yencode 1.2.1 → 1.2.3

package/binding.gyp

@@ -50,7 +50,7 @@
       }],
       ['OS!="win"', {
         "variables": {
-          "missing_memalign%": "<!(<!(echo ${CC_target:-${CC:-cc}}) -c src/test_alignalloc.c -o /dev/null -Werror 2>/dev/null || echo failed)",
+          "missing_memalign%": "<!(<!(echo ${CXX_target:-${CXX:-c++}}) -c src/test_alignalloc.cc -o /dev/null -Werror 2>/dev/null || echo failed)",
         },
         "conditions": [
           ['missing_memalign!=""', {
@@ -60,7 +60,7 @@
       }]
     ],
     "cflags": ["-Wno-unused-function"],
-    "cxxflags": ["-Wno-unused-function", "-std=c++03", "-D_POSIX_C_SOURCE=200112L"],
+    "cxxflags": ["-Wno-unused-function", "-std=c++03"],
     "xcode_settings": {
       "OTHER_CFLAGS": ["-Wno-unused-function"],
       "OTHER_CXXFLAGS": ["-Wno-unused-function"]

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "yencode",
-  "version": "1.2.1",
+  "version": "1.2.3",
   "description": "SIMD accelerated yEnc encoder/decoder and CRC32 calculator",
   "keywords": [
     "yenc",

package/src/crc.cc

@@ -206,7 +206,7 @@ namespace RapidYenc {
 # elif defined(__has_include)
 #  if __has_include(<sys/auxv.h>)
 #   include <sys/auxv.h>
-#   ifdef __FreeBSD__
+#   if defined(__FreeBSD__) || defined(__OpenBSD__)
 static unsigned long getauxval(unsigned long cap) {
 	unsigned long ret;
 	elf_aux_info(cap, &ret, sizeof(ret));

package/src/crc.h

@@ -23,7 +23,7 @@ static inline int crc32_isa_level() {
 
 // computes `n % 0xffffffff` (well, almost), using some bit-hacks
 static inline uint32_t crc32_powmod(uint64_t n) {
-#ifdef __GNUC__
+#if defined(__GNUC__) && (__GNUC__ >= 5 || (defined(__clang__) && (__clang_major__ > 3 || (__clang_major__ == 3 && __clang_minor__ > 3))))
 	unsigned res;
 	unsigned carry = __builtin_uadd_overflow(n >> 32, n, &res);
 	res += carry;

package/src/crc_arm_pmull.cc

@@ -207,7 +207,7 @@ static uint32_t crc32_shift_pmull(uint32_t crc1, uint32_t n) {
 void RapidYenc::crc_pmull_set_funcs() {
 	_crc32_multiply = &crc32_multiply_pmull;
 	_crc32_shift = &crc32_shift_pmull;
-	_crc32_isa &= ISA_FEATURE_PMULL;
+	_crc32_isa |= ISA_FEATURE_PMULL;
 }
 
 #else

package/src/decoder_rvv.cc

@@ -209,7 +209,8 @@ HEDLEY_ALWAYS_INLINE void do_decode_rvv(const uint8_t* src, long& len, unsigned
 			if(LIKELIHOOD(0.0001, RV(vcpop_m_b4)(RV(vmandn_mm_b4)(cmpEqShift1, cmp, vl2), vl2) != 0)) {
 				// replicate fix_eqMask, but in vector form
 				vbool4_t groupStart = RV(vmandn_mm_b4)(cmpEq, cmpEqShift1, vl2);
-				vbool4_t evenBits = RV_MASK_CAST(4, 8, RV(vmv_v_x_u8m1)(0x55, vl2));
+				vuint8m1_t evenBitsV = RV(vmv_v_x_u8m1)(0x55, vl2);
+				vbool4_t evenBits = RV_MASK_CAST(4, 8, evenBitsV);
 				vbool4_t evenStart = RV(vmand_mm_b4)(groupStart, evenBits, vl2);
 				
