Upgrade ANGLE to 2.1~99f075dade7c

This aligns with Chromium branch 2356.

This version brings more complete OpenGL ES 3 support as well as various
bug fixes and performance improvements.

The following changes were made to earlier patches:

-0000-General-fixes-for-ANGLE-2.1
  Removed. All changes are now handled elsewhere.
+0001-ANGLE-Improve-Windows-Phone-support
  Consolidated remaining parts from 0009/0010.
+0002-ANGLE-Fix-compilation-with-MinGW
  Remaining issues from patch 0016.
+0003-ANGLE-Fix-compilation-with-MSVC2010
  Remaining issues from patch 0015.
+0004-ANGLE-Dynamically-load-D3D-compiler-from-list
  Renamed from patch 0008.
+0005-ANGLE-Add-support-for-querying-platform-device
  Renamed from patch 0013.
-0004-Make-it-possible-to-link-ANGLE-statically-for-single
  Removed. Fixed by adding defines to project files.
-0008-ANGLE-Dynamically-load-D3D-compiler-from-a-list-or-t
  Renamed to patch 0005.
-0009-ANGLE-Support-WinRT
  Removed. Mostly fixed upstream; remaining parts in patch 0001.
-0010-ANGLE-Enable-D3D11-for-feature-level-9-cards
  Removed. Mostly fixed upstream; remaining parts in patch 0001.
-0012-ANGLE-fix-semantic-index-lookup
  Removed. Fixed upstream.
-0013-ANGLE-Add-support-for-querying-platform-device
  Renamed to patch 0005.
-0014-Let-ANGLE-use-multithreaded-devices-if-necessary
  Removed. No longer needed.
-0015-ANGLE-Fix-angle-d3d11-on-MSVC2010
  Moved remaining parts to patch 0003.
-0016-ANGLE-Fix-compilation-with-MinGW-D3D11
  Moved remaining parts to patch 0002.
-0017-ANGLE-Fix-compilation-with-D3D9
  Removed. Fixed upstream.
-0018-ANGLE-Fix-releasing-textures-after-we-kill-D3D11
  Removed. Fixed upstream.
-0019-ANGLE-Fix-handling-of-shader-source-with-fixed-lengt
  Removed. Fixed upstream.
-0020-ANGLE-Do-not-use-std-strlen
  Removed. Fixed upstream.
-0020-ANGLE-Fix-compilation-with-MSVC2013-Update4
  Removed. Fixed upstream.

[ChangeLog][Third-party libraries] ANGLE was updated to Chromium branch
2356 (2.1~99f075dade7c).

Change-Id: I32ccbfe95e10986bd94be7191dfd53445ea09158
Task-number: QTBUG-44815
Task-number: QTBUG-37660
Task-number: QTBUG-44694
Task-number: QTBUG-42443
Reviewed-by: Andrew Knight <qt@panimo.net>
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@theqtcompany.com>

