Running the k8/k10 asm code with no changes on the core2 machine sage we get
this
popcount,hamdist no popcount instruction
slowdowns---------
add,sub are 0.50x
rshift1,lshift1 0.70x
k8 lshift,rshift 0.91x
addmul_1,submul_1 0.89x but faster for <20 limbs...
speedups----------
and,ior,xor are 1.13x
nand,nior,xnor,andn,iorn are 1.50x
com is 2.00x
divebyff 1.40x although not better until 12limbs
diveby3 2.30x
addadd,addsub 1.50x
sumdiff 1.26x
addlsh1 1.50x
sublsh1 1.40x
k10 lshift,rshift 1.18x
mul_1 1.04x
for mul basecase we get
./speed -c -r -s 1-40 mpn_jaytest mpn_mul_basecase
overhead 6.12 cycles, precision 10000 units of 3.75e-10 secs, CPU freq 2666.76
MHz
mpn_jaytest mpn_mul_basecase
1 #9.21 2.3531
2 #21.43 2.0070
3 #56.00 1.5234
4 #91.36 1.4960
5 #136.17 1.4320
6 #195.54 1.3769
7 #261.22 1.3718
8 #336.56 1.3482
9 #419.62 1.3441
10 #527.14 1.3054
11 #634.44 1.3105
12 #744.00 1.3172
13 #873.85 1.2931
14 #1024.55 1.1088
15 #1169.00 1.0873
16 #1328.89 1.0704
17 #1492.50 1.0672
18 #1710.00 1.0317
19 #1880.00 1.0488
20 #2112.00 1.0246
21 #2288.00 1.0385
22 #2547.50 1.0128
23 #2787.50 1.0063
24 #3012.50 1.0108
25 #3212.50 1.0241
26 3556.67 #0.9953
27 3836.67 #0.9939
28 4106.67 #0.9935
29 4370.00 #0.9908
30 4700.00 #0.9617
31 4996.67 #0.9973
32 5380.00 #0.9944
33 5685.00 #0.9727
34 6105.00 #0.9853
35 6375.00 #0.9906
36 6775.00 #0.9764
37 7540.00 #0.9151
38 7515.00 #0.9714
39 #7955.00 1.0578
40 #8750.00 1.0086
This is all with no tweeking on
cpu family : 6
model : 29
model name : Intel(R) Xeon(R) CPU X7460 @ 2.66GHz
stepping : 1
--~--~---------~--~----~------------~-------~--~----~
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
-~----------~----~----~----~------~----~------~--~---
