Bug 774755 - 3/8 - ANGLE r1242: reapply/update existing patches - no review
authorBenoit Jacob <bjacob@mozilla.com>
Wed, 25 Jul 2012 12:13:45 -0400
changeset 100449 2592d31d2f0d97287df2f58e177b72f4160d9240
parent 100448 c4a4804b2420ffa82a27f8e98b0f4d768c9629dd
child 100450 41466cf61ec6b3b27670380a1e7d1831f55becae
push id23182
push useremorley@mozilla.com
push dateThu, 26 Jul 2012 12:04:56 +0000
treeherdermozilla-central@20db7c6d82cc [default view] [failures only]
perfherder[talos] [build metrics] [platform microbench] (compared to previous push)
bugs774755
milestone17.0a1
first release with
nightly linux32
nightly linux64
nightly mac
nightly win32
nightly win64
last release without
nightly linux32
nightly linux64
nightly mac
nightly win32
nightly win64
Bug 774755 - 3/8 - ANGLE r1242: reapply/update existing patches - no review
gfx/angle/src/compiler/Diagnostics.cpp
gfx/angle/src/compiler/DirectiveHandler.cpp
gfx/angle/src/compiler/MapLongVariableNames.cpp
gfx/angle/src/compiler/OutputGLSLBase.cpp
gfx/angle/src/compiler/OutputHLSL.cpp
gfx/angle/src/compiler/Types.h
gfx/angle/src/compiler/compilerdebug.cpp
gfx/angle/src/compiler/compilerdebug.h
gfx/angle/src/compiler/debug.cpp
gfx/angle/src/compiler/debug.h
gfx/angle/src/compiler/osinclude.h
gfx/angle/src/compiler/preprocessor/atom.c
gfx/angle/src/compiler/preprocessor/tokens.c
gfx/angle/src/compiler/spooky.cpp
gfx/angle/src/compiler/spooky.h
gfx/angle/src/compiler/translator_common.vcproj
gfx/angle/src/libGLESv2/Context.cpp
gfx/angle/src/libGLESv2/Context.h
gfx/angle/src/libGLESv2/libGLESv2.cpp
--- a/gfx/angle/src/compiler/Diagnostics.cpp
+++ b/gfx/angle/src/compiler/Diagnostics.cpp
@@ -1,17 +1,17 @@
 //
 // Copyright (c) 2012 The ANGLE Project Authors. All rights reserved.
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file.
 //
 
 #include "compiler/Diagnostics.h"
 
-#include "compiler/debug.h"
+#include "compiler/compilerdebug.h"
 #include "compiler/InfoSink.h"
 #include "compiler/preprocessor/new/SourceLocation.h"
 
 TDiagnostics::TDiagnostics(TInfoSink& infoSink) :
     mInfoSink(infoSink),
     mNumErrors(0),
     mNumWarnings(0)
 {
--- a/gfx/angle/src/compiler/DirectiveHandler.cpp
+++ b/gfx/angle/src/compiler/DirectiveHandler.cpp
@@ -3,17 +3,17 @@
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file.
 //
 
 #include "compiler/DirectiveHandler.h"
 
 #include <sstream>
 
-#include "compiler/debug.h"
+#include "compiler/compilerdebug.h"
 #include "compiler/Diagnostics.h"
 
 static TBehavior getBehavior(const std::string& str)
 {
     static const std::string kRequire("require");
     static const std::string kEnable("enable");
     static const std::string kDisable("disable");
     static const std::string kWarn("warn");
--- a/gfx/angle/src/compiler/MapLongVariableNames.cpp
+++ b/gfx/angle/src/compiler/MapLongVariableNames.cpp
@@ -1,29 +1,36 @@
 //
 // Copyright (c) 2002-2012 The ANGLE Project Authors. All rights reserved.
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file.
 //
 
 #include "compiler/MapLongVariableNames.h"
+#include "spooky.h"
 
 namespace {
 
 TString mapLongName(int id, const TString& name, bool isGlobal)
 {
     ASSERT(name.size() > MAX_SHORTENED_IDENTIFIER_SIZE);
     TStringStream stream;
-    stream << "webgl_";
-    if (isGlobal)
-        stream << "g";
-    stream << id;
-    if (name[0] != '_')
-        stream << "_";
-    stream << name.substr(0, MAX_SHORTENED_IDENTIFIER_SIZE - stream.str().size());
+    uint64 hash = SpookyHash::Hash64(name.data(), name.length(), 0);
+
+    // We want to avoid producing a string with a double underscore,
+    // which would be an illegal GLSL identifier. We can assume that the
+    // original identifier doesn't have a double underscore, otherwise
+    // it's illegal anyway.
+    stream << (name[0] == '_' ? "webgl" : "webgl_")
+           << name.substr(0, 9)
+           << (name[8] == '_' ? "" : "_")
+           << std::hex
+           << hash;
+    ASSERT(stream.str().length() <= MAX_SHORTENED_IDENTIFIER_SIZE);
+    ASSERT(stream.str().length() >= MAX_SHORTENED_IDENTIFIER_SIZE - 2);
     return stream.str();
 }
 
 LongNameMap* gLongNameMapInstance = NULL;
 
