Parallel calculations: Difference between revisions
→{{header|Phix}}
Thundergnat (talk | contribs) m (→{{header|Perl 6}}: Minor performance tweak. Clarify verbiage a bit, reduce redundancy) |
|||
Line 1,638:
In the future hyper operators, junctions and feeds will be candidates for autothreading.
=={{header|Phix}}==
{{libheader|mpfr}}
<lang Phix>-- demo/rosetta/ParallelCalculations.exw
include mpfr.e
sequence res
constant res_cs = init_cs() -- critical section
procedure athread()
mpz z = mpz_init()
while true do
integer found = 0
enter_cs(res_cs)
for i=1 to length(res) do
if integer(res[i]) then
found = i
res[i] = {}
exit
end if
end for
leave_cs(res_cs)
if not found then exit end if
mpz_ui_pow_ui(z,2,found)
mpz_add_ui(z,z,1)
sequence r = mpz_prime_factors(z, 1_000_000)
enter_cs(res_cs)
res[found] = r
leave_cs(res_cs)
end while
exit_thread(0)
end procedure
for nthreads=1 to 5 do
atom t0 = time()
res = tagset(100)
sequence threads = {}
for i=1 to nthreads do
threads = append(threads,create_thread(routine_id("athread"),{}))
end for
wait_thread(threads)
integer k = largest(res,true)
string e = elapsed(time()-t0)
printf(1,"largest is 2^%d+1 with smallest factor of %d (%d threads, %s)\n",
{k,res[k][1][1],nthreads,e})
end for
delete_cs(res_cs)</lang>
{{out}}
<pre>
largest is 2^64+1 with smallest factor of 274177 (1 threads, 9.5s)
largest is 2^64+1 with smallest factor of 274177 (2 threads, 7.1s)
largest is 2^64+1 with smallest factor of 274177 (3 threads, 5.8s)
largest is 2^64+1 with smallest factor of 274177 (4 threads, 4.5s)
largest is 2^64+1 with smallest factor of 274177 (5 threads, 4.5s)
</pre>
IE: Checking the first 1 million primes as factors of 2^(1..100)+1 takes 1 core 9.5s and less when spread over multiple cores.<br>
Note however that I added quite a bit of locking to mpz_prime_factors(), specifically around get_prime() and mpz_probable_prime(),
[happy to leave it in since the effect on the single-thread case was neglible] so it is really only the mpz_divisible_ui_p() calls
and some control-flow scaffolding that is getting parallelised. I only got 4 cores so the 5th thread was not expected to help.
=={{header|PicoLisp}}==
| |||