use optimized implementations from XKCP (#8)

* src/libkeccak: Use XKCP's optimized implementations

It looks like we were using the reference C implementation, which is not
designed for good performance.

Switch to using either the 32-bit or 64-bit optimized C implementations,
depending on the Node "arch" variable.

* Remove submodule KeccakCodePackage

* Bump minimum node version to 6.0.0

* Small improvements in src/README.md

* Add preprocessor in src/addon.cc

* Add submodle github.com/XKCP/XKCP for test vectors

.gitmodules

@@ -1,3 +1,3 @@
-[submodule "keccak-src"]
-	path = util/KeccakCodePackage
-	url = https://github.com/gvanas/KeccakCodePackage.git
+[submodule "util/XKCP"]
+	path = util/XKCP
+	url = https://github.com/XKCP/XKCP.git

binding.gyp

@@ -1,19 +1,39 @@
 {
+  "variables": {
+    "arch": "<!(node -p 'process.arch')",
+  },
   "targets": [{
     "target_name": "keccak",
     "sources": [
       "./src/addon.cc",
-      "./src/libkeccak/KeccakSponge.c",
-      "./src/libkeccak/KeccakP-1600-reference.c"
+    ],
+    "conditions": [
+      ["arch in ('arm64','ppc64','x64')",
+        # For known 64-bit architectures, use the implementation optimized for 64-bit CPUs.
+        {
+          "sources": [
+            "./src/libkeccak-64/KeccakSpongeWidth1600.c",
+            "./src/libkeccak-64/KeccakP-1600-opt64.c",
+          ],
+          "defines": [
+            "LIBKECCAK=64",
+          ],
+        },
+        # Otherwise, use the implementation optimized for 32-bit CPUs.
+        {
+          "sources": [
+            "./src/libkeccak-32/KeccakSpongeWidth1600.c",
+            "./src/libkeccak-32/KeccakP-1600-inplace32BI.c",
+          ],
+          "defines": [
+            "LIBKECCAK=32",
+          ],
+        },
+      ],
     ],
     "include_dirs": [
       "<!(node -e \"require('nan')\")"
     ],
-    "defines": [
-      "KeccakP200_excluded=1",
-      "KeccakP400_excluded=1",
-      "KeccakP800_excluded=1"
-    ],
     "cflags": [
       "-Wall",
       "-Wno-maybe-uninitialized",

package.json

@@ -32,7 +32,6 @@
     "coverage": "nyc tape test/index.js",
     "coverage-lcov": "npm run coverage && nyc report -r lcov",
     "install": "npm run rebuild || echo \"Keccak bindings compilation fail. Pure JS implementation will be used.\"",
-    "libkeccak": "./util/libkeccak.sh",
     "lint": "standard",
     "rebuild": "node-gyp rebuild",
     "test": "npm run lint && npm run unit",
@@ -52,7 +51,7 @@
     "tape": "^4.5.1"
   },
   "engines": {
-    "node": ">=4.0.0"
+    "node": ">=6.0.0"
   },
   "gypfile": true,
   "browser": {

src/README.md

@@ -0,0 +1,28 @@
+# Importing Keccak C code
+
+The XKCP project contains various implementations of Keccak-related algorithms. These are the steps to select a specific implementation and import the code into our project.
+
+First, generate the source bundles in XKCP:
+
+```
+git clone https://github.com/XKCP/XKCP.git
+cd XKCP
+git checkout 58b20ec
+
+# Edit "Makefile.build".  After all the <fragment> tags, add the following two <target> tags:
+<target name="node32" inherits="KeccakSpongeWidth1600 inplace1600bi"/>
+<target name="node64" inherits="KeccakSpongeWidth1600 optimized1600ufull"/>
+
+make node32.pack node64.pack
+```
+
+The source files we need are now under XKCP's "bin/.pack/npm32/" and "bin/.pack/npm64/".
+- Copy those to our repo under "src/libkeccak-32" and "src/libkeccak-64".
+- Update our "binding.gyp" to point to the correct ".c" files.
+- Run `npm run rebuild`.
+
+## Implementation Choice
+
+Currently, we're using two of XKCP KeccakP[1600] implementations -- the generic 32-bit-optimized one and the generic 64-bit-optimized one.
+
+XKCP has implementations that use CPU-specific instructions (e.g. Intel AVR) and are likely much faster. It might be worth using those.

src/addon.cc

@@ -2,7 +2,13 @@
 #include <nan.h>
 
 extern "C" {
-  #include "libkeccak/KeccakSponge.h"
+#if LIBKECCAK == 32
+  #include "libkeccak-32/KeccakSpongeWidth1600.h"
+#elif LIBKECCAK == 64
+  #include "libkeccak-64/KeccakSpongeWidth1600.h"
+#else
+  #error "LIBKECCAK not defined correctly"
+#endif
 }
 
 class KeccakWrapper : public Nan::ObjectWrap {
@@ -34,8 +40,13 @@ class KeccakWrapper : public Nan::ObjectWrap {
 
   static NAN_METHOD(Initialize) {
     KeccakWrapper* obj = Nan::ObjectWrap::Unwrap<KeccakWrapper>(info.Holder());
+#if (NODE_MODULE_VERSION >= NODE_7_0_MODULE_VERSION)
+    unsigned int rate = info[0]->IntegerValue(info.GetIsolate()->GetCurrentContext()).ToChecked();
+    unsigned int capacity = info[1]->IntegerValue(info.GetIsolate()->GetCurrentContext()).ToChecked();
+#else
     unsigned int rate = info[0]->IntegerValue();
     unsigned int capacity = info[1]->IntegerValue();
+#endif
 
