Hash :
b16d69c3
Author :
Date :
2019-05-13T16:28:27
Vulkan: Add support for surface multisampling A multisample image is created for the surface if multisampling is enabled. Prior to present, this multisample image is resolved into the swapchain image. FramebufferVk::readPixelsImpl similarly has got the ability to resolve the region of interest into a temporary image prior to readback. Tests are added to render a point, line and a triangle on a 4x multisampled surface. Bug: angleproject:3204 Change-Id: I34aca502fa1918b5cbf000ff11521c350372e051 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1610188 Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Jamie Madill <jmadill@chromium.org> Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396
//
// Copyright 2018 The ANGLE Project Authors. All rights reserved.
// Use of this source code is governed by a BSD-style license that can be
// found in the LICENSE file.
//
// vk_utils:
// Helper functions for the Vulkan Caps.
//
#include "libANGLE/renderer/vulkan/vk_caps_utils.h"
#include <type_traits>
#include "common/utilities.h"
#include "libANGLE/Caps.h"
#include "libANGLE/formatutils.h"
#include "libANGLE/renderer/vulkan/DisplayVk.h"
#include "libANGLE/renderer/vulkan/RendererVk.h"
#include "vk_format_utils.h"
namespace
{
constexpr unsigned int kComponentsPerVector = 4;
} // anonymous namespace
namespace rx
{
void RendererVk::ensureCapsInitialized() const
{
if (mCapsInitialized)
return;
mCapsInitialized = true;
ASSERT(mCurrentQueueFamilyIndex < mQueueFamilyProperties.size());
const VkQueueFamilyProperties &queueFamilyProperties =
mQueueFamilyProperties[mCurrentQueueFamilyIndex];
mNativeExtensions.setTextureExtensionSupport(mNativeTextureCaps);
// Enable this for simple buffer readback testing, but some functionality is missing.
// TODO(jmadill): Support full mapBufferRange extension.
mNativeExtensions.mapBuffer = true;
mNativeExtensions.mapBufferRange = true;
mNativeExtensions.textureStorage = true;
mNativeExtensions.drawBuffers = true;
mNativeExtensions.fragDepth = true;
mNativeExtensions.framebufferBlit = true;
mNativeExtensions.copyTexture = true;
mNativeExtensions.copyCompressedTexture = true;
mNativeExtensions.debugMarker = true;
mNativeExtensions.robustness = true;
mNativeExtensions.textureBorderClamp = false; // not implemented yet
mNativeExtensions.translatedShaderSource = true;
mNativeExtensions.eglImage = true;
mNativeExtensions.eglImageExternal = true;
// TODO(geofflang): Support GL_OES_EGL_image_external_essl3. http://anglebug.com/2668
mNativeExtensions.eglImageExternalEssl3 = false;
mNativeExtensions.memoryObject = true;
mNativeExtensions.memoryObjectFd = getFeatures().supportsExternalMemoryFd.enabled;
mNativeExtensions.semaphore = true;
mNativeExtensions.semaphoreFd = getFeatures().supportsExternalSemaphoreFd.enabled;
// TODO: Enable this always and emulate instanced draws if any divisor exceeds the maximum
// supported. http://anglebug.com/2672
mNativeExtensions.instancedArraysANGLE = mMaxVertexAttribDivisor > 1;
// Only expose robust buffer access if the physical device supports it.
mNativeExtensions.robustBufferAccessBehavior = mPhysicalDeviceFeatures.robustBufferAccess;
mNativeExtensions.eglSync = true;
// We use secondary command buffers almost everywhere and they require a feature to be
// able to execute in the presence of queries. As a result, we won't support queries
// unless that feature is available.
mNativeExtensions.occlusionQueryBoolean =
vk::CommandBuffer::SupportsQueries(mPhysicalDeviceFeatures);
// From the Vulkan specs:
// > The number of valid bits in a timestamp value is determined by the
// > VkQueueFamilyProperties::timestampValidBits property of the queue on which the timestamp is
// > written. Timestamps are supported on any queue which reports a non-zero value for
// > timestampValidBits via vkGetPhysicalDeviceQueueFamilyProperties.
