--- /dev/null
- context.requireDeviceExtension("VK_EXT_blend_operation_advanced");
+/*------------------------------------------------------------------------
+ * Vulkan Conformance Tests
+ * ------------------------
+ *
+ * Copyright (c) 2019 Valve Corporation.
+ * Copyright (c) 2019 The Khronos Group Inc.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ *
+ *//*!
+ * \file
+ * \brief VK_EXT_blend_operation_advanced tests
+ *//*--------------------------------------------------------------------*/
+
+#include "vktPipelineBlendOperationAdvancedTests.hpp"
+#include "vktPipelineImageUtil.hpp"
+#include "vktPipelineReferenceRenderer.hpp"
+#include "vktTestCaseUtil.hpp"
+#include "vkCmdUtil.hpp"
+#include "vkImageUtil.hpp"
+#include "vkRefUtil.hpp"
+#include "vkQueryUtil.hpp"
+#include "vkTypeUtil.hpp"
+#include "vkBuilderUtil.hpp"
+#include "vkObjUtil.hpp"
+
+#include "tcuTestLog.hpp"
+#include "tcuImageCompare.hpp"
+
+namespace vkt
+{
+namespace pipeline
+{
+
+using namespace vk;
+
+namespace
+{
+using tcu::Vec3;
+using tcu::Vec4;
+
+const deUint32 widthArea = 32u;
+const deUint32 heightArea = 32u;
+
+static const float A1 = 0.750f; // Between 1 and 0.5
+static const float A2 = 0.375f; // Between 0.5 and 0.25
+static const float A3 = 0.125f; // Between 0.25 and 0.0
+
+const Vec4 srcColors[] = {
+ // Test that pre-multiplied is converted correctly.
+ // Should not test invalid premultiplied colours (1, 1, 1, 0).
+ { 1.000f, 0.750f, 0.500f, 1.00f },
+ { 0.250f, 0.125f, 0.000f, 1.00f },
+
+ // Test clamping.
+ { 1.000f, 0.750f, 0.500f, 1.00f },
+ { 0.250f, 0.125f, 0.000f, 1.00f },
+ { 1.000f, 0.750f, 0.500f, 1.00f },
+ { 0.250f, 0.125f, 0.000f, 1.00f },
+
+ // Combinations that test other branches of blend equations.
+ { 1.000f, 0.750f, 0.500f, 1.00f },
+ { 0.250f, 0.125f, 0.000f, 1.00f },
+ { 1.000f, 0.750f, 0.500f, 1.00f },
+ { 0.250f, 0.125f, 0.000f, 1.00f },
+ { 1.000f, 0.750f, 0.500f, 1.00f },
+ { 0.250f, 0.125f, 0.000f, 1.00f },
+ { 1.000f, 0.750f, 0.500f, 1.00f },
+ { 0.250f, 0.125f, 0.000f, 1.00f },
+ { 1.000f, 0.750f, 0.500f, 1.00f },
+ { 0.250f, 0.125f, 0.000f, 1.00f },
+
+ // Above block with few different pre-multiplied alpha values.
+ { 1.000f * A1, 0.750f * A1, 0.500f * A1, 1.00f * A1},
+ { 0.250f * A1, 0.125f * A1, 0.000f * A1, 1.00f * A1},
+ { 1.000f * A1, 0.750f * A1, 0.500f * A1, 1.00f * A1},
+ { 0.250f * A1, 0.125f * A1, 0.000f * A1, 1.00f * A1},
+ { 1.000f * A1, 0.750f * A1, 0.500f * A1, 1.00f * A1},
+ { 0.250f * A1, 0.125f * A1, 0.000f * A1, 1.00f * A1},
+ { 1.000f * A1, 0.750f * A1, 0.500f * A1, 1.00f * A1},
+ { 0.250f * A1, 0.125f * A1, 0.000f * A1, 1.00f * A1},
+ { 1.000f * A1, 0.750f * A1, 0.500f * A1, 1.00f * A1},
+ { 0.250f * A1, 0.125f * A1, 0.000f * A1, 1.00f * A1},
+
+ { 1.000f * A2, 0.750f * A2, 0.500f * A2, 1.00f * A2},
+ { 0.250f * A2, 0.125f * A2, 0.000f * A2, 1.00f * A2},
+ { 1.000f * A2, 0.750f * A2, 0.500f * A2, 1.00f * A2},
+ { 0.250f * A2, 0.125f * A2, 0.000f * A2, 1.00f * A2},
+ { 1.000f * A2, 0.750f * A2, 0.500f * A2, 1.00f * A2},
+ { 0.250f * A2, 0.125f * A2, 0.000f * A2, 1.00f * A2},
+ { 1.000f * A2, 0.750f * A2, 0.500f * A2, 1.00f * A2},
+ { 0.250f * A2, 0.125f * A2, 0.000f * A2, 1.00f * A2},
+ { 1.000f * A2, 0.750f * A2, 0.500f * A2, 1.00f * A2},
+ { 0.250f * A2, 0.125f * A2, 0.000f * A2, 1.00f * A2},
+
+ { 1.000f * A3, 0.750f * A3, 0.500f * A3, 1.00f * A3},
+ { 0.250f * A3, 0.125f * A3, 0.000f * A3, 1.00f * A3},
+ { 1.000f * A3, 0.750f * A3, 0.500f * A3, 1.00f * A3},
+ { 0.250f * A3, 0.125f * A3, 0.000f * A3, 1.00f * A3},
+ { 1.000f * A3, 0.750f * A3, 0.500f * A3, 1.00f * A3},
+ { 0.250f * A3, 0.125f * A3, 0.000f * A3, 1.00f * A3},
+ { 1.000f * A3, 0.750f * A3, 0.500f * A3, 1.00f * A3},
+ { 0.250f * A3, 0.125f * A3, 0.000f * A3, 1.00f * A3},
+ { 1.000f * A3, 0.750f * A3, 0.500f * A3, 1.00f * A3},
+ { 0.250f * A3, 0.125f * A3, 0.000f * A3, 1.00f * A3},
+
+ // Add some source colors with alpha component that is different than the respective destination color
+ { 0.750f, 0.750f, 0.500f, 0.750f },
+ { 0.250f, 0.500f, 0.500f, 0.750f },
+ { 0.250f, 0.125f, 0.000f, 0.500f },
+ { 0.250f, 0.250f, 0.500f, 0.500f },
+ { 0.250f, 0.125f, 0.000f, 0.250f },
+ { 0.125f, 0.125f, 0.125f, 0.250f }};
+
+const Vec4 dstColors[] = {
+ // Test that pre-multiplied is converted correctly.
+ // Should not test invalid premultiplied colours (1, 1, 1, 0).
+ { 0.000f, 0.000f, 0.000f, 0.00f },
+ { 0.000f, 0.000f, 0.000f, 0.00f },
+
+ // Test clamping.
+ { -0.125f, -0.125f, -0.125f, 1.00f },
+ { -0.125f, -0.125f, -0.125f, 1.00f },
+ { 1.125f, 1.125f, 1.125f, 1.00f },
+ { 1.125f, 1.125f, 1.125f, 1.00f },
+
+ // Combinations that test other branches of blend equations.
+ { 1.000f, 1.000f, 1.000f, 1.00f },
+ { 1.000f, 1.000f, 1.000f, 1.00f },
+ { 0.500f, 0.500f, 0.500f, 1.00f },
+ { 0.500f, 0.500f, 0.500f, 1.00f },
+ { 0.250f, 0.250f, 0.250f, 1.00f },
+ { 0.250f, 0.250f, 0.250f, 1.00f },
+ { 0.125f, 0.125f, 0.125f, 1.00f },
+ { 0.125f, 0.125f, 0.125f, 1.00f },
+ { 0.000f, 0.000f, 0.000f, 1.00f },
+ { 0.000f, 0.000f, 0.000f, 1.00f },
+
+ // Above block with few different pre-multiplied alpha values.
+ { 1.000f * A1, 1.000f * A1, 1.000f * A1, 1.00f * A1},
+ { 1.000f * A1, 1.000f * A1, 1.000f * A1, 1.00f * A1},
+ { 0.500f * A1, 0.500f * A1, 0.500f * A1, 1.00f * A1},
+ { 0.500f * A1, 0.500f * A1, 0.500f * A1, 1.00f * A1},
+ { 0.250f * A1, 0.250f * A1, 0.250f * A1, 1.00f * A1},
+ { 0.250f * A1, 0.250f * A1, 0.250f * A1, 1.00f * A1},
+ { 0.125f * A1, 0.125f * A1, 0.125f * A1, 1.00f * A1},
+ { 0.125f * A1, 0.125f * A1, 0.125f * A1, 1.00f * A1},
+ { 0.000f * A1, 0.000f * A1, 0.000f * A1, 1.00f * A1},
+ { 0.000f * A1, 0.000f * A1, 0.000f * A1, 1.00f * A1},
+
+ { 1.000f * A2, 1.000f * A2, 1.000f * A2, 1.00f * A2},
+ { 1.000f * A2, 1.000f * A2, 1.000f * A2, 1.00f * A2},
+ { 0.500f * A2, 0.500f * A2, 0.500f * A2, 1.00f * A2},
+ { 0.500f * A2, 0.500f * A2, 0.500f * A2, 1.00f * A2},
+ { 0.250f * A2, 0.250f * A2, 0.250f * A2, 1.00f * A2},
+ { 0.250f * A2, 0.250f * A2, 0.250f * A2, 1.00f * A2},
+ { 0.125f * A2, 0.125f * A2, 0.125f * A2, 1.00f * A2},
+ { 0.125f * A2, 0.125f * A2, 0.125f * A2, 1.00f * A2},
+ { 0.000f * A2, 0.000f * A2, 0.000f * A2, 1.00f * A2},
+ { 0.000f * A2, 0.000f * A2, 0.000f * A2, 1.00f * A2},
+
+ { 1.000f * A3, 1.000f * A3, 1.000f * A3, 1.00f * A3},
+ { 1.000f * A3, 1.000f * A3, 1.000f * A3, 1.00f * A3},
+ { 0.500f * A3, 0.500f * A3, 0.500f * A3, 1.00f * A3},
+ { 0.500f * A3, 0.500f * A3, 0.500f * A3, 1.00f * A3},
+ { 0.250f * A3, 0.250f * A3, 0.250f * A3, 1.00f * A3 },
+ { 0.250f * A3, 0.250f * A3, 0.250f * A3, 1.00f * A3 },
+ { 0.125f * A3, 0.125f * A3, 0.125f * A3, 1.00f * A3 },
+ { 0.125f * A3, 0.125f * A3, 0.125f * A3, 1.00f * A3 },
+ { 0.000f * A3, 0.000f * A3, 0.000f * A3, 1.00f * A3 },
+ { 0.000f * A3, 0.000f * A3, 0.000f * A3, 1.00f * A3 },
+
+ // Add some source colors with alpha component that is different than the respective source color
+ { 1.000f, 1.000f, 1.000f, 1.000f },
+ { 0.250f, 0.250f, 0.250f, 0.500f },
+ { 0.500f, 0.500f, 0.500f, 0.750f },
+ { 0.250f, 0.250f, 0.250f, 0.250f },
+ { 0.250f, 0.250f, 0.250f, 0.500f },
+ { 0.125f, 0.125f, 0.125f, 0.125f }};
+
+const Vec4 clearColorVec4 (1.0f, 1.0f, 1.0f, 1.0f);
+
+enum TestMode
+{
+ TEST_MODE_GENERIC = 0,
+ TEST_MODE_COHERENT = 1,
+};
+
+struct BlendOperationAdvancedParam
+{
+ TestMode testMode;
+ deUint32 testNumber;
+ std::vector<VkBlendOp> blendOps;
+ deBool coherentOperations;
+ deBool independentBlend;
+ deUint32 colorAttachmentsCount;
+ VkBool32 premultipliedSrcColor;
+ VkBool32 premultipliedDstColor;
+ VkBlendOverlapEXT overlap;
+};
+
+// helper functions
+const std::string generateTestName (struct BlendOperationAdvancedParam param)
+{
+ std::ostringstream result;
+
+ result << ((param.testMode == TEST_MODE_COHERENT && !param.coherentOperations) ? "barrier_" : "");
+ result << "color_attachments_" << param.colorAttachmentsCount;
+ result << "_" << de::toLower(getBlendOverlapEXTStr(param.overlap).toString().substr(3));
+ result << (!param.premultipliedSrcColor ? "_nonpremultipliedsrc" : "");
+ result << (!param.premultipliedDstColor ? "_nonpremultiplieddst" : "");
+ result << "_" << param.testNumber;
+ return result.str();
+}
+
+const std::string generateTestDescription ()
+{
+ std::string result("Test advanced blend operations");
+ return result;
+}
+
+Vec3 calculateWeightingFactors(BlendOperationAdvancedParam param,
+ float alphaSrc, float alphaDst)
+{
+ Vec3 p = Vec3(0.0f, 0.0f, 0.0f);
+ switch(param.overlap)
+ {
+ case VK_BLEND_OVERLAP_UNCORRELATED_EXT:
+ p.x() = alphaSrc * alphaDst;
+ p.y() = alphaSrc * (1.0f - alphaDst);
+ p.z() = alphaDst * (1.0f - alphaSrc);
+ break;
+ case VK_BLEND_OVERLAP_CONJOINT_EXT:
+ p.x() = deFloatMin(alphaSrc, alphaDst);
+ p.y() = deFloatMax(alphaSrc - alphaDst, 0.0f);
+ p.z() = deFloatMax(alphaDst - alphaSrc, 0.0f);
+ break;
+ case VK_BLEND_OVERLAP_DISJOINT_EXT:
+ p.x() = deFloatMax(alphaSrc + alphaDst - 1.0f, 0.0f);
+ p.y() = deFloatMin(alphaSrc, 1.0f - alphaDst);
+ p.z() = deFloatMin(alphaDst, 1.0f - alphaSrc);
+ break;
+ default:
+ DE_FATAL("Unsupported Advanced Blend Overlap Mode");
+ };
+ return p;
+}
+
+ Vec3 calculateXYZFactors(VkBlendOp op)
+{
+ Vec3 xyz = Vec3(0.0f, 0.0f, 0.0f);
+ switch (op)
+ {
+ case VK_BLEND_OP_ZERO_EXT:
+ xyz = Vec3(0.0f, 0.0f, 0.0f);
+ break;
+
+ case VK_BLEND_OP_DST_ATOP_EXT:
+ case VK_BLEND_OP_SRC_EXT:
+ xyz = Vec3(1.0f, 1.0f, 0.0f);
+ break;
+
+ case VK_BLEND_OP_DST_EXT:
+ xyz = Vec3(1.0f, 0.0f, 1.0f);
+ break;
+
+ case VK_BLEND_OP_HSL_LUMINOSITY_EXT:
+ case VK_BLEND_OP_HSL_COLOR_EXT:
+ case VK_BLEND_OP_HSL_SATURATION_EXT:
+ case VK_BLEND_OP_HSL_HUE_EXT:
+ case VK_BLEND_OP_HARDMIX_EXT:
+ case VK_BLEND_OP_PINLIGHT_EXT:
+ case VK_BLEND_OP_LINEARLIGHT_EXT:
+ case VK_BLEND_OP_VIVIDLIGHT_EXT:
+ case VK_BLEND_OP_LINEARBURN_EXT:
+ case VK_BLEND_OP_LINEARDODGE_EXT:
+ case VK_BLEND_OP_EXCLUSION_EXT:
+ case VK_BLEND_OP_DIFFERENCE_EXT:
+ case VK_BLEND_OP_SOFTLIGHT_EXT:
+ case VK_BLEND_OP_HARDLIGHT_EXT:
+ case VK_BLEND_OP_COLORBURN_EXT:
+ case VK_BLEND_OP_COLORDODGE_EXT:
+ case VK_BLEND_OP_LIGHTEN_EXT:
+ case VK_BLEND_OP_DARKEN_EXT:
+ case VK_BLEND_OP_OVERLAY_EXT:
+ case VK_BLEND_OP_SCREEN_EXT:
+ case VK_BLEND_OP_MULTIPLY_EXT:
+ case VK_BLEND_OP_SRC_OVER_EXT:
+ case VK_BLEND_OP_DST_OVER_EXT:
+ xyz = Vec3(1.0f, 1.0f, 1.0f);
+ break;
+
+ case VK_BLEND_OP_SRC_IN_EXT:
+ case VK_BLEND_OP_DST_IN_EXT:
+ xyz = Vec3(1.0f, 0.0f, 0.0f);
+ break;
+
+ case VK_BLEND_OP_SRC_OUT_EXT:
+ xyz = Vec3(0.0f, 1.0f, 0.0f);
+ break;
+
+ case VK_BLEND_OP_DST_OUT_EXT:
+ xyz = Vec3(0.0f, 0.0f, 1.0f);
+ break;
+
+ case VK_BLEND_OP_INVERT_RGB_EXT:
+ case VK_BLEND_OP_INVERT_EXT:
+ case VK_BLEND_OP_SRC_ATOP_EXT:
+ xyz = Vec3(1.0f, 0.0f, 1.0f);
+ break;
+
+ case VK_BLEND_OP_XOR_EXT:
+ xyz = Vec3(0.0f, 1.0f, 1.0f);
+ break;
+
+ default:
+ DE_FATAL("Unsupported f/X/Y/Z Advanced Blend Operations Mode");
+ };
+
+ return xyz;
+}
+
+float blendOpOverlay(float src, float dst)
+{
+ if (dst <= 0.5f)
+ return (2.0f * src * dst);
+ else
+ return (1.0f - (2.0f * (1.0f - src) * (1.0f - dst)));
+}
+
+float blendOpColorDodge(float src, float dst)
+{
+ if (dst <= 0.0f)
+ return 0.0f;
+ else if (src < 1.0f)
+ return deFloatMin(1.0f, (dst / (1.0f - src)));
+ else
+ return 1.0f;
+}
+
+float blendOpColorBurn(float src, float dst)
+{
+ if (dst >= 1.0f)
+ return 1.0f;
+ else if (src > 0.0f)
+ return 1.0f - deFloatMin(1.0f, (1.0f - dst) / src);
+ else
+ return 0.0f;
+}
+
+float blendOpHardlight(float src, float dst)
+{
+ if (src <= 0.5f)
+ return 2.0f * src * dst;
+ else
+ return 1.0f - (2.0f * (1.0f - src) * (1.0f - dst));
+}
+
+float blendOpSoftlight(float src, float dst)
+{
+ if (src <= 0.5f)
+ return dst - ((1.0f - (2.0f * src)) * dst * (1.0f - dst));
+ else if (dst <= 0.25f)
+ return dst + (((2.0f * src) - 1.0f) * dst * ((((16.0f * dst) - 12.0f) * dst) + 3.0f));
+ else
+ return dst + (((2.0f * src) - 1.0f) * (deFloatSqrt(dst) - dst));
+}
+
+float blendOpLinearDodge(float src, float dst)
+{
+ if ((src + dst) <= 1.0f)
+ return src + dst;
+ else
+ return 1.0f;
+}
+
+float blendOpLinearBurn(float src, float dst)
+{
+ if ((src + dst) > 1.0f)
+ return src + dst - 1.0f;
+ else
+ return 0.0f;
+}
+
+float blendOpVividLight(float src, float dst)
+{
+ if (src <= 0.0f)
+ return 0.0f;
+ if (src < 0.5f)
+ return 1.0f - (deFloatMin(1.0f, (1.0f - dst) / (2.0f * src)));
+ if (src < 1.0f)
+ return deFloatMin(1.0f, dst / (2.0f * (1.0f - src)));
+ else
+ return 1.0f;
+}
+
+float blendOpLinearLight(float src, float dst)
+{
+ if ((2.0f * src + dst) > 2.0f)
+ return 1.0f;
+ if ((2.0f * src + dst) <= 1.0f)
+ return 0.0f;
+ return (2.0f * src) + dst - 1.0f;
+}
+
+float blendOpPinLight(float src, float dst)
+{
+ if (((2.0f * src - 1.0f) > dst) && src < 0.5f)
+ return 0.0f;
+ if (((2.0f * src - 1.0f) > dst) && src >= 0.5f)
+ return 2.0f * src - 1.0f;
+ if (((2.0f * src - 1.0f) <= dst) && src < (0.5f * dst))
+ return 2.0f * src;
+ if (((2.0f * src - 1.0f) <= dst) && src >= (0.5f * dst))
+ return dst;
+ return 0.0f;
+}
+
+float blendOpHardmix(float src, float dst)
+{
+ if ((src + dst) < 1.0f)
+ return 0.0f;
+ else
+ return 1.0f;
+}
+
+float minv3(Vec3 c)
+{
+ return deFloatMin(deFloatMin(c.x(), c.y()), c.z());
+}
+
+float maxv3(Vec3 c)
+{
+ return deFloatMax(deFloatMax(c.x(), c.y()), c.z());
+}
+
+float lumv3(Vec3 c)
+{
+ return dot(c, Vec3(0.3f, 0.59f, 0.11f));
+}
+
+float satv3(Vec3 c)
+{
+ return maxv3(c) - minv3(c);
+}
+
+// If any color components are outside [0,1], adjust the color to
+// get the components in range.
