Fixes issue #75

[victoria-mcgrath]

Use actual tier sizes (rounded down to slab size and decreased by header size) when creating new memory pools

---

This change is 

[igchor]

Reviewable status: 0 of 2 files reviewed, 1 unresolved discussion (waiting on @igchor and @victoria-mcgrath)

---

cachelib/allocator/CacheAllocator-inl.h line 2391 at r1 (raw file):

      for (TierId tid = 0; (tid < numTiers_) && remainingPoolSize;
           tid++, remainingPoolSize--) {
        tierPoolSizes[tid]++;

Can this be off only be one?

[victoria-mcgrath]

cachelib/allocator/CacheAllocator-inl.h line 2391 at r1 (raw file):

Previously, igchor (Igor Chorążewicz) wrote…

Can this be off only be one?

I checked again manually and also added an extra check in tests. Could you review my last commit please?

[igchor]

Reviewed 1 of 3 files at r2, all commit messages.
Reviewable status: 1 of 3 files reviewed, all discussions resolved (waiting on @igchor)

[byrnedj]

totalCacheSize overflows with ratio 1:3 using graph cache leader fbobj config 8GB total memory

[igchor]

In such case, I think it would be good to add some more tests with different ratios. For example X:Y where X and Y are in range from 1 to 1000.

Reviewable status: 1 of 3 files reviewed, all discussions resolved (waiting on @igchor)

[igchor]

@victoria-mcgrath would you be able to add a few more tests with different ratios (e.g. all combinations from 1 to 100)?

[victoria-mcgrath]

yes, I'll work on test coverage before merging.

[victoria-mcgrath]

I reworked tests to make them go over a few more combinations of sizes and ratios.

[igchor]

rebased version on: #88