mNativeExtensions.disjointTimerQuery = queueFamilyProperties.timestampValidBits > 0;
mNativeExtensions.queryCounterBitsTimeElapsed = queueFamilyProperties.timestampValidBits;
mNativeExtensions.queryCounterBitsTimestamp = queueFamilyProperties.timestampValidBits;
mNativeExtensions.textureFilterAnisotropic =
mPhysicalDeviceFeatures.samplerAnisotropy &&
mPhysicalDeviceProperties.limits.maxSamplerAnisotropy > 1.0f;
mNativeExtensions.maxTextureAnisotropy =
mNativeExtensions.textureFilterAnisotropic
? mPhysicalDeviceProperties.limits.maxSamplerAnisotropy
: 0.0f;
// Vulkan natively supports non power-of-two textures
mNativeExtensions.textureNPOT = true;
// Vulkan natively supports standard derivatives
mNativeExtensions.standardDerivatives = true;
// https://vulkan.lunarg.com/doc/view/1.0.30.0/linux/vkspec.chunked/ch31s02.html
mNativeCaps.maxElementIndex = std::numeric_limits<GLuint>::max() - 1;
mNativeCaps.max3DTextureSize = mPhysicalDeviceProperties.limits.maxImageDimension3D;
mNativeCaps.max2DTextureSize = mPhysicalDeviceProperties.limits.maxImageDimension2D;
mNativeCaps.maxArrayTextureLayers = mPhysicalDeviceProperties.limits.maxImageArrayLayers;
mNativeCaps.maxLODBias = mPhysicalDeviceProperties.limits.maxSamplerLodBias;
mNativeCaps.maxCubeMapTextureSize = mPhysicalDeviceProperties.limits.maxImageDimensionCube;
mNativeCaps.maxRenderbufferSize = mNativeCaps.max2DTextureSize;
mNativeCaps.minAliasedPointSize =
std::max(1.0f, mPhysicalDeviceProperties.limits.pointSizeRange[0]);
mNativeCaps.maxAliasedPointSize = mPhysicalDeviceProperties.limits.pointSizeRange[1];
mNativeCaps.minAliasedLineWidth = 1.0f;
mNativeCaps.maxAliasedLineWidth = 1.0f;
mNativeCaps.maxDrawBuffers =
std::min<uint32_t>(mPhysicalDeviceProperties.limits.maxColorAttachments,
mPhysicalDeviceProperties.limits.maxFragmentOutputAttachments);
mNativeCaps.maxFramebufferWidth = mPhysicalDeviceProperties.limits.maxFramebufferWidth;
mNativeCaps.maxFramebufferHeight = mPhysicalDeviceProperties.limits.maxFramebufferHeight;
mNativeCaps.maxColorAttachments = mPhysicalDeviceProperties.limits.maxColorAttachments;
mNativeCaps.maxViewportWidth = mPhysicalDeviceProperties.limits.maxViewportDimensions[0];
mNativeCaps.maxViewportHeight = mPhysicalDeviceProperties.limits.maxViewportDimensions[1];
mNativeCaps.maxSampleMaskWords = mPhysicalDeviceProperties.limits.maxSampleMaskWords;
mNativeCaps.maxColorTextureSamples =
mPhysicalDeviceProperties.limits.sampledImageColorSampleCounts;
mNativeCaps.maxDepthTextureSamples =
mPhysicalDeviceProperties.limits.sampledImageDepthSampleCounts;
mNativeCaps.maxIntegerSamples =
mPhysicalDeviceProperties.limits.sampledImageIntegerSampleCounts;
mNativeCaps.maxVertexAttributes = mPhysicalDeviceProperties.limits.maxVertexInputAttributes;
mNativeCaps.maxVertexAttribBindings = mPhysicalDeviceProperties.limits.maxVertexInputBindings;
mNativeCaps.maxVertexAttribRelativeOffset =
mPhysicalDeviceProperties.limits.maxVertexInputAttributeOffset;
mNativeCaps.maxVertexAttribStride =
mPhysicalDeviceProperties.limits.maxVertexInputBindingStride;
mNativeCaps.maxElementsIndices = std::numeric_limits<GLuint>::max();
mNativeCaps.maxElementsVertices = std::numeric_limits<GLuint>::max();
// Looks like all floats are IEEE according to the docs here:
// https://www.khronos.org/registry/vulkan/specs/1.0-wsi_extensions/html/vkspec.html#spirvenv-precision-operation
mNativeCaps.vertexHighpFloat.setIEEEFloat();
mNativeCaps.vertexMediumpFloat.setIEEEFloat();
mNativeCaps.vertexLowpFloat.setIEEEFloat();
mNativeCaps.fragmentHighpFloat.setIEEEFloat();
mNativeCaps.fragmentMediumpFloat.setIEEEFloat();
mNativeCaps.fragmentLowpFloat.setIEEEFloat();
// Can't find documentation on the int precision in Vulkan.
