beneroth changed the topic of #picolisp to: PicoLisp language | The scalpel of software development | Channel Log: https://libera.irclog.whitequark.org/picolisp | Check www.picolisp.com for more information
seninha has quit [Quit: Leaving]
mario-goulart has quit [Ping timeout: 252 seconds]
genpaku has quit [Remote host closed the connection]
genpaku has joined #picolisp
seninha has joined #picolisp
<abu[m]> 'clang' does not seem to have an option for opaque pointers
<abu[m]> clang --help
<abu[m]> *If* I pass -opaque-pointers then I get the exact same .s code as without
<abu[m]> $ clang -opaque-pointers -S tst.c -emit-llvm -o - |vip
<abu[m]> But clang also does not complain if I pass -opaque-pointers
<abu[m]> This outputs normal typed pointers
<tankf33der> show me tst.c
<tankf33der> try:
<tankf33der> clang -Xclang -opaque-pointers ...
<abu[m]> same
<abu[m]> I see no 'ptr' but still i32*
<abu[m]> Also if I don't use a global pointer,.
<abu[m]> Gives: http://pb1n.de/?6905ca
<abu[m]> This is still all typed pointers: "store i32* @X, i32** %2, align 8"
<abu[m]> Or I misunderstand the whole issue
<abu[m]> I thought "ptr" must be used instead of types like "i32*"
<tankf33der> check how opt -S -opaque-pointers converts to pointers
<tankf33der> right?
<abu[m]> yes. is this llvm15?
<abu[m]> I have llvm14, there the opaque option seems to be ignored
<tankf33der> llvm14
<abu[m]> Ah, sorry! I did not see your "opt"
<abu[m]> ok, then clang does not need this option
<abu[m]> And: I don't need to change my code, right?
<tankf33der> Maybe
<tankf33der> push forward to know
<abu[m]> Cause what src/lib/llvm.l does is the same as what clang does, generate .ll
<abu[m]> i.e. before 'opt'
<abu[m]> Is it on llvm15 also so that clang does not output "ptr"?
<tankf33der> Will check in 5min
<tankf33der> lunch
<abu[m]> no hurry
<abu[m]> I'm on a walk
<abu[m]> This would mean that opaque pointers do not matter in the frontend. It is only opt (and perhaps the backend) which changes
<abu[m]> Pil is like clang a frontend
<tankf33der> output from llvm15
<abu[m]> clang? or opt?
<abu[m]> seems clang
<abu[m]> ok, so the frontend must change >= 15 too
<tankf33der> clang
<tankf33der> without opt
<abu[m]> We don't need that option for clang then
<abu[m]> It outputs the correct mode according to the version
seninha has quit [Quit: Leaving]
<abu[m]> I will change Makefile again. Remove $(OPQ) from the clang call
<abu[m]> And I wonder wether $(OPQ) is necessary for LLC and LINK
<abu[m]> In case of clang it is even harmful
<abu[m]> because "clang -opaque-pointers" interprets it as "-o" to a file "paque-pointers"
<abu[m]> Hmm, so only 'opt' and 'llc' document that option, but it is "--opaque-pointers" and not "-opaque-pointers"!
<abu[m]> A mess!
<abu[m]> I released an updated Makefile and give up. I postpone the issue until the situation is more clear.
kced has joined #picolisp
<kced> hello
aw- has quit [Ping timeout: 260 seconds]
aw- has joined #picolisp
chexum_ has quit [Ping timeout: 258 seconds]
chexum has joined #picolisp
<abu[m]> Hi kced!
chexum has quit [Quit: No Ping reply in 180 seconds.]
chexum has joined #picolisp
seninha has joined #picolisp
seninha has quit [Quit: Leaving]