yencode 1.2.0 → 1.2.1

package/src/platform.cc

@@ -17,7 +17,7 @@
 #   endif
 #  endif
 # endif
-bool cpu_supports_neon() {
+bool RapidYenc::cpu_supports_neon() {
 # if defined(AT_HWCAP)
 #  ifdef __FreeBSD__
 	unsigned long supported;
@@ -95,7 +95,7 @@ static inline int _GET_XCR() {
 // }
 
 
-int cpu_supports_isa() {
+int RapidYenc::cpu_supports_isa() {
 	int flags[4];
 	_cpuid1(flags);
 	int ret = 0;
@@ -132,29 +132,30 @@ int cpu_supports_isa() {
 					int cpuInfo[4];
 					_cpuidX(cpuInfo, 7, 0);
 					if((cpuInfo[1] & 0x128) == 0x128 && (ret & ISA_FEATURE_LZCNT)) { // BMI2 + AVX2 + BMI1
-						
-						// check AVX10
-						int cpuInfo2[4];
-						_cpuidX(cpuInfo2, 7, 1);
-						if(cpuInfo2[3] & 0x80000) {
-							_cpuidX(cpuInfo2, 0x24, 0);
-							if((cpuInfo2[1] & 0xff) >= 1 && ( // minimum AVX10.1
+						if((xcr & 0xE0) == 0xE0) { // AVX512 XSTATE (also applies to AVX10)
+							// check AVX10
+							int cpuInfo2[4];
+							_cpuidX(cpuInfo2, 7, 1);
+							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
+									cpuInfo2[1] & 0x10000 // AVX10/128
 #else
-								cpuInfo2[1] & 0x20000 // AVX10/256
+									cpuInfo2[1] & 0x20000 // AVX10/256
 #endif
-							)) {
-								if(((xcr & 0xE0) == 0xE0) && (cpuInfo2[1] & 0x40000)) ret |= ISA_FEATURE_EVEX512;
-								return ret | ISA_LEVEL_VBMI2;
+								)) {
+									if(cpuInfo2[1] & 0x40000) ret |= ISA_FEATURE_EVEX512;
+									return ret | ISA_LEVEL_VBMI2;
+								}
+							}
+							
+							if((cpuInfo[1] & 0xC0010000) == 0xC0010000) { // AVX512BW + AVX512VL + AVX512F
+								ret |= ISA_FEATURE_EVEX512;
+								if(cpuInfo[2] & 0x40)
+									return ret | ISA_LEVEL_VBMI2;
+								return ret | ISA_LEVEL_AVX3;
 							}
-						}
-						
-						if(((xcr & 0xE0) == 0xE0) && (cpuInfo[1] & 0xC0010000) == 0xC0010000) { // AVX512BW + AVX512VL + AVX512F
-							ret |= ISA_FEATURE_EVEX512;
-							if(cpuInfo[2] & 0x40)
-								return ret | ISA_LEVEL_VBMI2;
-							return ret | ISA_LEVEL_AVX3;
 						}
 						// AVX2 is beneficial even on Zen1
 						return ret | ISA_LEVEL_AVX2;
@@ -169,7 +170,7 @@ int cpu_supports_isa() {
 	return ret | ISA_LEVEL_SSE2;
 }
 
-int cpu_supports_crc_isa() {
+int RapidYenc::cpu_supports_crc_isa() {
 	int flags[4];
 	_cpuid1(flags);
 	
@@ -200,7 +201,7 @@ int cpu_supports_crc_isa() {
 #   endif
 #  endif
 # endif
-bool cpu_supports_rvv() {
+bool RapidYenc::cpu_supports_rvv() {
 # if defined(AT_HWCAP)
 	unsigned long ret;
 #  ifdef __FreeBSD__

package/src/yencode.cc

@@ -11,6 +11,7 @@
 #include "crc.h"
 
 using namespace v8;
+using namespace RapidYenc;
 
 static void free_buffer(char* data, void* _size) {
 #if !NODE_VERSION_AT_LEAST(0, 11, 0)
@@ -129,7 +130,7 @@ FUNC(Encode) {
 	size_t dest_len = YENC_MAX_SIZE(arg_len, line_size);
 	
 	unsigned char *result = (unsigned char*) malloc(dest_len);
-	size_t len = do_encode(line_size, &col, (const unsigned char*)node::Buffer::Data(args[0]), result, arg_len, true);
+	size_t len = encode(line_size, &col, (const unsigned char*)node::Buffer::Data(args[0]), result, arg_len, true);
 	result = (unsigned char*)realloc(result, len);
 	MARK_EXT_MEM(len);
 	RETURN_VAL( NEW_BUFFER((char*)result, len, free_buffer, (void*)len) );
@@ -164,7 +165,7 @@ FUNC(EncodeTo) {
 	if(node::Buffer::Length(args[1]) < dest_len)
 		RETURN_ERROR("Destination buffer does not have enough space (use `maxSize` to compute required space)");
 	