mNativeCaps.vertexHighpInt.setTwosComplementInt(32);
mNativeCaps.vertexMediumpInt.setTwosComplementInt(32);
mNativeCaps.vertexLowpInt.setTwosComplementInt(32);
mNativeCaps.fragmentHighpInt.setTwosComplementInt(32);
mNativeCaps.fragmentMediumpInt.setTwosComplementInt(32);
mNativeCaps.fragmentLowpInt.setTwosComplementInt(32);
// TODO(lucferron): This is something we'll need to implement custom in the back-end.
// Vulkan doesn't do any waiting for you, our back-end code is going to manage sync objects,
// and we'll have to check that we've exceeded the max wait timeout. Also, this is ES 3.0 so
// we'll defer the implementation until we tackle the next version.
// mNativeCaps.maxServerWaitTimeout
GLuint maxUniformVectors = mPhysicalDeviceProperties.limits.maxUniformBufferRange /
(sizeof(GLfloat) * kComponentsPerVector);
// Clamp the maxUniformVectors to 1024u, on AMD the maxUniformBufferRange is way too high.
maxUniformVectors = std::min(1024u, maxUniformVectors);
const GLuint maxUniformComponents = maxUniformVectors * kComponentsPerVector;
// Uniforms are implemented using a uniform buffer, so the max number of uniforms we can
// support is the max buffer range divided by the size of a single uniform (4X float).
mNativeCaps.maxVertexUniformVectors = maxUniformVectors;
mNativeCaps.maxShaderUniformComponents[gl::ShaderType::Vertex] = maxUniformComponents;
mNativeCaps.maxFragmentUniformVectors = maxUniformVectors;
mNativeCaps.maxShaderUniformComponents[gl::ShaderType::Fragment] = maxUniformComponents;
// A number of uniform buffers are reserved for internal use. There's one dynamic uniform
// buffer used per stage for default uniforms, and a single uniform buffer object used for
// ANGLE internal variables. ANGLE implements UBOs as uniform buffers, so the maximum number
// of uniform blocks is maxDescriptorSetUniformBuffers - 1:
const uint32_t maxUniformBuffers =
mPhysicalDeviceProperties.limits.maxDescriptorSetUniformBuffers -
kReservedDriverUniformBindingCount;
mNativeCaps.maxShaderUniformBlocks[gl::ShaderType::Vertex] = maxUniformBuffers;
mNativeCaps.maxShaderUniformBlocks[gl::ShaderType::Fragment] = maxUniformBuffers;
mNativeCaps.maxCombinedUniformBlocks = maxUniformBuffers;
mNativeCaps.maxUniformBufferBindings = maxUniformBuffers;
mNativeCaps.maxUniformBlockSize = mPhysicalDeviceProperties.limits.maxUniformBufferRange;
mNativeCaps.uniformBufferOffsetAlignment =
static_cast<GLuint>(mPhysicalDeviceProperties.limits.minUniformBufferOffsetAlignment);
// There is no additional limit to the combined number of components. We can have up to a
// maximum number of uniform buffers, each having the maximum number of components.
const uint32_t maxCombinedUniformComponents = maxUniformBuffers * maxUniformComponents;
for (gl::ShaderType shaderType : gl::kAllGraphicsShaderTypes)
{
mNativeCaps.maxCombinedShaderUniformComponents[shaderType] = maxCombinedUniformComponents;
}
// we use the same bindings on each stage, so the limitation is the same combined or not.
mNativeCaps.maxCombinedTextureImageUnits =
mPhysicalDeviceProperties.limits.maxPerStageDescriptorSamplers;
mNativeCaps.maxShaderTextureImageUnits[gl::ShaderType::Fragment] =
mPhysicalDeviceProperties.limits.maxPerStageDescriptorSamplers;
mNativeCaps.maxShaderTextureImageUnits[gl::ShaderType::Vertex] =
mPhysicalDeviceProperties.limits.maxPerStageDescriptorSamplers;
// The max vertex output components should not include gl_Position.
