g-gundam has quit [Remote host closed the connection]
semz has quit [Quit: ZNC 1.9.0+deb2build3 - https://znc.in]
parjanya has joined #commonlisp
<jasom>
gilberth: amd64 is not as register starved as x86, but neither is it as overfed on registers as Power. In testing I've done (by reserving registers on a Power machine), somewhere around 16 truly general registers (i.e. not including SP, PC, FP, LR) is where diminishing returns of having more registers start to kick in. POWER and MIPS easily have this, even with two stacks, but aarch32 and amd64 do not.
<jasom>
gilberth: There is a strong argument to be made that precise garbage collection is worth the performance hit that a dual-stack amd64 implementation would have, but I predict that there would definitely be a performance hit on real world code.
semz has joined #commonlisp
wohonajax has joined #commonlisp
spdegabrielle has quit [Quit: Connection closed for inactivity]
gko has joined #commonlisp
pabs3 has quit [Read error: Connection reset by peer]
pabs3 has joined #commonlisp
equwal has quit [Ping timeout: 258 seconds]
admich1 has quit [Remote host closed the connection]
admich1 has joined #commonlisp
Oladon has joined #commonlisp
Pixel_Outlaw has joined #commonlisp
elderK has quit [Quit: WeeChat 4.6.3]
<beach>
gilberth: I see, thank you. I may have questions later.
<beach>
gilberth: If the whole heap gets relocated, then every pointer in it must be modified, no?
<beach>
gilberth: I can see how a jump table to subroutines would solve the relocation of code, but I don't see how it will solve the relocation of data. When you take the CAR of a CONS cell, the result had better be a valid pointer into the heap, so if the heap is relocated, that pointer has to be updated, no?
<beach>
gilberth: Unless of course every heap access is indirect through some base pointer.
equwal has joined #commonlisp
_whitelogger has joined #commonlisp
<aeth>
jasom: looking up the number of registers, apparently Intel doubled the number of registers from 16 to 32 in a fairly recent extension (2023ish?)... https://en.wikipedia.org/wiki/X86#APX_(Advanced_Performance_Extensions)
<aeth>
So I guess compiles can assume it's present in 15+ years.
<aeth>
*compilers
<aeth>
Too bad they didn't do this in Haswell in 2013 because then we'd have this on nearly every Intel and AMD CPU by now.
kchanqvq29 has joined #commonlisp
kchanqvq62 has quit [Ping timeout: 250 seconds]
jon_atack has quit [Ping timeout: 248 seconds]
kchanqvq17 has joined #commonlisp
wohonajax has quit [Quit: leaving]
kchanqvq29 has quit [Ping timeout: 250 seconds]
admich1 has quit [Read error: Connection reset by peer]
admich1 has joined #commonlisp
kchanqvq19 has joined #commonlisp
kchanqvq17 has quit [Ping timeout: 250 seconds]
equwal has quit [Ping timeout: 260 seconds]
chiselfuse has quit [Remote host closed the connection]
anticomputer_ has quit [Remote host closed the connection]
triffid has quit [Remote host closed the connection]
ec_ has quit [Remote host closed the connection]
chiselfuse has joined #commonlisp
anticomputer has joined #commonlisp
triffid has joined #commonlisp
ec has joined #commonlisp
anticomputer has quit [Remote host closed the connection]
King_julian has quit [Read error: Connection reset by peer]
_whitelogger has joined #commonlisp
gilberth has quit [Ping timeout: 258 seconds]
aeth has joined #commonlisp
gilberth has joined #commonlisp
admich1 has quit [Ping timeout: 260 seconds]
admich1 has joined #commonlisp
bpanthi977 has quit [Ping timeout: 260 seconds]
viaken has joined #commonlisp
Posterdati has joined #commonlisp
gilberth has quit [Ping timeout: 258 seconds]
gilberth has joined #commonlisp
chomwitt has joined #commonlisp
dra has joined #commonlisp
dra has quit [Changing host]
dra has joined #commonlisp
kchanqvq19 has quit [Quit: Client closed]
kchanqvq19 has joined #commonlisp
rgherdt has joined #commonlisp
skippy8 has joined #commonlisp
bpanthi977 has joined #commonlisp
Oladon has quit [Quit: Leaving.]
bpanthi977 has quit [Ping timeout: 256 seconds]
<jackdaniel>
equwal: you may be interested in systems like external-program and, if you must, uiop that aim at providing full external program api compatibility layer
olivial has quit [Ping timeout: 260 seconds]
corvvs has joined #commonlisp
<drmeister>
Hey everyone. Is anyone using ocicl? How does it contrast to quicklisp?
chomwitt has quit [Quit: WeeChat 3.8]
chomwitt has joined #commonlisp
bpanthi977 has joined #commonlisp
bpanthi977 has quit [Ping timeout: 256 seconds]
zxcvz has joined #commonlisp
zxcvz has quit [Client Quit]
kchanqvq19 has quit [Quit: Client closed]
kchanqvq19 has joined #commonlisp
g-gundam has joined #commonlisp
reb` has quit [Remote host closed the connection]
alendvai__ has joined #commonlisp
yacin has quit [Ping timeout: 265 seconds]
<mesaoptimizer>
I still use quicklisp because I am lazy, I hope to eventually move to vend
<mesaoptimizer>
ocicl looks quite complex and not very lispy to me, somehow
leeb has joined #commonlisp
viaken has quit [Ping timeout: 258 seconds]
leeb_ has quit [Ping timeout: 258 seconds]
Th30n` has quit [Remote host closed the connection]
<gilberth>
beach: The jump table is in address space below 4GB for the sole reason that an instruction like "jmpq *1234" is used and that instruction only takes a 32-bit address.
<gilberth>
beach: For the heap. Even with ASLR it is still possible to mmap(2) pages at will provided that the pages aren't taken yet. ASLR only applies to the operating system's loader and dynamic linker putting your segments from the binary file at random addresses.
<gilberth>
beach: So even with ASLR you can map the heap where you want it. The routine in CCL to relocate the heap image if needed is there for robustness in case that the very address range we picked would not be available. And yes, it goes over the whole heap image and relocates pointers. GC must be able to that anyway.
<gilberth>
jasom: With AMD64 we would have enough registers to afford a second stack pointer. And modern CPUs don't care whether you use the PUSH instruction or the equivalent sequence of instructions using another register as the stack pointer. Note that CCL already partitions registers into pointer registers and immediate value registers. And in general the compiler cannot make good use of having plenty of registers available.
<gilberth>
jasom: Which imho is a bit overrated. It's only important for tight loops with no function calls at all.
apa_c has quit [Ping timeout: 265 seconds]
<gilberth>
jasom: Further. When coming from i386 and going to AMD64 you get eight extra registers. This switching %rsp back and forth is annoying and is a nightmare with signal handlers. It confuses both the kernel and me. And it confuses Rosetta. So I say that crafting the AMD64 from the PPC port with having two stack pointers would have been saner.
yitzi has joined #commonlisp
<gilberth>
jasom, beach: And wrt to that jump table I plan to have a register set aside for NIL. It would not only point to NIL but also have that jump table nearby. It'll actually be a kind of pointer to a bunch of global data.
berkowski has joined #commonlisp
X-Scale has joined #commonlisp
Lord_of_Life_ has joined #commonlisp
admich1 has quit [Read error: Connection reset by peer]
admich1 has joined #commonlisp
Lord_of_Life has quit [Ping timeout: 260 seconds]
Lord_of_Life_ is now known as Lord_of_Life
Guest47 has joined #commonlisp
<beach>
Hello xcombelle.
<beach>
gilberth: I see. Thanks.
<beach>
Oh, a register for NIL is a good idea.
yitzi has quit [Quit: yitzi]
yacin has joined #commonlisp
nullbath has quit [Ping timeout: 260 seconds]
yegor has joined #commonlisp
yitzi has joined #commonlisp
admich1 has quit [Read error: Connection reset by peer]
bpanthi977 has joined #commonlisp
cage has joined #commonlisp
bpanthi977 has quit [Ping timeout: 256 seconds]
apa_c has joined #commonlisp
apa_c has quit [Client Quit]
apa_c has joined #commonlisp
admich1 has joined #commonlisp
rgherdt has quit [Read error: Connection reset by peer]
apa_c has quit [Read error: Connection reset by peer]
gorignak has joined #commonlisp
berkowski has left #commonlisp [ERC 5.6.1-git (IRC client for GNU Emacs 31.0.50)]
yitzi has joined #commonlisp
sloanr`` has quit [Remote host closed the connection]
bpanthi977 has quit [Ping timeout: 256 seconds]
corvvs has quit [Ping timeout: 244 seconds]
corvvs has joined #commonlisp
bpanthi977 has joined #commonlisp
apa_c has joined #commonlisp
yitzi has quit [Quit: yitzi]
yaneko has quit [Quit: parting]
yaneko has joined #commonlisp
jonatack has joined #commonlisp
yaneko has quit [Client Quit]
yaneko has joined #commonlisp
cage has quit [Remote host closed the connection]
cage has joined #commonlisp
<xcombelle>
hello beach
yegor has left #commonlisp [WeeChat 4.6.3]
bpanthi977 has quit [Ping timeout: 250 seconds]
<beach>
xcombelle: Are you still learning more about Common Lisp?
bpanthi977 has joined #commonlisp
skippy8 has left #commonlisp [WeeChat 4.7.1]
<mesaoptimizer>
::notify fosskers you might want to get vend on https://awesome-cl.com/ eventually
<Colleen>
mesaoptimizer: Got it. I'll let fosskers know as soon as possible.
<ixelp>
awesome-cl
apa_c has quit [Quit: Konversation terminated!]
jonatack has quit [Quit: WeeChat 4.7.1]
<xcombelle>
beach: I'm currently doing advent of code 2024 in python
<beach>
Oh!
<xcombelle>
I'm too much fluent in python to make the switch
<beach>
I am sorry to hear that.
<xcombelle>
I have to find easier tas than aoc
jonatack has joined #commonlisp
<beach>
For Common Lisp or for Python?
<beach>
For Common Lisp, you can look at the exercises in one of the Common Lisp books. Perhaps ANSI Common Lisp, though I haven't checked that it has exercises.
yitzi has joined #commonlisp
Everything has quit [Read error: Connection reset by peer]
<yitzi>
xcombelle: There is also https://exercism.org/ They used to have a CL course.
bpanthi977 has quit [Remote host closed the connection]
bpanthi977 has joined #commonlisp
<mesaoptimizer>
there's also the Euler Project, but I think you cannot really avoid studying at least half a dozen chapters from one of the introductory CL textbooks
skin has joined #commonlisp
Everything has joined #commonlisp
bpanthi977 has quit [Ping timeout: 265 seconds]
<xcombelle>
I love programming languages my main misses are lisp, ocaml and haskell
cage has quit [Remote host closed the connection]
RDMengineer has joined #commonlisp
RDMengineer5 has joined #commonlisp
varjag has quit [Quit: ERC 5.5.0.29.1 (IRC client for GNU Emacs 29.3)]
RDMengineer has quit [Ping timeout: 258 seconds]
RDMengineer5 is now known as RDMengineer
rgherdt has quit [Ping timeout: 248 seconds]
<jasom>
gilberth: Very hot codepaths on SBCL benefit heavily from inlining (not just due to function-call overhead, but also elision of type checks) and that can stress the register allocator.
splatt9990 has joined #commonlisp
Fade_ is now known as Fade
Fade has quit [Changing host]
Fade has joined #commonlisp
bpanthi977 has joined #commonlisp
bpanthi977 has quit [Ping timeout: 256 seconds]
treflip has joined #commonlisp
<gilberth>
jasom: It doesn't matter for what I'm interested in. And it doesn't matter too much for CCL. So the extra eight registers are plenty, which is what I said. Besides the "B" in SBCL should stand for "batch". I don't like that I cannot reliably redefine a single function but must compile and load the whole source file again when I make changes.
treflip` has joined #commonlisp
treflip has quit [Read error: Connection reset by peer]
treflip` has quit [Changing host]
treflip` has joined #commonlisp
treflip` is now known as treflip
rgherdt has joined #commonlisp
yitzi has quit [Quit: yitzi]
svm has joined #commonlisp
msv has quit [Ping timeout: 248 seconds]
admich1 has quit [Read error: Connection reset by peer]
les has quit [Quit: Adios]
les has joined #commonlisp
gooba has quit [Remote host closed the connection]
gooba has joined #commonlisp
alendvai__ is now known as attila_lendvai
les has quit [Quit: Adios]
les has joined #commonlisp
<attila_lendvai>
phantomics_, hi, i'm pondering... does it have value to be able to switch the execution mode at runtime? or shall i just bite the bullet and emit the instructions in various isolated packages, e.g. 32/base/functional, 64/sse/macro, ... but i think this quickly blows up...
<attila_lendvai>
the question that brought me here is that some instructions (e.g. pextrb) needs to dispatch on the execution-mode when decoding the register name. now, in my current infrastructure that happens at macroexpand-time, which is too early. moving that code to runtime would be quite a reorg...
<attila_lendvai>
if it ever gets to that, then i may consider writing a simple partial evaluator and resolve this whole issue by always building a complete form for an instruction, and then based on the bound variables let the PE simplify the form that will then go into the insrtuction-emitter, be it a function or a macro.
bpanthi977 has joined #commonlisp
chomwitt has quit [Ping timeout: 250 seconds]
d-us-vb` has joined #commonlisp
<attila_lendvai>
while we are at it, does a functional interface have value? i.e. does the implementation of e.g. a JIT compiler have asm instructions where the operand is not a literal register name, but comes as a runtime variable?
bpanthi977 has quit [Ping timeout: 256 seconds]
d-us-vb has quit [Ping timeout: 256 seconds]
yitzi has joined #commonlisp
<attila_lendvai>
i believe it does have value, but i don't really know from experience
bpanthi977 has joined #commonlisp
equwal has joined #commonlisp
yitzi has quit [Quit: yitzi]
admich1 has joined #commonlisp
<phantomics_>
attila_lendvai, in my JIT system for April it creates internal variables, beyond that I'm not quite sure of your meaning there. Runtime execution mode switching is a property of x86 so my goal would be to facilitate it if desired
<attila_lendvai>
phantomics_, some instructions must emit different machine code depending on the targeted execution mode
<attila_lendvai>
...and some operands are only allowed in 64 bit mode. but then there are tricky ones like pextrb where the MC depends both on the register class of the operand, and the current execution mode. (it's the GR32orGR64 type in the TableGen json)
<attila_lendvai>
and what i meat is that in my initial experiment i generate macros as instruction emitters, and i think i'll need to turn them into functions for multiple reasons. then this whole question becomes moot, because macroexpand-time disappears.
<attila_lendvai>
then (speed 3) and inlining can be used to simplify away from the code body when the register name is known at the call site
treflip has quit [Ping timeout: 256 seconds]
bpanthi977 has quit [Read error: Connection reset by peer]
bpanthi978 has joined #commonlisp
<phantomics_>
I'm familiar, the width byte specification works differently depending on your mode. It could definitely cause problems if you emit instructions in the wrong mode
bpanthi978 is now known as bpanthi977
cercopith has quit [Read error: Connection reset by peer]
gooba has quit [Remote host closed the connection]
istewart has joined #commonlisp
gooba has joined #commonlisp
zxcvz has joined #commonlisp
zxcvz has quit [Client Quit]
Thermoriax has quit [Ping timeout: 260 seconds]
Thermoriax has joined #commonlisp
divlamir has quit [Read error: Connection reset by peer]
divlamir has joined #commonlisp
Oladon has joined #commonlisp
bpanthi977 has quit [Ping timeout: 265 seconds]
leeb_ has joined #commonlisp
leeb has quit [Ping timeout: 256 seconds]
cage has joined #commonlisp
<equwal>
:q
Lycurgus has joined #commonlisp
Alfr has quit [Ping timeout: 248 seconds]
lcn__ is now known as donleo
Everything has quit [Quit: leaving]
triffid has quit [Remote host closed the connection]
bpanthi977 has joined #commonlisp
Alfr has joined #commonlisp
skin has quit [Ping timeout: 258 seconds]
corvvs has quit [Remote host closed the connection]
corvvs has joined #commonlisp
skin has joined #commonlisp
bpanthi977 has quit [Ping timeout: 258 seconds]
parjanya has quit [Read error: Connection reset by peer]
triffid has joined #commonlisp
corvvs has quit [Remote host closed the connection]
donleo has quit [Quit: Leaving]
lcn_ has joined #commonlisp
chomwitt has joined #commonlisp
lcn_ has quit [Changing host]
lcn_ has joined #commonlisp
lcn_ has quit [Quit: Leaving]
lcn_ has joined #commonlisp
lcn_ has quit [Changing host]
lcn_ has joined #commonlisp
lcn_ has quit [Quit: Leaving]
lcn_ has joined #commonlisp
attila_lendvai_ has joined #commonlisp
cage has quit [Quit: rcirc on GNU Emacs 30.1]
lcn_ has quit [Quit: Leaving]
donleo has joined #commonlisp
attila_lendvai has quit [Ping timeout: 252 seconds]
donleo has quit [Quit: Leaving]
donleo has joined #commonlisp
donleo has quit [Changing host]
donleo has joined #commonlisp
donleo has quit [Client Quit]
donleo has joined #commonlisp
donleo has quit [Changing host]
donleo has joined #commonlisp
pve has quit [Quit: leaving]
pranav has quit [Read error: Connection reset by peer]
donlcn has joined #commonlisp
X-Scale has quit [Ping timeout: 248 seconds]
donlcn has quit [Client Quit]
donleo has quit [Quit: Leaving]
donleo has joined #commonlisp
mwnaylor has joined #commonlisp
Lycurgus has quit [Remote host closed the connection]
apac has joined #commonlisp
nullbath has joined #commonlisp
random-nick has joined #commonlisp
Lycurgus has joined #commonlisp
bpanthi977 has joined #commonlisp
X-Scale has joined #commonlisp
corvvs has joined #commonlisp
corvvs has quit [Client Quit]
corvvs has joined #commonlisp
attila_lendvai_ is now known as attila_lendvai
attila_lendvai has quit [Ping timeout: 248 seconds]
admich1 has quit [Ping timeout: 258 seconds]
akoana has joined #commonlisp
younder has joined #commonlisp
splatt9990 has quit [Quit: ERC 5.6.1-git (IRC client for GNU Emacs 31.0.50)]
akoana has quit [Read error: Connection reset by peer]
akoana has joined #commonlisp
akoana has quit [Client Quit]
Lycurgus has quit [Quit: alsoknownas.renjuan.org ( juan@acm.org )]
LainIwakura has joined #commonlisp
LainIwakura has quit [Client Quit]
kchanqvq4 has joined #commonlisp
LainIwakura has joined #commonlisp
chomwitt has quit [Ping timeout: 265 seconds]
admich1 has joined #commonlisp
tedwing has joined #commonlisp
triffid has quit [Remote host closed the connection]
triffid has joined #commonlisp
donlcn has joined #commonlisp
Alfr has quit [Killed (silver.libera.chat (Nickname regained by services))]
Alfr has joined #commonlisp
les_ has joined #commonlisp
gooba` has joined #commonlisp
skin_ has joined #commonlisp
d-us-vb`` has joined #commonlisp
Oladon has quit [*.net *.split]
random-nick has quit [*.net *.split]
donleo has quit [*.net *.split]
skin has quit [*.net *.split]
gooba has quit [*.net *.split]
d-us-vb` has quit [*.net *.split]
equwal has quit [*.net *.split]
les has quit [*.net *.split]
RDMengineer has quit [*.net *.split]
scymtym has quit [*.net *.split]
scymtym has joined #commonlisp
LainIwakura has quit [Quit: Client closed]
RDMengineer has joined #commonlisp
Oladon has joined #commonlisp
random-nick has joined #commonlisp
d-us-vb`` has quit [Quit: ERC 5.5.0.29.1 (IRC client for GNU Emacs 29.4)]