1 files changed, 125 insertions, 0 deletions

diff --git a/src/3rdparty/angle/src/libANGLE/renderer/d3d/loadimageSSE2.cpp b/src/3rdparty/angle/src/libANGLE/renderer/d3d/loadimageSSE2.cpp
new file mode 100644
index 0000000000..c87d35c82b
--- /dev/null
+++ b/src/3rdparty/angle/src/libANGLE/renderer/d3d/loadimageSSE2.cpp
@@ -0,0 +1,125 @@
+//
+// Copyright (c) 2002-2015 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.
+//
+
+// loadimageSSE2.cpp: Defines image loading functions. It's
+// in a separated file for GCC, which can enable SSE usage only per-file,
+// not for code blocks that use SSE2 explicitly.
+
+#include "libANGLE/renderer/d3d/loadimage.h"
+
+#include "common/platform.h"
+
+#ifdef ANGLE_USE_SSE
+#include <emmintrin.h>
+#endif
+
+namespace rx
+{
+
+void LoadA8ToBGRA8_SSE2(size_t width, size_t height, size_t depth,
+                        const uint8_t *input, size_t inputRowPitch, size_t inputDepthPitch,
+                        uint8_t *output, size_t outputRowPitch, size_t outputDepthPitch)
+{
+#if defined(ANGLE_USE_SSE)
+    __m128i zeroWide = _mm_setzero_si128();
+
+    for (size_t z = 0; z < depth; z++)
+    {
+        for (size_t y = 0; y < height; y++)
+        {
+            const uint8_t *source = OffsetDataPointer<uint8_t>(input, y, z, inputRowPitch, inputDepthPitch);
+            uint32_t *dest = OffsetDataPointer<uint32_t>(output, y, z, outputRowPitch, outputDepthPitch);
+
+            size_t x = 0;
+
+            // Make output writes aligned
+            for (; ((reinterpret_cast<intptr_t>(&dest[x]) & 0xF) != 0 && x < width); x++)
+            {
+                dest[x] = static_cast<uint32_t>(source[x]) << 24;
+            }
+
+            for (; x + 7 < width; x += 8)
+            {
+                __m128i sourceData = _mm_loadl_epi64(reinterpret_cast<const __m128i*>(&source[x]));
+                // Interleave each byte to 16bit, make the lower byte to zero
+                sourceData = _mm_unpacklo_epi8(zeroWide, sourceData);
+                // Interleave each 16bit to 32bit, make the lower 16bit to zero
+                __m128i lo = _mm_unpacklo_epi16(zeroWide, sourceData);
+                __m128i hi = _mm_unpackhi_epi16(zeroWide, sourceData);
+
+                _mm_store_si128(reinterpret_cast<__m128i*>(&dest[x]), lo);
+                _mm_store_si128(reinterpret_cast<__m128i*>(&dest[x + 4]), hi);
+            }
+
+            // Handle the remainder
+            for (; x < width; x++)
+            {
+                dest[x] = static_cast<uint32_t>(source[x]) << 24;
+            }
+        }
+    }
+#else
+    // Ensure that this function is reported as not implemented for ARM builds because
+    // the instructions below are not present for that architecture.
+    UNIMPLEMENTED();
+    return;
+#endif
+}
+
+void LoadRGBA8ToBGRA8_SSE2(size_t width, size_t height, size_t depth,
+                           const uint8_t *input, size_t inputRowPitch, size_t inputDepthPitch,
+                           uint8_t *output, size_t outputRowPitch, size_t outputDepthPitch)
+{
+#if defined(ANGLE_USE_SSE)
+    __m128i brMask = _mm_set1_epi32(0x00ff00ff);
+
+    for (size_t z = 0; z < depth; z++)
+    {
+        for (size_t y = 0; y < height; y++)
+        {
+            const uint32_t *source = OffsetDataPointer<uint32_t>(input, y, z, inputRowPitch, inputDepthPitch);
+            uint32_t *dest = OffsetDataPointer<uint32_t>(output, y, z, outputRowPitch, outputDepthPitch);
+
+            size_t x = 0;
+
+            // Make output writes aligned
+            for (; ((reinterpret_cast<intptr_t>(&dest[x]) & 15) != 0) && x < width; x++)
+            {
+                uint32_t rgba = source[x];
+                dest[x] = (ANGLE_ROTL(rgba, 16) & 0x00ff00ff) | (rgba & 0xff00ff00);
+            }
+
+            for (; x + 3 < width; x += 4)
+            {
+                __m128i sourceData = _mm_loadu_si128(reinterpret_cast<const __m128i*>(&source[x]));
+                // Mask out g and a, which don't change
+                __m128i gaComponents = _mm_andnot_si128(brMask, sourceData);
+                // Mask out b and r
+                __m128i brComponents = _mm_and_si128(sourceData, brMask);
+                // Swap b and r
+                __m128i brSwapped = _mm_shufflehi_epi16(_mm_shufflelo_epi16(brComponents, _MM_SHUFFLE(2, 3, 0, 1)), _MM_SHUFFLE(2, 3, 0, 1));
+                __m128i result = _mm_or_si128(gaComponents, brSwapped);
+                _mm_store_si128(reinterpret_cast<__m128i*>(&dest[x]), result);
+            }
+
+            // Perform leftover writes
+            for (; x < width; x++)
+            {
+                uint32_t rgba = source[x];
+                dest[x] = (ANGLE_ROTL(rgba, 16) & 0x00ff00ff) | (rgba & 0xff00ff00);
+            }
+        }
+    }
+#else
+    // Ensure that this function is reported as not implemented for ARM builds because
+    // the instructions below are not present for that architecture.
+    UNIMPLEMENTED();
+    return;
+#endif
+}
+
+}
+