 }  // anonymous namespace
 
 LongNameMap::LongNameMap()
--- a/gfx/angle/src/compiler/OutputGLSLBase.cpp
+++ b/gfx/angle/src/compiler/OutputGLSLBase.cpp
@@ -1,16 +1,16 @@
 //
 // Copyright (c) 2002-2011 The ANGLE Project Authors. All rights reserved.
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file.
 //
 
 #include "compiler/OutputGLSLBase.h"
-#include "compiler/debug.h"
+#include "compiler/compilerdebug.h"
 
 namespace
 {
 TString getTypeName(const TType& type)
 {
     TInfoSinkBase out;
     if (type.isMatrix())
     {
--- a/gfx/angle/src/compiler/OutputHLSL.cpp
+++ b/gfx/angle/src/compiler/OutputHLSL.cpp
@@ -2,17 +2,17 @@
 // Copyright (c) 2002-2012 The ANGLE Project Authors. All rights reserved.
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file.
 //
 
 #include "compiler/OutputHLSL.h"
 
 #include "common/angleutils.h"
-#include "compiler/debug.h"
+#include "compiler/compilerdebug.h"
 #include "compiler/InfoSink.h"
 #include "compiler/UnfoldShortCircuit.h"
 #include "compiler/SearchSymbol.h"
 #include "compiler/DetectDiscontinuity.h"
 
 #include <stdio.h>
 #include <algorithm>
 
--- a/gfx/angle/src/compiler/Types.h
+++ b/gfx/angle/src/compiler/Types.h
@@ -4,17 +4,17 @@
 // found in the LICENSE file.
 //
 
 #ifndef _TYPES_INCLUDED
 #define _TYPES_INCLUDED
 
 #include "compiler/BaseTypes.h"
 #include "compiler/Common.h"
-#include "compiler/debug.h"
+#include "compiler/compilerdebug.h"
 
 //
 // Need to have association of line numbers to types in a list for building structs.
 //
 class TType;
 struct TTypeLine {
     TType* type;
     int line;
rename from gfx/angle/src/compiler/debug.cpp
rename to gfx/angle/src/compiler/compilerdebug.cpp
--- a/gfx/angle/src/compiler/debug.cpp
+++ b/gfx/angle/src/compiler/compilerdebug.cpp
@@ -1,17 +1,17 @@
 //
 // Copyright (c) 2002-2010 The ANGLE Project Authors. All rights reserved.
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file.
 //
 
 // debug.cpp: Debugging utilities.
 
-#include "compiler/debug.h"
+#include "compiler/compilerdebug.h"
 
 #include <stdarg.h>
 #include <stdio.h>
 
 #include "compiler/InitializeParseContext.h"
 #include "compiler/ParseHelper.h"
 
 static const int kTraceBufferLen = 1024;
rename from gfx/angle/src/compiler/debug.h
rename to gfx/angle/src/compiler/compilerdebug.h
--- a/gfx/angle/src/compiler/debug.h
+++ b/gfx/angle/src/compiler/compilerdebug.h
@@ -1,15 +1,15 @@
 //
 // Copyright (c) 2002-2010 The ANGLE Project Authors. All rights reserved.
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file.
 //
 
-// debug.h: Debugging utilities.
+// compilerdebug.h: Debugging utilities.
 
 #ifndef COMPILER_DEBUG_H_
 #define COMPILER_DEBUG_H_
 
 #include <assert.h>
 
 #ifdef _DEBUG
 #define TRACE_ENABLED  // define to enable debug message tracing
--- a/gfx/angle/src/compiler/osinclude.h
+++ b/gfx/angle/src/compiler/osinclude.h
@@ -32,17 +32,17 @@
 #include <windows.h>
 #elif defined(ANGLE_OS_POSIX)
 #include <pthread.h>
 #include <semaphore.h>
 #include <errno.h>
 #endif  // ANGLE_USE_NSPR
 
 
-#include "compiler/debug.h"
+#include "compiler/compilerdebug.h"
 
 //
 // Thread Local Storage Operations
 //
 #if defined(ANGLE_USE_NSPR)
 typedef PRUintn OS_TLSIndex;
 #define OS_INVALID_TLS_INDEX 0xFFFFFFFF
 #elif defined(ANGLE_OS_WIN)
--- a/gfx/angle/src/compiler/preprocessor/atom.c
+++ b/gfx/angle/src/compiler/preprocessor/atom.c
@@ -46,17 +46,17 @@ NVIDIA HAS BEEN ADVISED OF THE POSSIBILI
 // atom.c
 //
 
 #include <stdlib.h>
 #include <stdio.h>
 #include <string.h>
 
 #include "common/angleutils.h"
-#include "compiler/debug.h"
+#include "compiler/compilerdebug.h"
 #include "compiler/preprocessor/slglobals.h"
 
 #undef malloc
 #undef realloc
 #undef free
 
 ///////////////////////////////////////////////////////////////////////////////////////////////
 ////////////////////////////////////////// String table: //////////////////////////////////////
@@ -327,22 +327,17 @@ static int GrowAtomTable(AtomTable *atab
             newmap = realloc(atable->amap, sizeof(int)*size);
             newrev = realloc(atable->arev, sizeof(int)*size);
         } else {
             newmap = malloc(sizeof(int)*size);
             newrev = malloc(sizeof(int)*size);
             atable->size = 0;
         }
         if (!newmap || !newrev) {
-            /* failed to grow -- error */
-            if (newmap)
-                atable->amap = newmap;
-            if (newrev)
-                atable->arev = newrev;
-            return -1;
+            abort();
         }
         memset(&newmap[atable->size], 0, (size - atable->size) * sizeof(int));
         memset(&newrev[atable->size], 0, (size - atable->size) * sizeof(int));
         atable->amap = newmap;
         atable->arev = newrev;
         atable->size = size;
     }
     return 0;
--- a/gfx/angle/src/compiler/preprocessor/tokens.c
+++ b/gfx/angle/src/compiler/preprocessor/tokens.c
@@ -46,17 +46,17 @@ NVIDIA HAS BEEN ADVISED OF THE POSSIBILI
 //
 
