skylake chip uses 64-entry level-1 (L1) TLBs and 12-way, 1,536-entry level-2 (L
125.
by dist1ll 3y ago
In general terms, any language aiming to be lower-level than C should - have an "abstract" machine that is more concrete than C (and by extension less portable) - be easier to lower into optimal assembly (especially loop ops) - gi
126.
by dist1ll 3y ago
Incremental type checking is still painfully slow on a large codebase.
127.
by dist1ll 3y ago
> If you are trying to maximize port utilization, sometimes the exact instruction ordering can make a big difference. Interesting. I thought you'd be at the mercy of the instruction scheduler for aggressive OoO cores.
128.
by dist1ll 3y ago
The main abstraction of intrinsics is register allocation, right? Is there anything else that can be gained by handwritten asm?
129.
by dist1ll 3y ago
Just to add some jargon for folks who want to do more research into this: The technique of having the interpreter dispatch duplicated in the postlude of each handler function is called "threading". If your dispatch uses an indirec
130.
by dist1ll 3y ago
I'm using Excalidraw :) [0] https://excalidraw.com/ [1] https://github.com/excalidraw/excalidraw
131.
by dist1ll 3y ago
Thanks! It's inspired by this https://edwardtufte.github.io/tufte-css/
132.
by dist1ll 3y ago
author here. That's a really good observation kornel. Rust's philosophy of avoiding post-monomorphization errors at all costs is usually sung in high praises (followed by a ridicule of C++ template errors) - but it comes at a cost
133.
by dist1ll 3y ago
Storing byte offset inline, great idea. Thanks for mentioning that. Just note one disadvantage: if the byte offset is stored in memory, you have a data dependence on the traversal. So even though it's cache-friendly, it may cause serio
134.
Memory-efficient enum arrays in Zig(alic.dev)
322 points by dist1ll 3y ago|239 comments
135.
Material for MkDocs – Documentation that simply works(github.com)
1 points by dist1ll 3y ago|0 comments
136.
by dist1ll 3y ago
Could you share what's the content/protocol of these requests? And what's the average payload size (for req and response each)?
137.
by dist1ll 3y ago
> CPU architectures other than AMD64 and ARM [..] no longer relevant cough RISC-V cough
138.
by dist1ll 3y ago
Is the attestation zero-knowledge? Or does the attestation carry metadata that can uniquely identify a device?
139.
by dist1ll 3y ago
What you're describing is very common amongst developers. So common in fact, that I've written a post about this https://alic.dev/blog/comparisons In short: recognizing your insecurities is the first step. Th
140.
by dist1ll 3y ago
Here https://docs.oxide.computer/guides/system/initial-rack-setup
141.
by dist1ll 3y ago
As someone who's only dealt with commodity server hardware, these specs make me salivate. And all these boot/management TUIs are just so satisfying to look at. (Sorry to be that guy, but just a friendly suggestion: high contrast d
142.
by dist1ll 3y ago
There are many domains in embedded. You're probably referring to safety-critical and hard real-time contexts? Because cooperative multitasking is bog-standard in OS kernels.
143.
by dist1ll 3y ago
Calling a compiler infrastructure project like MLIR an "ML framework" feels a bit disparaging.
144.
by dist1ll 3y ago
You might get a kick out of junkyard computing [1]. The idea is to turn a bunch of thrown away phones into a compute cluster. [1] https://dl.acm.org/doi/10.1145/3575693.3575710
145.
by dist1ll 3y ago
The most common example is DPDK [1]. It's a framework for building bespoke networking stacks that are usable from userspace, without involving the kernel. You'll find DPDK mentioned a lot in the networking/HPC/data cente
146.
by dist1ll 3y ago
This paper isn't aimed at random developers, and it's not a criticism of mmap in general. This is an appeal to core database engineers to stop using the wrong tool for the job.
147.
by dist1ll 3y ago
Many general-purpose OS abstractions start leaking when you're working on systems-like software. You notice it when web servers are doing kernel bypass to for zero-copy, low-latency networking, or database engines throw away the kernel
148.
by dist1ll 3y ago
> small AST nodes will be pointlessly padded to account for the size of the largest AST node. This is a great point, and something I mentioned in my blog post on custom-bitwidth integer types: https://alic.dev/blog/c
149.
by dist1ll 3y ago
I feel that for this to be really viable, Rust would have to allow references that consume less memory (like 8-bit or 16-bit pointers).
150.
by dist1ll 3y ago
Correct. Rust lifetimes become less powerful in certain niches of high-performance programming where you're avoiding the heap or don't have one in the first place. Games, databases, embedded, HPC-style batch workloads, and even co

More ›