We ported pbrt-v4 to Julia and built it into a Makie backend. Any Makie plot can now be rendered with physically-based path tracing. Julia compiles user-defined physics directly into GPU kernels, so anyone can extend the ray tracer with new materials and media - a black hole with gravitational lensing is ~200 lines of Julia. Runs on AMD, NVIDIA, and CPU via KernelAbstractions.jl, with Metal coming soon. Demo scenes: github.com/SimonDanisch/RayDemo |
This is why I wish Julia were the language for ML and sci comp in general, but Python is sucking all of the air out of the room.
Most "Python" applications are actually bindings to C, C++ and Fortran code doing the real work.
All those GPU targets are powered by libraries, that are not part of Julia itself (GPUCompiler.jl). The same goes for automatic differentiation. That's remarkable in my opinion.
So you're right, that many programming languages could do it, but it's no wonder, that other languages are lacking in this regard compared to Julia.
Re: the compilation latency discussion — it's a real tension. JIT gives you expressiveness but kills startup. AOT gives you instant start but limits flexibility. Interesting that most GPU languages went JIT when the GPU itself runs pre-compiled SPIR-V/PTX anyway.
I'm asking because I had a lot of trouble trying to describe interfaces between materials, only to find out that what I wanted to do was not possible in PBRT without modifying the code. Apparently, in PBRT a material can only have one other material touching it. So, for example rendering a glass filled with water and ice is not possible without hacks. From a user's point of view this is a bit of a let-down, of course.
https://blog.yiningkarlli.com/2019/05/nested-dielectrics.htm...
And I'm curious how you solve it.
>the reference implementation from Physically Based Rendering (Pharr, Jakob, Humphreys)
I'd like to know a little about the process you went through for the port. That book * sounds like an excellent resource to start from but what was it like using it and the code?
Then I found that pbrt moved away from the initial design and I used claude code to port large parts of the new C++ code to Julia. This lead to a pretty bad port and I had lots of back and forth to fix bugs, improve the GPU acceleration, make the code more concise and "Julian" and correct the AIs mistakes and bogus design decisions ;) This polish isn't really over yet, but it works well enough and is fast enough for a beta release!
It’s like calling a framework Mike
Waving around an outdated blogpost as if it would automatically invalidate everything is just silly at this point.
They were paid to do this cool thing and possibly to post it here as well but if you dislike "capitalist pigs", maybe you shouldn't browse yc.
It took me days to get that build to work; doing this compilation once in CI so you don't have to do it on every machine is trickier than it sounds in Julia. The "obvious" way (install packages in Docker, run container on target machine) does not work because Julia wants to see exactly the same machine that it was precompiled on. It ends up precompiling again every time you run the container on other machines. I nearly shed a tear the first time I got Julia not to precompile everything again on a new machine.
R and Python are done in five minutes on the standard worker and it was easy; it's just the amount of time it takes to download and extract the prebuilt binaries. Do that inside a Docker container and it's portable as expected. I maintain Linux and Windows environments for the three languages and Julia causes me the most headaches, by far. I absolutely do not care about the tiny improvement in performance from compiling for my particular microarch; I would opt into prebuilt x86_64 generic binaries if Julia had them. I'm very happy to take R's and Python's prebuilt binaries.
> I would opt into prebuilt x86_64 generic binaries if Julia had them
The environment varial JULIA_CPU_TARGET [1] is what you are looking for, it controls what micro-architecture Julia emits for and supports multi-versioning.
As an example Julia is built with [2]: generic;sandybridge,-xsaveopt,clone_all;haswell,-rdrnd,base(1)
[1] https://docs.julialang.org/en/v1/manual/environment-variable...
[2] https://github.com/JuliaCI/julia-buildkite/blob/9c9f7d324c94...
You may be interested in looking into AppBundler. Apart from the full application packaging it also offers ability to make Julia image bundles. While offering sysimage compilation option it also enables to bundle an application via compiled pkgimages which requires less RAM and is much faster to compile.
Elsewhere someone used the term "janky" and perhaps it's the fact that there are so many incredibly smart people around it that makes it so janky. By way of example, somebody needed to check disk space and the architect told him to shell out to Python.
Remember when LLVM first came out and it got kudos for the quality of its error messages? Well if you miss the old-school 1980s GCC experience the nonsense that eventually comes out of the Julia compiler after an hour will relight that flame.