 #include <stdlib.h>
 #include <stdio.h>
 #include <string.h>
 #include <ctype.h>
 
 #include "common/angleutils.h"
-#include "compiler/debug.h"
+#include "compiler/compilerdebug.h"
 #include "compiler/preprocessor/slglobals.h"
 #include "compiler/util.h"
 
 #if defined(_MSC_VER)
 #pragma warning(disable: 4054)
 #pragma warning(disable: 4152)
 #endif
 
new file mode 100644
--- /dev/null
+++ b/gfx/angle/src/compiler/spooky.cpp
@@ -0,0 +1,348 @@
+// Spooky Hash
+// A 128-bit noncryptographic hash, for checksums and table lookup
+// By Bob Jenkins.  Public domain.
+//   Oct 31 2010: published framework, disclaimer ShortHash isn't right
+//   Nov 7 2010: disabled ShortHash
+//   Oct 31 2011: replace End, ShortMix, ShortEnd, enable ShortHash again
+
+#include <memory.h>
+#include <string.h>
+#include "spooky.h"
+
+#define ALLOW_UNALIGNED_READS 1
+
+//
+// short hash ... it could be used on any message, 
+// but it's used by Spooky just for short messages.
+//
+void SpookyHash::Short(
+    const void *message,
+    size_t length,
+    uint64 *hash1,
+    uint64 *hash2)
+{
+    uint64 buf[sc_numVars];
+    union 
+    { 
+        const uint8 *p8; 
+        uint32 *p32;
+        uint64 *p64; 
+        size_t i; 
+    } u;
+
+    u.p8 = (const uint8 *)message;
+    
+    if (!ALLOW_UNALIGNED_READS && (u.i & 0x7))
+    {
+        memcpy(buf, message, length);
+        u.p64 = buf;
+    }
+
+    size_t remainder = length%32;
+    uint64 a=*hash1;
+    uint64 b=*hash2;
+    uint64 c=sc_const;
+    uint64 d=sc_const;
+
+    if (length > 15)
+    {
+        const uint64 *end = u.p64 + (length/32)*4;
+        
+        // handle all complete sets of 32 bytes
+        for (; u.p64 < end; u.p64 += 4)
+        {
+            c += u.p64[0];
+            d += u.p64[1];
+            ShortMix(a,b,c,d);
+            a += u.p64[2];
+            b += u.p64[3];
+        }
+        
+        //Handle the case of 16+ remaining bytes.
+        if (remainder >= 16)
+        {
+            c += u.p64[0];
+            d += u.p64[1];
+            ShortMix(a,b,c,d);
+            u.p64 += 2;
+            remainder -= 16;
+        }
+    }
+    
+    // Handle the last 0..15 bytes, and its length
+    d = ((uint64)length) << 56;
+    switch (remainder)
+    {
+    case 15:
+    d += ((uint64)u.p8[14]) << 48;
+    case 14:
+        d += ((uint64)u.p8[13]) << 40;
+    case 13:
+        d += ((uint64)u.p8[12]) << 32;
+    case 12:
+        d += u.p32[2];
+        c += u.p64[0];
+        break;
+    case 11:
+        d += ((uint64)u.p8[10]) << 16;
+    case 10:
+        d += ((uint64)u.p8[9]) << 8;
+    case 9:
+        d += (uint64)u.p8[8];
+    case 8:
+        c += u.p64[0];
+        break;
+    case 7:
+        c += ((uint64)u.p8[6]) << 48;
+    case 6:
+        c += ((uint64)u.p8[5]) << 40;
+    case 5:
+        c += ((uint64)u.p8[4]) << 32;
+    case 4:
+        c += u.p32[0];
+        break;
+    case 3:
+        c += ((uint64)u.p8[2]) << 16;
+    case 2:
+        c += ((uint64)u.p8[1]) << 8;
+    case 1:
+        c += (uint64)u.p8[0];
+        break;
+    case 0:
+        c += sc_const;
+        d += sc_const;
+    }
+    ShortEnd(a,b,c,d);
+    *hash1 = a;
+    *hash2 = b;
+}
+
+
+
+
+// do the whole hash in one call
+void SpookyHash::Hash128(
+    const void *message, 
+    size_t length, 
+    uint64 *hash1, 
+    uint64 *hash2)
+{
+    if (length < sc_bufSize)
+    {
+        Short(message, length, hash1, hash2);
+        return;
+    }
+
+    uint64 h0,h1,h2,h3,h4,h5,h6,h7,h8,h9,h10,h11;
+    uint64 buf[sc_numVars];
+    uint64 *end;
+    union 
+    { 
+        const uint8 *p8; 
+        uint64 *p64; 
+        size_t i; 
+    } u;
+    size_t remainder;
+    
+    h0=h3=h6=h9  = *hash1;
+    h1=h4=h7=h10 = *hash2;
+    h2=h5=h8=h11 = sc_const;
+    
+    u.p8 = (const uint8 *)message;
+    end = u.p64 + (length/sc_blockSize)*sc_numVars;
+
+    // handle all whole sc_blockSize blocks of bytes
+    if (ALLOW_UNALIGNED_READS || ((u.i & 0x7) == 0))
+    {
+        while (u.p64 < end)
+        { 
+            Mix(u.p64, h0,h1,h2,h3,h4,h5,h6,h7,h8,h9,h10,h11);
+	    u.p64 += sc_numVars;
+        }
+    }
+    else
+    {
+        while (u.p64 < end)
+        {
+            memcpy(buf, u.p64, sc_blockSize);
+            Mix(buf, h0,h1,h2,h3,h4,h5,h6,h7,h8,h9,h10,h11);
+	    u.p64 += sc_numVars;
+        }
+    }
+
+    // handle the last partial block of sc_blockSize bytes
+    remainder = (length - ((const uint8 *)end-(const uint8 *)message));
+    memcpy(buf, end, remainder);
+    memset(((uint8 *)buf)+remainder, 0, sc_blockSize-remainder);
+    ((uint8 *)buf)[sc_blockSize-1] = remainder;
+    Mix(buf, h0,h1,h2,h3,h4,h5,h6,h7,h8,h9,h10,h11);
+    
+    // do some final mixing 
+    End(h0,h1,h2,h3,h4,h5,h6,h7,h8,h9,h10,h11);
+    *hash1 = h0;
+    *hash2 = h1;
+}
+
+
+
+// init spooky state
+void SpookyHash::Init(uint64 seed1, uint64 seed2)
+{
+    m_length = 0;
+    m_remainder = 0;
+    m_state[0] = seed1;
+    m_state[1] = seed2;
+}
+
+
+// add a message fragment to the state
+void SpookyHash::Update(const void *message, size_t length)
+{
+    uint64 h0,h1,h2,h3,h4,h5,h6,h7,h8,h9,h10,h11;
+    size_t newLength = length + m_remainder;
+    uint8  remainder;
+    union 
+    { 
+        const uint8 *p8; 
+        uint64 *p64; 
+        size_t i; 
+    } u;
+    const uint64 *end;
+    
+    // Is this message fragment too short?  If it is, stuff it away.
+    if (newLength < sc_bufSize)
+    {
