a definition of
The overall entropy of the universe really depends on where the energy ends up and where it comes from.
Whenever you put energy out into the universe to roam at random in the form of wandering photons and neutrinos ... you've increased the entropy of the universe. If you collect that energy in one place, you decrease entropy. But you can't, in principle, re-collect any of that energy because more energy always leaves your location than what falls on your collector. Space is always colder than your location so it's a thermodynamic impossibility to collect more energy from intergalactic space then you eject, i.e. there is no heat sink below the temperature of intergalactic space for heat to escape to ... in principle.
There is no way to "attract" it to you except by gravity. But there isn't much gravity out there to pull in photons or neutrinos. If all the free-roaming neutrinos and photons in the universe intersected black holes and got stuck ... the entropy of the universe would decrease ... in theory. But the odds against that are ... astronomical ... so entropy is the direction of greatest probability. And that's where we are all going ... downhill.
That's the second law of thermodynamics. Once the energy is liberated into space in the form of photons and neutrinos, you can't recover it ... because the probability of that happening is nil. It's a functional impossibility ... but not a theoretical impossibility. So, the entropy of the universe might reverse itself in the future if some unknown factor weights the probability in that direction. Or, if you wait long enough, the improbable will eventually happen.
Before discussing entropy further, you must first define your "system". That is, are you talking about the whole universe? Or, are you talking about a finite, bounded volume of space?
In hydrogen fusion ... entropy increases (you're sending photons out of a finite box into space while nuclei go into a lower energy state). If the photons stay in the box, entropy also increases ... but, at some point enough photons will return to the formed helium to heat it up and break it back up into hydrogen ... so, the increase in entropy would then stop and you would have a stable measure of entropy.
In fission, entropy increases because you're releasing kinetic energy stored in the nucleus, i.e. it's a higher energy state. Again, assuming that energy keeps leaving the finite box.
When the fissile material is made by fusion, the nuclear force goes to a lower energy state but the electromagnetic interaction goes into a higher energy state which is greater than the nuclear is lower, i.e. energy is stored in the nucleus (charge repulsion) which is now heavier than the constituent particles were when they were apart. This decreases entropy inside the box.
In pair production ... energy turns into mass so you are lessening entropy by putting the universe into a less probable state ... but one piece is anti-matter so entropy quickly increases again when that anti-particle annihilates with its corresponding particle. In order to produce the anti-particles to annihilate particles with (thus facilitating total mass-to-energy conversion), you must liberate photons into the universe at large. I'm not talking about baryonic mass which can't all be turned completely into photons because there is no significant amount of anti-matter. We only see anti-matter in pair production. We can't convert energy into just, say, positrons without the corresponding electrons.
When we convert mass into energy we are basically talking about the small amount of mass that is lost when nuclei fuse into a new nucleus which is less massive than the two original nuclei taken as separated individuals. The difference is slight and consists of the photons (and neutrinos) which are liberated into the universe at random. Potential energy in the form of a little "extra" mass is converted into kinetic energy of moving particles or photons. If those particles or photons fall into another potential, the entropy of the universe decreases. But generally they just roam forever out in interstellar space. Thus, the total number of free-roaming photons and neutrinos is the truest measure of entropy.
I'll grant that gravity may not pull in photons but it certainly "redirects" them and when a photon escapes a gravitational well, it loses energy. So, gravity can in principle "suck dry" the photons energy. At any rate, it doesn't change anything. Whether it pulls them in or not, they're not coming back in any great numbers.
All energy potentials are like springs
If something goes down a potential it will simply come back out again because as the object is going down into the potential is converts that potential energy into kinetic energy and "flies back out the other side". It's like dropping into a hypothetical hole in the ground that goes all the way through the earth. You come back out on the other side.
You lose some energy in the hole ... like from scraping along the walls of the hole. Then, you wouldn't quite make it out of the hole on the other side. Same thing with matter falling into a potential well. Unless it gives off photons to lessen its kinetic energy, it will just come back out. If all the free-roaming photons and neutrinos came back to stars and galaxies, the stars and galaxies would heat up and expand back into the gases from which they condensed.
There is a cosmological problem called "breaking energy" which is about this. The problem is ... if galaxies are in clusters and big voids are left over where they came from ... how is it that the galaxies just stopped in the clusters. Where is the energy that they built up as they fell into clumps? It can be crudely calculated ... but where is it?