emacs28

5 ms·

1.
by emacs28 2y ago
First author here. The hardware architectures are realistic - we developed & evaluated real example hardware implementations for them, validated on FPGA, and they achieved state-of-the-art ResNet performance in a deep learning accelerat
2.
Karatsuba Matrix Multiplication and Its Efficient Hardware Implementations(arxiv.org)
139 points by emacs28 2y ago|22 comments
3.
Karatsuba Matrix Multiplication and Its Efficient Hardware Implementations(arxiv.org)
5 points by emacs28 2y ago|0 comments
4.
Strassen Multisystolic Array Hardware Architectures(arxiv.org)
1 points by emacs28 2y ago|0 comments
5.
Karatsuba Matrix Multiplication and Its Efficient Hardware Implementations(arxiv.org)
2 points by emacs28 2y ago|0 comments
6.
by emacs28 2y ago
It produces identical/bit-equivalent results as conventional/naive matrix multiplication for integer/fixed-point data types
7.
by emacs28 3y ago
For everyone discussing the reduced accuracy/numerical stability of the algorithms in floating-point, this is true. But note that the application of the algorithms in the work is explored for fixed-point MM/quantized integer NN in
8.
by emacs28 3y ago
> you have to build hardware that matches the dimensions of the algorithm Yes the benefits are realized in custom hardware designs as opposed to software, however, the hardware architectures work for multiplying matrices of arbitrary dim
9.
by emacs28 3y ago
Thanks, good summary. Regarding numerical stability, the application is for fixed-point arithmetic, and therefore numerical stability is not an issue (the result is identical compared to using the conventional inner-product)
10.
by emacs28 3y ago
IMHO, for fixed-point MM accelerators, there is no catch, I think it's an overlooked algorithm. It's based on an algorithm by Winograd who coincidentally also proposed another unrelated algorithm that later became very popular for
11.
Show HN: Matrix Multiplication with Half the Multiplications(github.com)
310 points by emacs28 3y ago|77 comments
12.
Show HN: AI with Half the Multiplications(github.com)
2 points by emacs28 3y ago|0 comments
13.
Matrix multiplication hardware architectures requiring half the multipliers(github.com)
1 points by emacs28 3y ago|0 comments
14.
Show HN: Matrix Multiplication with Half the Multiplications(github.com)
18 points by emacs28 3y ago|0 comments
15.
Systolic arrays(github.com)
1 points by emacs28 3y ago|0 comments
16.
Hardware research(github.com)
1 points by emacs28 3y ago|0 comments
17.
Fast Inner-Product Algorithms and Architectures for DNN Accelerators(ieeexplore.ieee.org)
1 points by emacs28 3y ago|0 comments
18.
Double the performance per MAC unit in ML accelerators(arxiv.org)
1 points by emacs28 3y ago|0 comments
19.
Fast DNN Accelerator Architectures(arxiv.org)
1 points by emacs28 3y ago|0 comments
20.
by emacs28 3y ago
Personally I love my overpriced Samsung z fold, I don't use a laptop anymore (just a desktop), I can easily read double-column research articles wherever I am, it's great for drawing diagrams, and all of that without having to rem
21.
by emacs28 4y ago
CCX stands for Core Complex CCD stands for Core Complex Die (and neither terms refer to the IO die)
22.
by emacs28 4y ago
This can be resolved by changing the cells' format from General to Text. This makes the cells display the text exactly as entered. Select the relevant cells -> right click on them -> Format Cells... -> Text -> Ok
23.
by emacs28 4y ago
One good approach could be to base the architecture on the TPU v1 from [1]. There are also open-source accelerators you could get inspiration from, for example [2][3]. If you want to do less work/not hand code the RTL yourself then you
24.
by emacs28 4y ago
The best productivity hack is to get a keyboard with programmable QMK firmware and remap the keys however/wherever you want.
25.
by emacs28 4y ago
When in Emacs I use a custom command that converts lowercase underscore-separated words to uppercase after typing the word out in lowercase so I don't have to use shift or capslock.
26.
by emacs28 4y ago
Yeah I haven't seen it done elsewhere, but I started doing that because those 2 inside keys on the third row are where my thumbs sit naturally if fully rest my hands.
27.
by emacs28 4y ago
I'm down to a 32-key layout, you can see my layout here https://configure.zsa.io/moonlander/layouts/wyyxP/latest/0 (The layout currently has more than 32 keys due to extra function keys on the sides
28.
by emacs28 4y ago
> Do you have a CI badge to show build/test status? This signal would better reach important parties by emailing maintainers or automatically creating an issue. I find a CI and coverage badge worthwhile because they provide a quick
29.
by emacs28 4y ago
There is a standardized API for package tools to expose and invoke editable-install functionality [1] which pip install follows, but Poetry didn't implement this API before (I see it does now though as of this past February [2]). In ot
30.
by emacs28 4y ago
What I mean is `pip install -e` installs it outside of the virtual environment also.

More ›