ANGLE: upgrade to 2.1~07d49ef5350a

This version of ANGLE provides partial ES3 support, numerous
bug fixes, and several potentially useful vendor extensions.

All patches have been rebased. The following changes are noted:

0000-General-fixes-for-ANGLE-2.1.patch
  contains compile fixes for the new ANGLE

0004-Make-it-possible-to-link-ANGLE-statically-for-single.patch
  has incorporated patch 0015.

0007-Make-DX9-DX11-mutually-exclusive.patch
  has been removed as it was fixed upstream.

0007-Fix-ANGLE-build-with-Microsoft-Visual-Studio-14-CTP.patch
  has been moved up to fill the patch number gap.

0010-ANGLE-Enable-D3D11-for-feature-level-9-cards.patch
  now contains patch 0014 and 0017.

0013-ANGLE-Allow-for-universal-program-binaries.patch
  has been removed as it is no longer relevant.

0014-ANGLE-D3D11-Fix-internal-index-buffer-for-level-9-ha.patch
  has been merged with patch 0010.

0015-ANGLE-Don-t-export-DLLMain-functions-for-static-buil.patch
  has been merged with patch 0004.

0016-ANGLE-WinRT-Call-Trim-when-application-suspends.patch
  has been removed and will be replaced by a follow-up patch using a
  different technique.

0017-ANGLE-D3D11-Don-t-use-mipmaps-in-level-9-textures.patch
  has been merged with patch 0010.

0018-ANGLE-WinRT-Create-swap-chain-using-physical-resolut.patch
  has been removed and will be replaced by a follow-up patch extending
  the EGL_ANGLE_window_fixed_size extension.

0019-Fix-ANGLE-build-with-Microsoft-Visual-Studio-14-CTP.patch
  is now patch 0007.

[ChangeLog][Third-party libraries] ANGLE has been upgraded to
version 2.1, bringing partial support for OpenGL ES3 over
Direct3D 11, numerous bug fixes, and several new vendor extensions.

Change-Id: I6d95ce1480462d67228d83c1e5c74a1706b5b21c
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@digia.com>

1 files changed, 113 insertions, 0 deletions

diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/loadimageSSE2.cpp b/src/3rdparty/angle/src/libGLESv2/renderer/loadimageSSE2.cpp
new file mode 100644
index 0000000000..dcf347d421
--- /dev/null
+++ b/src/3rdparty/angle/src/libGLESv2/renderer/loadimageSSE2.cpp
@@ -0,0 +1,113 @@
+#include "precompiled.h"
+//
+// Copyright (c) 2002-2014 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 "libGLESv2/renderer/loadimage.h"
+
+#if !defined(__SSE2__) && (defined(_M_X64) || _M_IX86_FP == 2)
+#define __SSE2__
+#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)
+{
+#ifdef __SSE2__
+    __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;
+            }
+        }
+    }
+#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)
+{
+#ifdef __SSE2__
+    __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] = (_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] = (_rotl(rgba, 16) & 0x00ff00ff) | (rgba & 0xff00ff00);
+            }
+        }
+    }
+#endif
+}
+
+}