Improved SHA-1 performance with Rusha library

app/index.html

@@ -56,6 +56,7 @@
 
   <script type="text/javascript" src="vendor/jsbn/jsbn_combined.js"></script>
   <script type="text/javascript" src="vendor/cryptoJS/crypto.js"></script>
+  <script type="text/javascript" src="vendor/rusha/rusha.js"></script>
   <script type="text/javascript" src="vendor/zlib/gunzip.min.js"></script>
   <script type="text/javascript" src="vendor/closure/long.js"></script>
 

app/js/lib/bin_utils.js

@@ -196,9 +196,10 @@ function uintToInt (val) {
 }
 
 function sha1Hash (bytes) {
-  // console.log('SHA-1 hash start');
+  this.rushaInstance = this.rushaInstance || new Rusha(1024 * 1024);
-  var hashBytes = sha1.hash(bytes, true);
+  // console.log(dT(), 'SHA-1 hash start', bytes.byteLength || bytes.length);
-  // console.log('SHA-1 hash finish');
+  var hashBytes = bytesFromArrayBuffer(rushaInstance.rawDigest(bytes).buffer);
+  // console.log(dT(), 'SHA-1 hash finish');
 
   return hashBytes;
 }

app/js/lib/crypto_worker.js

@@ -11,7 +11,8 @@ importScripts(
   '../../vendor/jsbn/jsbn_combined.js',
   '../../vendor/leemon_bigint/bigint.js',
   '../../vendor/closure/long.js',
-  '../../vendor/cryptoJS/crypto.js'
+  '../../vendor/cryptoJS/crypto.js',
+  '../../vendor/rusha/rusha.js'
 );
 
 onmessage = function (e) {

app/js/lib/tl_utils.js

@@ -36,7 +36,16 @@ TLSerialization.prototype.getBuffer = function () {
   return this.getArray().buffer;
 };
 
-TLSerialization.prototype.getBytes = function () {
+TLSerialization.prototype.getBytes = function (typed) {
+  if (typed) {
+    var resultBuffer = new ArrayBuffer(this.offset);
+    var resultArray  = new Uint8Array(resultBuffer);
+
+    resultArray.set(this.byteView.subarray(0, this.offset));
+
+    return resultArray;
+  }
+
   var bytes = [];
   for (var i = 0; i < this.offset; i++) {
     bytes.push(this.byteView[i]);
@@ -141,9 +150,8 @@ TLSerialization.prototype.storeString = function (s, field) {
 TLSerialization.prototype.storeBytes = function (bytes, field) {
   this.debug && console.log('>>>', bytesToHex(bytes), (field || '') + ':bytes');
 
-  this.checkLength(bytes.length + 8);
+  var len = bytes.byteLength || bytes.length;
-
+  this.checkLength(len + 8);
-  var len = bytes.length;
   if (len <= 253) {
     this.byteView[this.offset++] = len;
   } else {
@@ -152,9 +160,8 @@ TLSerialization.prototype.storeBytes = function (bytes, field) {
     this.byteView[this.offset++] = (len & 0xFF00) >> 8;
     this.byteView[this.offset++] = (len & 0xFF0000) >> 16;
   }
-  for (var i = 0; i < len; i++) {
+  this.byteView.set(bytes, this.offset);
-    this.byteView[this.offset++] = bytes[i];
+  this.offset += len;
-  }
 
   // Padding
   while (this.offset % 4) {
@@ -177,14 +184,13 @@ TLSerialization.prototype.storeIntBytes = function (bytes, bits, field) {
 };
 
 TLSerialization.prototype.storeRawBytes = function (bytes, field) {
-  var len = bytes.length;
+  var len = bytes.byteLength || bytes.length;
 
   this.debug && console.log('>>>', bytesToHex(bytes), (field || ''));
   this.checkLength(len);
 
-  for (var i = 0; i < len; i++) {
+  this.byteView.set(bytes, this.offset);
-    this.byteView[this.offset++] = bytes[i];
+  this.offset += len;
-  }
 };
 