+Vec3 clipColor(Vec3 color)
+{
+ float lum = lumv3(color);
+ float mincol = minv3(color);
+ float maxcol = maxv3(color);
+
+ if (mincol < 0.0)
+ {
+ color = lum + ((color - lum) * lum) / (lum - mincol);
+ }
+ if (maxcol > 1.0)
+ {
+ color = lum + ((color - lum) * (1.0f - lum)) / (maxcol - lum);
+ }
+ return color;
+}
+
+// Take the base RGB color <cbase> and override its luminosity
+// with that of the RGB color <clum>.
+Vec3 setLum(Vec3 cbase, Vec3 clum)
+{
+ float lbase = lumv3(cbase);
+ float llum = lumv3(clum);
+ float ldiff = llum - lbase;
+
+ Vec3 color = cbase + Vec3(ldiff);
+ return clipColor(color);
+}
+
+// Take the base RGB color <cbase> and override its saturation with
+// that of the RGB color <csat>. The override the luminosity of the
+// result with that of the RGB color <clum>.
+Vec3 setLumSat(Vec3 cbase, Vec3 csat, Vec3 clum)
+{
+ float minbase = minv3(cbase);
+ float sbase = satv3(cbase);
+ float ssat = satv3(csat);
+ Vec3 color;
+
+ if (sbase > 0)
+ {
+ // Equivalent (modulo rounding errors) to setting the
+ // smallest (R,G,B) component to 0, the largest to <ssat>,
+ // and interpolating the "middle" component based on its
+ // original value relative to the smallest/largest.
+ color = (cbase - minbase) * ssat / sbase;
+ } else {
+ color = Vec3(0.0f);
+ }
+ return setLum(color, clum);
+}
+
+Vec3 calculateFFunction(VkBlendOp op,
+ Vec3 src, Vec3 dst)
+{
+ Vec3 f = Vec3(0.0f, 0.0f, 0.0f);
+
+ switch (op)
+ {
+ case VK_BLEND_OP_XOR_EXT:
+ case VK_BLEND_OP_SRC_OUT_EXT:
+ case VK_BLEND_OP_DST_OUT_EXT:
+ case VK_BLEND_OP_ZERO_EXT:
+ f = Vec3(0.0f, 0.0f, 0.0f);
+ break;
+
+ case VK_BLEND_OP_SRC_ATOP_EXT:
+ case VK_BLEND_OP_SRC_IN_EXT:
+ case VK_BLEND_OP_SRC_OVER_EXT:
+ case VK_BLEND_OP_SRC_EXT:
+ f = src;
+ break;
+
+ case VK_BLEND_OP_DST_ATOP_EXT:
+ case VK_BLEND_OP_DST_IN_EXT:
+ case VK_BLEND_OP_DST_OVER_EXT:
+ case VK_BLEND_OP_DST_EXT:
+ f = dst;
+ break;
+
+ case VK_BLEND_OP_MULTIPLY_EXT:
+ f = src * dst;
+ break;
+
+ case VK_BLEND_OP_SCREEN_EXT:
+ f = src + dst - (src*dst);
+ break;
+
+ case VK_BLEND_OP_OVERLAY_EXT:
+ f.x() = blendOpOverlay(src.x(), dst.x());
+ f.y() = blendOpOverlay(src.y(), dst.y());
+ f.z() = blendOpOverlay(src.z(), dst.z());
+ break;
+
+ case VK_BLEND_OP_DARKEN_EXT:
+ f.x() = deFloatMin(src.x(), dst.x());
+ f.y() = deFloatMin(src.y(), dst.y());
+ f.z() = deFloatMin(src.z(), dst.z());
+ break;
+
+ case VK_BLEND_OP_LIGHTEN_EXT:
+ f.x() = deFloatMax(src.x(), dst.x());
+ f.y() = deFloatMax(src.y(), dst.y());
+ f.z() = deFloatMax(src.z(), dst.z());
+ break;
+
+ case VK_BLEND_OP_COLORDODGE_EXT:
+ f.x() = blendOpColorDodge(src.x(), dst.x());
+ f.y() = blendOpColorDodge(src.y(), dst.y());
+ f.z() = blendOpColorDodge(src.z(), dst.z());
+ break;
+
+ case VK_BLEND_OP_COLORBURN_EXT:
+ f.x() = blendOpColorBurn(src.x(), dst.x());
+ f.y() = blendOpColorBurn(src.y(), dst.y());
+ f.z() = blendOpColorBurn(src.z(), dst.z());
+ break;
+
+ case VK_BLEND_OP_HARDLIGHT_EXT:
+ f.x() = blendOpHardlight(src.x(), dst.x());
+ f.y() = blendOpHardlight(src.y(), dst.y());
+ f.z() = blendOpHardlight(src.z(), dst.z());
+ break;
+
+ case VK_BLEND_OP_SOFTLIGHT_EXT:
+ f.x() = blendOpSoftlight(src.x(), dst.x());
+ f.y() = blendOpSoftlight(src.y(), dst.y());
+ f.z() = blendOpSoftlight(src.z(), dst.z());
+ break;
+
+ case VK_BLEND_OP_DIFFERENCE_EXT:
+ f.x() = deFloatAbs(dst.x() - src.x());
+ f.y() = deFloatAbs(dst.y() - src.y());
+ f.z() = deFloatAbs(dst.z() - src.z());
+ break;
+
+
+ case VK_BLEND_OP_EXCLUSION_EXT:
+ f = src + dst - (2.0f * src * dst);
+ break;
+
+ case VK_BLEND_OP_INVERT_EXT:
+ f = 1.0f - dst;
+ break;
+
+ case VK_BLEND_OP_INVERT_RGB_EXT:
+ f = src * (1.0f - dst);
+ break;
+
+ case VK_BLEND_OP_LINEARDODGE_EXT:
+ f.x() = blendOpLinearDodge(src.x(), dst.x());
+ f.y() = blendOpLinearDodge(src.y(), dst.y());
+ f.z() = blendOpLinearDodge(src.z(), dst.z());
+ break;
+
+ case VK_BLEND_OP_LINEARBURN_EXT:
+ f.x() = blendOpLinearBurn(src.x(), dst.x());
+ f.y() = blendOpLinearBurn(src.y(), dst.y());
+ f.z() = blendOpLinearBurn(src.z(), dst.z());
+ break;
+
+ case VK_BLEND_OP_VIVIDLIGHT_EXT:
+ f.x() = blendOpVividLight(src.x(), dst.x());
+ f.y() = blendOpVividLight(src.y(), dst.y());
+ f.z() = blendOpVividLight(src.z(), dst.z());
+ break;
+
+ case VK_BLEND_OP_LINEARLIGHT_EXT:
+ f.x() = blendOpLinearLight(src.x(), dst.x());
+ f.y() = blendOpLinearLight(src.y(), dst.y());
+ f.z() = blendOpLinearLight(src.z(), dst.z());
+ break;
+
+ case VK_BLEND_OP_PINLIGHT_EXT:
+ f.x() = blendOpPinLight(src.x(), dst.x());
+ f.y() = blendOpPinLight(src.y(), dst.y());
+ f.z() = blendOpPinLight(src.z(), dst.z());
+ break;
+
+ case VK_BLEND_OP_HARDMIX_EXT:
+ f.x() = blendOpHardmix(src.x(), dst.x());
+ f.y() = blendOpHardmix(src.y(), dst.y());
+ f.z() = blendOpHardmix(src.z(), dst.z());
+ break;
+
+ case VK_BLEND_OP_HSL_HUE_EXT:
+ f = setLumSat(src, dst, dst);
+ break;
+
+ case VK_BLEND_OP_HSL_SATURATION_EXT:
+ f = setLumSat(dst, src, dst);
+ break;
+
+ case VK_BLEND_OP_HSL_COLOR_EXT:
+ f = setLum(src, dst);
+ break;
+
+ case VK_BLEND_OP_HSL_LUMINOSITY_EXT:
+ f = setLum(dst, src);
+ break;
+
+ default:
+ DE_FATAL("Unsupported f/X/Y/Z Advanced Blend Operations Mode");
+ };
+
+ return f;
+}
+
+Vec4 additionalRGBBlendOperations(VkBlendOp op,
+ Vec4 src, Vec4 dst)
+{
+ Vec4 res = Vec4(0.0f, 0.0f, 0.0f, 1.0f);
+
+ switch (op)
+ {
+ case VK_BLEND_OP_PLUS_EXT:
+ res = src + dst;
+ break;
+
+ case VK_BLEND_OP_PLUS_CLAMPED_EXT:
+ res.x() = deFloatMin(1.0f, src.x() + dst.x());
+ res.y() = deFloatMin(1.0f, src.y() + dst.y());
+ res.z() = deFloatMin(1.0f, src.z() + dst.z());
+ res.w() = deFloatMin(1.0f, src.w() + dst.w());
+ break;
+
+ case VK_BLEND_OP_PLUS_CLAMPED_ALPHA_EXT:
+ res.x() = deFloatMin(deFloatMin(1.0f, src.w() + dst.w()), src.x() + dst.x());
+ res.y() = deFloatMin(deFloatMin(1.0f, src.w() + dst.w()), src.y() + dst.y());
+ res.z() = deFloatMin(deFloatMin(1.0f, src.w() + dst.w()), src.z() + dst.z());
+ res.w() = deFloatMin(1.0f, src.w() + dst.w());
+ break;
+
+ case VK_BLEND_OP_PLUS_DARKER_EXT:
+ res.x() = deFloatMax(0.0f, deFloatMin(1.0f, src.w() + dst.w()) - ((src.w() - src.x()) + (dst.w() - dst.x())));
+ res.y() = deFloatMax(0.0f, deFloatMin(1.0f, src.w() + dst.w()) - ((src.w() - src.y()) + (dst.w() - dst.y())));
+ res.z() = deFloatMax(0.0f, deFloatMin(1.0f, src.w() + dst.w()) - ((src.w() - src.z()) + (dst.w() - dst.z())));
+ res.w() = deFloatMin(1.0f, src.w() + dst.w());
+ break;
+
+ case VK_BLEND_OP_MINUS_EXT:
+ res = dst - src;
+ break;
+
+ case VK_BLEND_OP_MINUS_CLAMPED_EXT:
+ res.x() = deFloatMax(0.0f, dst.x() - src.x());
+ res.y() = deFloatMax(0.0f, dst.y() - src.y());
+ res.z() = deFloatMax(0.0f, dst.z() - src.z());
+ res.w() = deFloatMax(0.0f, dst.w() - src.w());
+ break;
+
+ case VK_BLEND_OP_CONTRAST_EXT:
+ res.x() = (dst.w() / 2.0f) + 2.0f * (dst.x() - (dst.w() / 2.0f)) * (src.x() - (src.w() / 2.0f));
+ res.y() = (dst.w() / 2.0f) + 2.0f * (dst.y() - (dst.w() / 2.0f)) * (src.y() - (src.w() / 2.0f));
+ res.z() = (dst.w() / 2.0f) + 2.0f * (dst.z() - (dst.w() / 2.0f)) * (src.z() - (src.w() / 2.0f));
+ res.w() = dst.w();
+ break;
+
+ case VK_BLEND_OP_INVERT_OVG_EXT:
+ res.x() = src.w() * (1.0f - dst.x()) + (1.0f - src.w()) * dst.x();
+ res.y() = src.w() * (1.0f - dst.y()) + (1.0f - src.w()) * dst.y();
+ res.z() = src.w() * (1.0f - dst.z()) + (1.0f - src.w()) * dst.z();
+ res.w() = src.w() + dst.w() - src.w() * dst.w();
+ break;
+
+ case VK_BLEND_OP_RED_EXT:
+ res = dst;
+ res.x() = src.x();
+ break;
+
+ case VK_BLEND_OP_GREEN_EXT:
+ res = dst;
+ res.y() = src.y();
+ break;
+
+ case VK_BLEND_OP_BLUE_EXT:
+ res = dst;
+ res.z() = src.z();
+ break;
+
+ default:
+ DE_FATAL("Unsupported blend operation");
+ };
+ return res;
+}
+
+Vec4 calculateFinalColor(BlendOperationAdvancedParam param, VkBlendOp op,
+ Vec4 source, Vec4 destination)
+{
+ Vec4 result = Vec4(0.0f, 0.0f, 0.0f, 1.0f);
+ Vec3 srcColor = source.xyz();
+ Vec3 dstColor = destination.xyz();
+
+ // Calculate weighting factors
+ Vec3 p = calculateWeightingFactors(param, source.w(), destination.w());
+
+ if (op > VK_BLEND_OP_MAX && op < VK_BLEND_OP_PLUS_EXT)
+ {
+ {
+ // If srcPremultiplied is set to VK_TRUE, the fragment color components
+ // are considered to have been premultiplied by the A component prior to
+ // blending. The base source color (Rs',Gs',Bs') is obtained by dividing
+ // through by the A component.
+ if (param.premultipliedSrcColor)
+ {
+ if (source.w() != 0.0f)
+ srcColor = srcColor / source.w();
+ else
+ srcColor = Vec3(0.0f, 0.0f, 0.0f);
+ }
+ // If dstPremultiplied is set to VK_TRUE, the destination components are
+ // considered to have been premultiplied by the A component prior to
+ // blending. The base destination color (Rd',Gd',Bd') is obtained by dividing
+ // through by the A component.
+ if (param.premultipliedDstColor)
+ {
+ if (destination.w() != 0.0f)
+ dstColor = dstColor / destination.w();
+ else
+ dstColor = Vec3(0.0f, 0.0f, 0.0f);
+ }
+ }
+
+ // Calculate X, Y, Z terms of the equation
+ Vec3 xyz = calculateXYZFactors(op);
+ Vec3 fSrcDst = calculateFFunction(op, srcColor, dstColor);
+
+ result.x() = fSrcDst.x() * p.x() + xyz.y() * srcColor.x() * p.y() + xyz.z() * dstColor.x() * p.z();
+ result.y() = fSrcDst.y() * p.x() + xyz.y() * srcColor.y() * p.y() + xyz.z() * dstColor.y() * p.z();
+ result.z() = fSrcDst.z() * p.x() + xyz.y() * srcColor.z() * p.y() + xyz.z() * dstColor.z() * p.z();
+ result.w() = xyz.x() * p.x() + xyz.y() * p.y() + xyz.z() * p.z();
+ }
+ else if (op >= VK_BLEND_OP_PLUS_EXT && op < VK_BLEND_OP_MAX_ENUM)
+ {
+ // Premultiply colors for additional RGB blend operations. The formula is different than the rest of operations.
+ {
+ if (!param.premultipliedSrcColor)
+ {
+ srcColor = srcColor * source.w();
+ }
+
+ if (!param.premultipliedDstColor)
+ {
+ dstColor = dstColor * destination.w();
+ }
+
+ }
+ Vec4 src = Vec4(srcColor.x(), srcColor.y(), srcColor.z(), source.w());
+ Vec4 dst = Vec4(dstColor.x(), dstColor.y(), dstColor.z(), destination.w());
+ result = additionalRGBBlendOperations(op, src, dst);
+ }
+ else
+ {
+ DE_FATAL("Unsupported Blend Operation");
+ }
+ return result;
+}
+
+static inline void getCoordinates (deUint32 index, deInt32 &x, deInt32 &y)
+{
+ x = index % widthArea;
+ y = index / heightArea;
+}
+
+static inline std::vector<Vec4> createPoints (void)
+{
+ std::vector<Vec4> vertices;
+ vertices.push_back(Vec4(-1.0f, -1.0f, 0.0f, 1.0f));
+ vertices.push_back(Vec4( 1.0f, 1.0f, 0.0f, 1.0f));
+ vertices.push_back(Vec4(-1.0f, 1.0f, 0.0f, 1.0f));
+ vertices.push_back(Vec4(-1.0f, -1.0f, 0.0f, 1.0f));
+ vertices.push_back(Vec4( 1.0f, 1.0f, 0.0f, 1.0f));
+ vertices.push_back(Vec4( 1.0f, -1.0f, 0.0f, 1.0f));
+ return vertices;
+}
+
+template <class Test>
+vkt::TestCase* newTestCase (tcu::TestContext& testContext,
+ const BlendOperationAdvancedParam testParam)
+{
+ return new Test(testContext,
+ generateTestName(testParam).c_str(),
+ generateTestDescription().c_str(),
+ testParam);
+}
+
+Move<VkRenderPass> makeTestRenderPass (BlendOperationAdvancedParam param,
+ const DeviceInterface& vk,
+ const VkDevice device,
+ const VkFormat colorFormat,
+ VkAttachmentLoadOp colorLoadOp = VK_ATTACHMENT_LOAD_OP_CLEAR)
+{
+ const VkAttachmentDescription colorAttachmentDescription =
+ {
+ (VkAttachmentDescriptionFlags)0, // VkAttachmentDescriptionFlags flags
+ colorFormat, // VkFormat format
+ VK_SAMPLE_COUNT_1_BIT, // VkSampleCountFlagBits samples
+ colorLoadOp, // VkAttachmentLoadOp loadOp
+ VK_ATTACHMENT_STORE_OP_STORE, // VkAttachmentStoreOp storeOp
+ VK_ATTACHMENT_LOAD_OP_DONT_CARE, // VkAttachmentLoadOp stencilLoadOp
+ VK_ATTACHMENT_STORE_OP_DONT_CARE, // VkAttachmentStoreOp stencilStoreOp
+ (colorLoadOp == VK_ATTACHMENT_LOAD_OP_LOAD) ?
+ VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL :
+ VK_IMAGE_LAYOUT_UNDEFINED, // VkImageLayout initialLayout
+ VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL // VkImageLayout finalLayout
+ };
+
+ std::vector<VkAttachmentDescription> attachmentDescriptions;
+ std::vector<VkAttachmentReference> colorAttachmentRefs;
+
+
+ for (deUint32 i = 0; i < param.colorAttachmentsCount; i++)
+ {
+ attachmentDescriptions.push_back(colorAttachmentDescription);
+ const VkAttachmentReference colorAttachmentRef =
+ {
+ i, // deUint32 attachment
+ VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL // VkImageLayout layout
+ };
+
+ colorAttachmentRefs.push_back(colorAttachmentRef);
+ }
+
+ const VkSubpassDescription subpassDescription =
+ {
+ (VkSubpassDescriptionFlags)0, // VkSubpassDescriptionFlags flags
+ VK_PIPELINE_BIND_POINT_GRAPHICS, // VkPipelineBindPoint pipelineBindPoint
+ 0u, // deUint32 inputAttachmentCount
+ DE_NULL, // const VkAttachmentReference* pInputAttachments
+ param.colorAttachmentsCount, // deUint32 colorAttachmentCount
+ colorAttachmentRefs.data(), // const VkAttachmentReference* pColorAttachments
+ DE_NULL, // const VkAttachmentReference* pResolveAttachments
+ DE_NULL, // const VkAttachmentReference* pDepthStencilAttachment
+ 0u, // deUint32 preserveAttachmentCount
+ DE_NULL // const deUint32* pPreserveAttachments
+ };
+
+ const VkRenderPassCreateInfo renderPassInfo =
+ {
+ VK_STRUCTURE_TYPE_RENDER_PASS_CREATE_INFO, // VkStructureType sType
+ DE_NULL, // const void* pNext
+ (VkRenderPassCreateFlags)0, // VkRenderPassCreateFlags flags
+ (deUint32)attachmentDescriptions.size(), // deUint32 attachmentCount
+ attachmentDescriptions.data(), // const VkAttachmentDescription* pAttachments
+ 1u, // deUint32 subpassCount
+ &subpassDescription, // const VkSubpassDescription* pSubpasses
+ 0u, // deUint32 dependencyCount
+ DE_NULL // const VkSubpassDependency* pDependencies
+ };
+
+ return createRenderPass(vk, device, &renderPassInfo, DE_NULL);
+}
+
+Move<VkBuffer> createBufferAndBindMemory (Context& context, VkDeviceSize size, VkBufferUsageFlags usage, de::MovePtr<Allocation>* pAlloc)
+{
+ const DeviceInterface& vk = context.getDeviceInterface();
+ const VkDevice vkDevice = context.getDevice();
+ const deUint32 queueFamilyIndex = context.getUniversalQueueFamilyIndex();
+
+ const VkBufferCreateInfo vertexBufferParams =
+ {
+ VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO, // VkStructureType sType;
+ DE_NULL, // const void* pNext;
+ 0u, // VkBufferCreateFlags flags;
+ size, // VkDeviceSize size;
+ usage, // VkBufferUsageFlags usage;
+ VK_SHARING_MODE_EXCLUSIVE, // VkSharingMode sharingMode;
+ 1u, // deUint32 queueFamilyCount;
+ &queueFamilyIndex // const deUint32* pQueueFamilyIndices;
+ };
+
+ Move<VkBuffer> vertexBuffer = createBuffer(vk, vkDevice, &vertexBufferParams);
+
+ *pAlloc = context.getDefaultAllocator().allocate(getBufferMemoryRequirements(vk, vkDevice, *vertexBuffer), MemoryRequirement::HostVisible);
+ VK_CHECK(vk.bindBufferMemory(vkDevice, *vertexBuffer, (*pAlloc)->getMemory(), (*pAlloc)->getOffset()));
+
+ return vertexBuffer;
+}
+
+Move<VkImage> createImage2DAndBindMemory (Context& context,
+ VkFormat format,
+ deUint32 width,
+ deUint32 height,
+ VkImageUsageFlags usage,
+ VkSampleCountFlagBits sampleCount,
+ de::details::MovePtr<Allocation>* pAlloc)
+{
+ const DeviceInterface& vk = context.getDeviceInterface();
+ const VkDevice vkDevice = context.getDevice();
+ const deUint32 queueFamilyIndex = context.getUniversalQueueFamilyIndex();
+
+ const VkImageCreateInfo colorImageParams =
+ {
+ VK_STRUCTURE_TYPE_IMAGE_CREATE_INFO, // VkStructureType sType;
+ DE_NULL, // const void* pNext;
+ 0u, // VkImageCreateFlags flags;
+ VK_IMAGE_TYPE_2D, // VkImageType imageType;
+ format, // VkFormat format;
+ { width, height, 1u }, // VkExtent3D extent;
+ 1u, // deUint32 mipLevels;
+ 1u, // deUint32 arraySize;
+ sampleCount, // deUint32 samples;
+ VK_IMAGE_TILING_OPTIMAL, // VkImageTiling tiling;
+ usage, // VkImageUsageFlags usage;
+ VK_SHARING_MODE_EXCLUSIVE, // VkSharingMode sharingMode;
+ 1u, // deUint32 queueFamilyCount;
+ &queueFamilyIndex, // const deUint32* pQueueFamilyIndices;
+ VK_IMAGE_LAYOUT_UNDEFINED, // VkImageLayout initialLayout;
+ };
+
+ Move<VkImage> image = createImage(vk, vkDevice, &colorImageParams);
+
+ *pAlloc = context.getDefaultAllocator().allocate(getImageMemoryRequirements(vk, vkDevice, *image), MemoryRequirement::Any);
+ VK_CHECK(vk.bindImageMemory(vkDevice, *image, (*pAlloc)->getMemory(), (*pAlloc)->getOffset()));
+
+ return image;
+}
+
+// Test Classes
+class BlendOperationAdvancedTestInstance : public vkt::TestInstance
+{
+public:
+ BlendOperationAdvancedTestInstance (Context& context,
+ const BlendOperationAdvancedParam param);
+ virtual ~BlendOperationAdvancedTestInstance (void);
+ virtual tcu::TestStatus iterate (void);
+protected:
+ void prepareRenderPass (VkFramebuffer framebuffer, VkPipeline pipeline) const;
+ void prepareCommandBuffer (void) const;
+ void buildPipeline (VkBool32 premultiplySrc, VkBool32 premultiplyDst);
+ void bindShaderStage (VkShaderStageFlagBits stage,
+ const char* sourceName,
+ const char* entryName);
+ deBool verifyTestResult (void);
+protected:
+ const BlendOperationAdvancedParam m_param;
+ const tcu::UVec2 m_renderSize;
+ const VkFormat m_colorFormat;
+ Move<VkPipelineLayout> m_pipelineLayout;
+
+ Move<VkBuffer> m_vertexBuffer;
+ de::MovePtr<Allocation> m_vertexBufferMemory;
+ std::vector<Vec4> m_vertices;
+
+ Move<VkRenderPass> m_renderPass;
+ Move<VkCommandPool> m_cmdPool;
+ Move<VkCommandBuffer> m_cmdBuffer;
+ std::vector<Move<VkImage>> m_colorImages;
+ std::vector<Move<VkImageView>> m_colorAttachmentViews;
+ std::vector<de::MovePtr<Allocation>> m_colorImageAllocs;
+ std::vector<VkImageMemoryBarrier> m_imageLayoutBarriers;
+ Move<VkFramebuffer> m_framebuffer;
+ Move<VkPipeline> m_pipeline;
+
+ Move<VkShaderModule> m_shaderModules[2];
+ deUint32 m_shaderStageCount;
+ VkPipelineShaderStageCreateInfo m_shaderStageInfo[2];
+};
+
+void BlendOperationAdvancedTestInstance::bindShaderStage (VkShaderStageFlagBits stage,
+ const char* sourceName,
+ const char* entryName)
+{
+ const DeviceInterface& vk = m_context.getDeviceInterface();
+ const VkDevice vkDevice = m_context.getDevice();
+
+ // Create shader module
+ deUint32* code = (deUint32*)m_context.getBinaryCollection().get(sourceName).getBinary();
+ deUint32 codeSize = (deUint32)m_context.getBinaryCollection().get(sourceName).getSize();
+
+ const VkShaderModuleCreateInfo moduleCreateInfo =
+ {
+ VK_STRUCTURE_TYPE_SHADER_MODULE_CREATE_INFO, // VkStructureType sType;
+ DE_NULL, // const void* pNext;
+ 0u, // VkShaderModuleCreateFlags flags;
+ codeSize, // deUintptr codeSize;
+ code, // const deUint32* pCode;
+ };
+
+ m_shaderModules[m_shaderStageCount] = createShaderModule(vk, vkDevice, &moduleCreateInfo);
+
+ // Prepare shader stage info
+ m_shaderStageInfo[m_shaderStageCount].sType = VK_STRUCTURE_TYPE_PIPELINE_SHADER_STAGE_CREATE_INFO;
+ m_shaderStageInfo[m_shaderStageCount].pNext = DE_NULL;
+ m_shaderStageInfo[m_shaderStageCount].flags = 0u;
+ m_shaderStageInfo[m_shaderStageCount].stage = stage;
+ m_shaderStageInfo[m_shaderStageCount].module = *m_shaderModules[m_shaderStageCount];
+ m_shaderStageInfo[m_shaderStageCount].pName = entryName;
+ m_shaderStageInfo[m_shaderStageCount].pSpecializationInfo = DE_NULL;
+
+ m_shaderStageCount++;
+}
+
+void BlendOperationAdvancedTestInstance::buildPipeline (VkBool32 srcPremultiplied,
+ VkBool32 dstPremultiplied)
+{
+ const DeviceInterface& vk = m_context.getDeviceInterface();
+ const VkDevice vkDevice = m_context.getDevice();
+
+ // Create pipeline
+ const VkVertexInputBindingDescription vertexInputBindingDescription =
+ {
+ 0u, // deUint32 binding;
+ sizeof(Vec4), // deUint32 strideInBytes;
+ VK_VERTEX_INPUT_RATE_VERTEX, // VkVertexInputRate inputRate;
+ };
+
+ const VkVertexInputAttributeDescription vertexInputAttributeDescription =
+ {
+ 0u, // deUint32 location;
+ 0u, // deUint32 binding;
+ VK_FORMAT_R32G32B32A32_SFLOAT, // VkFormat format;
+ 0u // deUint32 offsetInBytes;
+ };
+
+ const VkPipelineVertexInputStateCreateInfo vertexInputStateParams =
+ {
+ VK_STRUCTURE_TYPE_PIPELINE_VERTEX_INPUT_STATE_CREATE_INFO, // VkStructureType sType;
+ DE_NULL, // const void* pNext;
+ 0u, // VkPipelineVertexInputStateCreateFlags flags;
+ 1u, // deUint32 vertexBindingDescriptionCount;
+ &vertexInputBindingDescription, // const VkVertexInputBindingDescription* pVertexBindingDescriptions;
+ 1u, // deUint32 vertexAttributeDescriptionCount;
+ &vertexInputAttributeDescription, // const VkVertexInputAttributeDescription* pVertexAttributeDescriptions;
+ };
+
+ const VkPipelineInputAssemblyStateCreateInfo inputAssemblyStateParams =
+ {
+ VK_STRUCTURE_TYPE_PIPELINE_INPUT_ASSEMBLY_STATE_CREATE_INFO, // VkStructureType sType;
+ DE_NULL, // const void* pNext;
+ 0u, // VkPipelineInputAssemblyStateCreateFlags flags;
+ VK_PRIMITIVE_TOPOLOGY_TRIANGLE_LIST, // VkPrimitiveTopology topology;
+ VK_FALSE, // VkBool32 primitiveRestartEnable;
+ };
+
+ const VkRect2D scissor = makeRect2D(m_renderSize);
+ VkViewport viewport = makeViewport(m_renderSize);
+
+ const VkPipelineViewportStateCreateInfo viewportStateParams =
+ {
+ VK_STRUCTURE_TYPE_PIPELINE_VIEWPORT_STATE_CREATE_INFO, // VkStructureType sType;
+ DE_NULL, // const void* pNext;
+ 0u, // VkPipelineViewportStateCreateFlags flags;
+ 1u, // deUint32 viewportCount;
+ &viewport, // const VkViewport* pViewports;
+ 1u, // deUint32 scissorCount;
+ &scissor // const VkRect2D* pScissors;
+ };
+
+ const VkPipelineRasterizationStateCreateInfo rasterStateParams =
+ {
+ VK_STRUCTURE_TYPE_PIPELINE_RASTERIZATION_STATE_CREATE_INFO, // VkStructureType sType;
+ DE_NULL, // const void* pNext;
+ 0u, // VkPipelineRasterizationStateCreateFlags flags;
+ VK_FALSE, // VkBool32 depthClampEnable;
+ VK_FALSE, // VkBool32 rasterizerDiscardEnable;
+ VK_POLYGON_MODE_FILL, // VkPolygonMode polygonMode;
+ VK_CULL_MODE_NONE, // VkCullModeFlags cullMode;
+ VK_FRONT_FACE_COUNTER_CLOCKWISE, // VkFrontFace frontFace;
+ VK_FALSE, // VkBool32 depthBiasEnable;
+ 0.0f, // float depthBiasConstantFactor;
+ 0.0f, // float depthBiasClamp;
+ 0.0f, // float depthBiasSlopeFactor;
+ 1.0f, // float lineWidth;
+ };
+
+ const VkPipelineColorBlendAdvancedStateCreateInfoEXT blendAdvancedStateParams =
+ {
+ VK_STRUCTURE_TYPE_PIPELINE_COLOR_BLEND_ADVANCED_STATE_CREATE_INFO_EXT, // VkStructureType sType;
+ DE_NULL, // const void* pNext;
+ srcPremultiplied, // VkBool32 srcPremultiplied;
+ dstPremultiplied, // VkBool32 dstPremultiplied;
+ m_param.overlap, // VkBlendOverlapEXT blendOverlap;
+ };
+
+ std::vector<VkPipelineColorBlendAttachmentState> colorBlendAttachmentStates;
+
+ for (deUint32 i = 0; i < m_param.colorAttachmentsCount; i++)
+ {
+ const VkPipelineColorBlendAttachmentState colorBlendAttachmentState =
+ {
+ VK_TRUE, // VkBool32 blendEnable;
+ VK_BLEND_FACTOR_ONE, // VkBlendFactor srcColorBlendFactor;
+ VK_BLEND_FACTOR_ONE, // VkBlendFactor dstColorBlendFactor;
+ m_param.blendOps[i], // VkBlendOp colorBlendOp;
+ VK_BLEND_FACTOR_ONE, // VkBlendFactor srcAlphaBlendFactor;
+ VK_BLEND_FACTOR_ONE, // VkBlendFactor dstAlphaBlendFactor;
+ m_param.blendOps[i], // VkBlendOp alphaBlendOp;
+ VK_COLOR_COMPONENT_R_BIT |
+ VK_COLOR_COMPONENT_G_BIT |
+ VK_COLOR_COMPONENT_B_BIT |
+ VK_COLOR_COMPONENT_A_BIT // VkColorComponentFlags colorWriteMask;
+ };
+ colorBlendAttachmentStates.emplace_back(colorBlendAttachmentState);
+ }
+
+ const VkPipelineColorBlendStateCreateInfo colorBlendStateParams =
+ {
+ VK_STRUCTURE_TYPE_PIPELINE_COLOR_BLEND_STATE_CREATE_INFO, // VkStructureType sType;
+ &blendAdvancedStateParams, // const void* pNext;
+ 0u, // VkPipelineColorBlendStateCreateFlags flags;
+ VK_FALSE, // VkBool32 logicOpEnable;
+ VK_LOGIC_OP_COPY, // VkLogicOp logicOp;
+ (deUint32)colorBlendAttachmentStates.size(), // deUint32 attachmentCount;
+ colorBlendAttachmentStates.data(), // const VkPipelineColorBlendAttachmentState* pAttachments;
+ { 0.0f, 0.0f, 0.0f, 0.0f }, // float blendConst[4];
+ };
+
+ const VkPipelineMultisampleStateCreateInfo multisampleStateParams =
+ {
+ VK_STRUCTURE_TYPE_PIPELINE_MULTISAMPLE_STATE_CREATE_INFO, // VkStructureType sType;
+ DE_NULL, // const void* pNext;
+ 0u, // VkPipelineMultisampleStateCreateFlags flags;
+ VK_SAMPLE_COUNT_1_BIT, // VkSampleCountFlagBits rasterizationSamples;
+ VK_FALSE, // VkBool32 sampleShadingEnable;
+ 0.0f, // float minSampleShading;
+ DE_NULL, // const VkSampleMask* pSampleMask;
+ VK_FALSE, // VkBool32 alphaToCoverageEnable;
+ VK_FALSE, // VkBool32 alphaToOneEnable;
+ };
+
+ VkPipelineDepthStencilStateCreateInfo depthStencilStateParams =
+ {
+ VK_STRUCTURE_TYPE_PIPELINE_DEPTH_STENCIL_STATE_CREATE_INFO, // VkStructureType sType;
+ DE_NULL, // const void* pNext;
+ 0u, // VkPipelineDepthStencilStateCreateFlags flags;
+ VK_FALSE, // VkBool32 depthTestEnable;
+ VK_FALSE, // VkBool32 depthWriteEnable;
+ VK_COMPARE_OP_NEVER, // VkCompareOp depthCompareOp;
+ VK_FALSE, // VkBool32 depthBoundsTestEnable;
+ VK_FALSE, // VkBool32 stencilTestEnable;
+ // VkStencilOpState front;
+ {
+ VK_STENCIL_OP_KEEP, // VkStencilOp failOp;
+ VK_STENCIL_OP_KEEP, // VkStencilOp passOp;
+ VK_STENCIL_OP_KEEP, // VkStencilOp depthFailOp;
+ VK_COMPARE_OP_NEVER, // VkCompareOp compareOp;
+ 0u, // deUint32 compareMask;
+ 0u, // deUint32 writeMask;
+ 0u, // deUint32 reference;
+ },
+ // VkStencilOpState back;
+ {
+ VK_STENCIL_OP_KEEP, // VkStencilOp failOp;
+ VK_STENCIL_OP_KEEP, // VkStencilOp passOp;
+ VK_STENCIL_OP_KEEP, // VkStencilOp depthFailOp;
+ VK_COMPARE_OP_NEVER, // VkCompareOp compareOp;
+ 0u, // deUint32 compareMask;
+ 0u, // deUint32 writeMask;
+ 0u, // deUint32 reference;
+ },
+ 0.0f, // float minDepthBounds;
+ 1.0f, // float maxDepthBounds;
+ };
+
+ const VkDynamicState dynamicState = VK_DYNAMIC_STATE_SCISSOR;
+ const VkPipelineDynamicStateCreateInfo dynamicStateParams =
+ {
+ VK_STRUCTURE_TYPE_PIPELINE_DYNAMIC_STATE_CREATE_INFO, // VkStructureType sType;
+ DE_NULL, // const void* pNext;
+ 0u, // VkPipelineDynamicStateCreateFlags flags;
+ 1u, // uint32_t dynamicStateCount;
+ &dynamicState // const VkDynamicState* pDynamicStates;
+ };
+
+ const VkGraphicsPipelineCreateInfo graphicsPipelineParams =
+ {
+ VK_STRUCTURE_TYPE_GRAPHICS_PIPELINE_CREATE_INFO, // VkStructureType sType;
+ DE_NULL, // const void* pNext;
+ 0u, // VkPipelineCreateFlags flags;
+ m_shaderStageCount, // deUint32 stageCount;
+ m_shaderStageInfo, // const VkPipelineShaderStageCreateInfo* pStages;
+ &vertexInputStateParams, // const VkPipelineVertexInputStateCreateInfo* pVertexInputState;
+ &inputAssemblyStateParams, // const VkPipelineInputAssemblyStateCreateInfo* pInputAssemblyState;
+ DE_NULL, // const VkPipelineTessellationStateCreateInfo* pTessellationState;
+ &viewportStateParams, // const VkPipelineViewportStateCreateInfo* pViewportState;
+ &rasterStateParams, // const VkPipelineRasterizationStateCreateInfo* pRasterState;
+ &multisampleStateParams, // const VkPipelineMultisampleStateCreateInfo* pMultisampleState;
+ &depthStencilStateParams, // const VkPipelineDepthStencilStateCreateInfo* pDepthStencilState;
+ &colorBlendStateParams, // const VkPipelineColorBlendStateCreateInfo* pColorBlendState;
+ &dynamicStateParams, // const VkPipelineDynamicStateCreateInfo* pDynamicState;
+ *m_pipelineLayout, // VkPipelineLayout layout;
+ *m_renderPass, // VkRenderPass renderPass;
+ 0u, // deUint32 subpass;
+ DE_NULL, // VkPipeline basePipelineHandle;
+ 0u, // deInt32 basePipelineIndex;
+ };
+
+ m_pipeline = createGraphicsPipeline(vk, vkDevice, DE_NULL, &graphicsPipelineParams);
+}
+
+void BlendOperationAdvancedTestInstance::prepareRenderPass (VkFramebuffer framebuffer, VkPipeline pipeline) const
+{
+ const DeviceInterface& vk = m_context.getDeviceInterface();
+
+ std::vector<VkClearValue> attachmentClearValues;
+
+ for (deUint32 i = 0; i < m_param.colorAttachmentsCount; i++)
+ attachmentClearValues.emplace_back(makeClearValueColor(clearColorVec4));
+
+ beginRenderPass(vk, *m_cmdBuffer, *m_renderPass, framebuffer, makeRect2D(0, 0, m_renderSize.x(), m_renderSize.y()),
+ m_param.colorAttachmentsCount, attachmentClearValues.data());
+ vk.cmdBindPipeline(*m_cmdBuffer, VK_PIPELINE_BIND_POINT_GRAPHICS, pipeline);
+ VkDeviceSize offsets = 0u;
+ vk.cmdBindVertexBuffers(*m_cmdBuffer, 0u, 1u, &m_vertexBuffer.get(), &offsets);
+
+ // Draw all colors
+ deUint32 skippedColors = 0u;
+ for (deUint32 color = 0; color < DE_LENGTH_OF_ARRAY(srcColors); color++)
+ {
+ // Skip ill-formed colors when we have non-premultiplied destination colors.
+ if (m_param.premultipliedDstColor == VK_FALSE)
+ {
+ deBool skipColor = false;
+ for (deUint32 i = 0; i < m_param.colorAttachmentsCount; i++)
+ {
+ Vec4 calculatedColor = calculateFinalColor(m_param, m_param.blendOps[i], srcColors[color], dstColors[color]);
+ if (calculatedColor.w() <= 0.0f && calculatedColor != Vec4(0.0f))
+ {
+ // Skip ill-formed colors, because the spec says the result is undefined.
+ skippedColors++;
+ skipColor = true;
+ break;
+ }
+ }
+ if (skipColor)
+ continue;
+ }
+
+ deInt32 x = 0;
+ deInt32 y = 0;
+ getCoordinates(color, x, y);
+
+ // Set source color as push constant
+ vk.cmdPushConstants(*m_cmdBuffer, *m_pipelineLayout, VK_SHADER_STAGE_FRAGMENT_BIT, 0u, sizeof(Vec4), &srcColors[color]);
+
+ VkRect2D scissor = makeRect2D(x, y, 1u, 1u);
+ vk.cmdSetScissor(*m_cmdBuffer, 0u, 1u, &scissor);
+
+ // To set destination color, we do clear attachment restricting the area to the respective pixel of each color attachment.
+ {
+ // Set destination color as push constant.
+ std::vector<VkClearAttachment> attachments;
+ VkClearValue clearValue = vk::makeClearValueColorVec4(dstColors[color]);
+
+ for (deUint32 i = 0; i < m_param.colorAttachmentsCount; i++)
+ {
+ VkClearAttachment attachment =
+ {
+ VK_IMAGE_ASPECT_COLOR_BIT,
+ i,
+ clearValue
+ };
+ attachments.emplace_back(attachment);
+ }
+
+ const VkClearRect rect =
+ {
+ scissor,
+ 0u,
+ 1u
+ };
+ vk.cmdClearAttachments(*m_cmdBuffer, (deUint32)attachments.size(), attachments.data(), 1u, &rect);
+ }
+
+ // Draw
+ vk.cmdDraw(*m_cmdBuffer, (deUint32)m_vertices.size(), 1u, 0u, 0u);
+ }
+
+ // If we break this assert, then we are not testing anything in this test.
