update comments

[unalmis]

Adds some comments to objectives

[rahulgaur104]

Full optimization performance benchmarks were never performed. For example, when optimizing an equilibrium, the new bounce integral from ku/bounce (merged) took almost exactly the same time as the old bounce integration routines. So compute maybe faster but optimization definitely was not. Maybe sparse pullback will change that, maybe not.

I don't actually care at this point, so approving this PR.

[unalmis]

Full optimization performance benchmarks were never performed. For example, when optimizing an equilibrium, the new bounce integral from ku/bounce (merged) took almost exactly the same time as the old bounce integration routines. So compute maybe faster but optimization definitely was not. Maybe sparse pullback will change that, maybe not.
I don't actually care at this point, so approving this PR.

@rahulgaur104 's comment purpose is not clear to me, nor why he feels the need to advertise he does not care.

Focusing on the technical/work relevant content of that comment, his statements in reference to my work are verifiably false.

To reduce spread of misinformation for others who are reading, note that

• ku/bounce (link: New inverse stream map to accelerate convergence #1919 ) is a branch that primarily focused on correcting bugs in DESC and improving convergence; not performance.
• Still, the performance benefit from that PR is significant if one benchmarks it properly - e.g. compare how much faster it is when a low resolution solver is used (because the convergence improvements make the new low resolution solver as accurate as the old high resolution solver).
• All of this is shown in detail in New inverse stream map to accelerate convergence #1919 .

Next, ku/sparse_pullback (link: #2170 ) and the following PRs reduces memory and improves performance by a factor of the number of flux surfaces the optimization is done with.

• This was benchmarked and proven (in math and in practice). One should not diminish the value of my work there, which is actually the largest performance boost in terms of memory+compute in desc to date. One performance benchmark demonstrating this (among many others that were shared with visualization) is available here (https://github.com/unalmis/DESC/blob/master/publications/unalmis2025/effective_ripple_profile.py).
• My later PRs (outward superbanana #2220 and Single contangent pullback through compute pipeline #2171) continue this improvement by extending the sparse pullback through more of the compute pipeline.

I suggest making effort to become better informed before making negative claims on others' work.

[unalmis]

@YigitElma @ddudt @f0uriest @dpanici it still won't let me merge. Please click merge.