     // ignore return code, rate & capacity always will right because internal object
     KeccakWidth1600_SpongeInitialize(&obj->sponge, rate, capacity);
@@ -53,15 +64,23 @@ class KeccakWrapper : public Nan::ObjectWrap {
 
   static NAN_METHOD(AbsorbLastFewBits) {
     KeccakWrapper* obj = Nan::ObjectWrap::Unwrap<KeccakWrapper>(info.Holder());
+#if (NODE_MODULE_VERSION >= NODE_7_0_MODULE_VERSION)
+    unsigned char bits = info[0]->IntegerValue(info.GetIsolate()->GetCurrentContext()).ToChecked();
+#else
     unsigned char bits = info[0]->IntegerValue();
+#endif
 
     // ignore return code, bcause internal object
     KeccakWidth1600_SpongeAbsorbLastFewBits(&obj->sponge, bits);
   }
 
   static NAN_METHOD(Squeeze) {
     KeccakWrapper* obj = Nan::ObjectWrap::Unwrap<KeccakWrapper>(info.Holder());
+#if (NODE_MODULE_VERSION >= NODE_7_0_MODULE_VERSION)
+    size_t length = info[0]->IntegerValue(info.GetIsolate()->GetCurrentContext()).ToChecked();
+#else
     size_t length = info[0]->IntegerValue();
+#endif
 
     v8::Local<v8::Object> buffer = Nan::NewBuffer(length).ToLocalChecked();
     unsigned char* data = (unsigned char*) node::Buffer::Data(buffer);
@@ -72,7 +91,11 @@ class KeccakWrapper : public Nan::ObjectWrap {
 
   static NAN_METHOD(Copy) {
     KeccakWrapper* from = Nan::ObjectWrap::Unwrap<KeccakWrapper>(info.Holder());
+#if (NODE_MODULE_VERSION >= NODE_7_0_MODULE_VERSION)
+    KeccakWrapper* to = Nan::ObjectWrap::Unwrap<KeccakWrapper>(info[0]->ToObject(info.GetIsolate()->GetCurrentContext()).ToLocalChecked());
+#else
     KeccakWrapper* to = Nan::ObjectWrap::Unwrap<KeccakWrapper>(info[0]->ToObject());
+#endif
 
     memcpy(&to->sponge, &from->sponge, sizeof(KeccakWidth1600_SpongeInstance));
   }

src/libkeccak/KeccakP-1600-SnP.h → src/libkeccak-32/KeccakP-1600-SnP.h

@@ -1,16 +1,16 @@
 /*
-Implementation by the Keccak, Keyak and Ketje Teams, namely, Guido Bertoni,
-Joan Daemen, Michaël Peeters, Gilles Van Assche and Ronny Van Keer, hereby
-denoted as "the implementer".
+Implementation by Ronny Van Keer, hereby denoted as "the implementer".
 
-For more information, feedback or questions, please refer to our websites:
-http://keccak.noekeon.org/
-http://keyak.noekeon.org/
-http://ketje.noekeon.org/
+For more information, feedback or questions, please refer to our website:
+https://keccak.team/
 
 To the extent possible under law, the implementer has waived all copyright
 and related or neighboring rights to the source code in this file.
 http://creativecommons.org/publicdomain/zero/1.0/
+
+---
+
+Please refer to SnP-documentation.h for more details.
 */
 
 #ifndef _KeccakP_1600_SnP_h_
@@ -19,20 +19,17 @@ and related or neighboring rights to the source code in this file.
 /** For the documentation, see SnP-documentation.h.
  */
 
-#define KeccakP1600_implementation      "64-bit reference implementation"
+#define KeccakP1600_implementation      "in-place 32-bit optimized implementation"
 #define KeccakP1600_stateSizeInBytes    200
 #define KeccakP1600_stateAlignment      8
 
-#ifdef KeccakReference
-void KeccakP1600_StaticInitialize( void );
-#else
 #define KeccakP1600_StaticInitialize()
-#endif
 void KeccakP1600_Initialize(void *state);
 void KeccakP1600_AddByte(void *state, unsigned char data, unsigned int offset);
 void KeccakP1600_AddBytes(void *state, const unsigned char *data, unsigned int offset, unsigned int length);
 void KeccakP1600_OverwriteBytes(void *state, const unsigned char *data, unsigned int offset, unsigned int length);
 void KeccakP1600_OverwriteWithZeroes(void *state, unsigned int byteCount);
+void KeccakP1600_Permute_Nrounds(void *state, unsigned int nrounds);
 void KeccakP1600_Permute_12rounds(void *state);
 void KeccakP1600_Permute_24rounds(void *state);
 void KeccakP1600_ExtractBytes(const void *state, unsigned char *data, unsigned int offset, unsigned int length);