babashka-sci-dev

borkdude 2023-09-28T09:50:58.184939Z

This is interesting, on GHA with babashka dev (master) I get:

Run bash <(curl ) --dev-build
  % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
                                 Dload  Upload   Total   Spent    Left  Speed

  0     0    0     0    0     0      0      0 --:--:-- --:--:-- --:--:--     0
100  5010  100  5010    0     0   150k      0 --:--:-- --:--:-- --:--:--  152k
Downloading  to /var/folders/24/8k48jl6d249_n_qfxwsl6xvm0000gn/T/tmp.fJv954EH
Moving /usr/local/bin/bb to /usr/local/bin/bb.old
Successfully installed bb in /usr/local/bin
The current machine does not support all of the following CPU features that are required by the image: [CX8, CMOV, FXSR, MMX, SSE, SSE2, SSE3, SSSE3, SSE4_1, SSE4_2, POPCNT, LZCNT, AVX, AVX2, BMI1, BMI2, FMA].

borkdude 2023-09-28T09:52:41.637879Z

I'll try and see if the -march=compatibility argument will work

borkdude 2023-09-28T10:10:13.034009Z

That fixed it. Geeze, it looks like this could be breaking a lot of images

borkdude 2023-09-28T10:10:20.039279Z

cc @ericdallo

ericdallo 2023-09-28T11:07:07.682179Z

Interesting, I'll try the pgo on the weekend, will probably need to do that as well

lispyclouds 2023-09-28T22:11:19.468119Z

interesting observation: the locally PGO compiled binary without the -march=comaptibilty seems to be ~100ms faster than the CI compiled binary. Maybe the special instructions at play here?

lispyclouds 2023-09-28T22:13:46.109529Z

CI:
bb -e '(time (loop [val 0 cnt 10000000] (if (pos? cnt) (recur (inc val) (dec cnt)) val)))'
"Elapsed time: 334.463748 msecs"
10000000

local:
bb -e '(time (loop [val 0 cnt 10000000] (if (pos? cnt) (recur (inc val) (dec cnt)) val)))'
"Elapsed time: 259.723079 msecs"
10000000

borkdude 2023-10-11T13:09:33.378379Z

I ran into issues with PGO on aarch64 when bumping http-kit. I guess the iprof file should be updated when bumping deps. Weirdly enough I didn't get trouble on other systems

borkdude 2023-10-11T13:45:25.944099Z

something is fishy, none of the builds work anymore with my locally produced iprof file

lispyclouds 2023-10-11T17:50:55.041779Z

whats the kind of issues youre seeing?

lispyclouds 2023-10-11T17:51:09.709889Z

the build itself is failing?

borkdude 2023-10-11T19:56:01.359089Z

yes, see the branch bump-http-kit

borkdude 2023-09-29T07:12:29.364599Z

Yeah I suspected. Pity

lispyclouds 2023-09-29T07:13:36.642599Z

well now the my linux/amd64 is same perf as your mac M1 😛

lispyclouds 2023-09-29T07:14:12.278099Z

maybe your mac has gotten worse too 😉

borkdude 2023-09-29T07:14:57.863889Z

Yes my Mac has gotten worse too

lispyclouds 2023-10-12T09:51:51.373919Z

i see that this is merged now, what was the fix?

borkdude 2023-10-12T09:52:07.492459Z

removing PGO

borkdude 2023-10-12T09:52:16.858429Z

as you can see

lispyclouds 2023-10-12T09:52:48.923339Z

ah yeah, next ones wont be PGO anymore?

borkdude 2023-10-12T09:53:32.503299Z

apparently the shared pgo file doesn't work after all, so no PGO anymore

lispyclouds 2023-10-12T09:53:57.278069Z

yeah i guess we'd have run into this sooner or later