Hi
Some timings for bobcat in svn size=1000 limbs
./benchmpn
cpu bobcat
add_n 2538
sub_n 2528
mul_1 5064 this is the 1-way unrolled core2 code
addmul_1 5308 again this is core2 code
submul_1 5292
mul_2
addmul_2
submul_2
addadd_n 3715
addsub_n 3696
subadd_n 3719
lshift 3784
rshift 3546
lshift2 2536
rshift2 2776
lshift1 1912
rshift1 3546
addlsh1_n 3293
sublsh1_n 3541
addlsh_n 4913
sublsh_n 4918
inclsh_n 4926
declsh_n 4915
rsh1add_n 3027
rsh1sub_n 3033
sumdiff_n 4715
store 1157
copyi 1786
copyd 1659
rsblsh1_n
addlsh2_n
rsblsh2_n
popcount 5876 nehalem code
hamdist 6784 k10 code
com 1789
not 1118
and_n 2531
xor_n 2530
ior_n 2542
nand_n 2537
nior_n 2544
xnor_n 2546
andn_n 2533
iorn_n 2534
lshiftc 4543
divexact_byff 2039
divexact_byfobm1 6261
divexact_by3 6246
divexact_1 15062
modexact_1c_odd 15084
add_err1_n 4810
sub_err1_n 4802
divrem_euclidean_qr_1 19548
divrem_euclidean_qr_2 40166
divrem_euclidean_r_1 8261
divrem_1 21478
divrem_2 28162
divrem_hensel_qr_1 12568
divrem_hensel_qr_1_1 14104
divrem_hensel_qr_1_2 12606
divrem_hensel_r_1 14064
divrem_hensel_rsh_qr_1 16603
rsh_divrem_hensel_qr_1 13170
rsh_divrem_hensel_qr_1_1 14112
rsh_divrem_hensel_qr_1_2 13176
mod_1_1 12003
mod_1_2 7556
mod_1_3 8062 C code was faster than asm ,
but
slower than the the asm mod_1_2 !!!
mod_1_4
mod_34lsub1 3045
mul_basecase runs at about 6c/l so clearly from the above we can do better
A quick test gives us 6c for rax , 7c for rdx , thruput of 5c for mul ,
although these maximum values , could be better than this. I've got some more
on this...
Jason
--
You received this message because you are subscribed to the Google Groups
"mpir-devel" group.
To post to this group, send email to mpir-devel@googlegroups.com.
To unsubscribe from this group, send email to
mpir-devel+unsubscr...@googlegroups.com.
For more options, visit this group at
http://groups.google.com/group/mpir-devel?hl=en.