-	size_t len = do_encode(line_size, &col, (const unsigned char*)node::Buffer::Data(args[0]), (unsigned char*)node::Buffer::Data(args[1]), arg_len, true);
+	size_t len = encode(line_size, &col, (const unsigned char*)node::Buffer::Data(args[0]), (unsigned char*)node::Buffer::Data(args[1]), arg_len, true);
 	RETURN_VAL( Integer::New(ISOLATE len) );
 }
 
@@ -220,7 +221,7 @@ FUNC(EncodeIncr) {
 		result = (unsigned char*) malloc(dest_len);
 	}
 	
-	size_t len = do_encode(line_size, &col, (const unsigned char*)node::Buffer::Data(args[0]), result, arg_len, false);
+	size_t len = encode(line_size, &col, (const unsigned char*)node::Buffer::Data(args[0]), result, arg_len, false);
 	
 	SET_OBJ(ret, "written", Integer::New(ISOLATE len));
 	if(allocResult) {
@@ -247,7 +248,7 @@ FUNC(Decode) {
 		isRaw = ARG_TO_BOOL(args[1]);
 	
 	unsigned char *result = (unsigned char*) malloc(arg_len);
-	size_t len = do_decode(isRaw, (const unsigned char*)node::Buffer::Data(args[0]), result, arg_len, NULL);
+	size_t len = decode(isRaw, (const unsigned char*)node::Buffer::Data(args[0]), result, arg_len, NULL);
 	result = (unsigned char*)realloc(result, len);
 	MARK_EXT_MEM(len);
 	RETURN_VAL( NEW_BUFFER((char*)result, len, free_buffer, (void*)len) );
@@ -271,7 +272,7 @@ FUNC(DecodeTo) {
 	if (args.Length() > 2)
 		isRaw = ARG_TO_BOOL(args[2]);
 	
-	size_t len = do_decode(isRaw, (const unsigned char*)node::Buffer::Data(args[0]), (unsigned char*)node::Buffer::Data(args[1]), arg_len, NULL);
+	size_t len = decode(isRaw, (const unsigned char*)node::Buffer::Data(args[0]), (unsigned char*)node::Buffer::Data(args[1]), arg_len, NULL);
 	RETURN_VAL( Integer::New(ISOLATE len) );
 }
 
@@ -310,7 +311,7 @@ FUNC(DecodeIncr) {
 	
 	if(allocResult) result = (unsigned char*) malloc(arg_len);
 	unsigned char* dp = result;
-	YencDecoderEnd ended = do_decode_end(&sp, &dp, arg_len, &state);
+	YencDecoderEnd ended = RapidYenc::decode_end((const void**)&sp, (void**)&dp, arg_len, &state);
 	size_t len = dp - result;
 	if(allocResult) result = (unsigned char*)realloc(result, len);
 	
@@ -363,7 +364,7 @@ FUNC(CRC32) {
 			RETURN_ERROR("Second argument must be a 4 byte buffer");
 		crc = read_crc32(args[1]);
 	}
-	crc = do_crc32(
+	crc = crc32(
 		(const void*)node::Buffer::Data(args[0]),
 		node::Buffer::Length(args[0]),
 		crc
@@ -447,7 +448,7 @@ FUNC(CRC32Shift) {
 static void init_all() {
 	encoder_init();
 	decoder_init();
-	crc_init();	
+	crc32_init();	
 }
 
 #if NODE_VERSION_AT_LEAST(10, 7, 0)

package/test/_speedbase.js

@@ -81,12 +81,14 @@ module.exports = {
 			module.exports._benchAsync(fn, cb, trials, []);
 		}, asyncWait);
 	},
-	run: function(name, fn, sz2) {
+	run: function(name, fn, sz2, fn2) {
 		var time = module.exports.bench(fn);
+		var time2 = fn2 ? module.exports.bench(fn2) : null;
 		console.log(
 			(name+'                         ').substring(0, 25) + ':'
 			+ fmtSpeed(sz*rounds, time)
-			+ (sz2 ? (' ' + fmtSpeed(sz2*rounds, time)) : '')
+			+ ' ' + (sz2 ? fmtSpeed(sz2*rounds, time) : '                 ')
+			+ ' ' + (fn2 ? fmtSpeed(sz*rounds, time2) : '                 ')
 		);
 	},
 	

package/test/speeddec.js

@@ -30,7 +30,7 @@ _.bufAvg2x.forEach(function(buf, i) {
 
 _.parseArgs('Syntax: node test/speeddec [-a|--average-only] [{-s|--sleep}=msecs(0)] [{-m|--methods}=clean,raw,rawincr]');
 
