Some code improvements.

[besser82]

Change 'uint512_u' to be a type, not a symbol.

OSX / XCode complains about duplicate symbols.

---

Rename 'uint512_u' to 'gost34112012_uint512_u' to avoid clashes.

---

Rename the 'ALIGN' macro to 'GOST3411_ALIGN'.

macOS may define 'ALIGN' with a different meaning if some system
headers are included.

---

Fix '-Wstringop-truncation'.

---

Use the unaligned version of '_mm_store_si128'.

This is needed for ix86 architectures, as for some unidentified
reason the memory address may not be 16 bytes aligned properly.

The information from the Intel Intrinsics Guide shows no speed
penalty between using the aligned or the unaligned version of
this function.

---

Remove trailing whitespace.

[MikhailKasimov]

@besser82 Fix '-Wstringop-truncation' --> O, this is actual problem. E.g. for Fedora 28:

[0] https://build.opensuse.org/package/live_build_log/home:k_mikhail/GOST34.11-2012/Fedora_28/x86_64
[1] https://build.opensuse.org/package/live_build_log/home:k_mikhail/GOST34.11-2012/Fedora_28/i586
[2] https://build.opensuse.org/package/live_build_log/home:k_mikhail/GOST34.11-2012/Fedora_28/s390x
[3] https://build.opensuse.org/package/live_build_log/home:k_mikhail/GOST34.11-2012/Fedora_28/armv7l
[4] https://build.opensuse.org/package/live_build_log/home:k_mikhail/GOST34.11-2012/Fedora_28/aarch64
[5] https://build.opensuse.org/package/live_build_log/home:k_mikhail/GOST34.11-2012/Fedora_28/ppc64le

[adegtyarev]

Have you used a CLI tool for syntax checking?

I'd like to integrate that kind of checking to the CI/CD process with Travis. Any ideas?

[adegtyarev]

abb9919 merged via 3cbee1f with rebase of the source branch. Thanks!

[db4]

Once you're doing this, please also remove

streebog/gost3411-2012-core.c

Lines 119 to 120 in 8e41694

	/* Restore the Floating-point status on the CPU */
	_mm_empty();

This command is useless in SSE2 mode (it's only needed to switch from MMX to x87), and it becomes an undefined function when MSVC compiler is used.

[db4]

Also _mm_cvtsi64_si128 should not be used in 32-bit mode:

streebog/gost3411-2012-sse2.h

Lines 127 to 128 in 8e41694

	xmm4 = _mm_cvtsi64_si128((long long) r0); \
	tmm4 = _mm_cvtsi64_si128((long long) r1); \

that translates to 64-bit movq, that is only available in x64 mode. Somehow GCC is able to translate it to several SSE instructions, but that's clearly an extension (MSVC refuses to do that). So the code compiles for Win64 but fails for Win32.