+        memcpy(&((uint8 *)m_data)[m_remainder], message, length);
+        m_length = length + m_length;
+        m_remainder = (uint8)newLength;
+        return;
+    }
+    
+    // init the variables
+    if (m_length < sc_bufSize)
+    {
+        h0=h3=h6=h9  = m_state[0];
+        h1=h4=h7=h10 = m_state[1];
+        h2=h5=h8=h11 = sc_const;
+    }
+    else
+    {
+        h0 = m_state[0];
+        h1 = m_state[1];
+        h2 = m_state[2];
+        h3 = m_state[3];
+        h4 = m_state[4];
+        h5 = m_state[5];
+        h6 = m_state[6];
+        h7 = m_state[7];
+        h8 = m_state[8];
+        h9 = m_state[9];
+        h10 = m_state[10];
+        h11 = m_state[11];
+    }
+    m_length = length + m_length;
+    
+    // if we've got anything stuffed away, use it now
+    if (m_remainder)
+    {
+        uint8 prefix = sc_bufSize-m_remainder;
+        memcpy(&(((uint8 *)m_data)[m_remainder]), message, prefix);
+        u.p64 = m_data;
+        Mix(u.p64, h0,h1,h2,h3,h4,h5,h6,h7,h8,h9,h10,h11);
+        Mix(&u.p64[sc_numVars], h0,h1,h2,h3,h4,h5,h6,h7,h8,h9,h10,h11);
+        u.p8 = ((const uint8 *)message) + prefix;
+        length -= prefix;
+    }
+    else
+    {
+        u.p8 = (const uint8 *)message;
+    }
+    
+    // handle all whole blocks of sc_blockSize bytes
+    end = u.p64 + (length/sc_blockSize)*sc_numVars;
+    remainder = (uint8)(length-((const uint8 *)end-u.p8));
+    if (ALLOW_UNALIGNED_READS || (u.i & 0x7) == 0)
+    {
+        while (u.p64 < end)
+        { 
+            Mix(u.p64, h0,h1,h2,h3,h4,h5,h6,h7,h8,h9,h10,h11);
+	    u.p64 += sc_numVars;
+        }
+    }
+    else
+    {
+        while (u.p64 < end)
+        { 
+            memcpy(m_data, u.p8, sc_blockSize);
+            Mix(m_data, h0,h1,h2,h3,h4,h5,h6,h7,h8,h9,h10,h11);
+	    u.p64 += sc_numVars;
+        }
+    }
+
+    // stuff away the last few bytes
+    m_remainder = remainder;
+    memcpy(m_data, end, remainder);
+    
+    // stuff away the variables
+    m_state[0] = h0;
+    m_state[1] = h1;
+    m_state[2] = h2;
+    m_state[3] = h3;
+    m_state[4] = h4;
+    m_state[5] = h5;
+    m_state[6] = h6;
+    m_state[7] = h7;
+    m_state[8] = h8;
+    m_state[9] = h9;
+    m_state[10] = h10;
+    m_state[11] = h11;
+}
+
+
+// report the hash for the concatenation of all message fragments so far
+void SpookyHash::Final(uint64 *hash1, uint64 *hash2)
+{
+    // init the variables
+    if (m_length < sc_bufSize)
+    {
+        Short( m_data, m_length, hash1, hash2);
+        return;
+    }
+    
+    const uint64 *data = (const uint64 *)m_data;
+    uint8 remainder = m_remainder;
+    
+    uint64 h0 = m_state[0];
+    uint64 h1 = m_state[1];
+    uint64 h2 = m_state[2];
+    uint64 h3 = m_state[3];
+    uint64 h4 = m_state[4];
+    uint64 h5 = m_state[5];
+    uint64 h6 = m_state[6];
+    uint64 h7 = m_state[7];
+    uint64 h8 = m_state[8];
+    uint64 h9 = m_state[9];
+    uint64 h10 = m_state[10];
+    uint64 h11 = m_state[11];
+
+    if (remainder >= sc_blockSize)
+    {
+        // m_data can contain two blocks; handle any whole first block
+        Mix(data, h0,h1,h2,h3,h4,h5,h6,h7,h8,h9,h10,h11);
+	data += sc_numVars;
+	remainder -= sc_blockSize;
+    }
+
+    // mix in the last partial block, and the length mod sc_blockSize
+    memset(&((uint8 *)data)[remainder], 0, (sc_blockSize-remainder));
+
+    ((uint8 *)data)[sc_blockSize-1] = remainder;
+    Mix(data, h0,h1,h2,h3,h4,h5,h6,h7,h8,h9,h10,h11);
+    
+    // do some final mixing
+    End(h0,h1,h2,h3,h4,h5,h6,h7,h8,h9,h10,h11);
+
+    *hash1 = h0;
+    *hash2 = h1;
+}
+
new file mode 100644
--- /dev/null
+++ b/gfx/angle/src/compiler/spooky.h
@@ -0,0 +1,293 @@
+//
+// SpookyHash: a 128-bit noncryptographic hash function
+// By Bob Jenkins, public domain
+//   Oct 31 2010: alpha, framework + SpookyHash::Mix appears right
+//   Oct 31 2011: alpha again, Mix only good to 2^^69 but rest appears right
+//   Dec 31 2011: beta, improved Mix, tested it for 2-bit deltas
+//   Feb  2 2012: production, same bits as beta
+//   Feb  5 2012: adjusted definitions of uint* to be more portable
+// 
+// Up to 4 bytes/cycle for long messages.  Reasonably fast for short messages.
+// All 1 or 2 bit deltas achieve avalanche within 1% bias per output bit.
+//
+// This was developed for and tested on 64-bit x86-compatible processors.
+// It assumes the processor is little-endian.  There is a macro
+// controlling whether unaligned reads are allowed (by default they are).
+// This should be an equally good hash on big-endian machines, but it will
+// compute different results on them than on little-endian machines.
+//
+// Google's CityHash has similar specs to SpookyHash, and CityHash is faster
+// on some platforms.  MD4 and MD5 also have similar specs, but they are orders
+// of magnitude slower.  CRCs are two or more times slower, but unlike 
+// SpookyHash, they have nice math for combining the CRCs of pieces to form 
+// the CRCs of wholes.  There are also cryptographic hashes, but those are even 
+// slower than MD5.
+//
+
+#include <stddef.h>
+
+#ifdef _MSC_VER
+# define INLINE __forceinline
+  typedef  unsigned __int64 uint64;
+  typedef  unsigned __int32 uint32;
+  typedef  unsigned __int16 uint16;
+  typedef  unsigned __int8  uint8;
+#else
+# include <stdint.h>
+# define INLINE inline
+  typedef  uint64_t  uint64;
+  typedef  uint32_t  uint32;
+  typedef  uint16_t  uint16;
+  typedef  uint8_t   uint8;
+#endif
+
+
+class SpookyHash
+{
+public:
+    //
+    // SpookyHash: hash a single message in one call, produce 128-bit output
+    //
+    static void Hash128(
+        const void *message,  // message to hash
+        size_t length,        // length of message in bytes
+        uint64 *hash1,        // in/out: in seed 1, out hash value 1
+        uint64 *hash2);       // in/out: in seed 2, out hash value 2
+
+    //
+    // Hash64: hash a single message in one call, return 64-bit output
+    //
+    static uint64 Hash64(
+        const void *message,  // message to hash
+        size_t length,        // length of message in bytes
+        uint64 seed)          // seed
+    {
+        uint64 hash1 = seed;
+        Hash128(message, length, &hash1, &seed);
+        return hash1;
+    }
+
+    //
+    // Hash32: hash a single message in one call, produce 32-bit output
+    //
+    static uint32 Hash32(
+        const void *message,  // message to hash
+        size_t length,        // length of message in bytes
+        uint32 seed)          // seed
+    {
+        uint64 hash1 = seed, hash2 = seed;
+        Hash128(message, length, &hash1, &hash2);
+        return (uint32)hash1;
+    }
+
+    //
+    // Init: initialize the context of a SpookyHash
+    //
+    void Init(
+        uint64 seed1,       // any 64-bit value will do, including 0
+        uint64 seed2);      // different seeds produce independent hashes
+    
+    //
+    // Update: add a piece of a message to a SpookyHash state
+    //
+    void Update(
+        const void *message,  // message fragment
+        size_t length);       // length of message fragment in bytes
+
+
+    //
+    // Final: compute the hash for the current SpookyHash state
+    //
+    // This does not modify the state; you can keep updating it afterward
+    //
+    // The result is the same as if SpookyHash() had been called with
+    // all the pieces concatenated into one message.
+    //
+    void Final(
+        uint64 *hash1,    // out only: first 64 bits of hash value.
+        uint64 *hash2);   // out only: second 64 bits of hash value.
+
+    //
+    // left rotate a 64-bit value by k bytes
+    //
+    static INLINE uint64 Rot64(uint64 x, int k)
+    {
+        return (x << k) | (x >> (64 - k));
+    }
+
+    //
+    // This is used if the input is 96 bytes long or longer.
+    //
+    // The internal state is fully overwritten every 96 bytes.
+    // Every input bit appears to cause at least 128 bits of entropy
+    // before 96 other bytes are combined, when run forward or backward
+    //   For every input bit,
+    //   Two inputs differing in just that input bit
+    //   Where "differ" means xor or subtraction
+    //   And the base value is random
+    //   When run forward or backwards one Mix
+    // I tried 3 pairs of each; they all differed by at least 212 bits.
+    //
+    static INLINE void Mix(
+        const uint64 *data, 
+        uint64 &s0, uint64 &s1, uint64 &s2, uint64 &s3,
+        uint64 &s4, uint64 &s5, uint64 &s6, uint64 &s7,
+        uint64 &s8, uint64 &s9, uint64 &s10,uint64 &s11)
+    {
+      s0 += data[0];    s2 ^= s10;    s11 ^= s0;    s0 = Rot64(s0,11);    s11 += s1;
+      s1 += data[1];    s3 ^= s11;    s0 ^= s1;    s1 = Rot64(s1,32);    s0 += s2;
+      s2 += data[2];    s4 ^= s0;    s1 ^= s2;    s2 = Rot64(s2,43);    s1 += s3;
+      s3 += data[3];    s5 ^= s1;    s2 ^= s3;    s3 = Rot64(s3,31);    s2 += s4;
+      s4 += data[4];    s6 ^= s2;    s3 ^= s4;    s4 = Rot64(s4,17);    s3 += s5;
+      s5 += data[5];    s7 ^= s3;    s4 ^= s5;    s5 = Rot64(s5,28);    s4 += s6;
+      s6 += data[6];    s8 ^= s4;    s5 ^= s6;    s6 = Rot64(s6,39);    s5 += s7;
+      s7 += data[7];    s9 ^= s5;    s6 ^= s7;    s7 = Rot64(s7,57);    s6 += s8;
+      s8 += data[8];    s10 ^= s6;    s7 ^= s8;    s8 = Rot64(s8,55);    s7 += s9;
+      s9 += data[9];    s11 ^= s7;    s8 ^= s9;    s9 = Rot64(s9,54);    s8 += s10;
+      s10 += data[10];    s0 ^= s8;    s9 ^= s10;    s10 = Rot64(s10,22);    s9 += s11;
+      s11 += data[11];    s1 ^= s9;    s10 ^= s11;    s11 = Rot64(s11,46);    s10 += s0;
+    }
+
+    //
+    // Mix all 12 inputs together so that h0, h1 are a hash of them all.
+    //
+    // For two inputs differing in just the input bits
+    // Where "differ" means xor or subtraction
+    // And the base value is random, or a counting value starting at that bit
+    // The final result will have each bit of h0, h1 flip
+    // For every input bit,
+    // with probability 50 +- .3%
+    // For every pair of input bits,
+    // with probability 50 +- 3%
+    //
+    // This does not rely on the last Mix() call having already mixed some.
+    // Two iterations was almost good enough for a 64-bit result, but a
+    // 128-bit result is reported, so End() does three iterations.
+    //
+    static INLINE void EndPartial(
+        uint64 &h0, uint64 &h1, uint64 &h2, uint64 &h3,
+        uint64 &h4, uint64 &h5, uint64 &h6, uint64 &h7, 
+        uint64 &h8, uint64 &h9, uint64 &h10,uint64 &h11)
+    {
+        h11+= h1;    h2 ^= h11;   h1 = Rot64(h1,44);
+	h0 += h2;    h3 ^= h0;    h2 = Rot64(h2,15);
+	h1 += h3;    h4 ^= h1;    h3 = Rot64(h3,34);
+	h2 += h4;    h5 ^= h2;    h4 = Rot64(h4,21);
+	h3 += h5;    h6 ^= h3;    h5 = Rot64(h5,38);
+	h4 += h6;    h7 ^= h4;    h6 = Rot64(h6,33);
+	h5 += h7;    h8 ^= h5;    h7 = Rot64(h7,10);
+	h6 += h8;    h9 ^= h6;    h8 = Rot64(h8,13);
+	h7 += h9;    h10^= h7;    h9 = Rot64(h9,38);