// The gles2.0 section 2.10 states that "gl_Position is not a varying variable and does
// not count against this limit.", but the Vulkan spec has no such mention in its Built-in
// vars section. It is implicit that we need to actually reserve it for Vulkan in that case.
// TODO(lucferron): AMD has a weird behavior when we edge toward the maximum number of varyings
// and can often crash. Reserving an additional varying just for them bringing the total to 2.
// http://anglebug.com/2483
constexpr GLint kReservedVaryingCount = 2;
mNativeCaps.maxVaryingVectors =
(mPhysicalDeviceProperties.limits.maxVertexOutputComponents / 4) - kReservedVaryingCount;
mNativeCaps.maxVertexOutputComponents = mNativeCaps.maxVaryingVectors * 4;
mNativeCaps.subPixelBits = mPhysicalDeviceProperties.limits.subPixelPrecisionBits;
}
namespace egl_vk
{
namespace
{
EGLint ComputeMaximumPBufferPixels(const VkPhysicalDeviceProperties &physicalDeviceProperties)
{
// EGLints are signed 32-bit integers, it's fairly easy to overflow them, especially since
// Vulkan's minimum guaranteed VkImageFormatProperties::maxResourceSize is 2^31 bytes.
constexpr uint64_t kMaxValueForEGLint =
static_cast<uint64_t>(std::numeric_limits<EGLint>::max());
// TODO(geofflang): Compute the maximum size of a pbuffer by using the maxResourceSize result
// from vkGetPhysicalDeviceImageFormatProperties for both the color and depth stencil format and
// the exact image creation parameters that would be used to create the pbuffer. Because it is
// always safe to return out-of-memory errors on pbuffer allocation, it's fine to simply return
// the number of pixels in a max width by max height pbuffer for now. http://anglebug.com/2622
// Storing the result of squaring a 32-bit unsigned int in a 64-bit unsigned int is safe.
static_assert(std::is_same<decltype(physicalDeviceProperties.limits.maxImageDimension2D),
uint32_t>::value,
"physicalDeviceProperties.limits.maxImageDimension2D expected to be a uint32_t.");
const uint64_t maxDimensionsSquared =
static_cast<uint64_t>(physicalDeviceProperties.limits.maxImageDimension2D) *
static_cast<uint64_t>(physicalDeviceProperties.limits.maxImageDimension2D);
return static_cast<EGLint>(std::min(maxDimensionsSquared, kMaxValueForEGLint));
}
// Generates a basic config for a combination of color format, depth stencil format and sample
// count.
egl::Config GenerateDefaultConfig(const RendererVk *renderer,
const gl::InternalFormat &colorFormat,
const gl::InternalFormat &depthStencilFormat,
EGLint sampleCount)
{
const VkPhysicalDeviceProperties &physicalDeviceProperties =
renderer->getPhysicalDeviceProperties();
gl::Version maxSupportedESVersion = renderer->getMaxSupportedESVersion();
EGLint es2Support = (maxSupportedESVersion.major >= 2 ? EGL_OPENGL_ES2_BIT : 0);
EGLint es3Support = (maxSupportedESVersion.major >= 3 ? EGL_OPENGL_ES3_BIT : 0);
egl::Config config;
config.renderTargetFormat = colorFormat.internalFormat;
config.depthStencilFormat = depthStencilFormat.internalFormat;
config.bufferSize = colorFormat.pixelBytes * 8;
config.redSize = colorFormat.redBits;
config.greenSize = colorFormat.greenBits;
config.blueSize = colorFormat.blueBits;
config.alphaSize = colorFormat.alphaBits;
config.alphaMaskSize = 0;
config.bindToTextureRGB = colorFormat.format == GL_RGB;
config.bindToTextureRGBA = colorFormat.format == GL_RGBA || colorFormat.format == GL_BGRA_EXT;
config.colorBufferType = EGL_RGB_BUFFER;
config.configCaveat = EGL_NONE;
config.conformant = es2Support | es3Support;
config.depthSize = depthStencilFormat.depthBits;
config.stencilSize = depthStencilFormat.stencilBits;
config.level = 0;
config.matchNativePixmap = EGL_NONE;
config.maxPBufferWidth = physicalDeviceProperties.limits.maxImageDimension2D;
config.maxPBufferHeight = physicalDeviceProperties.limits.maxImageDimension2D;
config.maxPBufferPixels = ComputeMaximumPBufferPixels(physicalDeviceProperties);
config.maxSwapInterval = 1;
config.minSwapInterval = 0;
config.nativeRenderable = EGL_TRUE;
config.nativeVisualID = 0;
config.nativeVisualType = EGL_NONE;
config.renderableType = es2Support | es3Support;
config.sampleBuffers = (sampleCount > 0) ? 1 : 0;
config.samples = sampleCount;
config.surfaceType = EGL_WINDOW_BIT | EGL_PBUFFER_BIT;
// Vulkan surfaces use a different origin than OpenGL, always prefer to be flipped vertically if
// possible.
