This allows us to to only include it from files which really need
it (of which there aren't many; currently only SVML), which is
important since it can pull in a lot of extra headers and dramatically
increase compilation time.

Fixes #693

For value outside of (INT32_MIN, INT32_MAX), as well as NaNs, x86
returns INT32_MIN.

I added a SIMDE_FAST_CONVERSION_RANGE macro (which is defined by
default if SIMDE_FAST_MATH is defined) to opt-out of the more accurate
behavior since it's also substantially slower on many platforms.

Fixes #685

Fixes #682

These are based on some StackOverflow answers linked to in this here:
https://stackoverflow.com/questions/65441496/what-is-the-most-efficient-way-to-do-unsigned-64-bit-comparison-on-sse2

Oops, forgot this for the 0.6.0 release!

Add _mm{,256}_{,mask_,mask2_,maskz_}permutex2var_epi{8,16,32,64} intrinsics to permutex2var.[ch]
Add _mm{,256}_{,mask_,mask2_,maskz_}permutex2var_{pd,ps} intrinsics to permutex2var.[ch]

Add AVX, SSE, NEON, Altivec and Wasm translations for most intrinsics.

* operationalize clang bug 44589
* remove unused bug references

This is basically the same false positive as #657.  As of GCC 10 the
diagnostic is no longer emitted in C mode, but it is still emitted in
C++ mode, so I'm removing the version check from the ld3 code, too.

Fixes #676

_mm_prefetch is supposed to take an int, but when optimization enabled
GCC uses an enum _mm_hint instead.  Without optimization _mm_prefetch
is defined as a macro (as it is on clang) since there is no constant
propagation.

Since we can't rely on all hints being defined on all platforms we
define our own, which are not part of the enum _mm_hint.  So, on GCC
we cast each of the hints to enum _mm_hint to avoid a warning about
an invalid conversion.

IMHO this is technically a bug in the GCC API, but I appreciate the
effort at getting the compiler to validate the input so I'm not going
to complain to the GCC folks as long as this doesn't cause additional
breakage.

Fixes #673

Add the following intrinsics to permutexvar.[ch] :
- _mm_{,mask_,maskz_}permutexvar_epi{8,16}
- _mm256_{,mask_,maskz_}permutexvar_epi{8,16,32,64}
- _mm256_{,mask_,maskz_}permutexvar_{pd,ps}

Add translations to AVX, SSE, NEON, Altivec and WASM

Fixes #670

This should just be slightly easier (less difficult) for compilers
since they may be able to convert this to a gather operation on
some architectures.  I'm not going to hold my breath, though.

Fixes #657

Fixes #656

Remove useless for-loops
Remove redundant to_private/from_private
Remove redundant defined(SIMDE_ARM_NEON_A32V7_NATIVE) when we already
  check for SIMDE_NATURAL_VECTOR_SIZE_LE(128)

Add _mm512_{,mask_,mask2_,maskz_}permutex2var_epi{8,16} intrinsics to permutex2var.[ch]

Add NEON translations for _mm512_permutex2var_epi{8,16,32}
This could easily be done for _mm512_permutex2var_epi64 too.
However it may not be faster than the portable version.
There are decreasing returns as the size increases.

Add _mm512_{,mask_,maskz_}permutexvar_epi{8,16} intrinsics to permutexvar.[ch]

Add NEON translations for _mm512_permutexvar_epi{8,16,32}
This could easily be done for _mm512_permutexvar_epi64 too.
However it may not be faster than the portable version.
There are decreasing returns as the size increases.

Add defines for AVX512VBMI to simde-arch.h and simde-features.h

warning C4701: potentially uninitialized local variable 'r_' used

Fixes #654

XL C/C++ supports __builtin_shufflevector, but only when using AltiVec
types, not on GCC-style vector extensions.  We could add alternate
implementations which use the same code but (for example) altivec_f32
instead of f32, but that introduces a lot of bloat into the code for
a compiler very few use *and* rewards IBM for a stupid decision which
breaks compatibility with *lots* of code.