<Stat_headcrabed>
Here is uboot binary compiled by myself
Stat_headcrabed has quit [Quit: Stat_headcrabed]
test925 has quit [Ping timeout: 255 seconds]
ball has left #riscv [#riscv]
<naoki>
mmm, I'm a fool :(
khem has joined #riscv
coldfeet has joined #riscv
Leopold has joined #riscv
jacklsw has quit [Ping timeout: 264 seconds]
BootLayer has quit [Quit: Leaving]
test924 has joined #riscv
mlw has joined #riscv
<conchuod>
Stat_headcrabbed I'll grab my logs tonight. SPL didn't even print any logs
<conchuod>
naoki: also, welcome. Good to have someone from milkv contributing
coldfeet has quit [Remote host closed the connection]
coldfeet has joined #riscv
<courmisch>
now that B is ratified (Zba+Zbb+Zbs, no Zbc), should the kernel set bit 1 in hwcap when applicable'
<courmisch>
?
jobol has joined #riscv
coldfeet has quit [Quit: Lost terminal]
mlw has quit [Ping timeout: 256 seconds]
jacklsw has joined #riscv
ldevulder has joined #riscv
Guest7963 is now known as chripo
<conchuod>
I suppose it should
<conchuod>
And set b in whatever other interfaces we have
alperak has joined #riscv
mlw has joined #riscv
nelnire has quit [Ping timeout: 252 seconds]
<LucasVella[m]>
why did conventions end up being that in static programs jumps use auipc to load the address, but data uses lui?
<LucasVella[m]>
everything else being equal, I would expect auipc to use more power
<sorear>
data benefits from rematerialization (if you access the same global several times or in a loop, the lui can be hoisted), jumps don't in the simple case (you'd need to use a call-saved register, which means adding instructions to save it, which costs more than it benefits; interprocedural register allocation gets around this but isn't standard)
<sorear>
there's also a long-standing tradition on almost all architectures that jumps and calls are always PC-relative
<sorear>
simplifies linkers since R_RISCV_CALL relaxation only needs to handle the AUIPC + JALR case, no separate LUI + JALR case
jacklsw has quit [Ping timeout: 272 seconds]
<LucasVella[m]>
It makes no difference to whether lui or auipc is used to set the base address in the register for the purposes of hoisting.
<sorear>
that's correct on every other architecture with an auipc-like instruction, but riscv's decision to use the _full_ PC instead of masking off the low 12 bits means that you can't have two auipc instructions in different places to generate the same value
<sorear>
which greatly complicates what you can do with them in a compiler
<sorear>
when I was working on the Go port a few years ago I wound up having it never hoist auipc instructions because there wasn't enough information in that pass to determine if the hoisted instruction would eventually need to be duplicated or spilled; gcc at the same time also never hoisted auipc, presumably for the same reason
<sorear>
there are a lot more people working on gcc so they might have found a solution, that one optimization problem costed a percent or so on SPEC if you build it with -mcmodel=medany
<sorear>
(which forces auipc for data access regardless of PIC/PIE/PDE)
jmcgnh has quit [Excess Flood]
Andre_Z has joined #riscv
jmcgnh has joined #riscv
bitoff has quit [Remote host closed the connection]
alperak has quit [Quit: Connection closed for inactivity]
fuwei has quit [Quit: Konversation terminated!]
coldfeet has quit [Remote host closed the connection]
Andre_Z has quit [Quit: Leaving.]
jobol has quit [Quit: Leaving]
mlw has quit [Ping timeout: 240 seconds]
mlw has joined #riscv
<ja_02>
conchuod have you heard any news on if SpacemiT K1 is getting mainlined soon. i dont want to step on someones shoes if they are.
<conchuod>
ja_02: No patches yet. There's a bunch of boards at Rivos and Esmil and smaeul have boards too, but if you want to submit, noone has yet committed to doing anything.
<conchuod>
If you want to work on anything for it, be my guest!
<conchuod>
Or maybe you are one of the Rivos people, given your name is Jesse...
mlw has quit [Ping timeout: 264 seconds]
<ja_02>
yeah i am :)
<ja_02>
im the 20yr old goofball spending her freetime writing patches for the kernel
<ja_02>
sry about any chaos i cause
meowray has quit [Ping timeout: 256 seconds]
<conchuod>
ja_02: Well, fire away with any spacemit k1 stuff. You know where to find me if you run into difficulties.
prabhakalad has quit [Ping timeout: 264 seconds]
prabhakalad has joined #riscv
prabhakalad has quit [Ping timeout: 256 seconds]
prabhakalad has joined #riscv
<lu_zero>
if you do and want early testers poke me as well