compiler: Improve CIRE's cost model #2476
Conversation
Codecov ReportAll modified and coverable lines are covered by tests ✅
Additional details and impacted files@@ Coverage Diff @@
## master #2476 +/- ##
==========================================
+ Coverage 87.19% 87.22% +0.02%
==========================================
Files 238 238
Lines 45201 45258 +57
Branches 4012 4019 +7
==========================================
+ Hits 39415 39475 +60
+ Misses 5104 5103 -1
+ Partials 682 680 -2 ☔ View full report in Codecov by Sentry. |
FabioLuporini
force-pushed
the
tweak-cire-cm
branch
from
dea7217 to
94c9260
Compare
| nfunctions1 = len({i.function for i in indexeds1}) | ||
| ntemps = 0 | ||
| for sa in i.schedule: | ||
| if len(sa.writeto) < grid.dim: |
There was a problem hiding this comment.
sa.writeto.itdims? that comparison seems odd
There was a problem hiding this comment.
it's ok, grid.dim is a number (should have been called grid.ndim, legacy...) and writeto is an IterationSpace so its length is the number of Dimensions in it
| op1.apply(time_m=0, time_M=nt-2, dt=dt, u=u1, vx=vx1, vy=vy1) | ||
|
|
||
| assert np.allclose(u.data, u1.data, rtol=1e-5) | ||
| assert np.allclose(u.data, u1.data, rtol=1e-3) |
There was a problem hiding this comment.
highly unstable equations and input data, and now more aggressive refactorings
FabioLuporini
force-pushed
the
tweak-cire-cm
branch
from
94c9260 to
31344be
Compare
There was a problem hiding this comment.
I would expect to see some change in counted flops in other tests?
There was a problem hiding this comment.
Is this indeed affecting only the new variant you added?
There was a problem hiding this comment.
what we had before was good already, here I've been surgical...
3 participants
No description provided.