https://gcc.gnu.org/bugzilla/show_bug.cgi?id=107757
--- Comment #4 from Michael Meissner <meissner at gcc dot gnu.org> ---
Note, this code only shows up when the target CPU is power8.
For the following code:
vector long long lsb64()
{
return vec_splats(1LL);
}
Both power9 and power10 generate:
xxspltib 34,1
vextsb2d 2,2
Now, if you want to add code for power8, you would:
1) Add code in rs6000.c's easy_altivec_constant (and possibly vsplitis_shifted
and other functions) to handle the operation on power8.
2) Then add the code in rs6000.c's gen_easy_altivec_constant to generate the
insns for this. You might need to add explicit insns in altivec.md or vsx.md
for this optimization.
