Use D: drive for Windows CI

[zanieb]

When using the standard Windows runners (as opposed to the larger GitHub runners), an undocumented D: drive is available and performant. We can save some money on by using this on a standard runner instead of a larger runner with an ReFS drive. Switching to the D: drive was not acceptable for cargo test >25m runtime.

Inspired by pypa/pip#13129
See actions/runner-images#8755

Timings (grain of salt — GitHub is super noisy):

• clippy: 2m 18s -> 2m 11s
• build binary: 2m 3s -> 2m 35s
• trampoline check (x86-64): 2m 32s -> 1m 50s (other architectures similar)
• trampoline test (x86-64): 4m 12s -> 6m 7s
• trampoline test (i686): 6m 44s -> 5m 35s

[samypr100]

Interesting, I did a similar benchmark originally (not with TEMP) and found ReFS was much faster with uv than D drive. One thing we can definitely improve is moving the actual dev drive vhdx file to D: rather than C:

We can also try a mix of Dev Drive but keep TEMP in D: rather than in the Dev Drive.

[zanieb]

Interesting, I did a similar benchmark originally (not with TEMP) and found ReFS was much faster with uv than D drive. One thing we can definitely improve is moving the actual dev drive vhdx file to D: rather than C:

Interesting, pip's recent benchmarking showed otherwise and the numbers here are honestly pretty promising for some of the jobs?

We can also try a mix of Dev Drive but keep TEMP in D: rather than in the Dev Drive.

Unfortunately there is no D: drive on the larger runners. So it's sort of looking like D: drive on a small runner or ReFS on a large runner.

[samypr100]

Unfortunately there is no D: drive on the larger runners

Ah, I didn't realize that. I haven't tried the larger runners on my personal runs.

[zanieb]

It took me a while to figure out too :D

[zanieb]

@samypr100 if you're interested, this seems good to go.