IRC channel logs
2020-12-07.log
back to list of logs
<siraben>OriansJ: do you mind if I make a PR to blynn-compiler adjusting the readme to better reflect the purpose of the fork? <OriansJ>siraben: I never mind Pull Requests; the worst that happens is I reject them if they seem wrong and try to explain my reasoning for why. <siraben>OriansJ: I just realized that I was copying the wrong file to /share, it should be ./bin/raw not ./bin/vm <siraben>so that explains why the hashes seemed mismatched yesterday <siraben>9732a8852bf92b4097f275da4ceba3b718138a5e16190cbef43bbd4be42a27dd ./result/share/raw <siraben>I'll also add a test in the CI workflow to check that the hash is that <yt>OriansJ: I'm wondering if I could reduce the review load if I use fewer M1 macros. hex2 is bad if you want to review all macros by hand, but cc_aarch64 is going to be a *lot* worse still <yt>For example, I could replace all SET_X0_FROM_X1 with FROM_X0 SET_X1 (push+pop, or move through a temporary register) <yt>It'll result in worse machine code, which might be harder to debug, but that wouldn't matter too much as I can do those changes incrementally and mechanically <deesix>I guess those are already reviewed, if they are the same as in M2-Planet. Am I missing something? <yt>deesix: I think M2-Planet's set of instructions is even smaller than that, it only really uses X0, X1 and a stack machine <yt>I went more for ease of translation from GNU assembly to M1 <yt>but the number of macros got out of hand a bit :) <deesix>OK, we'll see what OriansJ thinks. Thanks for the details. <yt>deesix: sounds good, thanks for taking a look <OriansJ>siraben: I'll look at that Pull reguest later today; So you'll have time to do a rebase if you wish to hide that mistake. <siraben>Isn't it generally a bad idea to rewrite history on master? <siraben>But there may not be many clones around so should be fine <OriansJ>siraben: only if you have downstreams <OriansJ>otherwise you can rebase to your hearts content until it is merged upstream. <OriansJ>yt: opt for easier to debug and understand M1 code; The biggest time delay with reviews is dealing with my son. <OriansJ>I'd rather deal with 1000 DEFINEs than a block of hard to understand assembly. <siraben>after finals i'll take a look at using mescc-tools-seed for bootstrapping and see if I can replicate building blynn-compiler without gcc as a dependency <siraben>OriansJ: It appears that I can remove gcc as a build-time dependency to blynn-compiler. I'll edit the PR