 				// compute `cmpEq + evenStart` to obtain oddGroups

package/src/encoder_avx_base.h

@@ -217,12 +217,12 @@ HEDLEY_ALWAYS_INLINE void do_encode_avx2(int line_size, int* colOffset, const ui
 				// duplicate halves
 				data1A = _mm256_inserti128_si256(dataA, _mm256_castsi256_si128(dataA), 1);
 				data1B = _mm256_inserti128_si256(dataB, _mm256_castsi256_si128(dataB), 1);
-#if defined(__tune_znver2__) || defined(__tune_znver3__) || defined(__tune_znver4__)
-				data2A = _mm256_permute2x128_si256(dataA, dataA, 0x11);
-				data2B = _mm256_permute2x128_si256(dataB, dataB, 0x11);
-#else
+#if defined(__tune_znver1__) || defined(__tune_bdver4__)
 				data2A = _mm256_permute4x64_epi64(dataA, 0xee);
 				data2B = _mm256_permute4x64_epi64(dataB, 0xee);
+#else
+				data2A = _mm256_permute2x128_si256(dataA, dataA, 0x11);
+				data2B = _mm256_permute2x128_si256(dataB, dataB, 0x11);
 #endif
 				
 				shuf1A = _mm256_load_si256(lookupsAVX2->shufExpand + m1);

package/src/encoder_sse_base.h

@@ -351,7 +351,7 @@ HEDLEY_ALWAYS_INLINE void do_encode_sse(int line_size, int* colOffset, const uin
 #if defined(__POPCNT__) && !defined(__tune_btver1__)
 				if(use_isa & ISA_FEATURE_POPCNT) {
 					shuf2Len = popcnt32(maskA) + 16;
-# if defined(__tune_znver4__) || defined(__tune_znver3__) || defined(__tune_znver2__) || defined(__tune_znver1__) || defined(__tune_btver2__)
+# if defined(__tune_znver6__) || defined(__tune_znver5__) || defined(__tune_znver4__) || defined(__tune_znver3__) || defined(__tune_znver2__) || defined(__tune_znver1__) || defined(__tune_btver2__)
 					shuf1Len = popcnt32(m1) + 8;
 					shuf3Len = popcnt32(m3) + shuf2Len + 8;
 # else

package/src/platform.cc

@@ -19,7 +19,7 @@
 # endif
 bool RapidYenc::cpu_supports_neon() {
 # if defined(AT_HWCAP)
-#  ifdef __FreeBSD__
+#  if defined(__FreeBSD__) || defined(__OpenBSD__)
 	unsigned long supported;
 	elf_aux_info(AT_HWCAP, &supported, sizeof(supported));
 #   ifdef __aarch64__
@@ -139,11 +139,7 @@ int RapidYenc::cpu_supports_isa() {
 							if(cpuInfo2[3] & 0x80000) {
 								_cpuidX(cpuInfo2, 0x24, 0);
 								if((cpuInfo2[1] & 0xff) >= 1 && ( // minimum AVX10.1
-#ifdef YENC_DISABLE_AVX256
-									cpuInfo2[1] & 0x10000 // AVX10/128
-#else
-									cpuInfo2[1] & 0x20000 // AVX10/256
-#endif
+									cpuInfo2[1] & 0x20000 // AVX10/256 (AVX10/128 is now invalid)
 								)) {
 									if(cpuInfo2[1] & 0x40000) ret |= ISA_FEATURE_EVEX512;
 									return ret | ISA_LEVEL_VBMI2;
@@ -204,7 +200,7 @@ int RapidYenc::cpu_supports_crc_isa() {
 bool RapidYenc::cpu_supports_rvv() {
 # if defined(AT_HWCAP)
 	unsigned long ret;
-#  ifdef __FreeBSD__
+#  if defined(__FreeBSD__) || defined(__OpenBSD__)
 	elf_aux_info(AT_HWCAP, &ret, sizeof(ret));
 #  else
 	ret = getauxval(AT_HWCAP);