 

app/vendor/rusha/rusha.js

@@ -0,0 +1,392 @@
+/*
+ * Rusha, a JavaScript implementation of the Secure Hash Algorithm, SHA-1,
+ * as defined in FIPS PUB 180-1, tuned for high performance with large inputs.
+ * (http://github.com/srijs/rusha)
+ *
+ * Inspired by Paul Johnstons implementation (http://pajhome.org.uk/crypt/md5).
+ *
+ * Copyright (c) 2013 Sam Rijs (http://awesam.de).
+ * Released under the terms of the MIT license as follows:
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+(function (global) {
+    // If we'e running in Node.JS, export a module.
+    if (typeof module !== 'undefined') {
+        module.exports = Rusha;
+    }
+    // If we're running in a DOM context, export
+    // the Rusha object to toplevel.
+    if (typeof global !== 'undefined') {
+        global.Rusha = Rusha;
+    }
+    // If we're running in a webworker, accept
+    // messages containing a jobid and a buffer
+    // or blob object, and return the hash result.
+    if (typeof FileReaderSync !== 'undefined') {
+        var reader = new FileReaderSync(), hasher = new Rusha(4 * 1024 * 1024);
+        self.onmessage = function onMessage(event) {
+            var hash, data = event.data.data;
+            if (data instanceof Blob) {
+                try {
+                    data = reader.readAsBinaryString(data);
+                } catch (e) {
+                    self.postMessage({
+                        id: event.data.id,
+                        error: e.name
+                    });
+                    return;
+                }
+            }
+            hash = hasher.digest(data);
+            self.postMessage({
+                id: event.data.id,
+                hash: hash
+            });
+        };
+    }
+    var util = {
+            getDataType: function (data) {
+                if (typeof data === 'string') {
+                    return 'string';
+                }
+                if (data instanceof Array) {
+                    return 'array';
+                }
+                if (typeof global !== 'undefined' && global.Buffer && global.Buffer.isBuffer(data)) {
+                    return 'buffer';
+                }
+                if (data instanceof ArrayBuffer) {
+                    return 'arraybuffer';
+                }
+                if (data.buffer instanceof ArrayBuffer) {
+                    return 'view';
+                }
+                throw new Error('Unsupported data type.');
+            }
+        };
+    // The Rusha object is a wrapper around the low-level RushaCore.
+    // It provides means of converting different inputs to the
+    // format accepted by RushaCore as well as other utility methods.
+    function Rusha(chunkSize) {
+        'use strict';
+        // Private object structure.
+        var self$2 = { fill: 0 };
+        // Calculate the length of buffer that the sha1 routine uses
+        // including the padding.
+        var padlen = function (len) {
+            for (len += 9; len % 64 > 0; len += 1);
+            return len;
+        };
+        var padZeroes = function (bin, len) {
+            for (var i = len >> 2; i < bin.length; i++)
+                bin[i] = 0;
+        };
+        var padData = function (bin, chunkLen, msgLen) {
+            bin[chunkLen >> 2] |= 128 << 24 - (chunkLen % 4 << 3);
+            bin[((chunkLen >> 2) + 2 & ~15) + 15] = msgLen << 3;
+        };
+        // Convert a binary string and write it to the heap.
+        // A binary string is expected to only contain char codes < 256.
+        var convStr = function (H8, H32, start, len, off) {
+            var str = this, i, om = off % 4, lm = len % 4, j = len - lm;
+            if (j > 0) {
+                switch (om) {
+                case 0:
+                    H8[off + 3 | 0] = str.charCodeAt(start);
+                case 1:
+                    H8[off + 2 | 0] = str.charCodeAt(start + 1);
+                case 2:
+                    H8[off + 1 | 0] = str.charCodeAt(start + 2);
+                case 3:
+                    H8[off | 0] = str.charCodeAt(start + 3);
+                }