Want to use greek letters and other symbols that don't appear on your keyboard as variable names? You've found your people.
I think what happened is this: Julia got advertised as "Python syntax, C speed" but in practice it turns out to really be "Python syntax, 50% of C speed if you were willing to avoid some semi-well-documented gotchas, where avoiding said gotchas will take some non-trivial effort". Again, great if you are willing to work with it.
I am not saying that the Julia people are responsible for the "Python syntax, C speed" perception as much as that was what the prevalent perception became. And
I have talked to people in computational biology who tried Julia, and they said something or the other similar to "It just wasn't performant enough for me to give up Python," and if you really dig in, what really happened was when new people tried Julia with old mental models, they walked away thinking, "Heh, more MIT hypeware."
https://arxiv.org/abs/2309.17309
This paper in experimental high-energy physics is a good example of why Julia is popular for scientific calculations.
It shows that #julialang is over 100 times faster than Python and even faster than C++.
https://github.com/NumericalEarth/NumericalEarth.jl
My own take is that Julia didn’t since the two language problem as much as was defeated by it.
Julia didn’t attract the high-level Python data science crowd because of Julia’s latency issue, lack of package ecosystem, and the inconveniences that a high performance compiled language incurs, such as having parametric containers.
The research software engineer crowds didn’t buy in because Julia has no interfaces or automatically checkable behavior, poor static tooling, imprecise semantics which is hard to build abstraction on, and a complex performance model that makes it hard to ensure speed, and is hard to deploy.
So, where they tried to make a language that can span the gap, they succeeded in making a language that works for neither, and which no-one wants.
I like the language. But after having used it for eight years, I find it increasingly hard to argue against the point that it’s better to choose Rust for software engineering and Python for scripting.
Edit: I should say: I used it for eight years because it IS fine for my specific niche: High performance research software engineering. Where I care neither about the convenience of Python, nor need to write truly robust and maintainable code. Where my choice of language was personal and I didn’t need to convince a team of coworkers.
I made many critical comments about Julia here, on this website, but they mostly boil down to clumsiness of Julia.
Does it solve "two language problem"? Kinda, but through this it is less convenient to use than Python, and harder to reason about performance of the particular piece of code than C. Yes, there is a big chance that idiomatic, straightforward Julia code will run pretty fast, but there is also big chance that it will run unexpectedly slow, and you need to know fair bit to be able to debug this... so kinda like going from Python to C?
Is dynamism and interactivity useful? Immensely, but Julia pays for it with poor AOT support (yes, even with juliac and, still experimental, --trim option).
There is also stuff that I consider unacceptable. Debugger being a separate package you need to download, and it even cannot debug compiled code so it needs its custom interpreter? This piece of crap that is Base.@enum, so anyone that wanting proper enums need to install EnumX.jl? And why the hell StaticArrays.jl even exists as a separate package if Julia puts so much focus on numerical applications?
And then we come to tooling and IDE support. Oh, boy - Julia VSC extension is such a miserable experience, and there is not much else out there.
Julia is incredibly fun to play and tinker with solo, and some stuff from SciML is straight-up awesome, but overall it is wasted potential, killed by thousand cuts.
Rust and Python work in some scenarios but not all. Unless the native components are all developed already, you will need a Rust programmer on your team. Rust expertise may be available to organizations of a sufficient size and a narrow focus. In my experience, this sort of arrangement requires a team with dedicated resources.
What I encountered more frequently are attempts to use Python as a full stack language. People are not just using it to implement the scripting frontend interface but also trying to implement the backend that does the heavy processing and the numerical computation. As a developer this is a terrible experience. I'm in the middle of replacing for loops in Python with vectorized numpy code but most of my efficiency gains are erased because I still need Python tuples in the end. Yesterday, I had to consider whether exceptions I throw in Cython code can be caught properly in Python.
Research software engineering is one field where you really do need a full stack language. That kind of software engineering requires high performance with limited resources. Julia with some initial patience does deliver a better developer experience in this scenario than the equivalent Python experience for me partly because I do not need to play vectorizarion games, and I can do everything in one integrated environment.
While, yes, interfaces and static tooling could be better, I do think the situation has gotten better over time. There are interface schemes available and additional static tooling is available.
Julia could deliver a better user experience though. Admittedly, the Python tooling to deploy complex solutions has significantly improved lately via tools such as pixi. Julia tools could deliver generic binary code with packages to ease the initial user experience. Advanced users could re-precompile packages to optimize.
