Thursday, 2022-12-22

*** ghostmansd[m] <ghostmansd[m]!> has quit IRC05:53
*** ghostmansd[m] <ghostmansd[m]!~ghostmans@> has joined #libre-soc05:54
*** ghostmansd[m] <ghostmansd[m]!~ghostmans@> has quit IRC07:34
*** ghostmansd[m] <ghostmansd[m]!> has joined #libre-soc07:34
*** Ritish <Ritish!~Ritish@2401:4900:627d:50a2:8af6:53d2:9841:6ccf> has joined #libre-soc08:09
*** Ritish <Ritish!~Ritish@2401:4900:627d:50a2:8af6:53d2:9841:6ccf> has quit IRC08:11
*** Ritish <Ritish!~Ritish@2401:4900:627d:50a2:8af6:53d2:9841:6ccf> has joined #libre-soc08:13
*** Ritish <Ritish!~Ritish@2401:4900:627d:50a2:8af6:53d2:9841:6ccf> has quit IRC08:23
*** ghostmansd[m] <ghostmansd[m]!> has quit IRC08:40
*** ghostmansd[m] <ghostmansd[m]!~ghostmans@> has joined #libre-soc08:41
*** ghostmansd[m] <ghostmansd[m]!~ghostmans@> has quit IRC10:51
*** ghostmansd[m] <ghostmansd[m]!> has joined #libre-soc10:52
*** ghostmansd[m] <ghostmansd[m]!> has quit IRC11:23
*** ghostmansd[m] <ghostmansd[m]!> has joined #libre-soc11:27
*** ghostmansd[m] <ghostmansd[m]!> has quit IRC11:31
*** ghostmansd[m] <ghostmansd[m]!> has joined #libre-soc12:16
*** ghostmansd[m] <ghostmansd[m]!> has quit IRC12:47
*** ghostmansd[m] <ghostmansd[m]!> has joined #libre-soc12:51
*** lxo <lxo!> has quit IRC12:58
openpowerbot[mattermost] <lkcl>
openpowerbot[mattermost] <lkcl> i started RFC ls005 - it's real basic (1 page).14:12
openpowerbot[mattermost] <lkcl> it's intended to be submitted as a... a... "radio-button-2", here
*** lxo <lxo!~lxo@gateway/tor-sasl/lxo> has joined #libre-soc15:12
*** ghostmansd[m] <ghostmansd[m]!> has quit IRC15:43
*** ghostmansd[m] <ghostmansd[m]!~ghostmans@> has joined #libre-soc15:43
*** ghostmansd[m] <ghostmansd[m]!~ghostmans@> has quit IRC17:00
*** ghostmansd[m] <ghostmansd[m]!~ghostmans@> has joined #libre-soc17:03
*** lxo <lxo!~lxo@gateway/tor-sasl/lxo> has quit IRC17:03
*** ghostmansd[m] <ghostmansd[m]!~ghostmans@> has quit IRC17:14
*** ghostmansd[m] <ghostmansd[m]!> has joined #libre-soc17:14
*** bagiyal8989 <bagiyal8989!~bagiyal89@2409:4053:2d0c:1e85:4d3e:bd61:a492:a815> has joined #libre-soc17:41
*** bagiyal <bagiyal!~bagiyal@2409:4053:2d0c:1e85:4d3e:bd61:a492:a815> has joined #libre-soc17:48
*** bagiyal8989 <bagiyal8989!~bagiyal89@2409:4053:2d0c:1e85:4d3e:bd61:a492:a815> has quit IRC17:49
*** bagiyal <bagiyal!~bagiyal@2409:4053:2d0c:1e85:4d3e:bd61:a492:a815> has quit IRC18:01
*** lxo <lxo!~lxo@gateway/tor-sasl/lxo> has joined #libre-soc18:49
programmerjakelkcl: i expanded on the differences of fp16 vs bf16:;a=commitdiff;h=674ea72e5fb3aae98d97bd19c883e183b7d0f67720:01
programmerjakebf8 is e5m2 in -- they call it fp8 along with e4m3. that paper is by nvidia, intel, and arm:
lkclprogrammerjake, i'm attempting to keep the RFC to 1 page22:21
lkclthings like that are really good for going into the "discussion" page22:23
programmerjakeah, ok. i don't have my desktop unpacked so idk how many pages the pdf is, the footnote shouldn't be on page 2, that's just ikiwiki not understanding latex22:23
lkclit's on page 2 because the last line is migrated to page 222:24
programmerjakek, note the order shown by ikiwiki is not the order in the original markdown22:25
lkclcorrect. doesn't matter. the PDF as a submission to the OPF ISA WG is what matters.22:26
lkcli've recorded the links (and also discussion here)22:26
lkclkeeping this one *real* brief is a priority, sparking discussion, and linking to the things that toshywoshy mentioned22:27
lkclwhich he can't talk about :)22:27
lkclbut are definitely relevant22:27
lkclgood links to wikipedia. makes things clear.22:28
lkclmmmm.... don't IBM use "bfloat16"?22:28
lkclwe discussed this before for another RFC22:28
* lkcl ponders...22:28
programmerjakek, just trying to make sure you didn't mistake the incorrect ikiwiki rendering as meaning guaranteed-2-pages22:28
lkclnot a priority.22:29
lkclah yes, ls002.22:29
lkclnggh have to retain consistency sigh22:29
* lkcl will update FP32 to BFP3222:30
programmerjakefeel free to replace bf16 with bfloat16, and fp8/16/32/64 with bfp8/16/32/6422:30
programmerjakeimho having both bf16 and bfp16 is too confusing so should use bfloat16 if possible22:31
programmerjakeactually, leave fp8 as fp8, since it matches the intel, arm, nvidia paper22:32
programmerjakeand since it isn't ieee22:32
* lkcl brainmelting22:32
lkcli'm on semi-vague-holiday-shutdown :)22:33
programmerjakei can do the search/replace22:34
lkclbut figured it'd be sensible to (slowly, no pressure) do this before things restart next year22:34
lkclremember the conversation with toshaan, he recommended getting this done... soon.22:35
programmerjakeso, did you push the last of your changes?22:35
programmerjakei'm assuming so22:36
lkcli'm looking to reduce lines further, but it's a different location22:37
lkclok i'm stopping22:37
programmerjakesorry it's all 1 commit, too annoying to use the web interface for multiple commits when copying to/from my text editor22:50

Generated by 2.17.1 by Marius Gedminas - find it at!