+            }
+            for (i = om; i < j; i = i + 4 | 0) {
+                H32[off + i >> 2] = str.charCodeAt(start + i) << 24 | str.charCodeAt(start + i + 1) << 16 | str.charCodeAt(start + i + 2) << 8 | str.charCodeAt(start + i + 3);
+            }
+            switch (lm) {
+            case 3:
+                H8[off + j + 1 | 0] = str.charCodeAt(start + j + 2);
+            case 2:
+                H8[off + j + 2 | 0] = str.charCodeAt(start + j + 1);
+            case 1:
+                H8[off + j + 3 | 0] = str.charCodeAt(start + j);
+            }
+        };
+        // Convert a buffer or array and write it to the heap.
+        // The buffer or array is expected to only contain elements < 256.
+        var convBuf = function (H8, H32, start, len, off) {
+            var buf = this, i, om = off % 4, lm = len % 4, j = len - lm;
+            if (j > 0) {
+                switch (om) {
+                case 0:
+                    H8[off + 3 | 0] = buf[start];
+                case 1:
+                    H8[off + 2 | 0] = buf[start + 1];
+                case 2:
+                    H8[off + 1 | 0] = buf[start + 2];
+                case 3:
+                    H8[off | 0] = buf[start + 3];
+                }
+            }
+            for (i = 4 - om; i < j; i = i += 4 | 0) {
+                H32[off + i >> 2] = buf[start + i] << 24 | buf[start + i + 1] << 16 | buf[start + i + 2] << 8 | buf[start + i + 3];
+            }
+            switch (lm) {
+            case 3:
+                H8[off + j + 1 | 0] = buf[start + j + 2];
+            case 2:
+                H8[off + j + 2 | 0] = buf[start + j + 1];
+            case 1:
+                H8[off + j + 3 | 0] = buf[start + j];
+            }
+        };
+        var convFn = function (data) {
+            switch (util.getDataType(data)) {
+            case 'string':
+                return convStr.bind(data);
+            case 'array':
+                return convBuf.bind(data);
+            case 'buffer':
+                return convBuf.bind(data);
+            case 'arraybuffer':
+                return convBuf.bind(new Uint8Array(data));
+            case 'view':
+                return convBuf.bind(new Uint8Array(data.buffer));
+            }
+        };
+        var slice = function (data, offset) {
+            switch (util.getDataType(data)) {
+            case 'string':
+                return data.slice(offset);
+            case 'array':
+                return data.slice(offset);
+            case 'buffer':
+                return data.slice(offset);
+            case 'arraybuffer':
+                return data.slice(offset);
+            case 'view':
+                return data.buffer.slice(offset);
+            }
+        };
+        // Convert an ArrayBuffer into its hexadecimal string representation.
+        var hex = function (arrayBuffer) {
+            var i, x, hex_tab = '0123456789abcdef', res = [], binarray = new Uint8Array(arrayBuffer);
+            for (i = 0; i < binarray.length; i++) {
+                x = binarray[i];
+                res[i] = hex_tab.charAt(x >> 4 & 15) + hex_tab.charAt(x >> 0 & 15);
+            }
+            return res.join('');
+        };
+        var ceilHeapSize = function (v) {
+            // The asm.js spec says:
+            // The heap object's byteLength must be either
+            // 2^n for n in [12, 24) or 2^24 * n for n ≥ 1.
+            // Also, byteLengths smaller than 2^16 are deprecated.
+            var p;
+            // If v is smaller than 2^16, the smallest possible solution
+            // is 2^16.
+            if (v <= 65536)
+                return 65536;
+            // If v < 2^24, we round up to 2^n,
+            // otherwise we round up to 2^24 * n.
+            if (v < 16777216) {
+                for (p = 1; p < v; p = p << 1);
+            } else {
+                for (p = 16777216; p < v; p += 16777216);
+            }
+            return p;
+        };
+        // Initialize the internal data structures to a new capacity.
+        var init = function (size) {
+            if (size % 64 > 0) {
