Category: Uncategorized

more with the raspberry pi 5

whbeebe

I’ve had some 48 hours to look at the Raspberry Pi 5, and I’ve developed some respect for the hardware. This is in contrast to my “meh” attitude back in September when I commented on its initial announcement. I’ve sorta kicked the tires a bit and taken out for a test drive, and I must admit to being satisfied, especially with the right distribution installed. Let me get this out of the way first: I tried to install Fedora 39 for AArch64 and it failed to install. No problem, as both the latest Raspberry Pi OS and Ubuntu 23.10 for the Raspberry Pi 5 (hereafter referred to as RPi 5) installed just fine, thank-you-very-much.

Raspberry Pi 5 in a Vilros Plexiglass sandwich case. Sandisk SDXC adapter shown for scale.

The latest RPi 5 is the same physical dimensions as all the other Pi SBCs that have come before it. The only major physical change is swapping the USB ports with the physical Ethernet port; they are now re-organized as they were on all Pi SBCs before the RPi 4 was released. Everything else remains the same as on the RPi 4, such as USB-C for power and the micro HDMI ports.

My RPi 5 is in a Vilros PI 5 Active Cooler Compatible Case because that was how I had to purchase the RPi 5 active cooler. Normally the active cooler is US$5, but they were all sold out, and this combination was purchased from Amazon for around US$12. The Vilros case was a reasonable compromise to protect the board and keep it from accidentally shorting out if the bottom of the board came into contact with any conductive surface, as well as providing physical protection for the active cooler with its fan. With the top in place there’s still enough room to plug in a cable to access the GPIO pins. As far as a hat is concerned, it won’t work with this style. Once the market for RPi 5 cases settles down a bit I will probably look for something else, but until then it’s all I have, so it’s good that it’s actually useful in a control office/home setting.

Raspberry Pi 5 power button next to power/status LED

One rather interesting feature is the pushbutton power switch on the same edge as the micro SDXC card socket. I’ve tried it out and it does work; press it when the board is on and running an operating system, and the operating system will immediately pop up a dialog saying it is shutting down in 60 seconds, and begins to count down. After shutdown the green power LED goes red, indicating that power is still supplied to the board. Press it again, and the board starts right back up.

Speaking of starting up, with Ubuntu 23.10 AArch64 installed it powers on and to a desktop in under five seconds. I’ve timed it. This fits in with the overall smooth and snappy performance of Ubuntu on this hardware. Right now I couldn’t be more pleased, but give it time, and I’m sure to find something to complain about.

If I could compare the overall performance at this time with other computers I own, I’d have to say it performs at the same level as my old Samsung R580 from 2010 and my current work Linux box, a Minis Forum U250. Both of these systems have 16GB of memory, and both are running an i5 or i5 equivalent from AMD. They’re multi-core, and for the Minis Forum, hyper-threaded as well. And yet if you sat me down in front of all three as a blind test with the same display and keyboard I don’t think I could tell them apart performance wise, without digging in with some utilities.

It will be interesting to see how this newest Raspberry Pi matures over time, and the type of Raspberry Pi comes in the future.

Links

the raspberry pi 5 — comment and opinion

raspberry pi (ancient Pi in plastic case February 2014) — /2014/02/26/new-bits-for-the-raspberry-pi/

raspberry pi (my first Pi from January 2014) — /2014/01/28/raspberry-pi/

little ai can’t write python code

whbeebe1 Comment

Two posts back I wrote about installing GPT4All on my M1 MacBook Pro. At first blush I was impressed, as it seemed to answer the softball questions I asked it correctly and with style. Then I got a little more serious and asked it this question:

write a python program that reads a directoryand prints out everything it finds

To which GPT4All Falcon gave this answer:

Here's an example Python program that uses the`os` module to read a directory and print outeverything it finds:
import os# Set the directory pathdir_path = '/path/to/directory'# Check if the directory existsif not os.path.exists(dir_path):print("Directory does not exist")return# Iterate through all files in the directoryfor file in os.listdir(dir_path):# Check if the file is a directoryif os.path.isdir(os.path.join(dir_path, file)):# If it's a directory, recursively call this functionprint("Directory:", file)print("Files:", os.listdir(os.path.join(dir_path, file)))# Check if the directory is emptyif not os.path.exists(dir_path):print("Directory does not exist")
This program uses the `os.listdir` functionto get a list of all files in the directoryand the `os.path.join` function to constructthe full path for each file.It then calls itself recursively on anydirectories it finds, printing out thedirectory name and its contents.

Let’s examine the Python code. There is a blatant error on line 9. Python does not allow the use of a return outside of a proceedure. If you need to exit a script due to some condition, then you use Python’s exit keyword.

The second problem are lines 19-21. They don’t do anything. The message on line 21 is the same as the message on line 8. If the block of code in lines 19-21 is to determine if the directory is empty then it should say that.

Let’s clean up the code enough to actually work.

#!/usr/bin/env python3import os# Set the directory pathdir_path = '.'# Check if the directory existsif not os.path.exists(dir_path):print("Directory does not exist")exit(1)# Iterate through all files in the directoryfor file in os.listdir(dir_path):# Check if the file is a directoryif os.path.isdir(os.path.join(dir_path, file)):# If it's a directory, recursively call this functionprint("Directory:", file)print("Files:", os.listdir(os.path.join(dir_path, file)))

Now let’s test it. I’ve created some dummy directories and files. Here’s what the utility tree shows.

├── foo│   ├── bar│   │   ├── five│   │   ├── seven│   │   └── six│   ├── four│   ├── one│   ├── three│   └── two├── foo2

Now let’s run our corrected AI code and look at the output.

Directory: fooFiles: ['three', 'one', 'two', 'bar', 'four']Directory: foo2Files: []

It’s unsurprisingly disappointing. There’s a comment on line 15 (16 in the corrected code) that says to “recursively call this function” if a directory is detected. It doesn’t in part because there are no functions, and because the test for a directory doesn’t do anything but print out if it’s a directory.

The code looks like during its learning phase it swept up someone’s Stack Overflow question, but didn’t know enough to look at any answers, assuming there were any. Whatever ideas I might have had to incorporate the Falcon model into anything have been pretty much squelched. I’ll install some others and give them a spin.

Links

ChatGPT Isn’t Coming for Your Coding Job — Wired — https://www.wired.com/story/chatgpt-coding-software-crisis/