Priyanka Panigrahi via llvm-dev
2020-Feb-12 10:35 UTC
[llvm-dev] tool options to generate spill code
Hello,
For the following test case, reg.c
#include <stdio.h>
int getinput()
{
static int u=10;
return u++;
}
int main()
{
int a,b,c,d,e,f,g;
a=getinput();
b=getinput();
c=getinput();
d=getinput();
e=getinput();
f=getinput();
g=getinput();
printf("%d %d %d %d %d %d %d\n",a,b,c,d,e,f,g);
a=b=c=d=e=f=g=0;
return 0;
}
*1. $clang reg.c -Xclang -disable-O0-optnone -S -emit-llvm -o reg.ll &&
opt
-mem2reg -S reg.ll -o reg.ll && llc --regalloc=greedy reg.ll -o reg.s*
*2. $clang reg.c -Xclang -disable-llvm-passes -S -emit-llvm -o reg.ll &&
opt -mem2reg -S reg.ll -o reg.ll && llc --regalloc=greedy reg.ll -o
reg.s*
*3. $clang reg.c -S -emit-llvm -o reg.ll && opt -mem2reg -S reg.ll -o
reg.ll && llc --regalloc=greedy reg.ll -o reg.s*
Only 1. gives the spill code but it deletes the dead stores., 2 and 3 give
only allocas. Why so??
How -mem2reg works and what it depends on?
How to generate the spill code with dead stores?
Thank you.
Regards,
Priyanka
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20200212/f7bdcec1/attachment.html>
Aaron Smith via llvm-dev
2020-Feb-12 15:55 UTC
[llvm-dev] tool options to generate spill code
If you use -O0 to generate LLVM IR you must pass -disable-O0-optnone or it cannot be optimized. If you look at the LLVM IR for #2 and #3 it will say "optnone" [1] in the attributes which will prevent any optimization like mem2reg [2] from running. [1] https://clang.llvm.org/docs/AttributeReference.html#optnone [2] http://llvm.org/docs/Passes.html#mem2reg-promote-memory-to-register On Wed, Feb 12, 2020 at 2:36 AM Priyanka Panigrahi via llvm-dev < llvm-dev at lists.llvm.org> wrote:> > Hello, > > For the following test case, reg.c > > #include <stdio.h> > > int getinput() > { > static int u=10; > return u++; > } > > int main() > { > > int a,b,c,d,e,f,g; > > a=getinput(); > b=getinput(); > c=getinput(); > d=getinput(); > e=getinput(); > f=getinput(); > g=getinput(); > > printf("%d %d %d %d %d %d %d\n",a,b,c,d,e,f,g); > a=b=c=d=e=f=g=0; > return 0; > } > > *1. $clang reg.c -Xclang -disable-O0-optnone -S -emit-llvm -o reg.ll && > opt -mem2reg -S reg.ll -o reg.ll && llc --regalloc=greedy reg.ll -o reg.s* > > *2. $clang reg.c -Xclang -disable-llvm-passes -S -emit-llvm -o reg.ll && > opt -mem2reg -S reg.ll -o reg.ll && llc --regalloc=greedy reg.ll -o reg.s* > > *3. $clang reg.c -S -emit-llvm -o reg.ll && opt -mem2reg -S reg.ll -o > reg.ll && llc --regalloc=greedy reg.ll -o reg.s* > > Only 1. gives the spill code but it deletes the dead stores., 2 and 3 give > only allocas. Why so?? > > How -mem2reg works and what it depends on? > > How to generate the spill code with dead stores? > > Thank you. > > Regards, > Priyanka > _______________________________________________ > LLVM Developers mailing list > llvm-dev at lists.llvm.org > https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev >-------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20200212/be967f04/attachment.html>
Priyanka Panigrahi via llvm-dev
2020-Feb-13 06:54 UTC
[llvm-dev] tool options to generate spill code
Thanks for the clarification. Why does mem2reg remove the dead stores? Regards, Priyanka On Wed, Feb 12, 2020 at 9:25 PM Aaron Smith <aaron.lee.smith at gmail.com> wrote:> If you use -O0 to generate LLVM IR you must pass -disable-O0-optnone or it > cannot be optimized. If you look at the LLVM IR for #2 and #3 it will say > "optnone" [1] in the attributes which will prevent any optimization like > mem2reg [2] from running. > > [1] https://clang.llvm.org/docs/AttributeReference.html#optnone > [2] http://llvm.org/docs/Passes.html#mem2reg-promote-memory-to-register > > > On Wed, Feb 12, 2020 at 2:36 AM Priyanka Panigrahi via llvm-dev < > llvm-dev at lists.llvm.org> wrote: > >> >> Hello, >> >> For the following test case, reg.c >> >> #include <stdio.h> >> >> int getinput() >> { >> static int u=10; >> return u++; >> } >> >> int main() >> { >> >> int a,b,c,d,e,f,g; >> >> a=getinput(); >> b=getinput(); >> c=getinput(); >> d=getinput(); >> e=getinput(); >> f=getinput(); >> g=getinput(); >> >> printf("%d %d %d %d %d %d %d\n",a,b,c,d,e,f,g); >> a=b=c=d=e=f=g=0; >> return 0; >> } >> >> *1. $clang reg.c -Xclang -disable-O0-optnone -S -emit-llvm -o reg.ll && >> opt -mem2reg -S reg.ll -o reg.ll && llc --regalloc=greedy reg.ll -o reg.s* >> >> *2. $clang reg.c -Xclang -disable-llvm-passes -S -emit-llvm -o reg.ll && >> opt -mem2reg -S reg.ll -o reg.ll && llc --regalloc=greedy reg.ll -o reg.s* >> >> *3. $clang reg.c -S -emit-llvm -o reg.ll && opt -mem2reg -S reg.ll -o >> reg.ll && llc --regalloc=greedy reg.ll -o reg.s* >> >> Only 1. gives the spill code but it deletes the dead stores., 2 and 3 >> give only allocas. Why so?? >> >> How -mem2reg works and what it depends on? >> >> How to generate the spill code with dead stores? >> >> Thank you. >> >> Regards, >> Priyanka >> _______________________________________________ >> LLVM Developers mailing list >> llvm-dev at lists.llvm.org >> https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev >> >-------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20200213/ade0bfed/attachment.html>