+ DE_ASSERT(skippedColors < DE_LENGTH_OF_ARRAY(srcColors));
+
+ // Log number of skipped colors
+ if (skippedColors != 0u)
+ {
+ tcu::TestLog& log = m_context.getTestContext().getLog();
+ log << tcu::TestLog::Message << "Skipped " << skippedColors << " out of " << DE_LENGTH_OF_ARRAY(srcColors) << " color cases due to ill-formed colors" << tcu::TestLog::EndMessage;
+ }
+ endRenderPass(vk, *m_cmdBuffer);
+}
+
+void BlendOperationAdvancedTestInstance::prepareCommandBuffer () const
+{
+ const DeviceInterface& vk = m_context.getDeviceInterface();
+
+
+ beginCommandBuffer(vk, *m_cmdBuffer, 0u);
+
+ vk.cmdPipelineBarrier(*m_cmdBuffer, VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT, VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT | VK_PIPELINE_STAGE_LATE_FRAGMENT_TESTS_BIT, (VkDependencyFlags)0,
+ 0u, DE_NULL, 0u, DE_NULL, (deUint32)m_imageLayoutBarriers.size(), m_imageLayoutBarriers.data());
+
+ prepareRenderPass(*m_framebuffer, *m_pipeline);
+
+ endCommandBuffer(vk, *m_cmdBuffer);
+}
+
+BlendOperationAdvancedTestInstance::BlendOperationAdvancedTestInstance (Context& context,
+ const BlendOperationAdvancedParam param)
+ : TestInstance (context)
+ , m_param (param)
+ , m_renderSize (tcu::UVec2(widthArea, heightArea))
+ , m_colorFormat (VK_FORMAT_R16G16B16A16_SFLOAT)
+ , m_shaderStageCount (0)
+{
+ const DeviceInterface& vk = m_context.getDeviceInterface();
+ const VkDevice vkDevice = m_context.getDevice();
+ const deUint32 queueFamilyIndex = context.getUniversalQueueFamilyIndex();
+
+ // Create vertex buffer and upload data
+ {
+ // Load vertices into vertex buffer
+ m_vertices = createPoints();
+ DE_ASSERT((deUint32)m_vertices.size() == 6);
+
+ m_vertexBuffer = createBufferAndBindMemory(m_context, m_vertices.size() * sizeof(Vec4), VK_BUFFER_USAGE_VERTEX_BUFFER_BIT, &m_vertexBufferMemory);
+ deMemcpy(m_vertexBufferMemory->getHostPtr(), m_vertices.data(), m_vertices.size() * sizeof(Vec4));
+ flushAlloc(vk, vkDevice, *m_vertexBufferMemory);
+ }
+
+ // Create render pass
+ m_renderPass = makeTestRenderPass(param, vk, vkDevice, m_colorFormat);
+
+ const VkComponentMapping componentMappingRGBA = { VK_COMPONENT_SWIZZLE_R, VK_COMPONENT_SWIZZLE_G, VK_COMPONENT_SWIZZLE_B, VK_COMPONENT_SWIZZLE_A};
+
+ // Create color images
+ for (deUint32 i = 0; i < param.colorAttachmentsCount; i++)
+ {
+ de::MovePtr<Allocation> colorImageAlloc;
+ m_colorImageAllocs.emplace_back(colorImageAlloc);
+
+ Move<VkImage> colorImage = createImage2DAndBindMemory(m_context,
+ m_colorFormat,
+ m_renderSize.x(),
+ m_renderSize.y(),
+ VK_IMAGE_USAGE_COLOR_ATTACHMENT_BIT | VK_IMAGE_USAGE_TRANSFER_SRC_BIT,
+ VK_SAMPLE_COUNT_1_BIT,
+ &m_colorImageAllocs.back());
+ m_colorImages.emplace_back(colorImage);
+
+ // Set up image layout transition barriers
+ {
+ VkImageMemoryBarrier colorImageBarrier =
+ {
+ VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER, // VkStructureType sType;
+ DE_NULL, // const void* pNext;
+ 0u, // VkAccessFlags srcAccessMask;
+ (VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT |
+ VK_ACCESS_COLOR_ATTACHMENT_READ_NONCOHERENT_BIT_EXT), // VkAccessFlags dstAccessMask;
+ VK_IMAGE_LAYOUT_UNDEFINED, // VkImageLayout oldLayout;
+ VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL, // VkImageLayout newLayout;
+ VK_QUEUE_FAMILY_IGNORED, // deUint32 srcQueueFamilyIndex;
+ VK_QUEUE_FAMILY_IGNORED, // deUint32 dstQueueFamilyIndex;
+ *m_colorImages.back(), // VkImage image;
+ { VK_IMAGE_ASPECT_COLOR_BIT, 0u, 1u, 0u, 1u }, // VkImageSubresourceRange subresourceRange;
+ };
+
+ m_imageLayoutBarriers.emplace_back(colorImageBarrier);
+ }
+
+ // Create color attachment view
+ {
+ VkImageViewCreateInfo colorAttachmentViewParams =
+ {
+ VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO, // VkStructureType sType;
+ DE_NULL, // const void* pNext;
+ 0u, // VkImageViewCreateFlags flags;
+ *m_colorImages.back(), // VkImage image;
+ VK_IMAGE_VIEW_TYPE_2D, // VkImageViewType viewType;
+ m_colorFormat, // VkFormat format;
+ componentMappingRGBA, // VkComponentMapping components;
+ { VK_IMAGE_ASPECT_COLOR_BIT, 0u, 1u, 0u, 1u }, // VkImageSubresourceRange subresourceRange;
+ };
+
+ m_colorAttachmentViews.emplace_back(createImageView(vk, vkDevice, &colorAttachmentViewParams));
+ }
+ }
+
+ // Create framebuffer
+ {
+ std::vector<VkImageView> imageViews;
+
+ for (auto& movePtr : m_colorAttachmentViews)
+ imageViews.push_back(movePtr.get());
+
+ const VkFramebufferCreateInfo framebufferParams =
+ {
+ VK_STRUCTURE_TYPE_FRAMEBUFFER_CREATE_INFO, // VkStructureType sType;
+ DE_NULL, // const void* pNext;
+ 0u, // VkFramebufferCreateFlags flags;
+ *m_renderPass, // VkRenderPass renderPass;
+ (deUint32)imageViews.size(), // deUint32 attachmentCount;
+ imageViews.data(), // const VkImageView* pAttachments;
+ (deUint32)m_renderSize.x(), // deUint32 width;
+ (deUint32)m_renderSize.y(), // deUint32 height;
+ 1u, // deUint32 layers;
+ };
+
+ m_framebuffer = createFramebuffer(vk, vkDevice, &framebufferParams);
+ }
+
+ // Bind shader stages
+ {
+ bindShaderStage(VK_SHADER_STAGE_VERTEX_BIT, "vert", "main");
+ bindShaderStage(VK_SHADER_STAGE_FRAGMENT_BIT, "frag", "main");
+ }
+
+
+ // Create pipeline layout
+ {
+ const VkPushConstantRange pushConstantRange =
+ {
+ VK_SHADER_STAGE_FRAGMENT_BIT, // VkShaderStageFlags stageFlags
+ 0, // deUint32 offset
+ sizeof(Vec4) // deUint32 size
+ };
+
+ const VkPipelineLayoutCreateInfo pipelineLayoutParams =
+ {
+ VK_STRUCTURE_TYPE_PIPELINE_LAYOUT_CREATE_INFO, // VkStructureType sType;
+ DE_NULL, // const void* pNext;
+ 0u, // VkPipelineLayoutCreateFlags flags;
+ 0u, // deUint32 setLayoutCount;
+ DE_NULL, // const VkDescriptorSetLayout* pSetLayouts;
+ 1u, // deUint32 pushConstantRangeCount;
+ &pushConstantRange // const VkPushConstantRange* pPushConstantRanges;
+ };
+
+ m_pipelineLayout = createPipelineLayout(vk, vkDevice, &pipelineLayoutParams);
+ }
+
+ // Create pipeline
+ buildPipeline(m_param.premultipliedSrcColor, m_param.premultipliedDstColor);
+
+ // Create command pool
+ m_cmdPool = createCommandPool(vk, vkDevice, VK_COMMAND_POOL_CREATE_TRANSIENT_BIT | VK_COMMAND_POOL_CREATE_RESET_COMMAND_BUFFER_BIT, queueFamilyIndex);
+
+ // Create command buffer
+ m_cmdBuffer = allocateCommandBuffer(vk, vkDevice, *m_cmdPool, VK_COMMAND_BUFFER_LEVEL_PRIMARY);
+}
+
+BlendOperationAdvancedTestInstance::~BlendOperationAdvancedTestInstance (void)
+{
+}
+
+tcu::TestStatus BlendOperationAdvancedTestInstance::iterate (void)
+{
+ const DeviceInterface& vk = m_context.getDeviceInterface();
+ const VkDevice vkDevice = m_context.getDevice();
+ const VkQueue queue = m_context.getUniversalQueue();
+ tcu::TestLog& log = m_context.getTestContext().getLog();
+
+ // Log the blend operations to test
+ {
+ if (m_param.independentBlend)
+ {
+ for (deUint32 i = 0; (i < m_param.colorAttachmentsCount); i++)
+ log << tcu::TestLog::Message << "Color attachment " << i << " uses depth op: "<< de::toLower(getBlendOpStr(m_param.blendOps[i]).toString().substr(3)) << tcu::TestLog::EndMessage;
+
+ }
+ else
+ {
+ log << tcu::TestLog::Message << "All color attachments use depth op: " << de::toLower(getBlendOpStr(m_param.blendOps[0]).toString().substr(3)) << tcu::TestLog::EndMessage;
+
+ }
+ }
+ prepareCommandBuffer();
+ submitCommandsAndWait(vk, vkDevice, queue, m_cmdBuffer.get());
+
+ if (verifyTestResult() == DE_FALSE)
+ return tcu::TestStatus::fail("Image mismatch");
+
+ return tcu::TestStatus::pass("Result images matches references");
+}
+
+deBool BlendOperationAdvancedTestInstance::verifyTestResult ()
+{
+ deBool compareOk = DE_TRUE;
+ const DeviceInterface& vk = m_context.getDeviceInterface();
+ const VkDevice vkDevice = m_context.getDevice();
+ const VkQueue queue = m_context.getUniversalQueue();
+ const deUint32 queueFamilyIndex = m_context.getUniversalQueueFamilyIndex();
+ Allocator& allocator = m_context.getDefaultAllocator();
+ std::vector<tcu::TextureLevel> referenceImages;
+
+ for (deUint32 colorAtt = 0; colorAtt < m_param.colorAttachmentsCount; colorAtt++)
+ {
+ tcu::TextureLevel refImage (vk::mapVkFormat(m_colorFormat), 32, 32);
+ tcu::clear(refImage.getAccess(), clearColorVec4);
+ referenceImages.emplace_back(refImage);
+ }
+
+ for (deUint32 color = 0; color < DE_LENGTH_OF_ARRAY(srcColors); color++)
+ {
+ deBool skipColor = DE_FALSE;
+
+ // Check if any color attachment will generate an ill-formed color. If that's the case, skip that color in the verification.
+ for (deUint32 colorAtt = 0; colorAtt < m_param.colorAttachmentsCount; colorAtt++)
+ {
+ Vec4 rectColor = calculateFinalColor(m_param, m_param.blendOps[colorAtt], srcColors[color], dstColors[color]);
+ if (m_param.premultipliedDstColor == VK_FALSE)
+ {
+ if (rectColor.w() > 0.0f)
+ {
+ rectColor.x() = rectColor.x() / rectColor.w();
+ rectColor.y() = rectColor.y() / rectColor.w();
+ rectColor.z() = rectColor.z() / rectColor.w();
+ }
+ else
+ {
+ // Skip the color check if it is ill-formed.
+ if (rectColor != Vec4(0.0f))
+ {
+ skipColor = DE_TRUE;
+ break;
+ }
+ }
+ }
+ }
+
+ // Skip ill-formed colors that appears in any color attachment.
+ if (skipColor)
+ continue;
+
+ // If we reach this point, the final color for all color attachment is not ill-formed.
+ for (deUint32 colorAtt = 0; colorAtt < m_param.colorAttachmentsCount; colorAtt++)
+ {
+ Vec4 rectColor = calculateFinalColor(m_param, m_param.blendOps[colorAtt], srcColors[color], dstColors[color]);
+ if (m_param.premultipliedDstColor == VK_FALSE)
+ {
+ if (rectColor.w() > 0.0f)
+ {
+ rectColor.x() = rectColor.x() / rectColor.w();
+ rectColor.y() = rectColor.y() / rectColor.w();
+ rectColor.z() = rectColor.z() / rectColor.w();
+ }
+ else
+ {
+ // Ill-formed colors were already skipped
+ DE_ASSERT(rectColor == Vec4(0.0f));
+ }
+ }
+ deInt32 x = 0;
+ deInt32 y = 0;
+ getCoordinates(color, x, y);
+ tcu::clear(tcu::getSubregion(referenceImages[colorAtt].getAccess(), x, y, 1u, 1u), rectColor);
+ }
+ }
+
+ for (deUint32 colorAtt = 0; colorAtt < m_param.colorAttachmentsCount; colorAtt++)
+ {
+ // Compare image
+ de::MovePtr<tcu::TextureLevel> result = vkt::pipeline::readColorAttachment(vk, vkDevice, queue, queueFamilyIndex, allocator, *m_colorImages[colorAtt], m_colorFormat, m_renderSize);
+ std::ostringstream name;
+ name << "Image comparison. Color attachment: " << colorAtt << ". Depth op: " << de::toLower(getBlendOpStr(m_param.blendOps[colorAtt]).toString().substr(3));
+
+ compareOk = tcu::floatThresholdCompare(m_context.getTestContext().getLog(),
+ "FloatImageCompare",
+ name.str().c_str(),
+ referenceImages[colorAtt].getAccess(),
+ result->getAccess(),
+ Vec4(0.01f, 0.01f, 0.01f, 0.01f),
+ tcu::COMPARE_LOG_RESULT);
+ if (!compareOk)
+ return DE_FALSE;
+ }
+ return DE_TRUE;
+}
+
+class BlendOperationAdvancedTest : public vkt::TestCase
+{
+public:
+ BlendOperationAdvancedTest (tcu::TestContext& testContext,
+ const std::string& name,
+ const std::string& description,
+ const BlendOperationAdvancedParam param)
+ : vkt::TestCase (testContext, name, description)
+ , m_param (param)
+ { }
+ virtual ~BlendOperationAdvancedTest (void) { }
+ virtual void initPrograms (SourceCollections& programCollection) const;
+ virtual TestInstance* createInstance (Context& context) const;
+ virtual void checkSupport (Context& context) const;
+
+protected:
+ const BlendOperationAdvancedParam m_param;
+};
+
+void BlendOperationAdvancedTest::checkSupport(Context& context) const
+{
+ const InstanceInterface& vki = context.getInstanceInterface();
+
++ context.requireDeviceFunctionality("VK_EXT_blend_operation_advanced");
+
+ VkPhysicalDeviceBlendOperationAdvancedPropertiesEXT blendProperties;
+ blendProperties.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_BLEND_OPERATION_ADVANCED_PROPERTIES_EXT;
+ blendProperties.pNext = DE_NULL;
+
+ VkPhysicalDeviceProperties2 properties2;
+ properties2.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_PROPERTIES_2;
+ properties2.pNext = &blendProperties;
+ vki.getPhysicalDeviceProperties2(context.getPhysicalDevice(), &properties2);
+
+ if (!blendProperties.advancedBlendAllOperations)
+ {
+ throw tcu::NotSupportedError("Unsupported all advanced blend operations");
+ }
+
+ if (m_param.colorAttachmentsCount > blendProperties.advancedBlendMaxColorAttachments)
+ {
+ std::ostringstream error;
+ error << "Unsupported number of color attachments (" << blendProperties.advancedBlendMaxColorAttachments << " < " << m_param.colorAttachmentsCount;
+ throw tcu::NotSupportedError(error.str().c_str());
+ }
+
+ if (m_param.overlap != VK_BLEND_OVERLAP_UNCORRELATED_EXT && !blendProperties.advancedBlendCorrelatedOverlap)
+ {
+ throw tcu::NotSupportedError("Unsupported blend correlated overlap");
+ }
+
+ if (m_param.colorAttachmentsCount > 1 && m_param.independentBlend && !blendProperties.advancedBlendIndependentBlend)
+ {
+ throw tcu::NotSupportedError("Unsupported independent blend");
+ }
+
+ if (!m_param.premultipliedSrcColor && !blendProperties.advancedBlendNonPremultipliedSrcColor)
+ {
+ throw tcu::NotSupportedError("Unsupported non-premultiplied source color");
+ }
+
+ if (!m_param.premultipliedDstColor && !blendProperties.advancedBlendNonPremultipliedDstColor)
+ {
+ throw tcu::NotSupportedError("Unsupported non-premultiplied destination color");
+ }
+
+ const VkPhysicalDeviceBlendOperationAdvancedFeaturesEXT blendFeatures = context.getBlendOperationAdvancedFeatures();
+ if (m_param.coherentOperations && !blendFeatures.advancedBlendCoherentOperations)
+ {
+ throw tcu::NotSupportedError("Unsupported required coherent operations");
+ }
+}
+
+void BlendOperationAdvancedTest::initPrograms (SourceCollections& programCollection) const
+{
+ programCollection.glslSources.add("vert") << glu::VertexSource(
+ "#version 310 es\n"
+ "layout(location = 0) in vec4 position;\n"
+ "void main (void)\n"
+ "{\n"
+ " gl_Position = position;\n"
+ "}\n");
+
+ std::ostringstream fragmentSource;
+ fragmentSource << "#version 310 es\n";
+ fragmentSource << "layout(push_constant) uniform Color { highp vec4 color; };\n";
+ for (deUint32 i = 0; i < m_param.colorAttachmentsCount; i++)
+ fragmentSource << "layout(location = "<< i <<") out highp vec4 fragColor" << i <<";\n";
+ fragmentSource << "void main (void)\n";
+ fragmentSource << "{\n";
+ for (deUint32 i = 0; i < m_param.colorAttachmentsCount; i++)
+ fragmentSource << " fragColor" << i <<" = color;\n";
+ fragmentSource << "}\n";
+ programCollection.glslSources.add("frag") << glu::FragmentSource(fragmentSource.str().c_str());
+}
+
+class BlendOperationAdvancedTestCoherentInstance : public vkt::TestInstance
+{
+public:
+ BlendOperationAdvancedTestCoherentInstance (Context& context,
+ const BlendOperationAdvancedParam param);
+ virtual ~BlendOperationAdvancedTestCoherentInstance (void);
+ virtual tcu::TestStatus iterate (void);
+protected:
+ void prepareRenderPass (VkFramebuffer framebuffer, VkPipeline pipeline,
+ VkRenderPass renderpass, deBool secondDraw);
+ virtual void prepareCommandBuffer (void);
+ virtual void buildPipeline (void);
+ virtual void bindShaderStage (VkShaderStageFlagBits stage,
+ const char* sourceName,
+ const char* entryName);
+ virtual tcu::TestStatus verifyTestResult (void);
+
+protected:
+ const BlendOperationAdvancedParam m_param;
+ const tcu::UVec2 m_renderSize;
+ const VkFormat m_colorFormat;
+ Move<VkPipelineLayout> m_pipelineLayout;
+
+ Move<VkBuffer> m_vertexBuffer;
+ de::MovePtr<Allocation> m_vertexBufferMemory;
+ std::vector<Vec4> m_vertices;
+
+ std::vector<Move<VkRenderPass>> m_renderPasses;
+ Move<VkCommandPool> m_cmdPool;
+ Move<VkCommandBuffer> m_cmdBuffer;
+ Move<VkImage> m_colorImage;
+ Move<VkImageView> m_colorAttachmentView;
+ de::MovePtr<Allocation> m_colorImageAlloc;
+ std::vector<VkImageMemoryBarrier> m_imageLayoutBarriers;
+ std::vector<Move<VkFramebuffer>> m_framebuffers;
+ std::vector<Move<VkPipeline>> m_pipelines;
+
+ Move<VkShaderModule> m_shaderModules[2];
+ deUint32 m_shaderStageCount;
+ VkPipelineShaderStageCreateInfo m_shaderStageInfo[2];
+};
+
+BlendOperationAdvancedTestCoherentInstance::~BlendOperationAdvancedTestCoherentInstance (void)
+{
+}
+
+void BlendOperationAdvancedTestCoherentInstance::bindShaderStage (VkShaderStageFlagBits stage,
+ const char* sourceName,
+ const char* entryName)
+{
+ const DeviceInterface& vk = m_context.getDeviceInterface();
+ const VkDevice vkDevice = m_context.getDevice();
+
+ // Create shader module
+ deUint32* code = (deUint32*)m_context.getBinaryCollection().get(sourceName).getBinary();
+ deUint32 codeSize = (deUint32)m_context.getBinaryCollection().get(sourceName).getSize();
+
+ const VkShaderModuleCreateInfo moduleCreateInfo =
+ {
+ VK_STRUCTURE_TYPE_SHADER_MODULE_CREATE_INFO, // VkStructureType sType;
+ DE_NULL, // const void* pNext;
+ 0u, // VkShaderModuleCreateFlags flags;
+ codeSize, // deUintptr codeSize;
+ code, // const deUint32* pCode;
+ };
+
+ m_shaderModules[m_shaderStageCount] = createShaderModule(vk, vkDevice, &moduleCreateInfo);
+
+ // Prepare shader stage info
+ m_shaderStageInfo[m_shaderStageCount].sType = VK_STRUCTURE_TYPE_PIPELINE_SHADER_STAGE_CREATE_INFO;
+ m_shaderStageInfo[m_shaderStageCount].pNext = DE_NULL;
+ m_shaderStageInfo[m_shaderStageCount].flags = 0u;
+ m_shaderStageInfo[m_shaderStageCount].stage = stage;
+ m_shaderStageInfo[m_shaderStageCount].module = *m_shaderModules[m_shaderStageCount];
+ m_shaderStageInfo[m_shaderStageCount].pName = entryName;
+ m_shaderStageInfo[m_shaderStageCount].pSpecializationInfo = DE_NULL;
+
+ m_shaderStageCount++;
+}
+
+void BlendOperationAdvancedTestCoherentInstance::buildPipeline ()
+{
+ const DeviceInterface& vk = m_context.getDeviceInterface();
+ const VkDevice vkDevice = m_context.getDevice();
+
+ // Create pipeline
+ const VkVertexInputBindingDescription vertexInputBindingDescription =
+ {
+ 0u, // deUint32 binding;
+ sizeof(Vec4) , // deUint32 strideInBytes;
+ VK_VERTEX_INPUT_RATE_VERTEX, // VkVertexInputRate inputRate;
+ };
+
+ const VkVertexInputAttributeDescription vertexInputAttributeDescription =
+ {
+ 0u, // deUint32 location;
+ 0u, // deUint32 binding;
+ VK_FORMAT_R32G32B32A32_SFLOAT, // VkFormat format;
+ 0u // deUint32 offsetInBytes;
+ };
+
+ const VkPipelineVertexInputStateCreateInfo vertexInputStateParams =
+ {
+ VK_STRUCTURE_TYPE_PIPELINE_VERTEX_INPUT_STATE_CREATE_INFO, // VkStructureType sType;
+ DE_NULL, // const void* pNext;
+ 0u, // VkPipelineVertexInputStateCreateFlags flags;
+ 1u, // deUint32 vertexBindingDescriptionCount;
+ &vertexInputBindingDescription, // const VkVertexInputBindingDescription* pVertexBindingDescriptions;
+ 1u, // deUint32 vertexAttributeDescriptionCount;
+ &vertexInputAttributeDescription, // const VkVertexInputAttributeDescription* pVertexAttributeDescriptions;
+ };
+
+ const VkPipelineInputAssemblyStateCreateInfo inputAssemblyStateParams =
+ {
+ VK_STRUCTURE_TYPE_PIPELINE_INPUT_ASSEMBLY_STATE_CREATE_INFO, // VkStructureType sType;
+ DE_NULL, // const void* pNext;
+ 0u, // VkPipelineInputAssemblyStateCreateFlags flags;
+ VK_PRIMITIVE_TOPOLOGY_TRIANGLE_LIST, // VkPrimitiveTopology topology;
+ VK_FALSE, // VkBool32 primitiveRestartEnable;
+ };
+
+ const VkRect2D scissor = makeRect2D(m_renderSize);
+ VkViewport viewport = makeViewport(m_renderSize);
+
+ const VkPipelineViewportStateCreateInfo viewportStateParams =
+ {
+ VK_STRUCTURE_TYPE_PIPELINE_VIEWPORT_STATE_CREATE_INFO, // VkStructureType sType;
+ DE_NULL, // const void* pNext;
+ 0u, // VkPipelineViewportStateCreateFlags flags;
+ 1u, // deUint32 viewportCount;
+ &viewport, // const VkViewport* pViewports;
+ 1u, // deUint32 scissorCount;
+ &scissor // const VkRect2D* pScissors;
+ };
+
+ const VkPipelineRasterizationStateCreateInfo rasterStateParams =
+ {
+ VK_STRUCTURE_TYPE_PIPELINE_RASTERIZATION_STATE_CREATE_INFO, // VkStructureType sType;
+ DE_NULL, // const void* pNext;
+ 0u, // VkPipelineRasterizationStateCreateFlags flags;
+ VK_FALSE, // VkBool32 depthClampEnable;
+ VK_FALSE, // VkBool32 rasterizerDiscardEnable;
+ VK_POLYGON_MODE_FILL, // VkPolygonMode polygonMode;
+ VK_CULL_MODE_NONE, // VkCullModeFlags cullMode;
+ VK_FRONT_FACE_COUNTER_CLOCKWISE, // VkFrontFace frontFace;
+ VK_FALSE, // VkBool32 depthBiasEnable;
+ 0.0f, // float depthBiasConstantFactor;
+ 0.0f, // float depthBiasClamp;
+ 0.0f, // float depthBiasSlopeFactor;
+ 1.0f, // float lineWidth;
+ };
+
+ const VkPipelineColorBlendAdvancedStateCreateInfoEXT blendAdvancedStateParams =
+ {
+ VK_STRUCTURE_TYPE_PIPELINE_COLOR_BLEND_ADVANCED_STATE_CREATE_INFO_EXT, // VkStructureType sType;
+ DE_NULL, // const void* pNext;
+ VK_TRUE, // VkBool32 srcPremultiplied;
+ VK_TRUE, // VkBool32 dstPremultiplied;
+ m_param.overlap, // VkBlendOverlapEXT blendOverlap;
+ };
+
+ std::vector<VkPipelineColorBlendAttachmentState> colorBlendAttachmentStates;
+
+ // One VkPipelineColorBlendAttachmentState for each pipeline, we only have one color attachment.
