00:00 <cmtptr> i was thinking something more software related

00:02 <remexre> have research or security experience?

00:02 <remexre> or nlp or planning

00:05 <mark4> cmtptr, where are you located?

00:08 <cmtptr> georgia

00:08 <cmtptr> and so are you

00:08 <mark4> i work for an agency, when they place me i generally make 50+ an hour and half of that is tax free (im back in NC now tho).

00:09 <cmtptr> oh

00:09 <mark4> if you are willing to travel for contract work i can put you in contact with my agency

00:09 <cmtptr> i don't want to move though

00:09 <mark4> dont have to MOVE

00:09 <cmtptr> or travel

00:09 <mark4> oh. thats not helpful :)

00:09 <cmtptr> i want my dream employment to come to me

00:10 <cmtptr> in fact could you just pay me to do whatever i want to do

00:12 <cmtptr> mark4, do you spend most of your time writing new code, or do you spend a lot of time in existing codebases that suck

00:15 <mark4> mostly the latter lol

00:15 <mark4> but 50+ an hour man

00:30 <cmtptr> i make that doing what i'm doing now

00:32 <mark4> what do you do now?

00:32 <mark4> and when i earn it half of it is not taxed

00:32 <mark4> cant be taxed

00:33 <cmtptr> i'd be willing to take some pay cut, but i don't ever want to have to fix some poorly designed pos written by an inferior developer with more responsibility in the last weeks before release

00:33 <cmtptr> again

00:33 <cmtptr> dodging the taxes is nice

00:34 <cmtptr> right now i work for a telecom equipment vendor

00:35 <cmtptr> we make the things that send you netflix via lasers

00:37 <mark4> lol

00:37 <mark4> how do you watch netflix with your eyesa burned out!

00:40 <cmtptr> very quickly

00:43 <mark4> lol

00:49 <lispmacs[work]> cmtptr: terrestrial links?

01:00 <cmtptr> versus?

01:01 <cmtptr> we don't put fiber in space if that's what you're asking

01:04 <cmtptr> we have products that are deployed to transmit over submarine fibers, i guess

02:06 <dave0> maw

09:43 <KipIngram> create, var, const, :, and ; work now. IT'S ALIVE!!!

09:44 <KipIngram> The drawing board code for create was completely right as written. I made a couple of errors transcribing into my source, so I still got to do some debugging. But it's still nice that it was right to start with.

18:27 <KipIngram> So, I've been thinking today about profiling. It seems like a good use for plentiful memory. My idea is to allocate a memory region that assigns a 32-bit (or 64-bit if necessary) counter for every definition cell in my system. The offset into the table will be equal to or computed from the location of the cell in the system.

18:27 <KipIngram> Then I profile by modifying NEXT to include code to increment the counter associated with each cell it executes.

18:28 <KipIngram> So that will show me where I'm spending my time - I can go to one of those cells, see where it points, and cross correlate the information after collecting it if I want to.

18:28 <KipIngram> It's hard for me to see how there would be any information more detailed than that.

18:29 <KipIngram> Every time NEXT fetches the contents of a cell, that cell's counter gets a bump.

18:29 <Zarutian_HTC> an execute heatmap basically

18:29 <KipIngram> And my NEXT lives in a fixed location anyway - it would be trivial to slap a jump in there to an alternative NEXT and restore the original contents when I was done.

18:29 <KipIngram> Right.

18:29 <KipIngram> Good phrase.

18:30 <KipIngram> I like it because it seems EASY.

18:30 <KipIngram> And no overhead when I'm not profiling.

18:30 <Zarutian_HTC> you can do the same with @ for read heatmap and ! for write heatmap

18:31 <KipIngram> I would be able to know what ranges of cells corresponded to my "application," and I could study only those if I wanted to.

18:31 <KipIngram> Indeed.

18:31 <KipIngram> And BLOCK.

18:32 <KipIngram> If I had a mixture of SSD and spinning disk storage, I could gradually migrate hot storage regions to the SSD.

18:33 <Zarutian_HTC> do not where I saw it but there was an emu/simu-lator that had this kind of feature of colouring the memory view with all three or one of the aforesaid heatmaps

18:33 <KipIngram> The write heat map might reveal cache false sharing problems.

18:34 <KipIngram> Anyway, I will probably try this out.

18:45 <KipIngram> It feels to me like it will be ample raw data, but I'll need to postprocess in a way that's informative.

18:46 <KipIngram> But that should be straightforward - we of course know exactly where our definitions sit in memory.

21:50 <dave0> maw

21:52 <KipIngram> Afternoon, dave0.

21:52 <dave0> hey KipIngram

22:27 <KipIngram> So I had a little bug in the compiler. Remember that bit I added to the headers that triggered compilation of an offset back to the start of the word if it's set? Well, when it goes and finds the location of the start of word, it gets an offset. I needed to convert it to an address by adding base address. I failed to do the @ from the variable that's stored in.

22:27 <KipIngram> Go directly to segfault - do not pass go.

22:27 <KipIngram> I had the @ in the code on the drawing board; I just neglected to copy it into the assembly file.

22:28 <KipIngram> First thing I confirmed when troubleshooting it was that I could manually patch the right jump offset into that cell and then everything would work.