[V1] get_computed_blocks avoids recompute

[Abatom]

Use self.req_to_blocks[request.request_id] instead of recompute.

[github-actions]

👋 Hi! Thank you for contributing to the vLLM project.
Just a reminder: PRs would not trigger full CI run by default. Instead, it would only run fastcheck CI which starts running only a small and essential subset of CI tests to quickly catch errors. You can run other CI tests on top of those by going to your fastcheck build on Buildkite UI (linked in the PR checks section) and unblock them. If you do not have permission to unblock, ping simon-mo or khluu to add you in our Buildkite org.

Once the PR is approved and ready to go, your PR reviewer(s) can run CI to test the changes comprehensively before merging.

To run CI, PR reviewers can do one of these:

• Add ready label to the PR
• Enable auto-merge.

🚀

[comaniac]

I don't think this is correct. In this way you only share the blocks within a request.

[Abatom]

@comaniac Thank you, I will think about it again.

[Abatom]

@comaniac I have drawn a diagram to illustrate my modifications. I believe that the cache is not fully shared within a request.

[comaniac]

What I'm saying is you change makes blocks cannot be shared "across" requests.

[Abatom]

@comaniac
I think the orange blocks in the picture are shared across requests.

The green blocks have already been calculated,so there's no need to calculate them again，this is where the optimization comes in.

[comaniac]

How do you hit cached blocks for a new request, which hasn't been added to req_to_blocks[request.request_id]?

[comaniac]

Ok I see what you meant. Yes this saves hash computation, but I'm afraid that this makes the code complicate. Did you have benchmark numbers to show the overhead is negligible so we have to introduce this optimization?

[Abatom]

Alright, I'll think about how to conduct benchmark testing.