+ for (deUint32 i = 0; i < 2; i++)
+ {
+ const VkPipelineColorBlendAttachmentState colorBlendAttachmentState =
+ {
+ VK_TRUE, // VkBool32 blendEnable;
+ VK_BLEND_FACTOR_ONE, // VkBlendFactor srcColorBlendFactor;
+ VK_BLEND_FACTOR_ONE, // VkBlendFactor dstColorBlendFactor;
+ m_param.blendOps[i], // VkBlendOp colorBlendOp;
+ VK_BLEND_FACTOR_ONE, // VkBlendFactor srcAlphaBlendFactor;
+ VK_BLEND_FACTOR_ONE, // VkBlendFactor dstAlphaBlendFactor;
+ m_param.blendOps[i], // VkBlendOp alphaBlendOp;
+ VK_COLOR_COMPONENT_R_BIT |
+ VK_COLOR_COMPONENT_G_BIT |
+ VK_COLOR_COMPONENT_B_BIT |
+ VK_COLOR_COMPONENT_A_BIT // VkColorComponentFlags colorWriteMask;
+ };
+ colorBlendAttachmentStates.emplace_back(colorBlendAttachmentState);
+ }
+
+ std::vector<VkPipelineColorBlendStateCreateInfo> colorBlendStateParams;
+ VkPipelineColorBlendStateCreateInfo colorBlendStateParam =
+ {
+ VK_STRUCTURE_TYPE_PIPELINE_COLOR_BLEND_STATE_CREATE_INFO, // VkStructureType sType;
+ &blendAdvancedStateParams, // const void* pNext;
+ 0u, // VkPipelineColorBlendStateCreateFlags flags;
+ VK_FALSE, // VkBool32 logicOpEnable;
+ VK_LOGIC_OP_COPY, // VkLogicOp logicOp;
+ 1u, // deUint32 attachmentCount;
+ &colorBlendAttachmentStates[0], // const VkPipelineColorBlendAttachmentState* pAttachments;
+ { 0.0f, 0.0f, 0.0f, 0.0f }, // float blendConst[4];
+ };
+ colorBlendStateParams.emplace_back(colorBlendStateParam);
+
+ // For the second pipeline, the blendOp changed.
+ colorBlendStateParam.pAttachments = &colorBlendAttachmentStates[1];
+ colorBlendStateParams.emplace_back(colorBlendStateParam);
+
+ const VkPipelineMultisampleStateCreateInfo multisampleStateParams =
+ {
+ VK_STRUCTURE_TYPE_PIPELINE_MULTISAMPLE_STATE_CREATE_INFO, // VkStructureType sType;
+ DE_NULL, // const void* pNext;
+ 0u, // VkPipelineMultisampleStateCreateFlags flags;
+ VK_SAMPLE_COUNT_1_BIT, // VkSampleCountFlagBits rasterizationSamples;
+ VK_FALSE, // VkBool32 sampleShadingEnable;
+ 0.0f, // float minSampleShading;
+ DE_NULL, // const VkSampleMask* pSampleMask;
+ VK_FALSE, // VkBool32 alphaToCoverageEnable;
+ VK_FALSE, // VkBool32 alphaToOneEnable;
+ };
+
+ VkPipelineDepthStencilStateCreateInfo depthStencilStateParams =
+ {
+ VK_STRUCTURE_TYPE_PIPELINE_DEPTH_STENCIL_STATE_CREATE_INFO, // VkStructureType sType;
+ DE_NULL, // const void* pNext;
+ 0u, // VkPipelineDepthStencilStateCreateFlags flags;
+ VK_FALSE, // VkBool32 depthTestEnable;
+ VK_FALSE, // VkBool32 depthWriteEnable;
+ VK_COMPARE_OP_NEVER, // VkCompareOp depthCompareOp;
+ VK_FALSE, // VkBool32 depthBoundsTestEnable;
+ VK_FALSE, // VkBool32 stencilTestEnable;
+ // VkStencilOpState front;
+ {
+ VK_STENCIL_OP_KEEP, // VkStencilOp failOp;
+ VK_STENCIL_OP_KEEP, // VkStencilOp passOp;
+ VK_STENCIL_OP_KEEP, // VkStencilOp depthFailOp;
+ VK_COMPARE_OP_NEVER, // VkCompareOp compareOp;
+ 0u, // deUint32 compareMask;
+ 0u, // deUint32 writeMask;
+ 0u, // deUint32 reference;
+ },
+ // VkStencilOpState back;
+ {
+ VK_STENCIL_OP_KEEP, // VkStencilOp failOp;
+ VK_STENCIL_OP_KEEP, // VkStencilOp passOp;
+ VK_STENCIL_OP_KEEP, // VkStencilOp depthFailOp;
+ VK_COMPARE_OP_NEVER, // VkCompareOp compareOp;
+ 0u, // deUint32 compareMask;
+ 0u, // deUint32 writeMask;
+ 0u, // deUint32 reference;
+ },
+ 0.0f, // float minDepthBounds;
+ 1.0f, // float maxDepthBounds;
+ };
+
+ const VkDynamicState dynamicState = VK_DYNAMIC_STATE_SCISSOR;
+ const VkPipelineDynamicStateCreateInfo dynamicStateParams =
+ {
+ VK_STRUCTURE_TYPE_PIPELINE_DYNAMIC_STATE_CREATE_INFO, // VkStructureType sType;
+ DE_NULL, // const void* pNext;
+ 0u, // VkPipelineDynamicStateCreateFlags flags;
+ 1u, // uint32_t dynamicStateCount;
+ &dynamicState // const VkDynamicState* pDynamicStates;
+ };
+
+ VkGraphicsPipelineCreateInfo graphicsPipelineParams =
+ {
+ VK_STRUCTURE_TYPE_GRAPHICS_PIPELINE_CREATE_INFO, // VkStructureType sType;
+ DE_NULL, // const void* pNext;
+ 0u, // VkPipelineCreateFlags flags;
+ m_shaderStageCount, // deUint32 stageCount;
+ m_shaderStageInfo, // const VkPipelineShaderStageCreateInfo* pStages;
+ &vertexInputStateParams, // const VkPipelineVertexInputStateCreateInfo* pVertexInputState;
+ &inputAssemblyStateParams, // const VkPipelineInputAssemblyStateCreateInfo* pInputAssemblyState;
+ DE_NULL, // const VkPipelineTessellationStateCreateInfo* pTessellationState;
+ &viewportStateParams, // const VkPipelineViewportStateCreateInfo* pViewportState;
+ &rasterStateParams, // const VkPipelineRasterizationStateCreateInfo* pRasterState;
+ &multisampleStateParams, // const VkPipelineMultisampleStateCreateInfo* pMultisampleState;
+ &depthStencilStateParams, // const VkPipelineDepthStencilStateCreateInfo* pDepthStencilState;
+ &colorBlendStateParams[0], // const VkPipelineColorBlendStateCreateInfo* pColorBlendState;
+ &dynamicStateParams, // const VkPipelineDynamicStateCreateInfo* pDynamicState;
+ *m_pipelineLayout, // VkPipelineLayout layout;
+ m_renderPasses[0].get(), // VkRenderPass renderPass;
+ 0u, // deUint32 subpass;
+ DE_NULL, // VkPipeline basePipelineHandle;
+ 0u, // deInt32 basePipelineIndex;
+ };
+
+ // Create first pipeline
+ m_pipelines.emplace_back(createGraphicsPipeline(vk, vkDevice, DE_NULL, &graphicsPipelineParams));
+ // Create second pipeline
+ graphicsPipelineParams.pColorBlendState = &colorBlendStateParams[1];
+ graphicsPipelineParams.renderPass = m_renderPasses[1].get();
+ m_pipelines.emplace_back(createGraphicsPipeline(vk, vkDevice, DE_NULL, &graphicsPipelineParams));
+}
+
+void BlendOperationAdvancedTestCoherentInstance::prepareRenderPass (VkFramebuffer framebuffer, VkPipeline pipeline, VkRenderPass renderpass, deBool secondDraw)
+{
+ const DeviceInterface& vk = m_context.getDeviceInterface();
+
+ VkClearValue attachmentClearValue = makeClearValueColor(clearColorVec4);
+
+ beginRenderPass(vk, *m_cmdBuffer, renderpass, framebuffer, makeRect2D(0, 0, m_renderSize.x(), m_renderSize.y()),
+ (secondDraw ? 0u : 1u),
+ (secondDraw ? DE_NULL : &attachmentClearValue));
+
+ vk.cmdBindPipeline(*m_cmdBuffer, VK_PIPELINE_BIND_POINT_GRAPHICS, pipeline);
+ VkDeviceSize offsets = 0u;
+ vk.cmdBindVertexBuffers(*m_cmdBuffer, 0u, 1u, &m_vertexBuffer.get(), &offsets);
+
+ // There are two different renderpasses, each of them draw
+ // one half of the colors.
+ deBool skippedColors = 0u;
+ for (deUint32 color = 0; color < DE_LENGTH_OF_ARRAY(srcColors)/2; color++)
+ {
+ // Skip ill-formed colors when we have non-premultiplied destination colors.
+ if (m_param.premultipliedDstColor == VK_FALSE)
+ {
+ deBool skipColor = false;
+ for (deUint32 i = 0; i < m_param.colorAttachmentsCount; i++)
+ {
+ Vec4 calculatedColor = calculateFinalColor(m_param, m_param.blendOps[i], srcColors[color], dstColors[color]);
+ if (calculatedColor.w() <= 0.0f && calculatedColor != Vec4(0.0f))
+ {
+ // Skip ill-formed colors, because the spec says the result is undefined.
+ skippedColors++;
+ skipColor = true;
+ break;
+ }
+ }
+ if (skipColor)
+ continue;
+ }
+ deInt32 x = 0;
+ deInt32 y = 0;
+ getCoordinates(color, x, y);
+
+ deUint32 index = secondDraw ? (color + DE_LENGTH_OF_ARRAY(srcColors) / 2) : color;
+
+ // Set source color as push constant
+ vk.cmdPushConstants(*m_cmdBuffer, *m_pipelineLayout, VK_SHADER_STAGE_FRAGMENT_BIT, 0u, sizeof(Vec4), &srcColors[index]);
+ VkRect2D scissor = makeRect2D(x, y, 1u, 1u);
+ vk.cmdSetScissor(*m_cmdBuffer, 0u, 1u, &scissor);
+
+ // To set destination color, we do clear attachment restricting the area to the respective pixel of each color attachment.
+ // Only clear in the first draw, for the second draw the destination color is the result of the first draw's blend.
+ if (secondDraw == DE_FALSE)
+ {
+ std::vector<VkClearAttachment> attachments;
+ VkClearValue clearValue = vk::makeClearValueColorVec4(dstColors[index]);
+
+ const VkClearAttachment attachment =
+ {
+ VK_IMAGE_ASPECT_COLOR_BIT,
+ 0u,
+ clearValue
+ };
+
+ const VkClearRect rect =
+ {
+ scissor,
+ 0u,
+ 1u
+ };
+ vk.cmdClearAttachments(*m_cmdBuffer, 1u, &attachment, 1u, &rect);
+ }
+
+ // Draw
+ vk.cmdDraw(*m_cmdBuffer, (deUint32)m_vertices.size(), 1u, 0u, 0u);
+ }
+
+ // If we break this assert, then we are not testing anything in this test.