+	h8 += h10;   h11^= h8;    h10= Rot64(h10,53);
+	h9 += h11;   h0 ^= h9;    h11= Rot64(h11,42);
+	h10+= h0;    h1 ^= h10;   h0 = Rot64(h0,54);
+    }
+
+    static INLINE void End(
+        uint64 &h0, uint64 &h1, uint64 &h2, uint64 &h3,
+        uint64 &h4, uint64 &h5, uint64 &h6, uint64 &h7, 
+        uint64 &h8, uint64 &h9, uint64 &h10,uint64 &h11)
+    {
+        EndPartial(h0,h1,h2,h3,h4,h5,h6,h7,h8,h9,h10,h11);
+        EndPartial(h0,h1,h2,h3,h4,h5,h6,h7,h8,h9,h10,h11);
+        EndPartial(h0,h1,h2,h3,h4,h5,h6,h7,h8,h9,h10,h11);
+    }
+
+    //
+    // The goal is for each bit of the input to expand into 128 bits of 
+    //   apparent entropy before it is fully overwritten.
+    // n trials both set and cleared at least m bits of h0 h1 h2 h3
+    //   n: 2   m: 29
+    //   n: 3   m: 46
+    //   n: 4   m: 57
+    //   n: 5   m: 107
+    //   n: 6   m: 146
+    //   n: 7   m: 152
+    // when run forwards or backwards
+    // for all 1-bit and 2-bit diffs
+    // with diffs defined by either xor or subtraction
+    // with a base of all zeros plus a counter, or plus another bit, or random
+    //
+    static INLINE void ShortMix(uint64 &h0, uint64 &h1, uint64 &h2, uint64 &h3)
+    {
+        h2 = Rot64(h2,50);  h2 += h3;  h0 ^= h2;
+        h3 = Rot64(h3,52);  h3 += h0;  h1 ^= h3;
+        h0 = Rot64(h0,30);  h0 += h1;  h2 ^= h0;
+        h1 = Rot64(h1,41);  h1 += h2;  h3 ^= h1;
+        h2 = Rot64(h2,54);  h2 += h3;  h0 ^= h2;
+        h3 = Rot64(h3,48);  h3 += h0;  h1 ^= h3;
+        h0 = Rot64(h0,38);  h0 += h1;  h2 ^= h0;
+        h1 = Rot64(h1,37);  h1 += h2;  h3 ^= h1;
+        h2 = Rot64(h2,62);  h2 += h3;  h0 ^= h2;
+        h3 = Rot64(h3,34);  h3 += h0;  h1 ^= h3;
+        h0 = Rot64(h0,5);   h0 += h1;  h2 ^= h0;
+        h1 = Rot64(h1,36);  h1 += h2;  h3 ^= h1;
+    }
+
+    //
+    // Mix all 4 inputs together so that h0, h1 are a hash of them all.
+    //
+    // For two inputs differing in just the input bits
+    // Where "differ" means xor or subtraction
+    // And the base value is random, or a counting value starting at that bit
+    // The final result will have each bit of h0, h1 flip
+    // For every input bit,
+    // with probability 50 +- .3% (it is probably better than that)
+    // For every pair of input bits,
+    // with probability 50 +- .75% (the worst case is approximately that)
+    //
+    static INLINE void ShortEnd(uint64 &h0, uint64 &h1, uint64 &h2, uint64 &h3)
+    {
+        h3 ^= h2;  h2 = Rot64(h2,15);  h3 += h2;
+        h0 ^= h3;  h3 = Rot64(h3,52);  h0 += h3;
+        h1 ^= h0;  h0 = Rot64(h0,26);  h1 += h0;
+        h2 ^= h1;  h1 = Rot64(h1,51);  h2 += h1;
+        h3 ^= h2;  h2 = Rot64(h2,28);  h3 += h2;
+        h0 ^= h3;  h3 = Rot64(h3,9);   h0 += h3;
+        h1 ^= h0;  h0 = Rot64(h0,47);  h1 += h0;
+        h2 ^= h1;  h1 = Rot64(h1,54);  h2 += h1;
+        h3 ^= h2;  h2 = Rot64(h2,32);  h3 += h2;
+        h0 ^= h3;  h3 = Rot64(h3,25);  h0 += h3;
+        h1 ^= h0;  h0 = Rot64(h0,63);  h1 += h0;
+    }
+    
+private:
+
+    //
+    // Short is used for messages under 192 bytes in length
+    // Short has a low startup cost, the normal mode is good for long
+    // keys, the cost crossover is at about 192 bytes.  The two modes were
+    // held to the same quality bar.
+    // 
+    static void Short(
+        const void *message,
+        size_t length,
+        uint64 *hash1,
+        uint64 *hash2);
+
+    // number of uint64's in internal state
+    static const size_t sc_numVars = 12;
+
+    // size of the internal state
+    static const size_t sc_blockSize = sc_numVars*8;
+
+    // size of buffer of unhashed data, in bytes
+    static const size_t sc_bufSize = 2*sc_blockSize;
+
+    //
+    // sc_const: a constant which:
+    //  * is not zero
+    //  * is odd
+    //  * is a not-very-regular mix of 1's and 0's
+    //  * does not need any other special mathematical properties
+    //
+    static const uint64 sc_const = 0xdeadbeefdeadbeefLL;
+
+    uint64 m_data[2*sc_numVars];   // unhashed data, for partial messages
+    uint64 m_state[sc_numVars];  // internal state of the hash
+    size_t m_length;             // total length of the input so far
+    uint8  m_remainder;          // length of unhashed data stashed in m_data
+};
+
+
+
--- a/gfx/angle/src/compiler/translator_common.vcproj
+++ b/gfx/angle/src/compiler/translator_common.vcproj
@@ -580,17 +580,17 @@
 				RelativePath=".\Common.h"
 				>
 			</File>
 			<File
 				RelativePath=".\ConstantUnion.h"
 				>
 			</File>
 			<File
-				RelativePath=".\debug.h"
+				RelativePath=".\compilerdebug.h"
 				>
 			</File>
 			<File
 				RelativePath=".\DetectRecursion.h"
 				>
 			</File>
 			<File
 				RelativePath=".\Diagnostics.h"
--- a/gfx/angle/src/libGLESv2/Context.cpp
+++ b/gfx/angle/src/libGLESv2/Context.cpp
@@ -2611,16 +2611,17 @@ void Context::readPixels(GLint x, GLint 
     {
         if (desc.Format == D3DFMT_A8R8G8B8 &&
             format == GL_BGRA_EXT &&
             type == GL_UNSIGNED_BYTE)
         {
             // Fast path for EXT_read_format_bgra, given
             // an RGBA source buffer.  Note that buffers with no
             // alpha go through the slow path below.
+            // Note that this is also the combo exposed by IMPLEMENTATION_COLOR_READ_TYPE/FORMAT
             memcpy(dest + j * outputPitch,
                    source + j * inputPitch,
                    (rect.right - rect.left) * 4);
             continue;
         }
 
         for (int i = 0; i < rect.right - rect.left; i++)
         {
@@ -2758,20 +2759,20 @@ void Context::readPixels(GLint x, GLint 
                         ((unsigned short)(     a + 0.5f) << 15) |
                         ((unsigned short)(31 * r + 0.5f) << 10) |
                         ((unsigned short)(31 * g + 0.5f) << 5) |
                         ((unsigned short)(31 * b + 0.5f) << 0);
                     break;
                   default: UNREACHABLE();
                 }
                 break;
-              case GL_RGB:   // IMPLEMENTATION_COLOR_READ_FORMAT
+              case GL_RGB:
                 switch (type)
                 {
-                  case GL_UNSIGNED_SHORT_5_6_5:   // IMPLEMENTATION_COLOR_READ_TYPE
+                  case GL_UNSIGNED_SHORT_5_6_5:
                     dest16[i + j * outputPitch / sizeof(unsigned short)] = 
                         ((unsigned short)(31 * b + 0.5f) << 0) |
                         ((unsigned short)(63 * g + 0.5f) << 5) |
                         ((unsigned short)(31 * r + 0.5f) << 11);
                     break;
                   default: UNREACHABLE();
                 }
                 break;
--- a/gfx/angle/src/libGLESv2/Context.h
+++ b/gfx/angle/src/libGLESv2/Context.h
@@ -70,18 +70,18 @@ enum
     MAX_VARYING_VECTORS_SM3 = 10,
     MAX_TEXTURE_IMAGE_UNITS = 16,
     MAX_VERTEX_TEXTURE_IMAGE_UNITS_VTF = 4,   // For devices supporting vertex texture fetch
     MAX_COMBINED_TEXTURE_IMAGE_UNITS_VTF = MAX_TEXTURE_IMAGE_UNITS + MAX_VERTEX_TEXTURE_IMAGE_UNITS_VTF,    
     MAX_FRAGMENT_UNIFORM_VECTORS_SM2 = 32 - 3,    // Reserve space for dx_Coord, dx_Depth, and dx_DepthRange. dx_PointOrLines and dx_FrontCCW use separate bool registers.
     MAX_FRAGMENT_UNIFORM_VECTORS_SM3 = 224 - 3,
     MAX_DRAW_BUFFERS = 1,
 
-    IMPLEMENTATION_COLOR_READ_FORMAT = GL_RGB,
-    IMPLEMENTATION_COLOR_READ_TYPE = GL_UNSIGNED_SHORT_5_6_5
+    IMPLEMENTATION_COLOR_READ_FORMAT = GL_BGRA_EXT,
+    IMPLEMENTATION_COLOR_READ_TYPE = GL_UNSIGNED_BYTE
 };
 
 enum QueryType
 {
     QUERY_ANY_SAMPLES_PASSED,
     QUERY_ANY_SAMPLES_PASSED_CONSERVATIVE,
 
     QUERY_TYPE_COUNT
--- a/gfx/angle/src/libGLESv2/libGLESv2.cpp
+++ b/gfx/angle/src/libGLESv2/libGLESv2.cpp
@@ -231,27 +231,16 @@ bool validReadFormatType(GLenum format, 
         switch (type)
         {
           case GL_UNSIGNED_BYTE:
             break;
           default:
             return false;
         }
         break;
-      case GL_BGRA_EXT:
-        switch (type)
-        {
-          case GL_UNSIGNED_BYTE:
-          case GL_UNSIGNED_SHORT_4_4_4_4_REV_EXT:
-          case GL_UNSIGNED_SHORT_1_5_5_5_REV_EXT:
-            break;
-          default:
-            return false;
-        }
-        break;
       case gl::IMPLEMENTATION_COLOR_READ_FORMAT:
         switch (type)
         {
           case gl::IMPLEMENTATION_COLOR_READ_TYPE:
             break;
           default:
             return false;
         }