config.optimalOrientation = EGL_SURFACE_ORIENTATION_INVERT_Y_ANGLE;
config.transparentType = EGL_NONE;
config.transparentRedValue = 0;
config.transparentGreenValue = 0;
config.transparentBlueValue = 0;
config.colorComponentType =
gl_egl::GLComponentTypeToEGLColorComponentType(colorFormat.componentType);
return config;
}
} // anonymous namespace
egl::ConfigSet GenerateConfigs(const GLenum *colorFormats,
size_t colorFormatsCount,
const GLenum *depthStencilFormats,
size_t depthStencilFormatCount,
DisplayVk *display)
{
ASSERT(colorFormatsCount > 0);
ASSERT(display != nullptr);
gl::SupportedSampleSet colorSampleCounts;
gl::SupportedSampleSet depthStencilSampleCounts;
gl::SupportedSampleSet sampleCounts;
const VkPhysicalDeviceLimits &limits =
display->getRenderer()->getPhysicalDeviceProperties().limits;
const uint32_t depthStencilSampleCountsLimit =
limits.framebufferDepthSampleCounts & limits.framebufferStencilSampleCounts;
vk_gl::AddSampleCounts(limits.framebufferColorSampleCounts, &colorSampleCounts);
vk_gl::AddSampleCounts(depthStencilSampleCountsLimit, &depthStencilSampleCounts);
// Always support 0 samples
colorSampleCounts.insert(0);
depthStencilSampleCounts.insert(0);
std::set_intersection(colorSampleCounts.begin(), colorSampleCounts.end(),
depthStencilSampleCounts.begin(), depthStencilSampleCounts.end(),
std::inserter(sampleCounts, sampleCounts.begin()));
egl::ConfigSet configSet;
for (size_t colorFormatIdx = 0; colorFormatIdx < colorFormatsCount; colorFormatIdx++)
{
const gl::InternalFormat &colorFormatInfo =
gl::GetSizedInternalFormatInfo(colorFormats[colorFormatIdx]);
ASSERT(colorFormatInfo.sized);
for (size_t depthStencilFormatIdx = 0; depthStencilFormatIdx < depthStencilFormatCount;
depthStencilFormatIdx++)
{
const gl::InternalFormat &depthStencilFormatInfo =
gl::GetSizedInternalFormatInfo(depthStencilFormats[depthStencilFormatIdx]);
ASSERT(depthStencilFormats[depthStencilFormatIdx] == GL_NONE ||
depthStencilFormatInfo.sized);
const gl::SupportedSampleSet *configSampleCounts = &sampleCounts;
// If there is no depth/stencil buffer, use the color samples set.
if (depthStencilFormats[depthStencilFormatIdx] == GL_NONE)
{
configSampleCounts = &colorSampleCounts;
}
// If there is no color buffer, use the depth/stencil samples set.
else if (colorFormats[colorFormatIdx] == GL_NONE)
{
configSampleCounts = &depthStencilSampleCounts;
}
for (EGLint sampleCount : *configSampleCounts)
{
egl::Config config = GenerateDefaultConfig(display->getRenderer(), colorFormatInfo,
depthStencilFormatInfo, sampleCount);
if (display->checkConfigSupport(&config))
{
configSet.add(config);
}
}
}
}
return configSet;
}
} // namespace egl_vk
} // namespace rx