+ DE_ASSERT(skippedColors < (DE_LENGTH_OF_ARRAY(srcColors) / 2));
+
+ // Log number of skipped colors
+ if (skippedColors != 0u)
+ {
+ tcu::TestLog& log = m_context.getTestContext().getLog();
+ log << tcu::TestLog::Message << "Skipped " << skippedColors << " out of " << (DE_LENGTH_OF_ARRAY(srcColors) / 2) << " color cases due to ill-formed colors" << tcu::TestLog::EndMessage;
+ }
+ endRenderPass(vk, *m_cmdBuffer);
+}
+
+void BlendOperationAdvancedTestCoherentInstance::prepareCommandBuffer ()
+{
+ const DeviceInterface& vk = m_context.getDeviceInterface();
+
+ beginCommandBuffer(vk, *m_cmdBuffer, 0u);
+
+ vk.cmdPipelineBarrier(*m_cmdBuffer, VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT, VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT | VK_PIPELINE_STAGE_LATE_FRAGMENT_TESTS_BIT, (VkDependencyFlags)0,
+ 0u, DE_NULL, 0u, DE_NULL, (deUint32)m_imageLayoutBarriers.size(), m_imageLayoutBarriers.data());
+
+ prepareRenderPass(m_framebuffers[0].get(), m_pipelines[0].get(), m_renderPasses[0].get(), false);
+
+ if (m_param.coherentOperations == DE_FALSE)
+ {
+ const VkImageMemoryBarrier colorImageBarrier =
+ {
+ VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER, // VkStructureType sType;
+ DE_NULL, // const void* pNext;
+ (VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT |
+ VK_ACCESS_COLOR_ATTACHMENT_READ_NONCOHERENT_BIT_EXT), // VkAccessFlags srcAccessMask;
+ (VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT |
+ VK_ACCESS_COLOR_ATTACHMENT_READ_NONCOHERENT_BIT_EXT), // VkAccessFlags dstAccessMask;
+ VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL, // VkImageLayout oldLayout;
+ VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL, // VkImageLayout newLayout;
+ VK_QUEUE_FAMILY_IGNORED, // deUint32 srcQueueFamilyIndex;
+ VK_QUEUE_FAMILY_IGNORED, // deUint32 dstQueueFamilyIndex;
+ *m_colorImage, // VkImage image;
+ { VK_IMAGE_ASPECT_COLOR_BIT, 0u, 1u, 0u, 1u }, // VkImageSubresourceRange subresourceRange;
+ };
+ vk.cmdPipelineBarrier(*m_cmdBuffer, VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT | VK_PIPELINE_STAGE_LATE_FRAGMENT_TESTS_BIT,
+ VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT | VK_PIPELINE_STAGE_LATE_FRAGMENT_TESTS_BIT, (VkDependencyFlags)0,
+ 0u, DE_NULL, 0u, DE_NULL, 1u, &colorImageBarrier);
+ }
+
+ prepareRenderPass(m_framebuffers[1].get(), m_pipelines[1].get(), m_renderPasses[1].get(), true);
+
+ endCommandBuffer(vk, *m_cmdBuffer);
+}
+
+BlendOperationAdvancedTestCoherentInstance::BlendOperationAdvancedTestCoherentInstance (Context& context,
+ const BlendOperationAdvancedParam param)
+ : TestInstance (context)
+ , m_param (param)
+ , m_renderSize (tcu::UVec2(widthArea, heightArea))
+ , m_colorFormat (VK_FORMAT_R16G16B16A16_SFLOAT)
+ , m_shaderStageCount (0)
+{
+ const DeviceInterface& vk = m_context.getDeviceInterface();
+ const VkDevice vkDevice = m_context.getDevice();
+ const deUint32 queueFamilyIndex = context.getUniversalQueueFamilyIndex();
+
+ // Create vertex buffer
+ {
+ m_vertices = createPoints();
+ DE_ASSERT((deUint32)m_vertices.size() == 6);
+
+ m_vertexBuffer = createBufferAndBindMemory(m_context, m_vertices.size() * sizeof(Vec4), VK_BUFFER_USAGE_VERTEX_BUFFER_BIT, &m_vertexBufferMemory);
+ // Load vertices into vertex buffer
+ deMemcpy(m_vertexBufferMemory->getHostPtr(), m_vertices.data(), m_vertices.size() * sizeof(Vec4));
+ flushAlloc(vk, vkDevice, *m_vertexBufferMemory);
+ }
+
+ // Create render passes
+ m_renderPasses.emplace_back(makeTestRenderPass(param, vk, vkDevice, m_colorFormat, VK_ATTACHMENT_LOAD_OP_CLEAR));
+ m_renderPasses.emplace_back(makeTestRenderPass(param, vk, vkDevice, m_colorFormat, VK_ATTACHMENT_LOAD_OP_LOAD));
+
+ const VkComponentMapping componentMappingRGBA = { VK_COMPONENT_SWIZZLE_R, VK_COMPONENT_SWIZZLE_G, VK_COMPONENT_SWIZZLE_B, VK_COMPONENT_SWIZZLE_A};
+
+ // Create color image
+ m_colorImage = createImage2DAndBindMemory(m_context,
+ m_colorFormat,
+ m_renderSize.x(),
+ m_renderSize.y(),
+ VK_IMAGE_USAGE_COLOR_ATTACHMENT_BIT | VK_IMAGE_USAGE_TRANSFER_SRC_BIT,
+ VK_SAMPLE_COUNT_1_BIT,
+ &m_colorImageAlloc);
+ // Set up image layout transition barriers
+ {
+ VkImageMemoryBarrier colorImageBarrier =
+ {
+ VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER, // VkStructureType sType;
+ DE_NULL, // const void* pNext;
+ 0u, // VkAccessFlags srcAccessMask;
+ (VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT |
+ VK_ACCESS_COLOR_ATTACHMENT_READ_NONCOHERENT_BIT_EXT), // VkAccessFlags dstAccessMask;
+ VK_IMAGE_LAYOUT_UNDEFINED, // VkImageLayout oldLayout;
+ VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL, // VkImageLayout newLayout;
+ VK_QUEUE_FAMILY_IGNORED, // deUint32 srcQueueFamilyIndex;
+ VK_QUEUE_FAMILY_IGNORED, // deUint32 dstQueueFamilyIndex;
+ *m_colorImage, // VkImage image;
+ { VK_IMAGE_ASPECT_COLOR_BIT, 0u, 1u, 0u, 1u }, // VkImageSubresourceRange subresourceRange;
+ };
+
+ m_imageLayoutBarriers.emplace_back(colorImageBarrier);
+ }
+
+ // Create color attachment view
+ {
+ VkImageViewCreateInfo colorAttachmentViewParams =
+ {
+ VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO, // VkStructureType sType;
+ DE_NULL, // const void* pNext;
+ 0u, // VkImageViewCreateFlags flags;
+ *m_colorImage, // VkImage image;
+ VK_IMAGE_VIEW_TYPE_2D, // VkImageViewType viewType;
+ m_colorFormat, // VkFormat format;
+ componentMappingRGBA, // VkComponentMapping components;
+ { VK_IMAGE_ASPECT_COLOR_BIT, 0u, 1u, 0u, 1u }, // VkImageSubresourceRange subresourceRange;
+ };
+
+ m_colorAttachmentView = createImageView(vk, vkDevice, &colorAttachmentViewParams);
+ }
+
+ // Create framebuffers
+ {
+ VkFramebufferCreateInfo framebufferParams =
+ {
+ VK_STRUCTURE_TYPE_FRAMEBUFFER_CREATE_INFO, // VkStructureType sType;
+ DE_NULL, // const void* pNext;
+ 0u, // VkFramebufferCreateFlags flags;
+ m_renderPasses[0].get(), // VkRenderPass renderPass;
+ 1u, // deUint32 attachmentCount;
+ &m_colorAttachmentView.get(), // const VkImageView* pAttachments;
+ (deUint32)m_renderSize.x(), // deUint32 width;
+ (deUint32)m_renderSize.y(), // deUint32 height;
+ 1u, // deUint32 layers;
+ };
+
+ m_framebuffers.emplace_back(createFramebuffer(vk, vkDevice, &framebufferParams));
+ framebufferParams.renderPass = m_renderPasses[1].get();
+ m_framebuffers.emplace_back(createFramebuffer(vk, vkDevice, &framebufferParams));
+ }
+
+ // Bind shader stages
+ {
+ bindShaderStage(VK_SHADER_STAGE_VERTEX_BIT, "vert", "main");
+ bindShaderStage(VK_SHADER_STAGE_FRAGMENT_BIT, "frag", "main");
+ }
+
+
+ // Create pipeline layout
+ {
+ const VkPushConstantRange pushConstantRange =
+ {
+ VK_SHADER_STAGE_FRAGMENT_BIT, // VkShaderStageFlags stageFlags
+ 0, // deUint32 offset
+ sizeof(Vec4) // deUint32 size
+ };
+
+ const VkPipelineLayoutCreateInfo pipelineLayoutParams =
+ {
+ VK_STRUCTURE_TYPE_PIPELINE_LAYOUT_CREATE_INFO, // VkStructureType sType;
+ DE_NULL, // const void* pNext;
+ 0u, // VkPipelineLayoutCreateFlags flags;
+ 0u, // deUint32 setLayoutCount;
+ DE_NULL, // const VkDescriptorSetLayout* pSetLayouts;
+ 1u, // deUint32 pushConstantRangeCount;
+ &pushConstantRange // const VkPushConstantRange* pPushConstantRanges;
+ };
+
+ m_pipelineLayout = createPipelineLayout(vk, vkDevice, &pipelineLayoutParams);
+ }
+
+ // Create pipeline
+ buildPipeline();
+
+ // Create command pool
+ m_cmdPool = createCommandPool(vk, vkDevice, VK_COMMAND_POOL_CREATE_TRANSIENT_BIT | VK_COMMAND_POOL_CREATE_RESET_COMMAND_BUFFER_BIT, queueFamilyIndex);
+
+ // Create command buffer
+ m_cmdBuffer = allocateCommandBuffer(vk, vkDevice, *m_cmdPool, VK_COMMAND_BUFFER_LEVEL_PRIMARY);
+}
+
+tcu::TestStatus BlendOperationAdvancedTestCoherentInstance::iterate (void)
+{
+ const DeviceInterface& vk = m_context.getDeviceInterface();
+ const VkDevice vkDevice = m_context.getDevice();
+ const VkQueue queue = m_context.getUniversalQueue();
+ tcu::TestLog& log = m_context.getTestContext().getLog();
+
+ // Log the blend operations to test
+ {
+ DE_ASSERT(m_param.blendOps.size() == 2u);
+ log << tcu::TestLog::Message << "First depth op: " << de::toLower(getBlendOpStr(m_param.blendOps[0]).toString().substr(3)) << tcu::TestLog::EndMessage;
+ log << tcu::TestLog::Message << "Second depth op: " << de::toLower(getBlendOpStr(m_param.blendOps[1]).toString().substr(3)) << tcu::TestLog::EndMessage;
+
+ }
+
+ prepareCommandBuffer();
+
+ submitCommandsAndWait(vk, vkDevice, queue, m_cmdBuffer.get());
+ return verifyTestResult();
+}
+
+tcu::TestStatus BlendOperationAdvancedTestCoherentInstance::verifyTestResult (void)
+{
+ deBool compareOk = DE_TRUE;
+ const DeviceInterface& vk = m_context.getDeviceInterface();
+ const VkDevice vkDevice = m_context.getDevice();
+ const VkQueue queue = m_context.getUniversalQueue();
+ const deUint32 queueFamilyIndex = m_context.getUniversalQueueFamilyIndex();
+ Allocator& allocator = m_context.getDefaultAllocator();
+ tcu::TextureLevel refImage (vk::mapVkFormat(m_colorFormat), 32, 32);
+
+ tcu::clear(refImage.getAccess(), clearColorVec4);
+
+ // Generate reference image
+ for (deUint32 color = 0; color < DE_LENGTH_OF_ARRAY(srcColors)/2; color++)
+ {
+ deUint32 secondDrawColorIndex = color + DE_LENGTH_OF_ARRAY(srcColors)/2;
+ // Calculate first draw final color
+ Vec4 rectColorTmp = calculateFinalColor(m_param, m_param.blendOps[0], srcColors[color], dstColors[color]);
+
+ if (m_param.premultipliedDstColor == VK_FALSE)
+ {
+ if (rectColorTmp.w() > 0.0f)
+ {
+ rectColorTmp.x() = rectColorTmp.x() / rectColorTmp.w();
+ rectColorTmp.y() = rectColorTmp.y() / rectColorTmp.w();
+ rectColorTmp.z() = rectColorTmp.z() / rectColorTmp.w();
+ }
+ else
+ {
+ // Skip the color check if it is ill-formed.
+ if (rectColorTmp != Vec4(0.0f))
+ continue;
+ }
+ }
+ // Calculate second draw final color
+ Vec4 rectColor = calculateFinalColor(m_param, m_param.blendOps[1], srcColors[secondDrawColorIndex], rectColorTmp);
+ if (m_param.premultipliedDstColor == VK_FALSE)
+ {
+ if (rectColor.w() > 0.0f)
+ {
+ rectColor.x() = rectColor.x() / rectColor.w();
+ rectColor.y() = rectColor.y() / rectColor.w();
+ rectColor.z() = rectColor.z() / rectColor.w();
+ }
+ else
+ {
+ // Skip the color check if it is ill-formed.
+ if (rectColor != Vec4(0.0f))
+ continue;
+ }
+ }
+
+ deInt32 x = 0;
+ deInt32 y = 0;
+ getCoordinates(color, x, y);
+ tcu::clear(tcu::getSubregion(refImage.getAccess(), x, y, 1u, 1u), rectColor);
+ }
+
+ de::MovePtr<tcu::TextureLevel> result = vkt::pipeline::readColorAttachment(vk, vkDevice, queue, queueFamilyIndex, allocator, *m_colorImage, m_colorFormat, m_renderSize);
+ std::ostringstream name;
+ name << "Image comparison. Depth ops: " << de::toLower(getBlendOpStr(m_param.blendOps[0]).toString().substr(3)) << " and " << de::toLower(getBlendOpStr(m_param.blendOps[1]).toString().substr(3));
+ compareOk = tcu::floatThresholdCompare(m_context.getTestContext().getLog(),
+ "FloatImageCompare",
+ name.str().c_str(),
+ refImage.getAccess(),
+ result->getAccess(),
+ Vec4(0.01f, 0.01f, 0.01f, 0.01f),
+ tcu::COMPARE_LOG_RESULT);
+ if (!compareOk)
+ return tcu::TestStatus::fail("Image mismatch");
+
+ return tcu::TestStatus::pass("Result images matches references");
+}
+
+TestInstance* BlendOperationAdvancedTest::createInstance (Context& context) const
+{
+ if (m_param.testMode == TEST_MODE_GENERIC)
+ return new BlendOperationAdvancedTestInstance(context, m_param);
+ else
+ return new BlendOperationAdvancedTestCoherentInstance(context, m_param);
+}
+
+} // anonymous
+
+tcu::TestCaseGroup* createBlendOperationAdvancedTests (tcu::TestContext& testCtx)
+{
+ enum nonpremultiplyEnum
+ {
+ PREMULTIPLY_SRC = 1u,
+ PREMULTIPLY_DST = 2u
+ };
+ deUint32 premultiplyModes[] = { 0u, PREMULTIPLY_SRC, PREMULTIPLY_DST, PREMULTIPLY_SRC | PREMULTIPLY_DST };
+ deUint32 colorAttachmentCounts[] = { 1u, 2u, 4u, 8u, 16u };
+ deBool coherentOps[] = { DE_FALSE, DE_TRUE };
+ VkBlendOp blendOps[] =
+ {
+ VK_BLEND_OP_ZERO_EXT, VK_BLEND_OP_SRC_EXT, VK_BLEND_OP_DST_EXT, VK_BLEND_OP_SRC_OVER_EXT, VK_BLEND_OP_DST_OVER_EXT,
+ VK_BLEND_OP_SRC_IN_EXT, VK_BLEND_OP_DST_IN_EXT, VK_BLEND_OP_SRC_OUT_EXT, VK_BLEND_OP_DST_OUT_EXT, VK_BLEND_OP_SRC_ATOP_EXT,
+ VK_BLEND_OP_DST_ATOP_EXT, VK_BLEND_OP_XOR_EXT, VK_BLEND_OP_MULTIPLY_EXT, VK_BLEND_OP_SCREEN_EXT, VK_BLEND_OP_OVERLAY_EXT,
+ VK_BLEND_OP_DARKEN_EXT, VK_BLEND_OP_LIGHTEN_EXT, VK_BLEND_OP_COLORDODGE_EXT, VK_BLEND_OP_COLORBURN_EXT, VK_BLEND_OP_HARDLIGHT_EXT,
+ VK_BLEND_OP_SOFTLIGHT_EXT, VK_BLEND_OP_DIFFERENCE_EXT, VK_BLEND_OP_EXCLUSION_EXT, VK_BLEND_OP_INVERT_EXT, VK_BLEND_OP_INVERT_RGB_EXT,
+ VK_BLEND_OP_LINEARDODGE_EXT, VK_BLEND_OP_LINEARBURN_EXT, VK_BLEND_OP_VIVIDLIGHT_EXT, VK_BLEND_OP_LINEARLIGHT_EXT, VK_BLEND_OP_PINLIGHT_EXT,
+ VK_BLEND_OP_HARDMIX_EXT, VK_BLEND_OP_HSL_HUE_EXT, VK_BLEND_OP_HSL_SATURATION_EXT, VK_BLEND_OP_HSL_COLOR_EXT, VK_BLEND_OP_HSL_LUMINOSITY_EXT,
+ VK_BLEND_OP_PLUS_EXT, VK_BLEND_OP_PLUS_CLAMPED_EXT, VK_BLEND_OP_PLUS_CLAMPED_ALPHA_EXT, VK_BLEND_OP_PLUS_DARKER_EXT, VK_BLEND_OP_MINUS_EXT,
+ VK_BLEND_OP_MINUS_CLAMPED_EXT, VK_BLEND_OP_CONTRAST_EXT, VK_BLEND_OP_INVERT_OVG_EXT, VK_BLEND_OP_RED_EXT, VK_BLEND_OP_GREEN_EXT, VK_BLEND_OP_BLUE_EXT,
+ };
+
+ de::MovePtr<tcu::TestCaseGroup> tests (new tcu::TestCaseGroup(testCtx, "blend_operation_advanced", "VK_EXT_blend_operation_advanced tests"));
+ de::Random rnd (deStringHash(tests->getName()));
+
+ de::MovePtr<tcu::TestCaseGroup> opsTests (new tcu::TestCaseGroup(testCtx, "ops", "Test each blend operation advance op"));
+
+
+ for (deUint32 colorAttachmentCount = 0u; colorAttachmentCount < DE_LENGTH_OF_ARRAY(colorAttachmentCounts); colorAttachmentCount++)
+ {
+ for (deUint32 overlap = 0; overlap <= VK_BLEND_OVERLAP_CONJOINT_EXT; overlap++)
+ {
+ for (deUint32 premultiply = 0u; premultiply < DE_LENGTH_OF_ARRAY(premultiplyModes); premultiply++)
+ {
+ deUint32 testNumber = 0u;
+ for (deUint64 blendOp = 0u; blendOp < DE_LENGTH_OF_ARRAY(blendOps); blendOp++)
+ {
+ deBool isAdditionalRGBBlendOp = blendOps[blendOp] >= VK_BLEND_OP_PLUS_EXT && blendOps[blendOp] < VK_BLEND_OP_MAX_ENUM;
+
+ // Additional RGB Blend operations are not affected by the blend overlap modes
+ if (isAdditionalRGBBlendOp && overlap != VK_BLEND_OVERLAP_UNCORRELATED_EXT)
+ continue;
+
+ BlendOperationAdvancedParam testParams;
+ testParams.testMode = TEST_MODE_GENERIC;
+ testParams.overlap = (VkBlendOverlapEXT) overlap;
+ testParams.coherentOperations = DE_FALSE;
+ testParams.colorAttachmentsCount = colorAttachmentCounts[colorAttachmentCount];
+ testParams.independentBlend = DE_FALSE;
+ testParams.premultipliedSrcColor = (premultiplyModes[premultiply] & PREMULTIPLY_SRC) ? VK_TRUE : VK_FALSE;
+ testParams.premultipliedDstColor = (premultiplyModes[premultiply] & PREMULTIPLY_DST) ? VK_TRUE : VK_FALSE;
+ testParams.testNumber = testNumber++;
+
+ for (deUint32 numColorAtt = 0; numColorAtt < colorAttachmentCounts[colorAttachmentCount]; numColorAtt++)
+ testParams.blendOps.push_back(blendOps[blendOp]);
+ opsTests->addChild(newTestCase<BlendOperationAdvancedTest>(testCtx, testParams));
+ }
+ }
+ }
+ }
+ tests->addChild(opsTests.release());
+
+ // Independent Blend Tests: test more than one color attachment.
+ de::MovePtr<tcu::TestCaseGroup> independentTests (new tcu::TestCaseGroup(testCtx, "independent", "Test independent blend feature"));
+ deUint32 testNumber = 0u;
+
+ for (deUint32 colorAttachmentCount = 1u; colorAttachmentCount < DE_LENGTH_OF_ARRAY(colorAttachmentCounts); colorAttachmentCount++)
+ {
+ BlendOperationAdvancedParam testParams;
+ testParams.testMode = TEST_MODE_GENERIC;
+ testParams.overlap = VK_BLEND_OVERLAP_UNCORRELATED_EXT;
+ testParams.coherentOperations = DE_FALSE;
+ testParams.colorAttachmentsCount = colorAttachmentCounts[colorAttachmentCount];
+ testParams.independentBlend = DE_TRUE;
+ testParams.premultipliedSrcColor = VK_TRUE;
+ testParams.premultipliedDstColor = VK_TRUE;
+ testParams.testNumber = testNumber++;
+
+ for (deUint32 numColorAtt = 0; numColorAtt < colorAttachmentCounts[colorAttachmentCount]; numColorAtt++)
+ {
+ deUint32 i = de::randomScalar<deUint32>(rnd, 0, DE_LENGTH_OF_ARRAY(blendOps) - 1);
+ testParams.blendOps.push_back(blendOps[i]);
+ }
+ independentTests->addChild(newTestCase<BlendOperationAdvancedTest>(testCtx, testParams));
+ }
+
+ tests->addChild(independentTests.release());
+
+ // Coherent tests, do two consecutive advanced blending operations on the same color attachment.
+ de::MovePtr<tcu::TestCaseGroup> coherentTests (new tcu::TestCaseGroup(testCtx, "coherent", "Test coherent memory"));
+ testNumber = 0u;
+
+ for (deUint32 coherent = 0u; coherent < DE_LENGTH_OF_ARRAY(coherentOps); coherent++)
+ {
+ BlendOperationAdvancedParam testParams;
+ testParams.testMode = TEST_MODE_COHERENT;
+ testParams.overlap = VK_BLEND_OVERLAP_UNCORRELATED_EXT;
+ testParams.coherentOperations = coherentOps[coherent];
+ testParams.colorAttachmentsCount = 1u;
+ testParams.independentBlend = DE_FALSE;
+ testParams.premultipliedSrcColor = VK_TRUE;
+ testParams.premultipliedDstColor = VK_TRUE;
+ testParams.testNumber = testNumber++;
+
+ // We do two consecutive advanced blending operations
+ deUint32 i = de::randomScalar<deUint32>(rnd, 0, DE_LENGTH_OF_ARRAY(blendOps) - 1);
+ testParams.blendOps.push_back(blendOps[i]);
+ i = de::randomScalar<deUint32>(rnd, 0, DE_LENGTH_OF_ARRAY(blendOps) - 1);
+ testParams.blendOps.push_back(blendOps[i]);
+
+ coherentTests->addChild(newTestCase<BlendOperationAdvancedTest>(testCtx, testParams));
+ }
+ tests->addChild(coherentTests.release());
+
+
+ return tests.release();
+}
+
+} // pipeline
+
+} // vkt
--- /dev/null
- if (!isDeviceExtensionSupported(context.getUsedApiVersion(), context.getDeviceExtensions(), extIter))
+/*-------------------------------------------------------------------------
+ * Vulkan Conformance Tests
+ * ------------------------
+ *
+ * Copyright (c) 2017 Google Inc.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ *
+ *//*!
