menaerus

51 ms·

961.
by menaerus 3y ago
I'd prefer discussing technical merits of given approach rather than who is who and who did what since that leads to appeal to authority fallacy. You're correct, results might be stale, although I wouldn't hold my breath for
962.
by menaerus 3y ago
"get statistics" is not what TPC-B does. Nor the invalid queries nor ... From https://www.tpc.org/tpcb/ , a TPC-B workload that pgbench runs by default: In August 1990, the TPC approved its second benchm
963.
by menaerus 3y ago
I was not even trying to make one. I was questioning the validity of "1% overhead" claim by providing the counter-example from respectable source.
964.
by menaerus 3y ago
While pgbench might be "simple" program, as in a test-runner, workloads that are run by it are far from it. It runs TPC-B by default but can also run your own arbitrary script that defines whatever the workload is? It also allows
965.
by menaerus 3y ago
pgbench is not a microbenchmark.
966.
by menaerus 3y ago
I believe it's a misrepresentation to say that "actual measured overhead is under 1%". I don't think such a claim can be universally applied because this depends on the very workload you're measuring the overhead wi
967.
by menaerus 3y ago
From https://fzn.fr/projects/frdwarf/frdwarf-oopsla19.pdf DWARF-based unwinding can be a bottleneck for time-sensitive program analysis tools. For instance the perf profiler is forced to copy the whole stack
968.
by menaerus 3y ago
Extremely light on the details, and also conflates it with the JIT which makes it harder to understand the point, so I was wondering about the same thing as well.
969.
by menaerus 3y ago
The fact that you will use a class or function template from STL, and compile it, doesn't change the fact that, regardless of that, you will have to link against the libstdc++ or libc++ which you didn't compile yourself. I am not
970.
by menaerus 3y ago
The problem as I understand it is not about the stale dependency artifacts in xyz directory but with the (transitive) (hell) dependency management itself. cargo makes it incredibly easy to pull in the shitload of dependencies of whom you do
971.
by menaerus 3y ago
ABI compatibility is needed in 99% of the cases since you basically never compile your own STL. You dynamically (or statically if you will) link against the one you have on a system since most software does not have a control over the machi
972.
by menaerus 3y ago
Great content from Jeff Dean slides. Since they are dating from 2009, I wonder what are the things he had changed his mind of.
973.
by menaerus 3y ago
This blog, and the corresponding paper, is mostly a bunch of hand-wavy biased statements, which google (job) security team seems to be very pervasive about for the past few years. Secure by Design™ - right. OTOH https://bughunter
974.
by menaerus 3y ago
Since the dataset is small enough so that it fits into the Linux kernel page cache, and since the benchmark is repeated for 5 consecutive times, first iteration of the benchmark will be bottlenecked by the disk I/O but the remaining 4
975.
by menaerus 3y ago
https://freebsdfoundation.org/blog/a-sneak-peek-simd-enhance...
976.
by menaerus 3y ago
I'd say that SIMD and even moreso CPU internals knowledge is not quite common and upmost performance is I think not among the highest priority goals in libc/libc++/libstdc++. The ones who need it will implement it themselves.
977.
by menaerus 3y ago
The distinction is whether or not you're rewriting something from scratch carrying no baggage from the thwarts of the existing system or you keep organically growing existing code to meet the new requirements. The latter is usually muc
978.
by menaerus 3y ago
Having to work and learn through the codebase to make some substantial improvements often requires substantial effort and even rewiring the code architecture itself. That's enough of the evidence for me.
979.
by menaerus 3y ago
C++ attempts were from scratch or were based on interventions on the existing codebase?
980.
by menaerus 3y ago
It's common sense what I said. It applies across the industry regardless of the programming languages used. On the contrary, where's the evidence suggesting that the Rust is what made Gecko rewrite succeed? Has there been any rewr
981.
by menaerus 3y ago
I did not mean "everything" in the broader context but in the context when it comes to writing "easy" multithreaded programs. Pretty much everything in that case becomes modeled through a shared-ownership or message-pass
982.
by menaerus 3y ago
I don't think this has anything to do with the language itself. If anything, you could claim the same for C++ since "easier to parallelize in Rust" is derived from the fact that Rust models pretty much everything as a shared-
983.
by menaerus 3y ago
Not that I was trying to make any of those claims but just trying to correlate the domain with what appears to be a common problem in it.
984.
by menaerus 3y ago
I was wondering about the same claim. However, I believe that JOIN's are a common weakness among OLAP database engines, and DataFusion is built on top of a columnar storage format - Apache Arrow.
985.
by menaerus 3y ago
> That might be true of C++, but I don't think its true in Rust And which was exactly my point. > both Scylla and Redpanda, two high performance C++ databases both use shared_ptr Such argument doesn't make sense since you wi
986.
by menaerus 3y ago
Not using SIMD at all as it seems. Thus relying on auto-vectorization of I'm not reading it indirectly. https://github.com/scipy/scipy/blob/main/scipy/spatial/src/d...
987.
by menaerus 3y ago
> I think the SIMD part has more to do with loop analysis than ILP. If you know how to rewrite the algorithm in such a way so that it makes close-to-ideal utilization of CPU ports through your SIMD then it is practically impossible to be
988.
by menaerus 3y ago
In my experience, compilers rarely know how to make use of ILP even in some for what you would expect to be the "simple" cases. Handwriting the SIMD, at least in my case, almost always proved to be several times faster than the au
989.
by menaerus 3y ago
Multi-threaded and/or async code can perfectly exist without shared-ptrs. And no, you don't need a shared-ptr to manage the lifetime of an object shared by two threads.
990.
by menaerus 3y ago
Something being async doesn't imply shared-ptr design. But perhaps this is what Rust makes you to to achieve its safety guarantees?

More ›