Intel SPMD: Compiler for High-Performance SIMD Programming

http://ispc.github.io/

by vmorgulis • 10 years ago

101 48 10 years ago

ispc.github.io

So this compiler is targeting the SIMD units of CPUs rather than GPUs. Can anyone contrast what the performance of this would be relative to Cuda or OpenCl for various applications, for example neural nets?

zurn• 10 years ago

SPMD is conceptually more like a job system, so this generates code that is not necessarily all executing same code in lockstep. SPMD/SIMD is "single program" vs "single instruction". The compiler does also exploit SIMD but that's orthogonal to the SPMD concept.

reference: https://ispc.github.io/ispc.html#the-ispc-parallel-execution...

alkonaut3• 10 years ago

For large data and trivial algorithms, such as multiplying matrices (which means then any problem you can express as a set of operations on large matrices) gpu's do really well, so it's hard to compete with something that has 1000 cores (Edit: "compute units", not "cores"). Neural nets is essentially matrix multiplication.

However a lot of interesting problems are seemingly parallel but highly branching and nonlinear. Take path tracing as an example: it's very little code and highly parallel as each Ray/pixel is independent, yet it's not an easy problem for a GPU: each time a ray bounces it will disperse and not do whatever the Ray next to it was doing in terms of which geometry it will hit etc.

It might seem like today if a problem can benefit from 8 CPU cores then it benefits 100x more from being run on a GPU but this is far from true. A great machine for general computing could do well with a board with 100 x86 CPUs apart from having a big gpu with a thousand cores for brute forcing the "simpler" problems.

vardump5• 10 years ago

... so it's hard to compete with something that has 1000 cores.

Any references to what has "1000 cores"? Nvidia GPUs usually have about 12 or so cores that can be compared to x86 cores, meaning they can independently branch.

For example high end Nvidia 980 GTX GPU has only 16 of such comparable SIMD execution cores. SMXs or whatever Nvidia calls them.

GPU marketing materials confusingly refer as cores to something like x86 CPU SIMD lanes (and that's being very generous to GPUs), that artificially inflates the numbers.

Or put differently, one CUDA core can compute up to 1 FMA per cycle @1196-1300 (?) MHz. One recent Intel X86 core can compute at least up to 16 FMAs per cycle @2800-4000 Mhz.

creshal2• 10 years ago

A great machine for general computing could do well with a board with 100 x86 CPUs apart from having a big gpu with a thousand cores for brute forcing the "simpler" problems.

Which is the idea behind Intel's Xeon Phi "GPU" with 70+ Pentium/Atom cores, which this compiler specifically targets.

[Long before that, Intel showcased an 80 core x86 CPU in 2007 (Polaris/Teraflops Research Chip) – and then promptly shelved it to focus on building programming languages and compilers that can actually make use of it, before introducing the Xeon Phi half a decade later.]

Symmetry2• 10 years ago

It's true that problems with diverging control flow work better on a number of cores than on a GPU. But by the same token a GPU's SIMT execution model does better with ray tracing than the SIMD units on a CPU. And the article is about targeting SIMD units.

pcwalton• 10 years ago

"Debunking the 100X GPU vs. CPU Myth: An Evaluation of Throughput Computing on CPU and GPU" is a good paper on this; obviously has an agenda, but rings true in my experience.

http://sbel.wisc.edu/Courses/ME964/Literature/LeeDebunkGPU20...

alkonaut3• 10 years ago

That's a good reference - however the drawback of comparing a GTX280 vs an i7 from the same period, is that now 8 years later the GPUs have scaled quite well for the same set of (simple) problems, with more units/bandwidth whereas CPU performance hasn't. The difference between todays biggest graphics cards and the GTX 280 is larger than the difference between the i7 from 2008 and the big desktop CPU from 2016. The "100x" is still a myth, but we are significantly closer today than 2008.

Athas2• 10 years ago

One significant benefit of CPU SIMD is that you also do not need to manage temperamental GPUs and GPU drivers. The CPU programming experience is much nicer, the infrastructure more robust, and you can generally expect some sort of SIMD support everywhere. It is not too hard to support SIMD-with-fallback code (and SIMD will generally just work, if supported). GPU support will often require configuration on the part of the user, especially if the system has several GPUs.