+ * \file
+ * \brief Testing compute shader writing to separate planes of a multiplanar format
+ *//*--------------------------------------------------------------------*/
+
+#include "vktYCbCrStorageImageWriteTests.hpp"
+#include "vktTestCaseUtil.hpp"
+#include "vktTestGroupUtil.hpp"
+#include "vktYCbCrUtil.hpp"
+#include "vkBuilderUtil.hpp"
+#include "vkObjUtil.hpp"
+#include "vkCmdUtil.hpp"
+#include "vkBarrierUtil.hpp"
+#include "vkImageUtil.hpp"
+#include "tcuTexVerifierUtil.hpp"
+#include "vkTypeUtil.hpp"
+#include "vkRefUtil.hpp"
+#include "vkQueryUtil.hpp"
+#include "tcuTestLog.hpp"
+
+namespace vkt
+{
+namespace ycbcr
+{
+namespace
+{
+
+using namespace vk;
+
+struct TestParameters
+{
+ VkFormat format;
+ tcu::UVec3 size;
+ VkImageCreateFlags flags;
+
+ TestParameters (VkFormat format_,
+ const tcu::UVec3& size_,
+ VkImageCreateFlags flags_)
+ : format (format_)
+ , size (size_)
+ , flags (flags_)
+ {
+ }
+
+ TestParameters (void)
+ : format (VK_FORMAT_UNDEFINED)
+ , flags (0u)
+ {
+ }
+};
+
+void checkSupport (Context& context, const TestParameters params)
+{
+ const bool disjoint = (params.flags & VK_IMAGE_CREATE_DISJOINT_BIT) != 0;
+ std::vector<std::string> reqExts;
+
+ if (disjoint)
+ {
+ if (!isCoreDeviceExtension(context.getUsedApiVersion(), "VK_KHR_bind_memory2"))
+ reqExts.push_back("VK_KHR_bind_memory2");
+ if (!isCoreDeviceExtension(context.getUsedApiVersion(), "VK_KHR_get_memory_requirements2"))
+ reqExts.push_back("VK_KHR_get_memory_requirements2");
+ }
+
+ for ( const auto& extIter : reqExts )
+ {
++ if (!context.isDeviceFunctionalitySupported(extIter))
+ TCU_THROW(NotSupportedError, (extIter + " is not supported").c_str());
+ }
+
+ {
+ const VkFormatProperties formatProperties = getPhysicalDeviceFormatProperties(context.getInstanceInterface(),
+ context.getPhysicalDevice(),
+ params.format);
+
+ if ((formatProperties.optimalTilingFeatures & VK_FORMAT_FEATURE_STORAGE_IMAGE_BIT) == 0)
+ TCU_THROW(NotSupportedError, "Storage images are not supported for this format");
+
+ if (disjoint && ((formatProperties.optimalTilingFeatures & VK_FORMAT_FEATURE_DISJOINT_BIT) == 0))
+ TCU_THROW(NotSupportedError, "Disjoint planes are not supported for this format");
+ }
+}
+
+template<typename T>
+inline de::SharedPtr<vk::Unique<T> > makeVkSharedPtr(vk::Move<T> vkMove)
+{
+ return de::SharedPtr<vk::Unique<T> >(new vk::Unique<T>(vkMove));
+}
+
+tcu::UVec3 computeWorkGroupSize(const VkExtent3D& planeExtent)
+{
+ const deUint32 maxComputeWorkGroupInvocations = 128u;
+ const tcu::UVec3 maxComputeWorkGroupSize = tcu::UVec3(128u, 128u, 64u);
+
+ const deUint32 xWorkGroupSize = std::min(std::min(planeExtent.width, maxComputeWorkGroupSize.x()), maxComputeWorkGroupInvocations);
+ const deUint32 yWorkGroupSize = std::min(std::min(planeExtent.height, maxComputeWorkGroupSize.y()), maxComputeWorkGroupInvocations / xWorkGroupSize);
+ const deUint32 zWorkGroupSize = std::min(std::min(planeExtent.depth, maxComputeWorkGroupSize.z()), maxComputeWorkGroupInvocations / (xWorkGroupSize*yWorkGroupSize));
+
+ return tcu::UVec3(xWorkGroupSize, yWorkGroupSize, zWorkGroupSize);
+}
+
+Move<VkPipeline> makeComputePipeline (const DeviceInterface& vk,
+ const VkDevice device,
+ const VkPipelineLayout pipelineLayout,
+ const VkShaderModule shaderModule,
+ const VkSpecializationInfo* specializationInfo)
+{
+ const VkPipelineShaderStageCreateInfo pipelineShaderStageParams =
+ {
+ VK_STRUCTURE_TYPE_PIPELINE_SHADER_STAGE_CREATE_INFO, // VkStructureType sType;
+ DE_NULL, // const void* pNext;
+ 0u, // VkPipelineShaderStageCreateFlags flags;
+ VK_SHADER_STAGE_COMPUTE_BIT, // VkShaderStageFlagBits stage;
+ shaderModule, // VkShaderModule module;
+ "main", // const char* pName;
+ specializationInfo, // const VkSpecializationInfo* pSpecializationInfo;
+ };
+ const VkComputePipelineCreateInfo pipelineCreateInfo =
+ {
+ VK_STRUCTURE_TYPE_COMPUTE_PIPELINE_CREATE_INFO, // VkStructureType sType;
+ DE_NULL, // const void* pNext;
+ 0u, // VkPipelineCreateFlags flags;
+ pipelineShaderStageParams, // VkPipelineShaderStageCreateInfo stage;
+ pipelineLayout, // VkPipelineLayout layout;
+ DE_NULL, // VkPipeline basePipelineHandle;
+ 0, // deInt32 basePipelineIndex;
+ };
+ return createComputePipeline(vk, device, DE_NULL , &pipelineCreateInfo);
+}
+
+vk::VkFormat getPlaneCompatibleFormatForWriting(const vk::PlanarFormatDescription& formatInfo, deUint32 planeNdx)
+{
+ DE_ASSERT(planeNdx < formatInfo.numPlanes);
+ vk::VkFormat result = formatInfo.planes[planeNdx].planeCompatibleFormat;
+
+ // redirect result for some of the YCbCr image formats
+ static const std::pair<vk::VkFormat, vk::VkFormat> ycbcrFormats[] =
+ {
+ { VK_FORMAT_G8B8G8R8_422_UNORM_KHR, VK_FORMAT_R8G8B8A8_UNORM },
+ { VK_FORMAT_G10X6B10X6G10X6R10X6_422_UNORM_4PACK16_KHR, VK_FORMAT_R16G16B16A16_UNORM },
+ { VK_FORMAT_G12X4B12X4G12X4R12X4_422_UNORM_4PACK16_KHR, VK_FORMAT_R16G16B16A16_UNORM },
+ { VK_FORMAT_G16B16G16R16_422_UNORM_KHR, VK_FORMAT_R16G16B16A16_UNORM },
+ { VK_FORMAT_B8G8R8G8_422_UNORM_KHR, VK_FORMAT_R8G8B8A8_UNORM },
+ { VK_FORMAT_B10X6G10X6R10X6G10X6_422_UNORM_4PACK16_KHR, VK_FORMAT_R16G16B16A16_UNORM },
+ { VK_FORMAT_B12X4G12X4R12X4G12X4_422_UNORM_4PACK16_KHR, VK_FORMAT_R16G16B16A16_UNORM },
+ { VK_FORMAT_B16G16R16G16_422_UNORM_KHR, VK_FORMAT_R16G16B16A16_UNORM }
+ };
+ auto it = std::find_if(std::begin(ycbcrFormats), std::end(ycbcrFormats), [result](const std::pair<vk::VkFormat, vk::VkFormat>& p) { return p.first == result; });
+ if (it != std::end(ycbcrFormats))
+ result = it->second;
+ return result;
+}
+
+tcu::TestStatus testStorageImageWrite (Context& context, TestParameters params)
+{
+ const DeviceInterface& vkd = context.getDeviceInterface();
+ const VkDevice device = context.getDevice();
+ const deUint32 queueFamilyIndex = context.getUniversalQueueFamilyIndex();
+ const VkQueue queue = context.getUniversalQueue();
+ const PlanarFormatDescription formatDescription = getPlanarFormatDescription(params.format);
+
+ VkImageCreateInfo imageCreateInfo =
+ {
+ VK_STRUCTURE_TYPE_IMAGE_CREATE_INFO,
+ DE_NULL,
+ params.flags,
+ VK_IMAGE_TYPE_2D,
+ params.format,
+ makeExtent3D(params.size.x(), params.size.y(), params.size.z()),
+ 1u, // mipLevels
+ 1u, // arrayLayers
+ VK_SAMPLE_COUNT_1_BIT,
+ VK_IMAGE_TILING_OPTIMAL,
+ VK_IMAGE_USAGE_TRANSFER_SRC_BIT | VK_IMAGE_USAGE_STORAGE_BIT,
+ VK_SHARING_MODE_EXCLUSIVE,
+ 0u,
+ (const deUint32*)DE_NULL,
+ VK_IMAGE_LAYOUT_UNDEFINED,
+ };
+
+ // check if we need to create VkImageView with different VkFormat than VkImage format
+ VkFormat planeCompatibleFormat0 = getPlaneCompatibleFormatForWriting(formatDescription, 0);
+ if (planeCompatibleFormat0 != getPlaneCompatibleFormat(formatDescription, 0))
+ {
+ imageCreateInfo.flags |= VK_IMAGE_CREATE_MUTABLE_FORMAT_BIT;
+ }
+
+ const Unique<VkImage> image (createImage(vkd, device, &imageCreateInfo));
+ // allocate memory for the whole image, or for each separate plane ( if the params.flags include VK_IMAGE_CREATE_DISJOINT_BIT )
+ const std::vector<AllocationSp> allocations (allocateAndBindImageMemory(vkd, device, context.getDefaultAllocator(), *image, params.format, params.flags, MemoryRequirement::Any));
+
+ // Create descriptor set layout
+ const Unique<VkDescriptorSetLayout> descriptorSetLayout (DescriptorSetLayoutBuilder()
+ .addSingleBinding(VK_DESCRIPTOR_TYPE_STORAGE_IMAGE, VK_SHADER_STAGE_COMPUTE_BIT)
+ .build(vkd, device));
+ const Unique<VkPipelineLayout> pipelineLayout (makePipelineLayout(vkd, device, *descriptorSetLayout));
+
+ // Create descriptor sets
+ const Unique<VkDescriptorPool> descriptorPool (DescriptorPoolBuilder()
+ .addType(VK_DESCRIPTOR_TYPE_STORAGE_IMAGE, 1u)
+ .build(vkd, device, VK_DESCRIPTOR_POOL_CREATE_FREE_DESCRIPTOR_SET_BIT, vk::PlanarFormatDescription::MAX_PLANES));
+
+ // Create command buffer for compute and transfer operations
+ const Unique<VkCommandPool> commandPool (makeCommandPool(vkd, device, queueFamilyIndex));
+ const Unique<VkCommandBuffer> commandBuffer (allocateCommandBuffer(vkd, device, *commandPool, VK_COMMAND_BUFFER_LEVEL_PRIMARY));
+
+ std::vector<de::SharedPtr<vk::Unique<vk::VkShaderModule>>> shaderModules;
+ std::vector<de::SharedPtr<vk::Unique<vk::VkPipeline>>> computePipelines;
+ std::vector<de::SharedPtr<vk::Unique<vk::VkDescriptorSet>>> descriptorSets;
+ std::vector<de::SharedPtr<vk::Unique<vk::VkImageView>>> imageViews;
+
+ deUint32 imageSizeInBytes = 0;
+ deUint32 planeOffsets[PlanarFormatDescription::MAX_PLANES];
+ deUint32 planeRowPitches[PlanarFormatDescription::MAX_PLANES];
+ void* planePointers[PlanarFormatDescription::MAX_PLANES];
+
+ {
+ // Start recording commands
+ beginCommandBuffer(vkd, *commandBuffer);
+
+ for (deUint32 planeNdx = 0; planeNdx < formatDescription.numPlanes; ++planeNdx)
+ {
+ const VkImageAspectFlags aspect = (formatDescription.numPlanes > 1) ? getPlaneAspect(planeNdx) : VK_IMAGE_ASPECT_COLOR_BIT;
+ const VkImageSubresourceRange subresourceRange = makeImageSubresourceRange(aspect, 0u, 1u, 0u, 1u);
+ VkFormat planeCompatibleFormat = getPlaneCompatibleFormatForWriting(formatDescription, planeNdx);
+ vk::PlanarFormatDescription compatibleFormatDescription = (planeCompatibleFormat != getPlaneCompatibleFormat(formatDescription, planeNdx)) ? getPlanarFormatDescription(planeCompatibleFormat) : formatDescription;
+ const tcu::UVec3 compatibleShaderGridSize ( params.size.x() / formatDescription.blockWidth, params.size.y() / formatDescription.blockHeight, params.size.z() / 1u);
+ VkExtent3D shaderExtent = getPlaneExtent(compatibleFormatDescription, VkExtent3D{ compatibleShaderGridSize.x(), compatibleShaderGridSize.y(), compatibleShaderGridSize.z() }, planeNdx, 0u);
+
+ // Create and bind compute pipeline
+ std::ostringstream shaderName;
+ shaderName << "comp" << planeNdx;
+ auto shaderModule = makeVkSharedPtr(createShaderModule(vkd, device, context.getBinaryCollection().get(shaderName.str()), DE_NULL));
+ shaderModules.push_back(shaderModule);
+ auto computePipeline = makeVkSharedPtr(makeComputePipeline(vkd, device, *pipelineLayout, shaderModule->get(), DE_NULL));
+ computePipelines.push_back(computePipeline);
+ vkd.cmdBindPipeline(*commandBuffer, VK_PIPELINE_BIND_POINT_COMPUTE, computePipeline->get());
+
+ auto descriptorSet = makeVkSharedPtr(makeDescriptorSet(vkd, device, *descriptorPool, *descriptorSetLayout));
+ descriptorSets.push_back(descriptorSet);
+
+ auto imageView = makeVkSharedPtr(makeImageView(vkd, device, *image, VK_IMAGE_VIEW_TYPE_2D, planeCompatibleFormat, subresourceRange));
+ imageViews.push_back(imageView);
+ const VkDescriptorImageInfo imageInfo = makeDescriptorImageInfo(DE_NULL, imageView->get(), VK_IMAGE_LAYOUT_GENERAL);
+
+ DescriptorSetUpdateBuilder()
+ .writeSingle(descriptorSet->get(), DescriptorSetUpdateBuilder::Location::binding(0u), VK_DESCRIPTOR_TYPE_STORAGE_IMAGE, &imageInfo)
+ .update(vkd, device);
+
+ vkd.cmdBindDescriptorSets(*commandBuffer, VK_PIPELINE_BIND_POINT_COMPUTE, *pipelineLayout, 0u, 1u, &descriptorSet->get(), 0u, DE_NULL);
+
+ {
+ const VkImageMemoryBarrier imageLayoutChangeBarrier = makeImageMemoryBarrier(0u, VK_ACCESS_SHADER_WRITE_BIT, VK_IMAGE_LAYOUT_UNDEFINED, VK_IMAGE_LAYOUT_GENERAL, *image, subresourceRange, VK_QUEUE_FAMILY_IGNORED, VK_QUEUE_FAMILY_IGNORED);
+ vkd.cmdPipelineBarrier(*commandBuffer, VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT, VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT, 0u, 0u, DE_NULL, 0u, DE_NULL, 1u, &imageLayoutChangeBarrier);
+ }
+
+ {
+ const tcu::UVec3 workGroupSize = computeWorkGroupSize(shaderExtent);
+
+ const deUint32 xWorkGroupCount = shaderExtent.width / workGroupSize.x() + (shaderExtent.width % workGroupSize.x() ? 1u : 0u);
+ const deUint32 yWorkGroupCount = shaderExtent.height / workGroupSize.y() + (shaderExtent.height % workGroupSize.y() ? 1u : 0u);
+ const deUint32 zWorkGroupCount = shaderExtent.depth / workGroupSize.z() + (shaderExtent.depth % workGroupSize.z() ? 1u : 0u);
+
+ const tcu::UVec3 maxComputeWorkGroupCount = tcu::UVec3(65535u, 65535u, 65535u);
+
+ if (maxComputeWorkGroupCount.x() < xWorkGroupCount ||
+ maxComputeWorkGroupCount.y() < yWorkGroupCount ||
+ maxComputeWorkGroupCount.z() < zWorkGroupCount)
+ {
+ TCU_THROW(NotSupportedError, "Image size is not supported");
+ }
+
+ vkd.cmdDispatch(*commandBuffer, xWorkGroupCount, yWorkGroupCount, zWorkGroupCount);
+ }
+
+ {
+ const VkImageMemoryBarrier imageTransferBarrier = makeImageMemoryBarrier(VK_ACCESS_SHADER_WRITE_BIT, VK_ACCESS_TRANSFER_READ_BIT, VK_IMAGE_LAYOUT_GENERAL, VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL, *image, subresourceRange);
+ vkd.cmdPipelineBarrier(*commandBuffer, VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT, VK_PIPELINE_STAGE_TRANSFER_BIT, 0u, 0u, DE_NULL, 0u, DE_NULL, 1u, &imageTransferBarrier);
+ }
+ }
+
+ for (deUint32 planeNdx = 0; planeNdx < formatDescription.numPlanes; ++planeNdx)
+ {
+ planeOffsets[planeNdx] = imageSizeInBytes;
+ const deUint32 planeW = imageCreateInfo.extent.width / (formatDescription.blockWidth * formatDescription.planes[planeNdx].widthDivisor);
+ planeRowPitches[planeNdx] = formatDescription.planes[planeNdx].elementSizeBytes * planeW;
+ imageSizeInBytes += getPlaneSizeInBytes(formatDescription, makeExtent3D( params.size.x(), params.size.y(), params.size.z()) , planeNdx, 0u, BUFFER_IMAGE_COPY_OFFSET_GRANULARITY);
+ }
+
+ const VkBufferCreateInfo outputBufferCreateInfo = makeBufferCreateInfo(imageSizeInBytes, VK_BUFFER_USAGE_TRANSFER_DST_BIT);
+ const Unique<VkBuffer> outputBuffer ( createBuffer(vkd, device, &outputBufferCreateInfo) );
+ const de::UniquePtr<Allocation> outputBufferAlloc ( bindBuffer(vkd, device, context.getDefaultAllocator(), *outputBuffer, MemoryRequirement::HostVisible) );
+ std::vector<VkBufferImageCopy> bufferImageCopy ( formatDescription.numPlanes );
+
+ for (deUint32 planeNdx = 0; planeNdx < formatDescription.numPlanes; ++planeNdx)
+ {
+ const VkImageAspectFlags aspect = (formatDescription.numPlanes > 1) ? getPlaneAspect(planeNdx) : VK_IMAGE_ASPECT_COLOR_BIT;
+
+ bufferImageCopy[planeNdx] =
+ {
+ planeOffsets[planeNdx], // VkDeviceSize bufferOffset;
+ 0u, // deUint32 bufferRowLength;
+ 0u, // deUint32 bufferImageHeight;
+ makeImageSubresourceLayers(aspect, 0u, 0u, 1u), // VkImageSubresourceLayers imageSubresource;
+ makeOffset3D(0, 0, 0), // VkOffset3D imageOffset;
+ getPlaneExtent(formatDescription, makeExtent3D(params.size.x(), params.size.y(), params.size.z()), planeNdx, 0u) // VkExtent3D imageExtent;
+ };
+ }
+ vkd.cmdCopyImageToBuffer(*commandBuffer, *image, VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL, *outputBuffer, static_cast<deUint32>(bufferImageCopy.size()), bufferImageCopy.data());
+
+ {
+ const VkBufferMemoryBarrier outputBufferHostReadBarrier = makeBufferMemoryBarrier
+ (
+ VK_ACCESS_TRANSFER_WRITE_BIT,
+ VK_ACCESS_HOST_READ_BIT,
+ *outputBuffer,
+ 0u,
+ imageSizeInBytes
+ );
+
+ vkd.cmdPipelineBarrier(*commandBuffer, VK_PIPELINE_STAGE_TRANSFER_BIT, VK_PIPELINE_STAGE_HOST_BIT, 0u, 0u, DE_NULL, 1u, &outputBufferHostReadBarrier, 0u, DE_NULL);
+ }
+
+ // End recording commands
+ endCommandBuffer(vkd, *commandBuffer);
+
+ // Submit commands for execution and wait for completion
+ submitCommandsAndWait(vkd, device, queue, *commandBuffer);
+
+ // Retrieve data from buffer to host memory
+ invalidateAlloc(vkd, device, *outputBufferAlloc);
+ deUint8* outputData = static_cast<deUint8*>(outputBufferAlloc->getHostPtr());
+
+ for (deUint32 planeNdx = 0; planeNdx < formatDescription.numPlanes; ++planeNdx)
+ planePointers[planeNdx] = outputData + static_cast<size_t>(planeOffsets[planeNdx]);
+ }
+
+ // write result images to log file
+ for (deUint32 channelNdx = 0; channelNdx < 4; ++channelNdx)
+ {
+ if (!formatDescription.hasChannelNdx(channelNdx))
