add support for leading_zeros, trailing_zeros and fix count_ones

[Firestar99]

In main leading_zeros and trailing_zeros require some obscure Intel Extension. This PR changes them to use the GLSL.std.450 functions FindILsb and FindUMsb, which do the exact same thing. Though they slightly mismatch when passing in 0, rust expects the bit count of the type but they return !0/-1. And they require to be used exclusively with 32bit integers, so some careful type and bitcasting is required to emulate it for all other types, like u8, u16 and u64.

Also fixes count_ones and bit_reverse, as VUID-StandaloneSpirv-Base-04781 also requires their args to always be 32bit integers as well but we were ignoring it.

While removing the Intel extension, some code managing enabled extensions was triggering an unused warning, so I removed it in 1fb0301 which we may revert if needed.

Closes #210 #215

[LegNeato]

Sweet! I actually looked at this too, just didn't put my WIP up. Differences between what I did:

• Kept the intel stuff, but made it queried / optional.

• Used bitshifts and masking to support rather than the glop (I would imagine the glop is faster?). I basically just looked at stack overflow (https://stackoverflow.com/questions/23856596/how-to-count-leading-zeros-in-a-32-bit-unsigned-integer) and hacker's delight.

I pushed a branch. I think the code is largely correct, but it is only lightly tested:

https://github.com/LegNeato/rust-gpu/tree/clz

[Firestar99]

This is ready for review, but completely untested beyond compile tests. Would love some integration tests or fuzzing right now, as it's quite easy to screw up bitcasts or misread documentation. But I guess some careful reviewing must do.

[LegNeato]

You could rebase on #216 ?

[Firestar99]

Sure, there you go a rebase, git could just do it automatically. I assume CI will now do run some difftests?

[LegNeato]

Yep! Now you can add a runtime diff test for this if you want. But I'll also be revamping a bit and squashing that PR, so I don't want to make your git stuff more annoying. Up to you if it is worth it or want to land as-is!

[Firestar99]

I've removed the differential testing harness again, as you've said it isn't ready for review yet, and it tests rust-gpu vs wgsl whereas I need rust-gpu vs standard rust on the cpu.

Just did some basic diff testing in my engine against a software impl I've been using as a replacement, and something isn't right with trailing / leading zeros. Will look at it at a later date.

[Firestar99]

I made a test in my bindless framework (shader cpu) and can confirm that u32 is working!

However, I'm running into issues testing out other unsigned types:

error: cannot load type u8 in an untyped buffer load
    --> /home/firestar99/.cargo/git/checkouts/rust-gpu-11142fd2aadc2318/57c7d9d/crates/spirv-std/src/byte_addressable_buffer.rs:97:9
     |
  97 |         buffer_load_intrinsic(self.data, byte_index)
     |         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  error: cannot store type struct count_leading_trailing::CountLeadingTrailingZerosTransfer<u8, u8, u8> { count: u32, leading: u32, trailing: u32, input: u8, swap_bytes: u8, reverse_bits: u8 } in an untyped buffer store
     --> /home/firestar99/.cargo/git/checkouts/rust-gpu-11142fd2aadc2318/57c7d9d/crates/spirv-std/src/byte_addressable_buffer.rs:150:9
      |
  150 |         buffer_store_intrinsic(self.data, byte_index, value);
      |         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
      |
      = note: due to containing type 49

I may rework this to cast them to u64 and back, though I'd of course prefer if we fix that one as well. Enabling Int8, Int16, Int64, StorageBufferInt8, StorageBufferInt16 didn't help.

[Firestar99]

u64 fails in the exact same way...

@LegNeato I say we just merge it for now and hope anything non-u32 works as well. At the very least I fixed u32.

[LegNeato]

Just a style thing, but I generally feel it is clearer to matched on signed as well, it makes all the cases clearer, similar to what you did in bitcast.

[LegNeato]

Doesn't this need to change depending on sign?

[LegNeato]

u64 fails in the exact same way...

@LegNeato I say we just merge it for now and hope anything non-u32 works as well. At the very least I fixed u32.

I'm not sure I parse this...the comment above was saying unsigned types except for u32 do not work, and then confirms u64 does not work, and then says let's land and "hope anything non-u32" works?

[Firestar99]

To correct things: I implemented it for u8, u16, u32 and u64 and today I wrote a test with my bindless stuff to debug why u32::trailing_zeros() was giving me crap values. But I was not able to easily adapt the test to anything other than u32, due to limitations with ByteAddresableBuffer not allowing you to read or write u8, u16 and u64 for whatever reason. (Which I was completely unaware of)

Anyways, I noticed that rustc actually does not call any of these methods on the signed types, as the rust std converts them to unsigned beforehand. So I can just remove all the signed conversions, except for [i8,i16,i32,i64]::leading_zeros() with non_zero: true. I have no idea how these are reachable, but they exist and we now support them by just converting to unsigned and "retrying".

Give me another iteration on this one, but I'd like to land it before we release.

[Firestar99]

@LegNeato Did another iteration, hopefully a lot cleaner now. Removed the signed-ness everywhere it's not needed, with just one extra branch for the leading_zeros() special case, see previous edited comment.

[LegNeato]

LGTM!