swung0x48/MobileGL
1
0
Fork 0
mirror of https://github.com/MobileGL-Dev/MobileGL synced 2026-09-11 13:48:30 +09:00
Code Issues Packages Projects Releases Wiki Activity

[Feat] (MG_Backend/DirectVulkan): native GPU indirect draws + draw parameters

Browse Source 
- Enable multiDrawIndirect and shaderDrawParameters device features when
  supported (the latter via VkPhysicalDeviceShaderDrawParametersFeatures on
  Vulkan 1.1+), so DrawIndex/BaseInstance SPIR-V builtins are valid and
  vkCmdDrawIndexedIndirect(Count) may draw more than one command.
- Plain glMultiDrawElementsIndirect no longer requires a GL_PARAMETER_BUFFER
  (it previously drew nothing for the standard Flywheel call); it now issues
  a native vkCmdDrawIndexedIndirect, with a per-command loop fallback when
  the multiDrawIndirect feature is unavailable.
- glDrawElementsIndirect / glDrawArraysIndirect / glMultiDrawArraysIndirect
  read the live GPU buffer via native indirect draws instead of the CPU
  shadow (which cannot see compute-written commands); the CPU path remains
  only for client-memory commands.
- Advertise the same five extensions as DirectGLES for Flywheel's
  capability probe.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
This commit is contained in:
swung0x48
2026-07-08 13:33:05 +00:00
co-authored by Claude Fable 5
parent 5ea49f3c50
commit 638999213e
4 changed files with 227 additions and 15 deletions
Download Patch File Download Diff File Expand all files Collapse all files
MobileGL/MG_Backend/DirectVulkan/BackendObject_DirectVulkan.cpp
+3 -1
View File
@@ -359,7 +359,9 @@ namespace MobileGL::MG_Backend::DirectVulkan {
E_GL_ARB_multi_draw_indirect, E_GL_ARB_indirect_parameters,
E_GL_EXT_framebuffer_object, E_GL_ARB_depth_texture, E_GL_ARB_buffer_storage,
E_GL_ARB_texture_storage, E_GL_ARB_direct_state_access,
E_GL_ARB_shader_draw_parameters, E_GL_ARB_gpu_shader_int64, E_GL_KHR_debug},
E_GL_ARB_shader_draw_parameters, E_GL_ARB_gpu_shader_int64, E_GL_KHR_debug,
E_GL_ARB_gpu_shader5, E_GL_ARB_multi_bind, E_GL_ARB_shading_language_420pack,
E_GL_ARB_vertex_attrib_binding, E_GL_ARB_shader_image_size},
.IsCompatibilityProfile = false
},
.StaticBackendCapability = {.AllowVSOnlyPrograms = false}} {}
MobileGL/MG_Backend/DirectVulkan/DirectVulkan.cpp
+29 -1
View File
@@ -409,7 +409,7 @@ namespace MobileGL::MG_Backend::DirectVulkan {
void MultiDrawElementsIndirect(GLenum mode, GLenum type, const void* indirect, GLsizei drawcount, GLsizei stride) {
MOBILEGL_ASSERT(pVulkanRenderer, "DirectVulkan::MultiDrawElementsIndirect called with null VulkanRenderer");
MOBILEGL_ASSERT(MG_State::pGLContext, "DirectVulkan::MultiDrawElementsIndirect called with null GL context");
pVulkanRenderer->MultiDrawElementsIndirectCount(mode, type, indirect, 0, drawcount, stride);
pVulkanRenderer->MultiDrawElementsIndirect(mode, type, indirect, drawcount, stride);
}
void MultiDrawArraysIndirect(GLenum mode, const void* indirect, GLsizei drawcount, GLsizei stride) {
MOBILEGL_ASSERT(pVulkanRenderer, "DirectVulkan::MultiDrawArraysIndirect called with null VulkanRenderer");
@@ -418,6 +418,15 @@ namespace MobileGL::MG_Backend::DirectVulkan {
if (drawcount <= 0) {
return;
}
// With a bound GL_DRAW_INDIRECT_BUFFER the command parameters may be GPU-written
// (e.g. by a compute shader), so consume them natively on the GPU.
