- Posts: 271
- Thank you received: 84
What once was no longer works.
- NWE
- Offline
- Elite Member
-
Less
More
1 month 5 days ago #349923
by NWE
Replied by NWE on topic What once was no longer works.
Oh, please don't delete it yet, this looks like it has a lot of potential.Or just delete it I suppose. Doesn't work anyway.
Please Log in or Create an account to join the conversation.
- Darrell
- Offline
- New Member
-
Less
More
- Posts: 6
- Thank you received: 5
1 month 2 days ago #350010
by Darrell
Replied by Darrell on topic What once was no longer works.
My initial plan was to create an "All In One" setup/GUI interface. I did a similar wire diagram as yo did. I was toying around with Unreal Engine years ago and got the graphical idea from that. under the hood it's goal was to simulate actual physical machine wiring.
I got bogged down in just trying to get the UI buttons to do what they should. lot's of safety concern logic, graphical rendering issues..
Along the way I learned that there is an "API." So here I am rewriting it.
I can kind of read C++ and do some coding in other languages as well, but AI is just so much faster than me at writing the actual code. Basically with a couple clicks a users with mesa cards should have a working tunable setup. That was the hope anyway. I've been working on this for the better part of 6 months or so. multiple Debian reinstallations. I am now trying to learn linux terminal (gross).
I had an old I5 pc that I am repurposing. I have an old GTX1070 I was trying to get to work with the RT kernel (fail), so the gcode preview wouldn't take so long to load with 5MB+ file sizes.. I get that users can run linuxcnc on PI and other micro controllers, but I thought I could just repurpose this old pc to do the job, and have some cool bigger "out of left field" features. haha
I got bogged down in just trying to get the UI buttons to do what they should. lot's of safety concern logic, graphical rendering issues..
Along the way I learned that there is an "API." So here I am rewriting it.
I can kind of read C++ and do some coding in other languages as well, but AI is just so much faster than me at writing the actual code. Basically with a couple clicks a users with mesa cards should have a working tunable setup. That was the hope anyway. I've been working on this for the better part of 6 months or so. multiple Debian reinstallations. I am now trying to learn linux terminal (gross).
I had an old I5 pc that I am repurposing. I have an old GTX1070 I was trying to get to work with the RT kernel (fail), so the gcode preview wouldn't take so long to load with 5MB+ file sizes.. I get that users can run linuxcnc on PI and other micro controllers, but I thought I could just repurpose this old pc to do the job, and have some cool bigger "out of left field" features. haha
The following user(s) said Thank You: NWE
Please Log in or Create an account to join the conversation.
- NWE
- Offline
- Elite Member
-
Less
More
- Posts: 271
- Thank you received: 84
4 weeks 1 day ago #350075
by NWE
I'm pretty sure I have a couple GTX1070 GPUs around here, I don't recall having much trouble with them, but have not recently run LinuxCNC with RT kernel on one. I might have live-booted it at some point but don't remember how it went. I will certainly try it out soon and if I make any progress on that I will share here.
You mention my project, I assume you are referring to halflow? That one got pushed back lately due to other high priority projects eating all my time. When I get back to it, I want to do a major rewrite of it. My new idea is to spend a lot more time planning the structure before the coding starts. AI/llm all the way. I can read and write bash, python and C++ but AI is just faster.
On several different recent projects I spent literally days planning and refining the whole thing before the coding started. I produce a big detailed text file with help from chatgpt, with the build split into stages and finally give it to claude and basically say go. Of course, it ends up needing lots of follow-up and direction. It produced a more coherent build. Back when I started with halflow I had only a vague idea what the end result was supposed to look like, so I simply plunged into building and trying stuff, eventually it became something of a spaghetti program for lack of a solid plan structure. To date, I've refused to buy anything more than Claude/Chatgpt's entry level $20usd/month plans. When they block me saying my quota is used up I say time for a break.
Replied by NWE on topic What once was no longer works.
Nice. I also enjoy using capable 'almost out-of-date by today's standards' PCs. There's lots of them that run great on linux.
I had an old I5 pc that I am repurposing. I have an old GTX1070 I was trying to get to work with the RT kernel (fail), so the gcode preview wouldn't take so long to load with 5MB+ file sizes.. I get that users can run linuxcnc on PI and other micro controllers, but I thought I could just repurpose this old pc to do the job, and have some cool bigger "out of left field" features. haha
I'm pretty sure I have a couple GTX1070 GPUs around here, I don't recall having much trouble with them, but have not recently run LinuxCNC with RT kernel on one. I might have live-booted it at some point but don't remember how it went. I will certainly try it out soon and if I make any progress on that I will share here.
You mention my project, I assume you are referring to halflow? That one got pushed back lately due to other high priority projects eating all my time. When I get back to it, I want to do a major rewrite of it. My new idea is to spend a lot more time planning the structure before the coding starts. AI/llm all the way. I can read and write bash, python and C++ but AI is just faster.
On several different recent projects I spent literally days planning and refining the whole thing before the coding started. I produce a big detailed text file with help from chatgpt, with the build split into stages and finally give it to claude and basically say go. Of course, it ends up needing lots of follow-up and direction. It produced a more coherent build. Back when I started with halflow I had only a vague idea what the end result was supposed to look like, so I simply plunged into building and trying stuff, eventually it became something of a spaghetti program for lack of a solid plan structure. To date, I've refused to buy anything more than Claude/Chatgpt's entry level $20usd/month plans. When they block me saying my quota is used up I say time for a break.
Please Log in or Create an account to join the conversation.
Time to create page: 0.141 seconds