The most promising success I have had with Julia is putting notebook interfaces in front of scientists or deploying web apps. My observation in this scenario is that Julia code can be presented as simple enough for a non-professional programmer to modify themselves. Previously, I have only seen that work well with MATLAB and Python. I have not seen this happen with Rust, and I do not expect that to change.
The other observation is that users appreciate the speed of Julia in two ways. 1. Advanced data visualizations are responsive to changes to data. In Python, I would typically need to precompute or prerender these visualizations. With Julia, this can be done dynamically on the fly. 2. Julia user interfaces respond quickly to user input.
While I think Julia has plenty of room to improve, I do think those improvements are possible. I have also greatly appreciated how much Julia has improved in the past five years.
"Data science" is an extremely broad term, so YMMV. That said, since you asked, Julia has absolutely replaced Python for me. I don't have anything new to add on the benefits of Julia; it's all been said before elsewhere. It's just a question of exactly what kind of stuff you want to do. Most of my recent work is math/algorithms flavored, and Python would be annoyingly verbose/inexpressive while also being substantially slower. Julia also tends to have many more high-quality packages of this kind that I can quickly use / build on.
Yes 1-based indexing is a mistake. It leads to significantly less elegant code - especially for generic code - and is no harder to understand than 1-based indexing for people capable of programming. Fight me.
Some would argue that 0-based indexing is significantly less elegant for numerical/scientific code, but that depends on whether they come from a MATLAB/Fortran or Python/C(++) background.
A decision was made to target the MATLAB/Fortran (and unhappy? Python/C++) crowd first, thus the choice of 1-based indexing and column-major order, but at the end of the day it's a matter of personal preference.
0-based indexing would have made it easier to reach a larger audience, however.
> and is no harder to understand than 1-based indexing for people capable of programming.
The same could be said the other way around ;-)
To reiterate, citing studies that show that smoking causes cancer in chain smokers does ... nothing. You are citing studies, but I am not the chain smoker; I am just the guy talking about chain smokers.
One more time, I wish we lived in a world where public perception was swayed by objective studies, but we don't.
Julia is fast, yes, but when a university sys-admin rolls their eyes at hearing its name, you have lost the battle for well and good.
Recently we got the public keyword, but even the PR there says:
"NOTE: This PR is not a complete solution to the "public interfaces are typically not well specified in Julia" problem. We would need to implement much than this to get to that point. Work on that problem is ongoing in Base and packages and contributions are welcome."
Use C, C++ or Fortran for the heavy lifting, and Python for UI/business logic/non-high perf stuff for rapid app development.
It gets the job done, but the existence of Cython, Numba, Pythran, PyPy and many, many, others are indication that this isn't a global optimum.
Anyway, I'm looking at this from the user's perspective. I wanted to do some physics-based ray-tracing with lenses and pbrt is what I ended up trying. As such, I really needed the multi-material aspect to work correctly. Also, it would be nice to be able to describe surfaces using a z=f(x,y) kind of formulation, or a way to place a hook in the renderer.
Keep in mind that it went with 1 based indexes to make the switch easy for Matlab types. I'm not sure if that was a good or bad move for the long term. I'm sure it got some people to move who otherwise wouldn't have but conversely there are also people like me who rejected it outright as a result (after suffering at the hands of 1 based indexing in Matlab I will never touch those again if I have any say in the matter).
I've considered switching to it a few times since seeing that they added variable indexes but Python works well enough. Honestly if I were going to the trouble of switching I'd much rather use Common Lisp or R5RS. The nearest miss for me is probably Chicken, where you can seamlessly inline C code but (fatally) not C++ templates.
If I ever encounter "Chicken, except Rust" I will probably switch to that for most things.
The syntax is a bit weird; python, swift, rust, and zig feel more parsimonious.
I absolutely love multimethods, but I think the language would have been better served by non-symmetric multimethods (rather than the symmetric multimethods which are used). The reason is that symmetric multimethods require a PHD-level compiler implementation. That, in turn, means a developer can't easily picture what the compiler is doing in any given situation. By contrast, had the language designers used asymmetric multimethods (where argument position affects type checking), compilation becomes trivial -- in particular, easily allowing separate compilation. You already know how: it's the draw shapes trick i.e., double-dispatch. So in this case, it's trivial to keep what compiler is "doing" in your head. (Of course, the compiler is free to use clever tricks, such as dispatch tables, to speed things up.)