(I say this as a GPU language developer. They are fast, but also a bit of a pain in the ass.)

CyberDildonics7• 10 years ago

Do you think using Vulkan for some general compute tasks will help this situation at all?

Ono-Sendai4• 10 years ago

I would estimate GPUs would be around 10x faster for neural nets.

CyberDildonics7• 10 years ago

That would depend heavily on the hardware and how the program was written. It is an apples and oranges comparison.

I will say this though, a program written well in ISPC with cache locality taken into account together with SIMD can run 100x faster than a naive C program.

scott_s3• 10 years ago

They link to a conference paper for the full technical details: "ispc: A SPMD Compiler for High-Performance CPU Programming", https://cloud.github.com/downloads/ispc/ispc/ispc_inpar_2012...

wyldfire• 10 years ago

GPU-oriented languages like OpenCL support SIMD but lack capabilities needed to achieve maximum efficiency on CPUs and suffer from GPU-driven constraints that impair ease of use on CPUs.

IMO these constraints are overstated and OpenCL offers a good abstraction for CPUs too.

Ono-Sendai4• 10 years ago

Maybe in theory but in practice we see poor performance using OpenCL to target CPU devices (much worse than our reference C++ code)

fsaintjacques2• 10 years ago

What benefits do we gain versus a standard compiler with intrinsic approach?

scott_s3• 10 years ago

Potentially significant, since they add new types and control structures. The new types inform the optimizer if data is local to SIMD computations or not, can optimize memory layout for SIMD computations, and the control structures allow for converged control-flow at runtime where only one branch is taken by all threads.

See the full paper I link to as a top-level comment for more details.

CyberDildonics7• 10 years ago

With intrinsics you have to write each instruction by hand. Not only is this a lot of work and compiler specific but if you aren't familiar with all of the instructions at your disposal it is unlikely you will get the same performance. Not only that but ISPC can be compiled to multiple different SIMD lane widths so that something doesn't need to be re-written with the width increases or decreases.

One example would be the n-body simulation of the computer language benchmarks game. The C++ version uses intrinsics but wouldn't benefit from anything that can do 4 doubles instead of only two at a time.

fsaintjacques2• 10 years ago

I would argue that learning the 'basics' of intrinsic is less of a burden than learning a new C extension and modifying your existing code base to include additional new build tools.

OTOH, I like the idea of automatic lane width detection.

a_e_k• 10 years ago

There's also the benefit of not having to worry about managing the scalar prologues and epilogues.

eggy3• 10 years ago

I wouldn't know how to utilize this yet, but if I write a C program that utilizes SIMD similar to the way you can in the Dart programmng language with its Float32x4 and Int32x4 types [1], that same program can be expanded to use the 4 cores with SPMD to do 4 (SIMD) vectors x 4 (SPDM) parallel tasks on my quad-core 2013 i7-4700MQ by using the Intel compiler and specially-written C code for a maximum 1600% speedup of 4x4 matrix operations? I am guessing it would be more like 800% to 1200% in reality if lucky, but still promising.

  [1] https://www.dartlang.org/articles/simd/

davidf18• 10 years ago

Not certain why the Mac binary is for Mavericks and not El Capitan.

billconan2• 10 years ago

I just tried using this to accelerate my nerual network. I got about 4.5x speedup. not bad

billconan2• 10 years ago

I thought this was a new compiler just been released, it turned out it's been around for a while. How come I have never heard of it!

eggy3• 10 years ago

I caught that too, and I was surprised, but like all closed research, it was there. I am guessing it will help to reinvigorate some uses for older Intel chips, and certainly newer ones. All good for Intel, and those who need the speedup.

eggy3• 10 years ago

How many cores and SIMD lanes? What type of ANN did you use if for? Was it hard to implement into your existing code? This is a big deal for me, since I gave up C for Erlang/LFE, and I am working my way through Gene Sher's book 'Handbook of Neuroevolution Through Erlang'. I like the conceptual fit, but I would be more interested in the present to try this on some old C coded NNs.

Intel SPMD: Compiler for High-Performance SIMD Programming

Related Stories