+ continue;
+ deUint32 planeNdx = formatDescription.channels[channelNdx].planeNdx;
+ vk::VkFormat planeCompatibleFormat = getPlaneCompatibleFormatForWriting(formatDescription, planeNdx);
+ vk::PlanarFormatDescription compatibleFormatDescription = (planeCompatibleFormat != getPlaneCompatibleFormat(formatDescription, planeNdx)) ? getPlanarFormatDescription(planeCompatibleFormat) : formatDescription;
+ const tcu::UVec3 compatibleShaderGridSize ( params.size.x() / formatDescription.blockWidth, params.size.y() / formatDescription.blockHeight, params.size.z() / 1u );
+ tcu::ConstPixelBufferAccess pixelBuffer = vk::getChannelAccess(compatibleFormatDescription, compatibleShaderGridSize, planeRowPitches, (const void* const*)planePointers, channelNdx);
+ std::ostringstream str;
+ str << "image" << channelNdx;
+ context.getTestContext().getLog() << tcu::LogImage(str.str(), str.str(), pixelBuffer);;
+ }
+
+ // verify data
+ const float epsilon = 1e-5f;
+ for (deUint32 channelNdx = 0; channelNdx < 4; ++channelNdx)
+ {
+ if (!formatDescription.hasChannelNdx(channelNdx))
+ continue;
+
+ deUint32 planeNdx = formatDescription.channels[channelNdx].planeNdx;
+ vk::VkFormat planeCompatibleFormat = getPlaneCompatibleFormatForWriting(formatDescription, planeNdx);
+ vk::PlanarFormatDescription compatibleFormatDescription = (planeCompatibleFormat != getPlaneCompatibleFormat(formatDescription, planeNdx)) ? getPlanarFormatDescription(planeCompatibleFormat) : formatDescription;
+ const tcu::UVec3 compatibleShaderGridSize ( params.size.x() / formatDescription.blockWidth, params.size.y() / formatDescription.blockHeight, params.size.z() / 1u );
+ VkExtent3D compatibleImageSize { imageCreateInfo.extent.width / formatDescription.blockWidth, imageCreateInfo.extent.height / formatDescription.blockHeight, imageCreateInfo.extent.depth / 1u };
+ tcu::ConstPixelBufferAccess pixelBuffer = vk::getChannelAccess(compatibleFormatDescription, compatibleShaderGridSize, planeRowPitches, (const void* const*)planePointers, channelNdx);
+ VkExtent3D planeExtent = getPlaneExtent(compatibleFormatDescription, compatibleImageSize, planeNdx, 0u);
+ tcu::IVec3 pixelDivider = pixelBuffer.getDivider();
+ float fixedPointError = tcu::TexVerifierUtil::computeFixedPointError(formatDescription.channels[channelNdx].sizeBits);
+
+ for (deUint32 offsetZ = 0u; offsetZ < planeExtent.depth; ++offsetZ)
+ for (deUint32 offsetY = 0u; offsetY < planeExtent.height; ++offsetY)
+ for (deUint32 offsetX = 0u; offsetX < planeExtent.width; ++offsetX)
+ {
+ deUint32 iReferenceValue;
+ float fReferenceValue;
+ switch (channelNdx)
+ {
+ case 0:
+ iReferenceValue = offsetX % 127u;
+ fReferenceValue = static_cast<float>(iReferenceValue) / 127.f;
+ break;
+ case 1:
+ iReferenceValue = offsetY % 127u;
+ fReferenceValue = static_cast<float>(iReferenceValue) / 127.f;
+ break;
+ case 2:
+ iReferenceValue = offsetZ % 127u;
+ fReferenceValue = static_cast<float>(iReferenceValue) / 127.f;
+ break;
+ case 3:
+ iReferenceValue = 0u;
+ fReferenceValue = 0.f;
+ break;
+ default: DE_FATAL("Unexpected channel index"); break;
+ }
+ float acceptableError = epsilon;
+
+ switch (formatDescription.channels[channelNdx].type)
+ {
+ case tcu::TEXTURECHANNELCLASS_SIGNED_INTEGER:
+ case tcu::TEXTURECHANNELCLASS_UNSIGNED_INTEGER:
+ {
+ tcu::UVec4 outputValue = pixelBuffer.getPixelUint(offsetX * pixelDivider.x(), offsetY * pixelDivider.y(), 0);
+
+ if (outputValue.x() != iReferenceValue)
+ return tcu::TestStatus::fail("Failed");
+
+ break;
+ }
+ case tcu::TEXTURECHANNELCLASS_UNSIGNED_FIXED_POINT:
+ case tcu::TEXTURECHANNELCLASS_SIGNED_FIXED_POINT:
+ {
+ acceptableError += fixedPointError;
+ tcu::Vec4 outputValue = pixelBuffer.getPixel(offsetX * pixelDivider.x(), offsetY * pixelDivider.y(), 0);
+
+ if (deAbs(outputValue.x() - fReferenceValue) > acceptableError)
+ return tcu::TestStatus::fail("Failed");
+
+ break;
+ }
+ case tcu::TEXTURECHANNELCLASS_FLOATING_POINT:
+ {
+ const tcu::Vec4 outputValue = pixelBuffer.getPixel(offsetX * pixelDivider.x(), offsetY * pixelDivider.y(), 0);
+
+ if (deAbs( outputValue.x() - fReferenceValue) > acceptableError)
+ return tcu::TestStatus::fail("Failed");
+
+ break;
+ }
+ default: DE_FATAL("Unexpected channel type"); break;
+ }
+ }
+ }
+ return tcu::TestStatus::pass("Passed");
+}
+
+std::string getShaderImageType (const vk::PlanarFormatDescription& description)
+{
+ std::string formatPart;
+
+ // all PlanarFormatDescription types have at least one channel ( 0 ) and all channel types are the same :
+ switch (description.channels[0].type)
+ {
+ case tcu::TEXTURECHANNELCLASS_SIGNED_INTEGER:
+ formatPart = "i";
+ break;
+ case tcu::TEXTURECHANNELCLASS_UNSIGNED_INTEGER:
+ formatPart = "u";
+ break;
+ case tcu::TEXTURECHANNELCLASS_UNSIGNED_FIXED_POINT:
+ case tcu::TEXTURECHANNELCLASS_SIGNED_FIXED_POINT:
+ case tcu::TEXTURECHANNELCLASS_FLOATING_POINT:
+ break;
+
+ default:
+ DE_FATAL("Unexpected channel type");
+ }
+
+ return formatPart + "image2D";
+}
+
+std::string getShaderImageDataType (const vk::PlanarFormatDescription& description)
+{
+ switch (description.channels[0].type)
+ {
+ case tcu::TEXTURECHANNELCLASS_UNSIGNED_INTEGER:
+ return "uvec4";
+ case tcu::TEXTURECHANNELCLASS_SIGNED_INTEGER:
+ return "ivec4";
+ case tcu::TEXTURECHANNELCLASS_UNSIGNED_FIXED_POINT:
+ case tcu::TEXTURECHANNELCLASS_SIGNED_FIXED_POINT:
+ case tcu::TEXTURECHANNELCLASS_FLOATING_POINT:
+ return "vec4";
+ default:
+ DE_FATAL("Unexpected channel type");
+ return "";
+ }
+}
+
+std::string getFormatValueString (const std::vector<std::pair<deUint32, deUint32>>& channelsOnPlane,
+ const std::vector<std::string>& formatValueStrings)
+{
+ std::string result = "( ";
+ deUint32 i;
+ for (i=0; i<channelsOnPlane.size(); ++i)
+ {
+ result += formatValueStrings[channelsOnPlane[i].first];
+ if (i < 3)
+ result += ", ";
+ }
+ for (; i < 4; ++i)
+ {
+ result += "0";
+ if (i < 3)
+ result += ", ";
+ }
+ result += " )";
+ return result;
+}
+
+std::string getShaderImageFormatQualifier (VkFormat format)
+{
+ switch (format)
+ {
+ case VK_FORMAT_R8_SINT: return "r8i";
+ case VK_FORMAT_R16_SINT: return "r16i";
+ case VK_FORMAT_R32_SINT: return "r32i";
+ case VK_FORMAT_R8_UINT: return "r8ui";
+ case VK_FORMAT_R16_UINT: return "r16ui";
+ case VK_FORMAT_R32_UINT: return "r32ui";
+ case VK_FORMAT_R8_SNORM: return "r8_snorm";
+ case VK_FORMAT_R16_SNORM: return "r16_snorm";
+ case VK_FORMAT_R8_UNORM: return "r8";
+ case VK_FORMAT_R16_UNORM: return "r16";
+
+ case VK_FORMAT_R8G8_SINT: return "rg8i";
+ case VK_FORMAT_R16G16_SINT: return "rg16i";
+ case VK_FORMAT_R32G32_SINT: return "rg32i";
+ case VK_FORMAT_R8G8_UINT: return "rg8ui";
+ case VK_FORMAT_R16G16_UINT: return "rg16ui";
+ case VK_FORMAT_R32G32_UINT: return "rg32ui";
+ case VK_FORMAT_R8G8_SNORM: return "rg8_snorm";
+ case VK_FORMAT_R16G16_SNORM: return "rg16_snorm";
+ case VK_FORMAT_R8G8_UNORM: return "rg8";
+ case VK_FORMAT_R16G16_UNORM: return "rg16";
+
+ case VK_FORMAT_R8G8B8A8_SINT: return "rgba8i";
+ case VK_FORMAT_R16G16B16A16_SINT: return "rgba16i";
+ case VK_FORMAT_R32G32B32A32_SINT: return "rgba32i";
+ case VK_FORMAT_R8G8B8A8_UINT: return "rgba8ui";
+ case VK_FORMAT_R16G16B16A16_UINT: return "rgba16ui";
+ case VK_FORMAT_R32G32B32A32_UINT: return "rgba32ui";
+ case VK_FORMAT_R8G8B8A8_SNORM: return "rgba8_snorm";
+ case VK_FORMAT_R16G16B16A16_SNORM: return "rgba16_snorm";
+ case VK_FORMAT_R8G8B8A8_UNORM: return "rgba8";
+ case VK_FORMAT_R16G16B16A16_UNORM: return "rgba16";
+
+ case VK_FORMAT_G8B8G8R8_422_UNORM: return "rgba8";
+ case VK_FORMAT_B8G8R8G8_422_UNORM: return "rgba8";
+ case VK_FORMAT_G8_B8_R8_3PLANE_420_UNORM: return "rgba8";
+ case VK_FORMAT_G8_B8R8_2PLANE_420_UNORM: return "rgba8";
+ case VK_FORMAT_G8_B8_R8_3PLANE_422_UNORM: return "rgba8";
+ case VK_FORMAT_G8_B8R8_2PLANE_422_UNORM: return "rgba8";
+ case VK_FORMAT_G8_B8_R8_3PLANE_444_UNORM: return "rgba8";
+ case VK_FORMAT_R10X6_UNORM_PACK16: return "r16";
+ case VK_FORMAT_R10X6G10X6_UNORM_2PACK16: return "rg16";
+ case VK_FORMAT_R10X6G10X6B10X6A10X6_UNORM_4PACK16: return "rgba16";
+ case VK_FORMAT_G10X6B10X6G10X6R10X6_422_UNORM_4PACK16: return "rgba16";
+ case VK_FORMAT_B10X6G10X6R10X6G10X6_422_UNORM_4PACK16: return "rgba16";
+ case VK_FORMAT_G10X6_B10X6_R10X6_3PLANE_420_UNORM_3PACK16: return "rgba16";
+ case VK_FORMAT_G10X6_B10X6R10X6_2PLANE_420_UNORM_3PACK16: return "rgba16";
+ case VK_FORMAT_G10X6_B10X6_R10X6_3PLANE_422_UNORM_3PACK16: return "rgba16";
+ case VK_FORMAT_G10X6_B10X6R10X6_2PLANE_422_UNORM_3PACK16: return "rgba16";
+ case VK_FORMAT_G10X6_B10X6_R10X6_3PLANE_444_UNORM_3PACK16: return "rgba16";
+ case VK_FORMAT_R12X4_UNORM_PACK16: return "r16";
+ case VK_FORMAT_R12X4G12X4_UNORM_2PACK16: return "rg16";
+ case VK_FORMAT_R12X4G12X4B12X4A12X4_UNORM_4PACK16: return "rgba16";
+ case VK_FORMAT_G12X4B12X4G12X4R12X4_422_UNORM_4PACK16: return "rgba16";
+ case VK_FORMAT_B12X4G12X4R12X4G12X4_422_UNORM_4PACK16: return "rgba16";
+ case VK_FORMAT_G12X4_B12X4_R12X4_3PLANE_420_UNORM_3PACK16: return "rgba16";
+ case VK_FORMAT_G12X4_B12X4R12X4_2PLANE_420_UNORM_3PACK16: return "rgba16";
+ case VK_FORMAT_G12X4_B12X4_R12X4_3PLANE_422_UNORM_3PACK16: return "rgba16";
+ case VK_FORMAT_G12X4_B12X4R12X4_2PLANE_422_UNORM_3PACK16: return "rgba16";
+ case VK_FORMAT_G12X4_B12X4_R12X4_3PLANE_444_UNORM_3PACK16: return "rgba16";
+ case VK_FORMAT_G16B16G16R16_422_UNORM: return "rgba16";
+ case VK_FORMAT_B16G16R16G16_422_UNORM: return "rgba16";
+ case VK_FORMAT_G16_B16_R16_3PLANE_420_UNORM: return "rgba16";
+ case VK_FORMAT_G16_B16R16_2PLANE_420_UNORM: return "rgba16";
+ case VK_FORMAT_G16_B16_R16_3PLANE_422_UNORM: return "rgba16";
+ case VK_FORMAT_G16_B16R16_2PLANE_422_UNORM: return "rgba16";
+ case VK_FORMAT_G16_B16_R16_3PLANE_444_UNORM: return "rgba16";
+
+ default:
+ DE_FATAL("Unexpected texture format");
+ return "error";
+ }
+}
+
+void initPrograms (SourceCollections& sourceCollections, TestParameters params)
+{
+ // Create compute program
+ const char* const versionDecl = glu::getGLSLVersionDeclaration(glu::GLSL_VERSION_440);
+ const PlanarFormatDescription formatDescription = getPlanarFormatDescription(params.format);
+ const std::string imageTypeStr = getShaderImageType(formatDescription);
+ const std::string formatDataStr = getShaderImageDataType(formatDescription);
+ const tcu::UVec3 shaderGridSize ( params.size.x(), params.size.y(), params.size.z() );
+
+ std::vector<std::string> formatValueStrings;
+ switch (formatDescription.channels[0].type)
+ {
+ case tcu::TEXTURECHANNELCLASS_SIGNED_INTEGER:
+ case tcu::TEXTURECHANNELCLASS_UNSIGNED_INTEGER:
+ formatValueStrings = {
+ "int(gl_GlobalInvocationID.x) % 127",
+ "int(gl_GlobalInvocationID.y) % 127",
+ "int(gl_GlobalInvocationID.z) % 127",
+ "1"
+ };
+ break;
+ case tcu::TEXTURECHANNELCLASS_UNSIGNED_FIXED_POINT:
+ case tcu::TEXTURECHANNELCLASS_SIGNED_FIXED_POINT:
+ case tcu::TEXTURECHANNELCLASS_FLOATING_POINT:
+ formatValueStrings = {
+ "float(int(gl_GlobalInvocationID.x) % 127) / 127.0" ,
+ "float(int(gl_GlobalInvocationID.y) % 127) / 127.0",
+ "float(int(gl_GlobalInvocationID.z) % 127) / 127.0",
+ "1.0"
+ };
+ break;
+ default: DE_ASSERT(false); break;
+ }
+
+ for (deUint32 planeNdx = 0; planeNdx < formatDescription.numPlanes; ++planeNdx)
+ {
+ VkFormat planeCompatibleFormat = getPlaneCompatibleFormatForWriting(formatDescription, planeNdx);
+ vk::PlanarFormatDescription compatibleFormatDescription = (planeCompatibleFormat != getPlaneCompatibleFormat(formatDescription, planeNdx)) ? getPlanarFormatDescription(planeCompatibleFormat) : formatDescription;
+ VkExtent3D compatibleShaderGridSize { shaderGridSize.x() / formatDescription.blockWidth, shaderGridSize.y() / formatDescription.blockHeight, shaderGridSize.z() / 1u };
+
+ std::vector<std::pair<deUint32, deUint32>> channelsOnPlane;
+ for (deUint32 channelNdx = 0; channelNdx < 4; ++channelNdx)
+ {
+ if (!formatDescription.hasChannelNdx(channelNdx))
+ continue;
+ if (formatDescription.channels[channelNdx].planeNdx != planeNdx)
+ continue;
+ channelsOnPlane.push_back({ channelNdx,formatDescription.channels[channelNdx].offsetBits });
+ }
+ // reorder channels for multi-planar images
+ if (formatDescription.numPlanes > 1)
+ std::sort(begin(channelsOnPlane), end(channelsOnPlane), [](const std::pair<deUint32, deUint32>& lhs, const std::pair<deUint32, deUint32>& rhs) { return lhs.second < rhs.second; });
+ std::string formatValueStr = getFormatValueString(channelsOnPlane, formatValueStrings);
+ VkExtent3D shaderExtent = getPlaneExtent(compatibleFormatDescription, compatibleShaderGridSize, planeNdx, 0);
+ const std::string formatQualifierStr = getShaderImageFormatQualifier(formatDescription.planes[planeNdx].planeCompatibleFormat);
+ const tcu::UVec3 workGroupSize = computeWorkGroupSize(shaderExtent);
+
+ std::ostringstream src;
+ src << versionDecl << "\n"
+ << "layout (local_size_x = " << workGroupSize.x() << ", local_size_y = " << workGroupSize.y() << ", local_size_z = " << workGroupSize.z() << ") in; \n"
+ << "layout (binding = 0, " << formatQualifierStr << ") writeonly uniform highp " << imageTypeStr << " u_image;\n"
+ << "void main (void)\n"
+ << "{\n"
+ << " if( gl_GlobalInvocationID.x < " << shaderExtent.width << " ) \n"
+ << " if( gl_GlobalInvocationID.y < " << shaderExtent.height << " ) \n"
+ << " if( gl_GlobalInvocationID.z < " << shaderExtent.depth << " ) \n"
+ << " {\n"
+ << " imageStore(u_image, ivec2( gl_GlobalInvocationID.x, gl_GlobalInvocationID.y ) ,"
+ << formatDataStr << formatValueStr << ");\n"
+ << " }\n"
+ << "}\n";
+ std::ostringstream shaderName;
+ shaderName << "comp" << planeNdx;
+ sourceCollections.glslSources.add(shaderName.str()) << glu::ComputeSource(src.str());
+ }
+}
+
+tcu::TestCaseGroup* populateStorageImageWriteFormatGroup (tcu::TestContext& testCtx, de::MovePtr<tcu::TestCaseGroup> testGroup)
+{
+ const std::vector<tcu::UVec3> availableSizes{ tcu::UVec3(512u, 512u, 1u), tcu::UVec3(1024u, 128u, 1u), tcu::UVec3(66u, 32u, 1u) };
+
+ for (int formatNdx = VK_YCBCR_FORMAT_FIRST; formatNdx < VK_YCBCR_FORMAT_LAST; formatNdx++)
+ {
+ const VkFormat format = (VkFormat)formatNdx;
+ tcu::UVec3 imageSizeAlignment = getImageSizeAlignment(format);
+ std::string formatName = de::toLower(de::toString(format).substr(10));
+ de::MovePtr<tcu::TestCaseGroup> formatGroup ( new tcu::TestCaseGroup(testCtx, formatName.c_str(), "") );
+
+ for (size_t sizeNdx = 0; sizeNdx < availableSizes.size(); sizeNdx++)
+ {
+ const tcu::UVec3 imageSize = availableSizes[sizeNdx];
+
+ // skip test for images with odd sizes for some YCbCr formats
+ if ((imageSize.x() % imageSizeAlignment.x()) != 0)
+ continue;
+ if ((imageSize.y() % imageSizeAlignment.y()) != 0)
+ continue;
+
+ std::ostringstream stream;
+ stream << imageSize.x() << "_" << imageSize.y() << "_" << imageSize.z();
+ de::MovePtr<tcu::TestCaseGroup> sizeGroup(new tcu::TestCaseGroup(testCtx, stream.str().c_str(), ""));
+
+ addFunctionCaseWithPrograms(sizeGroup.get(), "joint", "", checkSupport, initPrograms, testStorageImageWrite, TestParameters(format, imageSize, 0u));
+ addFunctionCaseWithPrograms(sizeGroup.get(), "disjoint", "", checkSupport, initPrograms, testStorageImageWrite, TestParameters(format, imageSize, (VkImageCreateFlags)VK_IMAGE_CREATE_DISJOINT_BIT));
+
+ formatGroup->addChild(sizeGroup.release());
+ }
+ testGroup->addChild(formatGroup.release());
+ }
+ return testGroup.release();
+}
+
+} // namespace
+
+tcu::TestCaseGroup* createStorageImageWriteTests (tcu::TestContext& testCtx)
+{
+ de::MovePtr<tcu::TestCaseGroup> testGroup(new tcu::TestCaseGroup(testCtx, "storage_image_write", "Writing to YCbCr storage images"));
+ return populateStorageImageWriteFormatGroup(testCtx, testGroup);
+}
+
+} // ycbcr
+} // vkt