rd | Reactive Distributed communication framework for .NET | Microservice library

I have implemented a Convolutional Neural Network in C and have been studying what parts of it have the longest latency.

Based on my research, the massive amounts of matricial multiplication required by CNNs makes running them on CPUs and even GPUs very inefficient. However, when I actually profiled my code (on an unoptimized build) I found out that something other than the multiplication itself was the bottleneck of the implementation.

After turning on optimization (-O3 -march=native -ffast-math, gcc cross compiler), the Gprof result was the following:

Clearly, the convolution2D function takes the largest amount of time to run, followed by the batch normalization and depthwise convolution functions.

The convolution function in question looks like this:

I have the following instruction: mov r1, r7 in my assembly code but after looking into disassembly, I've found that actual generated code was adds r1, r7, #0

I checked with ARMv6-M Architecture Reference Manual and I found out that there's MOVS , instruction (A6.7.40) which is different from ADDS.

While that's not a big issue, I'm still puzzled why assembler replaces code that I wrote by different op-codes. According to the book that I'm reading, all non-jump instructions take 1 cycle (and I'd prefer for assembler to be dumb rather than trying to optimize something for me).

I'm using Raspberry Pi Pico SDK which uses GNU Assembler, AFAIK.

All my code is written in helloworld.S, full source code is:

So I'm preparing for a technical interview, and one of my practice questions is the Kth smallest number. I know that I can do a sort for O(n * log(n)) time and use a heap for O(n * log(k)). However I also know I can partition it (similar to quicksort) for an average case of O(n).

The actual calculated average time complexity should be:

I've double checked this math using WolframAlpha, and it agrees.

So I've coded my solution, and then I calculated the actual average time complexity on random data sets. For small values of n, it's pretty close. For example n=5 might give me an actual of around 6.2 when I expect around 5.7. This slightly more error is consistent.

This only gets worse as I increase the value of n. For example, for n=5000, I get around 15,000 for my actual average time complexity, when it should be slightly less than 10,000.

So basically, my question is where are these extra iterations coming from? Is my code wrong, or is it my math? My code is below:

Can LaTeX escapes for accents be used in Rd files? I tried the standard \'e and many variants (\'{e}, {\'e}, \\'e, \\'{e}, {\\'e}, etc.), but none is rendered as an accented character in the PDF or HTML output.

I want my References section (i.e. \references{}) to be rendered with accented characters, but I do not want to type non-ASCII characters in my Rd files. Is there good/recommended practice? Should I simply replace non-ASCII characters with their ASCII equivalents (é → e, ø → o)?

To be clear, I know it is possible to type accented characters (e.g., é) directly in UTF-8-encoded files, but I would prefer to keep ASCII-encoded files.

This question is not about

• how to type special/accented letters in LaTeX
• how to use UTF-8 in LaTeX
• accents/special characters in R Markdown

or variants.

Package structure

My sample data start with values beginning with 700 so that there is nothing between 0 and 700. I want to cut that range out of the line plot but I also want to point the reader to that cut by visualizing it like this. This example picture is manipulated via a drawing software just to explain what I want.

Something like this is explained in the matplotlib documentation. But I do not understand what happens there in the example code. And I am not sure if it fits to my case because the example cutting data. And I just want to cut the area where is no data present.

Here is a minimal working example producing line figure where the y-axis starts at 700 not 0. Can we start from here?

This question pertains to the ARM assembly language.

My question is whether it is possible to use a macro to replace the immediate value in the ASM code to shift a register value so that I don't have to hard-code the number.

I'm not sure whether the above question makes sense, so I will provide an example with some asm codes:

So there exist few instructions such as ror instruction in the ARM (https://developer.arm.com/documentation/dui0473/m/arm-and-thumb-instructions/ror), where it is possible to use a register value to rotate the value as we wish:

I have a dataframe, let's say it's:

Before I begin let me say that I read thoroughly all the stack overflow posts and resources in the appendix, and could not find a solution to my problem.

I am trying to create, validate and connect a subdomain through Route53 and AWS Certificate Manager. The subdomain is challenge.sre.mycompany.com.

The terraform plan looks something like this:

I have downloaded the street abbreviations from USPS. Here is the data: