Add simd_bswap, simd_bitreverse, simd_ctlz, and simd_cttz intrinsics

[calebzulawski]

cc @workingjubilee

[rustbot]

r? @compiler-errors

(rustbot has picked a reviewer for you, use r? to override)

[workingjubilee]

Intrinsic has incorrect argument type!
<4 x i32> (<4 x i32>, i8)* @llvm.ctlz.v4i32
Intrinsic has incorrect argument type!
<4 x i32> (<4 x i32>, i8)* @llvm.ctlz.v4i32
Intrinsic has incorrect argument type!
<4 x i32> (<4 x i32>, i8)* @llvm.cttz.v4i32
Intrinsic has incorrect argument type!
<4 x i32> (<4 x i32>, i8)* @llvm.cttz.v4i32
LLVM ERROR: Broken module found, compilation aborted!

well
that won't do.

[calebzulawski]

Is that LLVM 14 specific? I'm pretty sure the bump to 15 is imminent.

[calebzulawski]

Ah, actually, the problem is that it's encoding the last argument as i8 instead of i1. Why does it work elsewhere?

[workingjubilee]

Is that LLVM 14 specific? I'm pretty sure the bump to 15 is imminent.

I don't thiiink so?

[Urgau]

Is that LLVM 14 specific? I'm pretty sure the bump to 15 is imminent.

I don't thiiink so?

#114148 8 hours ago ;-)

[workingjubilee]

@Urgau I mean that I don't think the issue is LLVM 14 specific, I could be wrong.

[Urgau]

@Urgau I mean that I don't think the issue is LLVM 14 specific, I could be wrong.

Oh, yeah sorry. As for the issue here, I would just note that type_bool() is indeed i8 not i1; to have i1 you should call type_i1().

[bjorn3]

cg_clif and cg_gcc will need these intrinsics too.

[calebzulawski]

Is that LLVM 14 specific? I'm pretty sure the bump to 15 is imminent.

I don't thiiink so?

It compiles on newer LLVM so at some point it became flexible enough to accept i8.

cg_clif and cg_gcc will need these intrinsics too.

@bjorn3 do you expect that to be part of this PR, or just making a note of it?

[bjorn3]

Mostly making a note. I can implement it on the cg_clif side myself. I don't know what @antoyo prefers for cg_gcc.

[antoyo]

I'm also taking note for cg_gcc and I'll implement it myself.

[calebzulawski]

@Urgau I mean that I don't think the issue is LLVM 14 specific, I could be wrong.

Oh, yeah sorry. As for the issue here, I would just note that type_bool() is indeed i8 not i1; to have i1 you should call type_i1().

It looks like this worked!

[compiler-errors]

@bors r+

[bors]

📌 Commit ce4a48f has been approved by compiler-errors

It is now in the queue for this repository.

[RalfJung]

Any chance someone could make a PR to also implement these in Miri? :) Let me know if you need any pointers for getting started. The existing SIMD intrinsics are probably a good template.

[bors]

⌛ Testing commit ce4a48f with merge 03a5725...

[bors]

☀️ Test successful - checks-actions
Approved by: compiler-errors
Pushing 03a5725 to master...

[rust-timer]

Finished benchmarking commit (03a5725): comparison URL.

Overall result: no relevant changes - no action needed

@rustbot label: -perf-regression

Instruction count

This benchmark run did not return any relevant results for this metric.

Max RSS (memory usage)

Results

This is a less reliable metric that may be of interest but was not used to determine the overall result at the top of this comment.

	mean	range	count
Regressions ❌ (primary)	1.6%	[0.8%, 5.6%]	16
Regressions ❌ (secondary)	4.7%	[1.4%, 9.0%]	5
Improvements ✅ (primary)	-1.8%	[-1.8%, -1.8%]	1
Improvements ✅ (secondary)	-2.6%	[-3.5%, -2.0%]	4
All ❌✅ (primary)	1.4%	[-1.8%, 5.6%]	17

Cycles

This benchmark run did not return any relevant results for this metric.

Binary size

This benchmark run did not return any relevant results for this metric.

Bootstrap: 652.746s -> 651.287s (-0.22%)