-console.log('    Test                     Output rate         Read rate   ');
+console.log('    Test                     Output rate         Read rate      Output + CRC   ');
 
 // warmup
 if(!_.sleep) {
@@ -43,51 +43,60 @@ if(!_.sleep) {
 	});
 }
 
+var decCrc = function(src, raw) {
+	var len = y.decodeTo(src, _.bufTarget, raw);
+	y.crc32(_.bufTarget.slice(0, len));
+};
+var decIncrCrc = function(src) {
+	var out = y.decodeIncr(src, 0, _.bufTarget);
+	y.crc32(_.bufTarget.slice(0, out.written));
+};
+
 setTimeout(function() {
 	if(!_.avgOnly) {
 		if(_.decMethods.clean) {
-			_.run('Clean worst (all escaping)', y.decodeTo.bind(null, mWorst, _.bufTarget), lenWorst);
-			_.run('Clean best (min escaping)',  y.decodeTo.bind(null, mBest, _.bufTarget), lenBest);
-			_.run('Clean pass (no escaping)',  y.decodeTo.bind(null, mBest2, _.bufTarget));
+			_.run('Clean worst (all escaping)', y.decodeTo.bind(null, mWorst, _.bufTarget), lenWorst, decCrc.bind(null, mWorst));
+			_.run('Clean best (min escaping)',  y.decodeTo.bind(null, mBest, _.bufTarget), lenBest, decCrc.bind(null, mBest));
+			_.run('Clean pass (no escaping)',  y.decodeTo.bind(null, mBest2, _.bufTarget), null, decCrc.bind(null, mBest2));
 		}
 		if(_.decMethods.raw) {
-			_.run('Raw worst', y.decodeTo.bind(null, mWorst, _.bufTarget, true), lenWorst);
-			_.run('Raw best',  y.decodeTo.bind(null, mBest, _.bufTarget, true), lenBest);
-			_.run('Raw pass',  y.decodeTo.bind(null, mBest2, _.bufTarget, true));
+			_.run('Raw worst', y.decodeTo.bind(null, mWorst, _.bufTarget, true), lenWorst, decCrc.bind(null, mWorst, true));
+			_.run('Raw best',  y.decodeTo.bind(null, mBest, _.bufTarget, true), lenBest, decCrc.bind(null, mBest, true));
+			_.run('Raw pass',  y.decodeTo.bind(null, mBest2, _.bufTarget, true), null, decCrc.bind(null, mBest2, true));
 		}
 		if(_.decMethods.rawincr) {
-			_.run('Raw-incr worst', y.decodeIncr.bind(null, mWorst, 0, _.bufTarget), lenWorst);
-			_.run('Raw-incr best',  y.decodeIncr.bind(null, mBest, 0, _.bufTarget), lenBest);
-			_.run('Raw-incr pass',  y.decodeIncr.bind(null, mBest2, 0, _.bufTarget));
+			_.run('Raw-incr worst', y.decodeIncr.bind(null, mWorst, 0, _.bufTarget), lenWorst, decIncrCrc.bind(null, mWorst));
+			_.run('Raw-incr best',  y.decodeIncr.bind(null, mBest, 0, _.bufTarget), lenBest, decIncrCrc.bind(null, mBest));
+			_.run('Raw-incr pass',  y.decodeIncr.bind(null, mBest2, 0, _.bufTarget), null, decIncrCrc.bind(null, mBest2));
 		}
 	}
 	
 	if(_.decMethods.clean)
 		mAvg.forEach(function(buf, i) {
-			_.run('Clean random ('+i+')',   y.decodeTo.bind(null, buf, _.bufTarget), lenAvg[i]);
+			_.run('Clean random ('+i+')',   y.decodeTo.bind(null, buf, _.bufTarget), lenAvg[i], decCrc.bind(null, buf));
 		});
 	if(_.decMethods.raw)
 		mAvg.forEach(function(buf, i) {
-			_.run('Raw random ('+i+')',   y.decodeTo.bind(null, buf, _.bufTarget, true), lenAvg[i]);
+			_.run('Raw random ('+i+')',   y.decodeTo.bind(null, buf, _.bufTarget, true), lenAvg[i], decCrc.bind(null, buf, true));
 		});
 	if(_.decMethods.rawincr) {
 		mAvg.forEach(function(buf, i) {
-			_.run('Raw-incr random ('+i+')',  y.decodeIncr.bind(null, buf, 0, _.bufTarget), lenAvg[i]);
+			_.run('Raw-incr random ('+i+')',  y.decodeIncr.bind(null, buf, 0, _.bufTarget), lenAvg[i], decIncrCrc.bind(null, buf));
 		});
 	}
 	