The aforementioned interacts sensitively with JIT compilation, with the net outcome that it's reportedly difficult to predict the performance of a snippet of Julia code.
1. I use the term "performance" slightly vaguely. It's comprised of two distinct things: the time it takes to compile the code, and the execution time. The issue is the compilation time: there are certain cases where it's exponential in the number of types which could unify with the callsite's type params.
2. IIRC, Julia compiler has heuristics to ensure things don't explode for common cases. If I'm not mistaken, not only do compile times explode, but certain (very common) things don't even typecheck. There's an excellent video about it by the designer of the language, Jeff Bezanson -- https://www.youtube.com/watch?v=TPuJsgyu87U . Note: Julia experts, please correct me if this has been fixed.
3. The difficulty in intuiting which combinations of types will unify at a given callsite isn't theoretical; there are reports of libraries which unexpectedly fail to work together. I want to qualify this statement: Julia is light years ahead of any language lacking multimethods when it comes to library composability. But my guess is that those problems would be reduced with non-symmetric multimethods.
4. The non-symmetric multimethod system I'm "proposing" isn't my idea. They are referred to variously as encapsulated or parasitic multimethods. See http://lucacardelli.name/Papers/Binary.pdf
I have huge respect for Jeff Bezanson, for the record!
In what way? It's more-or-less the same syntax as Ruby and Elixir, just with different keywords. Like as much as I love Zig, Zig's syntax is way weirder than Julia's IMO (and none of 'em hold a candle to the weirdness of, say, Erlang or Haskell or Forth or Lisp).
> My conclusion after using Julia for many years is that there are too many correctness and composability bugs throughout the ecosystem to justify using it in just about any context where correctness matters.
I am intimately familiar with JULIA_CPU_TARGET; it's part of configuring PackageCompiler and I had to spend a fair amount of time figuring it out. Mine is [0]. It's not related to what I was discussing there. I am looking for Julia to operate a package manager service like R's CRAN/Posit PPM or Python's PyPI/Conda that distributes compiled binaries for supported platforms. JuliaHub only distributes source code.
[0] generic;skylake-avx512,clone_all;cascadelake,clone_all;icelake-server,clone_all;sapphirerapids,clone_all;znver4,clone_all;znver2,clone_all
ARG JULIA_CPU_TARGET="generic;skylake-avx512,clone_all;cascadelake,clone_all;icelake-server,clone_all;sapphirerapids,clone_all;znver4,clone_all;znver2,clone_all"
ARG JULIA_PROJECT=[...]
ENV JULIA_PROJECT=[...]
RUN julia -e "using Pkg; Pkg.activate(\"[...]\"); Pkg.instantiate(); Pkg.precompile();"
Thank you so much for this! It's too late for me to edit my original post, but cutting the install time in half is a major win for me. Now it only needs to precompile, not also compile a sysimage.
The real indexing issue is whether arbitrary-base abstraction is too easily available.
# Correct, Vector is 1-based
function mysum(v::Vector{T}) where {T <: Integer}
s = zero(T)
for i in 1:length(v)
s += v[i]
end
return s
end
#Incorrect, AbstractVector is not necessarily one based
function mysum(v::AbstractVector{T}) where {T <: Integer)
s = zero(T)
for i in 1:length(v)
s += v[i]
end
return s
end
#Correct
function mysum(v::AbstractVector{T}) where {T <: Integer)
s = zero(T)
for e in v
s += e
end
return s
end
It really isn't. "Scientific code" isn't some separate thing.
The only way it can help is if you're trying to write code that matches equations in a paper that uses 1-based indexing. But that very minor advantage doesn't outweigh the disadvantages by a wide margin. Lean doesn't make this silly mistake.
> If you really need the first or last element
What if you need the Nth block of M elements? The number of times I've written arr[(n-1)m+1:nm] in MATLAB... I do not know how anyone can prefer that nonsense to e.g. nm..<(n+1)m
They way people reveal themselves is a pattern worthy of taking note.
I find it to be substantially worse. It's fine as long as you don't manipulate the indicies. But as soon as you start doing math on them 1 based becomes a headache (at least IME).
Meanwhile all you get in exchange (at least as far as I can tell) is ease of speaking about them in natural language. But I'm not usually conversing about indicies.
Concise range notations are a mixed bag. There's pros and cons to either scheme there as far as the syntax goes.