+                throw new Error('Chunk size must be a multiple of 128 bit');
+            }
+            self$2.maxChunkLen = size;
+            self$2.padMaxChunkLen = padlen(size);
+            // The size of the heap is the sum of:
+            // 1. The padded input message size
+            // 2. The extended space the algorithm needs (320 byte)
+            // 3. The 160 bit state the algoritm uses
+            self$2.heap = new ArrayBuffer(ceilHeapSize(self$2.padMaxChunkLen + 320 + 20));
+            self$2.h32 = new Int32Array(self$2.heap);
+            self$2.h8 = new Int8Array(self$2.heap);
+            self$2.core = RushaCore({
+                Int32Array: Int32Array,
+                DataView: DataView
+            }, {}, self$2.heap);
+            self$2.buffer = null;
+        };
+        // Iinitializethe datastructures according
+        // to a chunk siyze.
+        init(chunkSize || 64 * 1024);
+        var initState = function (heap, padMsgLen) {
+            var io = new Int32Array(heap, padMsgLen + 320, 5);
+            io[0] = 1732584193;
+            io[1] = -271733879;
+            io[2] = -1732584194;
+            io[3] = 271733878;
+            io[4] = -1009589776;
+        };
+        var padChunk = function (chunkLen, msgLen) {
+            var padChunkLen = padlen(chunkLen);
+            var view = new Int32Array(self$2.heap, 0, padChunkLen >> 2);
+            padZeroes(view, chunkLen);
+            padData(view, chunkLen, msgLen);
+            return padChunkLen;
+        };
+        // Write data to the heap.
+        var write = function (data, chunkOffset, chunkLen) {
+            convFn(data)(self$2.h8, self$2.h32, chunkOffset, chunkLen, 0);
+        };
+        // Initialize and call the RushaCore,
+        // assuming an input buffer of length len * 4.
+        var coreCall = function (data, chunkOffset, chunkLen, msgLen, finalize) {
+            var padChunkLen = chunkLen;
+            if (finalize) {
+                padChunkLen = padChunk(chunkLen, msgLen);
+            }
+            write(data, chunkOffset, chunkLen);
+            self$2.core.hash(padChunkLen, self$2.padMaxChunkLen);
+        };
+        var getRawDigest = function (heap, padMaxChunkLen) {
+            var io = new Int32Array(heap, padMaxChunkLen + 320, 5);
+            var out = new Int32Array(5);
+            var arr = new DataView(out.buffer);
+            arr.setInt32(0, io[0], false);
+            arr.setInt32(4, io[1], false);
+            arr.setInt32(8, io[2], false);
+            arr.setInt32(12, io[3], false);
+            arr.setInt32(16, io[4], false);
+            return out;
+        };
+        // Calculate the hash digest as an array of 5 32bit integers.
+        var rawDigest = this.rawDigest = function (str) {
+                var msgLen = str.byteLength || str.length;
+                initState(self$2.heap, self$2.padMaxChunkLen);
+                var chunkOffset = 0, chunkLen = self$2.maxChunkLen, last;
+                for (chunkOffset = 0; msgLen > chunkOffset + chunkLen; chunkOffset += chunkLen) {
+                    coreCall(str, chunkOffset, chunkLen, msgLen, false);
+                }
+                coreCall(str, chunkOffset, msgLen - chunkOffset, msgLen, true);
+                return getRawDigest(self$2.heap, self$2.padMaxChunkLen);
+            };
+        // The digest and digestFrom* interface returns the hash digest
+        // as a hex string.
+        this.digest = this.digestFromString = this.digestFromBuffer = this.digestFromArrayBuffer = function (str) {
+            return hex(rawDigest(str).buffer);
+        };
+    }
+    ;
+    // The low-level RushCore module provides the heart of Rusha,
+    // a high-speed sha1 implementation working on an Int32Array heap.
+    // At first glance, the implementation seems complicated, however
+    // with the SHA1 spec at hand, it is obvious this almost a textbook
+    // implementation that has a few functions hand-inlined and a few loops
+    // hand-unrolled.
+    function RushaCore(stdlib, foreign, heap) {
+        'use asm';
