Replace crc32 implementation

This is a properly constexpr version for modern c++. Also
added a fast runtime version of it.

src/crc32.cpp

@@ -0,0 +1,92 @@
+/* Copyright 2020-2022, Michele Santullo
+ * This file is part of wrenpp.
+ *
+ * Wrenpp is free software: you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation, either version 3 of the License, or
+ * (at your option) any later version.
+ *
+ * Wrenpp is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with wrenpp.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include "wrenpp/detail/crc32.hpp"
+#include "pvt_config.h"
+#if defined(WRENPP_WITH_SSE42)
+#	if defined(__aarch64__)
+#		include "sse2neon.h"
+#	else
+#		include <smmintrin.h>
+#		include <immintrin.h>
+#	endif
+#endif
+
+#if defined(WRENPP_WITH_SSE42)
+	// Byte-boundary alignment issues
+#	define CALC_CRC(op, crc, type, buf, len)							   \
+  do {																  \
+	for (; (len) >= sizeof (type); (len) -= sizeof(type), buf += sizeof (type)) { \
+	  (crc) = op((crc), *(type *) (buf));							   \
+	}																   \
+  } while(0)
+#endif
+
+namespace wren {
+	namespace {
+		//Some useful polynomials:
+		// zlib: 0x04C11DB7
+		// castagnoli (intel):  0x1EDC6F41
+
+#if defined(WRENPP_WITH_SSE42)
+		/* Compute CRC-32C using the Intel hardware instruction. */
+		/* for better parallelization with bigger buffers see
+		   http://www.drdobbs.com/parallel/fast-parallelized-crc-computation-using/229401411 */
+		std::uint32_t crc32c_hw(const void *input, std::size_t len, std::uint32_t crc)
+		{
+			//see https://github.com/rurban/smhasher/blob/master/crc32_hw.c
+			constexpr std::size_t align_size = alignof(std::uint64_t);
+			constexpr std::size_t align_mask = align_size - 1;
+
+			const char* buf = static_cast<const char*>(input);
+
+			// XOR the initial CRC with INT_MAX
+			crc ^= 0xFFFFFFFF;
+
+			// Align the input to the word boundary
+			for (; (len > 0) && (reinterpret_cast<std::uintptr_t>(buf) & align_mask); len--, buf++) {
+				crc = _mm_crc32_u8(crc, *buf);
+			}
+
+			// Blast off the CRC32 calculation
+#if defined(__x86_64__) || defined(__aarch64__)
+			CALC_CRC(_mm_crc32_u64, crc, std::uint64_t, buf, len);
+#endif
+			CALC_CRC(_mm_crc32_u32, crc, std::uint32_t, buf, len);
+			CALC_CRC(_mm_crc32_u16, crc, std::uint16_t, buf, len);
+			CALC_CRC(_mm_crc32_u8, crc, std::uint8_t, buf, len);
+
+			// Post-process the crc
+			return ~crc;
+		}
+#endif
+	} //unnamed namespace
+
+	[[gnu::const]]
+	std::uint32_t runtime_crc32c (const char* data, std::size_t size, std::uint32_t crc) {
+#if defined(WRENPP_WITH_SSE42)
+		//if (_may_i_use_cpu_feature(_FEATURE_SSE4_2)) {
+		if (__builtin_cpu_supports("sse4.2")) {
+			return crc32c_hw(data, size, crc);
+		}
+		else
+#endif
+		{
+			return detail::crc32<detail::g_castagnoli_polynomial>(data, size, crc);
+		}
+	}
+} //namespace wren

src/pvt_config.h.in

@@ -27,3 +27,5 @@ static_assert(sizeof(void*) == ASM_PTR_SIZE, "Build system reports an unexpected
 
 static_assert(sizeof(void(*)(int)) == ASM_FUNC_PTR_SIZE, "Build system reports an unexpected function pointer size, please ensure assembly code is correct");
 #endif
+
+#mesondefine WRENPP_WITH_SSE42