ARM public key benchmark
Niels Möller
nisse at lysator.liu.se
Thu Apr 4 09:40:33 CEST 2013
nisse at lysator.liu.se (Niels Möller) writes:
> I'll also try using fewer updates of the up pointer, that seems to save
> half a cycle, and could perhaps speed up addmul_1 too.
No speedup for addmul_1, unfortunately, but a saving for submul_1. Here
are new versions of both files (for mpn/arm/v6). I wonder if this
submul_1 complement trick is useful on some other platforms too, e.g.,
64-bit sparc?
Running at 3.25 and 3.9 c/l on A9:
$ GMP_CPU_FREQUENCY=1e9 ./speed -s1-1000 -f 1.2 -C mpn_addmul_1.17 mpn_submul_1.17
clock_gettime is 1.000ns accurate
overhead 8.98 cycles, precision 1000 units of 1.00e-06 secs, CPU freq 1000.00 MHz
mpn_addmul_1.17 mpn_submul_1.17
1 19.9985 #18.7576
2 10.9929 #10.8108
3 #7.9880 8.6664
4 #6.4099 6.9205
5 #5.9251 6.3792
6 #5.4384 6.1228
7 #5.2336 5.9634
8 #4.8648 5.4154
9 #4.8359 5.2633
10 #4.5423 5.2216
12 #4.3122 4.8634
14 #4.1876 4.8111
16 #4.0881 4.6616
19 #4.0045 4.5861
22 #3.8592 4.4916
26 #3.7191 4.4362
31 #3.7194 4.3459
37 #3.6437 4.2051
44 #3.8368 4.3953
52 #3.5039 4.1120
62 #3.6497 4.2448
74 #3.6015 4.1965
88 #3.5376 4.1343
105 #3.5087 4.0832
126 #3.4988 4.1222
151 #3.4397 3.9997
181 #3.3900 3.9654
217 #3.3483 3.9500
260 #3.2994 3.9175
312 #3.3313 3.8723
374 #3.3173 3.8869
448 #3.2657 3.9315
537 #3.2799 3.9140
644 #3.3077 3.8574
772 #3.2621 3.9059
926 #3.2673 3.8317
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: submul_1.asm
URL: <http://gmplib.org/list-archives/gmp-devel/attachments/20130404/f8f544d1/attachment-0002.ksh>
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: addmul_1.asm
URL: <http://gmplib.org/list-archives/gmp-devel/attachments/20130404/f8f544d1/attachment-0003.ksh>
-------------- next part --------------
Regards,
/Niels
--
Niels Möller. PGP-encrypted email is preferred. Keyid C0B98E26.
Internet email is subject to wholesale government surveillance.
More information about the gmp-devel
mailing list