+        var H = new stdlib.Int32Array(heap);
+        function hash(k, x) {
+            // k in bytes
+            k = k | 0;
+            x = x | 0;
+            var i = 0, j = 0, y0 = 0, z0 = 0, y1 = 0, z1 = 0, y2 = 0, z2 = 0, y3 = 0, z3 = 0, y4 = 0, z4 = 0, t0 = 0, t1 = 0;
+            y0 = H[x + 320 >> 2] | 0;
+            y1 = H[x + 324 >> 2] | 0;
+            y2 = H[x + 328 >> 2] | 0;
+            y3 = H[x + 332 >> 2] | 0;
+            y4 = H[x + 336 >> 2] | 0;
+            for (i = 0; (i | 0) < (k | 0); i = i + 64 | 0) {
+                z0 = y0;
+                z1 = y1;
+                z2 = y2;
+                z3 = y3;
+                z4 = y4;
+                for (j = 0; (j | 0) < 64; j = j + 4 | 0) {
+                    t1 = H[i + j >> 2] | 0;
+                    t0 = ((y0 << 5 | y0 >>> 27) + (y1 & y2 | ~y1 & y3) | 0) + ((t1 + y4 | 0) + 1518500249 | 0) | 0;
+                    y4 = y3;
+                    y3 = y2;
+                    y2 = y1 << 30 | y1 >>> 2;
+                    y1 = y0;
+                    y0 = t0;
+                    ;
+                    H[k + j >> 2] = t1;
+                }
+                for (j = k + 64 | 0; (j | 0) < (k + 80 | 0); j = j + 4 | 0) {
+                    t1 = (H[j - 12 >> 2] ^ H[j - 32 >> 2] ^ H[j - 56 >> 2] ^ H[j - 64 >> 2]) << 1 | (H[j - 12 >> 2] ^ H[j - 32 >> 2] ^ H[j - 56 >> 2] ^ H[j - 64 >> 2]) >>> 31;
+                    t0 = ((y0 << 5 | y0 >>> 27) + (y1 & y2 | ~y1 & y3) | 0) + ((t1 + y4 | 0) + 1518500249 | 0) | 0;
+                    y4 = y3;
+                    y3 = y2;
+                    y2 = y1 << 30 | y1 >>> 2;
+                    y1 = y0;
+                    y0 = t0;
+                    ;
+                    H[j >> 2] = t1;
+                }
+                for (j = k + 80 | 0; (j | 0) < (k + 160 | 0); j = j + 4 | 0) {
+                    t1 = (H[j - 12 >> 2] ^ H[j - 32 >> 2] ^ H[j - 56 >> 2] ^ H[j - 64 >> 2]) << 1 | (H[j - 12 >> 2] ^ H[j - 32 >> 2] ^ H[j - 56 >> 2] ^ H[j - 64 >> 2]) >>> 31;
+                    t0 = ((y0 << 5 | y0 >>> 27) + (y1 ^ y2 ^ y3) | 0) + ((t1 + y4 | 0) + 1859775393 | 0) | 0;
+                    y4 = y3;
+                    y3 = y2;
+                    y2 = y1 << 30 | y1 >>> 2;
+                    y1 = y0;
+                    y0 = t0;
+                    ;
+                    H[j >> 2] = t1;
+                }
+                for (j = k + 160 | 0; (j | 0) < (k + 240 | 0); j = j + 4 | 0) {
+                    t1 = (H[j - 12 >> 2] ^ H[j - 32 >> 2] ^ H[j - 56 >> 2] ^ H[j - 64 >> 2]) << 1 | (H[j - 12 >> 2] ^ H[j - 32 >> 2] ^ H[j - 56 >> 2] ^ H[j - 64 >> 2]) >>> 31;
+                    t0 = ((y0 << 5 | y0 >>> 27) + (y1 & y2 | y1 & y3 | y2 & y3) | 0) + ((t1 + y4 | 0) - 1894007588 | 0) | 0;
+                    y4 = y3;
+                    y3 = y2;
+                    y2 = y1 << 30 | y1 >>> 2;
+                    y1 = y0;
+                    y0 = t0;
+                    ;
+                    H[j >> 2] = t1;
+                }
+                for (j = k + 240 | 0; (j | 0) < (k + 320 | 0); j = j + 4 | 0) {
+                    t1 = (H[j - 12 >> 2] ^ H[j - 32 >> 2] ^ H[j - 56 >> 2] ^ H[j - 64 >> 2]) << 1 | (H[j - 12 >> 2] ^ H[j - 32 >> 2] ^ H[j - 56 >> 2] ^ H[j - 64 >> 2]) >>> 31;
+                    t0 = ((y0 << 5 | y0 >>> 27) + (y1 ^ y2 ^ y3) | 0) + ((t1 + y4 | 0) - 899497514 | 0) | 0;
+                    y4 = y3;
+                    y3 = y2;
+                    y2 = y1 << 30 | y1 >>> 2;
+                    y1 = y0;
+                    y0 = t0;
+                    ;
+                    H[j >> 2] = t1;
+                }
+                y0 = y0 + z0 | 0;
+                y1 = y1 + z1 | 0;
+                y2 = y2 + z2 | 0;
+                y3 = y3 + z3 | 0;
+                y4 = y4 + z4 | 0;
+            }
+            H[x + 320 >> 2] = y0;
+            H[x + 324 >> 2] = y1;
+            H[x + 328 >> 2] = y2;
+            H[x + 332 >> 2] = y3;
+            H[x + 336 >> 2] = y4;
+        }
+        return { hash: hash };
+    }
+}(this));

app/webogram.appcache

@@ -1,6 +1,6 @@
 CACHE MANIFEST
 
-# 39
+# 40
 
 NETWORK:
 *