 	if(!_.avgOnly) {
 		if(_.decMethods.clean)
 			mAvg2x.forEach(function(buf, i) {
-				_.run('Clean random 2xEsc ('+i+')',   y.decodeTo.bind(null, buf, _.bufTarget), lenAvg2x[i]);
+				_.run('Clean random 2xEsc ('+i+')',   y.decodeTo.bind(null, buf, _.bufTarget), lenAvg2x[i], decCrc.bind(null, buf));
 			});
 		if(_.decMethods.raw)
 			mAvg2x.forEach(function(buf, i) {
-				_.run('Raw random 2xEsc ('+i+')',   y.decodeTo.bind(null, buf, _.bufTarget, true), lenAvg2x[i]);
+				_.run('Raw random 2xEsc ('+i+')',   y.decodeTo.bind(null, buf, _.bufTarget, true), lenAvg2x[i], decCrc.bind(null, buf, true));
 			});
 		if(_.decMethods.rawincr)
 			mAvg2x.forEach(function(buf, i) {
-				_.run('Raw-incr random 2xEsc ('+i+')',  y.decodeIncr.bind(null, buf, 0, _.bufTarget), lenAvg2x[i]);
+				_.run('Raw-incr random 2xEsc ('+i+')',  y.decodeIncr.bind(null, buf, 0, _.bufTarget), lenAvg2x[i], decIncrCrc.bind(null, buf));
 			});
 	}
 }, _.sleep);

package/test/speedenc.js

@@ -5,7 +5,7 @@ var _ = require('./_speedbase');
 _.parseArgs('Syntax: node test/speedenc [-a|--average-only] [{-s|--sleep}=msecs(0)]');
 
 
-console.log('    Test                       Read rate       Output rate   ');
+console.log('    Test                       Read rate       Output rate        Read + CRC      ');
 
 var lenWorst = y.encodeTo(_.bufWorst, _.bufTarget);
 var lenBest = y.encodeTo(_.bufBest, _.bufTarget);
@@ -13,31 +13,35 @@ var lenAvg = Array(_.bufAvg.length);
 var lenAvg2x = Array(_.bufAvg2x.length);
 
 // warmup
-if(!_.sleep) {
-	_.bufAvg.forEach(function(buf, i) {
-		var p=process.hrtime();
-		for(var j=0;j<_.rounds;j+=1) lenAvg[i] = y.encodeTo(buf, _.bufTarget);
-		var t=process.hrtime(p);
-	});
-	_.bufAvg2x.forEach(function(buf, i) {
-		var p=process.hrtime();
-		for(var j=0;j<_.rounds;j+=1) lenAvg2x[i] = y.encodeTo(buf, _.bufTarget);
-		var t=process.hrtime(p);
-	});
-}
+var initRounds = _.sleep ? 1 : _.rounds;
+_.bufAvg.forEach(function(buf, i) {
+	var p=process.hrtime();
+	for(var j=0;j<initRounds;j+=1) lenAvg[i] = y.encodeTo(buf, _.bufTarget);
+	var t=process.hrtime(p);
+});
+_.bufAvg2x.forEach(function(buf, i) {
+	var p=process.hrtime();
+	for(var j=0;j<initRounds;j+=1) lenAvg2x[i] = y.encodeTo(buf, _.bufTarget);
+	var t=process.hrtime(p);
+});
+
+var encCrc = function(src) {
+	y.encodeTo(src, _.bufTarget);
+	y.crc32(src);
+};
 
 setTimeout(function() {
 	if(!_.avgOnly) {
-		_.run('Worst (all escaping)', y.encodeTo.bind(null, _.bufWorst, _.bufTarget), lenWorst);
-		_.run('Best (no escaping)', y.encodeTo.bind(null, _.bufBest, _.bufTarget), lenBest);
+		_.run('Worst (all escaping)', y.encodeTo.bind(null, _.bufWorst, _.bufTarget), lenWorst, encCrc.bind(null, _.bufWorst));
+		_.run('Best (no escaping)', y.encodeTo.bind(null, _.bufBest, _.bufTarget), lenBest, encCrc.bind(null, _.bufBest));
 	}
 	
 	_.bufAvg.forEach(function(buf, i) {
-		_.run('Random ('+i+')', y.encodeTo.bind(null, buf, _.bufTarget), lenAvg[i]);
+		_.run('Random ('+i+')', y.encodeTo.bind(null, buf, _.bufTarget), lenAvg[i], encCrc.bind(null, buf));
 	});
 	if(!_.avgOnly) {
 		_.bufAvg2x.forEach(function(buf, i) {
-			_.run('Random 2xEsc ('+i+')', y.encodeTo.bind(null, buf, _.bufTarget), lenAvg2x[i]);
+			_.run('Random 2xEsc ('+i+')', y.encodeTo.bind(null, buf, _.bufTarget), lenAvg2x[i], encCrc.bind(null, buf));
 		});
 	}