[libc-commits] [libc] [libc] Implement generic SIMD helper 'simd.h' and implement strlen (PR #152605)

[Joseph Huber via libc-commits]

jhuber6 wrote:

I updated this to provide a more comprehensive header and interface using some of the fixed I've landed in upstream clang. The generated ASM for AVX512 looks like this, so I think it's roughly equivalent to the 'hand-rolled' versions. This one uses the generic version if they are available.

```asm
strlen:                   # @srlen
	.cfi_startproc
# %bb.0:                                # %entry
	movq	%rdi, %rax
	andq	$-64, %rax
	vmovdqa64	(%rax), %zmm0
	subl	%eax, %edi
	vptestnmb	%zmm0, %zmm0, %k0
	kmovq	%k0, %rcx
	shrxq	%rdi, %rcx, %rcx
	testq	%rcx, %rcx
	jne	.LBB0_4
# %bb.1:                                # %for.cond.preheader
	addq	$64, %rax
	.p2align	4
.LBB0_2:                                # %for.cond
                                        # =>This Inner Loop Header: Depth=1
	vmovdqa64	(%rax), %zmm0
	addq	$64, %rax
	vptestnmb	%zmm0, %zmm0, %k0
	kortestq	%k0, %k0
	je	.LBB0_2
# %bb.3:                                # %cleanup.thread
	kmovq	%k0, %rcx
.LBB0_4:                                # %cleanup35
	tzcntq	%rcx, %rax
	vzeroupper
	retq
```

There's some stuff I could add, mostly helpers around shuffling and whatnot, but I'll save that for later.

https://github.com/llvm/llvm-project/pull/152605