size_t
As per #70 (closed), we will run into problems with more than 2bn particles per node as all the particle counters and loop variables are currently int
.
As a first step, switch the stored particle counts to size_t
.
The next series of commits will be to change all the loops over the particles. I'm breaking this into two pieces since these changes are highly disruptive, especially since I'm submitting them along with !113 (merged).
Merge request reports
Activity
Added 1 commit:
- 4070cda1 - %zi for size_t printf
I'm seeing a crash at:
#0 0x0000000000406a42 in space_gparts_sort (gparts=0x2ae196d8d040, ind=0x3979cf0, N=0, min=0, max=14399) at space.c:744 #1 0x00000000004058ca in space_rebuild (s=0x7fff98490040, cell_max=0, verbose=0) at space.c:441 #2 0x000000000043e1d9 in engine_rebuild (e=0x7fff98490220) at engine.c:1201 #3 0x000000000043e30f in engine_prepare (e=0x7fff98490220) at engine.c:1250 #4 0x000000000043eb02 in engine_init_particles (e=0x7fff98490220) at engine.c:1419 #5 0x0000000000403eb8 in main (argc=15, argv=0x7fff98490548) at main.c:492
with the following variables:
qstack = 0x414a100 qstack_size = 28808 first = 1 last = 1 waiting = 1 pivot = 7199 i = 0 ii = 1024191 j = 18446744073709551615 jj = 18446744073709342492 temp_i = 957597 qid = 0
Some of those look unlikely! Does the sorting require a signed int?
mentioned in merge request !113 (merged)
mentioned in commit 1f9e985c
Please register or sign in to reply