arlen has quit [Remote host closed the connection]
frost has quit [Quit: Connection closed]
linkliu61 has joined #riscv
hendursa1 has quit [Quit: hendursa1]
hendursaga has joined #riscv
jacklsw has joined #riscv
<jrtc27>
jemarch: the sudden sound of your voice on the youtube livestream startled me :')
arlen has joined #riscv
<jemarch>
heh sorry :D
aburgess has quit [Ping timeout: 260 seconds]
jellydonut has quit [Ping timeout: 265 seconds]
gordonDrogon has quit [Ping timeout: 252 seconds]
gordonDrogon has joined #riscv
<palmer1>
jimw, kito: do we need slides for this BOF?
[itchyjunk] has quit [Remote host closed the connection]
kaji has quit [Read error: Connection reset by peer]
pho has quit [Read error: Connection reset by peer]
EmanuelLoos[m] has quit [Write error: Connection reset by peer]
kaddkaka[m] has quit [Write error: Connection reset by peer]
JosefHolzmayr[m] has quit [Remote host closed the connection]
CarlosEDP has quit [Write error: Connection reset by peer]
AhmedCharles[m] has quit [Read error: Connection reset by peer]
psydroid has quit [Write error: Connection reset by peer]
pierce has quit [Remote host closed the connection]
khem has quit [Read error: Connection reset by peer]
psydroid has joined #riscv
<jimwilson_>
I've never seen Kito here, but I suspect he has slides. I don't.
<palmer1>
he just sent email
<palmer1>
we're good ;)
khem has joined #riscv
kaji has joined #riscv
CarlosEDP has joined #riscv
pho has joined #riscv
pierce has joined #riscv
EmanuelLoos[m] has joined #riscv
AhmedCharles[m] has joined #riscv
kaddkaka[m] has joined #riscv
JosefHolzmayrThe has joined #riscv
hendursaga has quit [Ping timeout: 276 seconds]
<jrtc27>
if anyone wants to relay feedback to the corev talk: 1. relaxation uses paired relocations in a special way, that's not existing infrastructure you can reuse 2. this only works for .o files, it doesn't work for dynamic relocations since we need to support both rela and rel that has nowhere to put that addend
<jrtc27>
custom dynamic relocations should be much rarer (though *we* have them) but is something to bear in mind...
<jrtc27>
and having paired relocations for loaders is much more painful than static linkers where things are already really complex
<jrtc27>
(most loaders, though perhaps not glibc, are deliberately very dumb)
hendursaga has joined #riscv
<jrtc27>
(using rel rather than rela for .o files is a no go for many reasons, but fuschia wants to use rel for linked objects)
<jrtc27>
doesn't mean I know of a better solution though... 32-bit ELF _sucks_
<jrtc27>
and it doesn't help that there are some RV64-only relocations allocated in the 0-255 range that waste encoding space
<jrtc27>
(and you can't use ELFCLASS64 for RV32 since, at least currently, the class is tied to XLEN)
bgamari_ is now known as bgamari
vagrantc has joined #riscv
<jrtc27>
only vague alternative (but it's ugly) that comes to mind is having an R_RISCV_PAIR that indicates the next two relocations constitute a single relocation
<jrtc27>
then you avoid the addend issue as you just get 16 bits of relocation type which is still tight but probably good enough
<jrtc27>
(and perhaps you can reserve a way to do length extension in future if that turns out to be false)
<jrtc27>
that lets us handle both rel and rela, and also lets us make it look less vendor-y and more just like how you'd do it on RV64 with its big relocation space
<jrtc27>
hmm, if I'd known this BoF was happening I would have registered for the conference...
* jrtc27
did not realise it was about things that affect other toolchains, rather than just being about the state of RISC-V GNU tools
<jrtc27>
thanks jimwilson_
<jrtc27>
(I assume I am persona non grata as a non-registered person for the matrix chat...)
cmuellner has joined #riscv
<jimwilson_>
there has been a RISC-V BOF at every GNU Cauldron for about 3 years now, though this one isn't very GNU specific
<jrtc27>
yeah, the psABI implications is the bit I'm concerned about
<jrtc27>
whether GCC supports Zks or whatever isn't so much of interest to me :D