Closed abelsiqueira closed 2 years ago
Merging #166 (8c2015c) into main (9f316a3) will increase coverage by
0.46%
. The diff coverage is100.00%
.
@@ Coverage Diff @@
## main #166 +/- ##
==========================================
+ Coverage 90.64% 91.11% +0.46%
==========================================
Files 8 8
Lines 171 180 +9
==========================================
+ Hits 155 164 +9
Misses 16 16
Impacted Files | Coverage Δ | |
---|---|---|
src/linesearch/line_model.jl | 100.00% <100.00%> (ø) |
Continue to review full report at Codecov.
Legend - Click here to learn more
Δ = absolute <relative> (impact)
,ø = not affected
,? = missing data
Powered by Codecov. Last update 9f316a3...8c2015c. Read the comment docs.
@dpo, I've updated with allocations tests, but there are a few issues.
First, the allocation of the LineModel API and of Armijo is different on 1.3, so it fails.
The allocation of Armijo is not really well defined. I initially thought it was (8 + nbk) * 32 + 16
, but that actually only works for bk_max >= 4
, which sets nbk = 4
. However, if I change bk_max < 4
, then it becomes (7 + 1.5nbk) * 32
. But this doesn't work for nbk = 4
. So I used min(12.5 * 32 + 16, 7 + 1.5nbk) * 32
. Furthermore, if I use a loop for the test, then it allocates and additional 208.
Any suggestions?
Does the armijo_wolfe
in #146 allocate?
Does the
armijo_wolfe
in #146 allocate?
Probably, I made no efforts to remove allocations on #146.
Maybe restrict allocation tests to Julia > 1.3?
Stores
xt
inside LineModel to prevent allocations.WIP because I can't predict the allocations of
armijo_wolfe
yet. Testing manually with this script gives a different result.