I don't think one is better than the other but my mind is currently wired to see indexing with base 1.
Then there's Option Base 1 in VBA if you don't like the default behavior. Perfect for creating subtle off-by-one bugs.
Second, null denominations usually introduce names -- whether for variables, types, functions, lifetimes, macros, etc. One exception to this are free-standing value expressions (a bit weird; less so when they're the last expression in a block indicating the value returned by it). Another other exception would be directive type constructs - eg directives to import names from another module, directives to give hints to the compiler, etc. The last two exceptions are the most common ones: variable assignment and function invocation.
The golden rule of good language design, as I see it, is this: null denominations must begin with a fixed and unique token. The only permissible exceptions should be for assignment and function invocation; exceptions which exist because those use-cases appear so often in a typical program that requiring a prefix would be insufferable.
Julia breaks this rule for global variables. (Fair enough, Python also commits this error, but it's a mistake and a source of bugs!) But wait, Julia also has "const" and "local" binding constructs, where it follows the golden rule -- but now your syntax isn't consistent. So now you need to keep in your head these nuances -- and know the difference between a soft and hard scope -- when you want to write a function which modifies a function using macrology.
(As a point of taste on the choice of prefix token: introduction of variables through "local" is just as weird as C++'s "auto" -- and at least Bjarne Stroustroup had an excuse for that choice. Anyone who introduces a global variable in a local scope should be punished imho, so there's no need to say "this is a local variable", it's obvious from the fact that the name is introduced inside a function. Instead, my personal preference is to introduce constants through "let", and variables through "var". The former is well-known to anyone numerate, and the latter is ubiquitous in software engineering. Both read well; they're as close as possible as you can get to constructs in English.)
Julia breaks the golden rule again with its succinct, Mathematica-style notation for function definition. I get that it wants to appeal to Mathematica users, but Python already proved you don't need to do that. This is a programming language; brainy types, like mathematicians and physicists, aren't going to be flummoxed by an unfamiliar notation for function definition, or irritated by having to type a few extra characters.
I mention macrology; it's not just that. Let's say you want write a syntax highlighter -- you need to take into account all that weirdness. If null denominations have a fixed & unique prefix, parsing is easy-peasy. Want to add a capability to "inline" HTML code within Julia, react-style? You're going to run into similar issues. And so on...
arr[n..=m]
> arr[1:m] and arr[(m+1):end]
arr[0..m], arr[m..]
Much nicer.
> Arrays are (conceptually) not pointer arithmetic.
Look at a ruler. Does it start at 1?
so you just need to overload the syntax of intervals even more to make it work
> arr[0..m], arr[m..]
now `m` refers to different things depending on which side of the interval it's on. less characters doesn't mean nicer
I get it though, I was skeptical about 1-based indexing when I started Julia. By the nature of indices vs length there will always be an off-by-one problem: either you have elements [n, m - 1] with length (m - n) or [n, m] with length (m - n + 1). Unless you're doing a bunch of pointer arithmetic type stuff, I find the symmetry of a inclusive-inclusive interval to be a better default.
As a final rebuttal I offer: range(n - 1, -1, -1)
As a neat bonus, in Julia 1:5 is just the iterator for the numbers 1 to 5. So slicing is typically not some special syntax either. It all works rather nicely.
Which is more natural? The ruler is exactly the right mental image if an array to you is a partitioned region of memory starting at a specific pointer location. If an array to you is an ordered collection of objects, you would never invent 0-based indexing or inclusive-exclusive slicing.
Either way, it's not a big deal. I have lived in both worlds, I have come to think Julia is a bit more natural and easier to teach. But it ls really the silliest bike shedding complaint, given that the language has considerable real trade offs.
Yes, of course distances are measured starting from 0. But we count discrete things starting at 1. You can do mental gymnastics to enumerate from zero and many programmers are (unfortunately IMO) taught to do so. It's a hard thing to learn that way, so for the folks that have done so, it often becomes a point of pride and a shibboleth.
As a classic example, a four story building has four floors. But you only need to go up three flights to get to the top. You can legitimately call the top floor either 3 or 4, and folks are similarly tribal about their own cultural norms around this one, too.
No I disagree entirely. One is simply better.
> It's a hard thing to learn that way, so for the folks that have done so, it often becomes a point of pride and a shibboleth.
It is not hard. It's not better because it's hard-won knowledge. It's better because it leads to simpler, more elegant code. Simple as.