Remnant44

9 ms·

61.
by Remnant44 2y ago
Hi Raph, first of all thank you for all of your contributions and writings - I've learned a ton from reading your blog! A minor quibble amidst your good comparison above ;) For a zen5 core, we have 16-wide SIMD with 4 pipes; 2 are FMA
62.
Zen5's AVX512 Uplift(phoronix.com)
6 points by Remnant44 2y ago|0 comments
63.
by Remnant44 2y ago
I agree. I work in a similar field, and the value of AVX512 is clearly there - it just hasn't been worth implementing for the tiny percentage of market penetration. This is directly due to the market segmentation strategy Intel applied
64.
by Remnant44 2y ago
Dynamic dispatch adds headaches to the build process; they are surmountable for sure, but in my experience the build wrangling to make it all happen is harder than the original work of rewriting your code with intrinsics! The other major pr
65.
by Remnant44 2y ago
Setting aside the complexity of "detachable" ALU clusters, the bigger problem is that essentially, we can currently drive all those ALUs _because_ they're SIMD and executing the same instruction. For example, if you could un
66.
by Remnant44 2y ago
The short answer is "Yes, sometimes". Clever hand-written SIMD code is still consistently better, sometimes dramatically better. But generally speaking, I've found Clang to be pretty good at auto-vectorizing code when I'
67.
by Remnant44 2y ago
Agreed. My current best-case-scenario hope is that the success of the Zen4/5/etc processors will force Intel to adapt their strategy towards AMDs, and finally move us out of the avx512 mess they've segmented us into.
68.
by Remnant44 2y ago
AMD's avx512 implementation is just lovely and they seem to be firing on all cylinders for it. Zen4 was already great, 'double pumped' or no. It looks like Zen5's support is essentially the dream - all EUs and load/
69.
by Remnant44 2y ago
Yes, the figure was 250GFLOP for 4 cores instead of per core, I misread. Still impressive but more reasonable
70.
by Remnant44 2y ago
I honestly didn't realize how performant the decades-old 2013 Haswell architecture is on vector workloads. 250GFLOP/core is no joke - He also cross-compared to an M1 Pro, that when not using the secret matrix coprocessor achieves
71.
by Remnant44 2y ago
Which is also why it's so attractive. :) Those large shuffles are really powerful for things like lookup tables. Large tables are suddenly way more feasible in-register, letting you replace a costly gather with an in-register permute.
72.
by Remnant44 2y ago
I agree. It's not even that we over-weight being technically correct. Correctness and kindness are not mutually exclusive at all. I suspect you're right and it boils down to the fact that a lot of us tech types have substantial pr
73.
by Remnant44 2y ago
I agree. I still hesitate to ship code that requires even AVX1, even though it was first introduced in 2011(!). AVX512 really improves the instruction set. Not just from masking but from some really big holes filled in terms of instructions
74.
by Remnant44 2y ago
Neat and performant code like the article makes me very curious how the competition will shake out between AMD's AVX512 implementation and Intel's upcoming AVX10. The entire point of AVX10 seems to be to resolve Intel's P vs
75.
by Remnant44 2y ago
I've run into the same behavior with clang and intrinsics. Well, I appreciate the fact that they're trying to optimize the intrinsics usage, there really does need to be a flag or pragma you can pass that says more along the lines
76.
by Remnant44 2y ago
Because most code is not running anywhere near saturation of the available resources, and the problem is only getting worse as cores get wider. I mean, look at the Zen5 block diagram - there are 6 ALUs and 4 AGUs on the integer side alone!
77.
by Remnant44 2y ago
For a fully compute-bound workload, you're certainly correct. That's rare though. All it takes is a couple stalls waiting on memory where a second thread is able to make progress to make that "ideal speedup" certainly be
78.
by Remnant44 2y ago
Those are totally reasonable caveats to deal with currently. Thanks!
79.
by Remnant44 2y ago
This looks great - I've been looking for a sustainable, cross-platform-and-vendor GPU compute solution, and the alternatives are not really great. CUDA is nvidia only, Metal is apple only, etc etc. OpenCL has been the closest match but
80.
by Remnant44 2y ago
This is definitely a problem; if the control circuitry is up for it you could PWM the pixel color, basically dithering in time instead of space to achieve white or arbitrary non-spectral colors.
81.
by Remnant44 2y ago
Very nice! I was struggling to come up with a realistic real world scenario for value speculation, but this is a perfect example. Any kind of multi-thread/contention related code seems like an ideal target for making a guess-based fast
82.
by Remnant44 2y ago
The linear node layout is not the point at all. It's serving two purposes here: 1) Providing us with a correct "guess" of what the next node is. 2) Ensuring that in all cases we're running from the L1 cache. In real worl
83.
by Remnant44 2y ago
It's rare to need to work at this level of optimization, but this is a really neat trick! Modern cores are quite wide - capable of running 6-8 instructions at once, as long as there are no dependencies. Something as simple and common a
84.
by Remnant44 2y ago
A huge percentage of America. In most more rural regions, this is the only kind of coffee shop you will find.
85.
by Remnant44 2y ago
agreed! The basic categories remain, just blurring a bit at the edges.
86.
by Remnant44 2y ago
I think the principle things that have changed since this article was written is mostly each category taking inspiration from the other. For example, SIMD instructions gained gather/scatter and even masking of instructions for divergen
87.
by Remnant44 2y ago
Rather the point is that depression is... more of a shared symptom really. We're mostly grasping at straws with our treatments, but we do know that some cases (and we assume, causes..) are amenable to lifestyle factors. Others are assu
88.
by Remnant44 2y ago
I've done a fair amount of floating point simd programming for what the article calls type 1 - iterate large amounts of data as quickly as possible on the CPU. I've been increasingly finding lately that integer-simd programming i
89.
by Remnant44 2y ago
I also disagree strongly with that. I live in Seattle (a city well known for being not social and welcoming - aka the seattle freeze), and didn't recover from being a shy introvert until around 30. The skills and mindset for human conn
90.
by Remnant44 2y ago
That's a pretty strong claim to make completely unsupported by evidence. In fact, I think you'll find more people than ever before are settling down in their 30s and starting families, etc.

More ›