auto drawBuffer = MG_State::pGLContext->GetBufferBindingSlot(BufferTarget::DrawIndirect).GetBoundObject();
if (drawBuffer) {
pVulkanRenderer->MultiDrawArraysIndirect(mode, indirect, drawcount, stride);
return;
}
if (stride == 0) {
stride = sizeof(DrawArraysIndirectCommand);
}
@@ -427,6 +436,7 @@ namespace MobileGL::MG_Backend::DirectVulkan {
return;
}
// No indirect buffer bound: the pointer refers to client memory.
const auto* commandBytes = ResolveIndirectCommandBytes(
indirect,
static_cast<SizeT>(stride) * static_cast<SizeT>(drawcount - 1) + sizeof(DrawArraysIndirectCommand),
@@ -541,6 +551,15 @@ namespace MobileGL::MG_Backend::DirectVulkan {
return;
}
// With a bound GL_DRAW_INDIRECT_BUFFER the command parameters may be GPU-written
// (e.g. by a compute shader), so consume them natively on the GPU.
auto drawBuffer = MG_State::pGLContext->GetBufferBindingSlot(BufferTarget::DrawIndirect).GetBoundObject();
if (drawBuffer) {
pVulkanRenderer->MultiDrawElementsIndirect(mode, type, indirect, 1, 0);
return;
}
// No indirect buffer bound: the pointer refers to client memory.
const auto* commandBytes =
ResolveIndirectCommandBytes(indirect, sizeof(DrawElementsIndirectCommand), "DrawElementsIndirect");
if (!commandBytes) {
@@ -585,6 +604,15 @@ namespace MobileGL::MG_Backend::DirectVulkan {
MOBILEGL_ASSERT(pVulkanRenderer, "DirectVulkan::DrawArraysIndirect called with null VulkanRenderer");
MOBILEGL_ASSERT(MG_State::pGLContext, "DirectVulkan::DrawArraysIndirect called with null GL context");
// With a bound GL_DRAW_INDIRECT_BUFFER the command parameters may be GPU-written
// (e.g. by a compute shader), so consume them natively on the GPU.
auto drawBuffer = MG_State::pGLContext->GetBufferBindingSlot(BufferTarget::DrawIndirect).GetBoundObject();
if (drawBuffer) {
pVulkanRenderer->MultiDrawArraysIndirect(mode, indirect, 1, 0);
return;
}
// No indirect buffer bound: the pointer refers to client memory.
const auto* commandBytes =
ResolveIndirectCommandBytes(indirect, sizeof(DrawArraysIndirectCommand), "DrawArraysIndirect");
if (!commandBytes) {
MobileGL/MG_Backend/DirectVulkan/Renderer/VulkanRenderer.cpp
+189 -13
View File
@@ -5225,7 +5225,10 @@ void main() {
}
MOBILEGL_ASSERT(frame.isCommandRecording, "%s: frame recording was not started", __func__);
if (m_drawIndirectCountExtensionEnabled && s_vkCmdDrawIndexedIndirectCount) {
// vkCmdDrawIndexedIndirectCount with maxDrawCount > 1 additionally requires the
// multiDrawIndirect device feature; fall back to the CPU readback loop otherwise.
if (m_drawIndirectCountExtensionEnabled && s_vkCmdDrawIndexedIndirectCount &&
(m_multiDrawIndirectFeatureEnabled || maxdrawcount == 1)) {
MGLOG_D("DirectVulkan: glMultiDrawElementsIndirectCountARB(max=%d stride=%d)", maxdrawcount, stride);
s_vkCmdDrawIndexedIndirectCount(frame.commandBuffer,
drawSlice.buffer,
@@ -5254,6 +5257,149 @@ void main() {
}
}
void VulkanRenderer::MultiDrawElementsIndirect(GLenum mode, GLenum type, const void* indirect,
GLsizei drawcount, GLsizei stride) {
auto& frame = m_frameContext.GetCurrent();
if (drawcount <= 0) {
return;
}
if (stride == 0) {
stride = sizeof(DrawIndexedCmdParam);
}
if (stride < static_cast<GLsizei>(sizeof(DrawIndexedCmdParam))) {
MGLOG_E("MultiDrawElementsIndirect skipped: stride %d is smaller than command size %zu",
stride, sizeof(DrawIndexedCmdParam));
return;
}
const SizeT indexSize = MG_Util::GetGLTypeSize(type);
if (indexSize == 0) {
MGLOG_E("MultiDrawElementsIndirect skipped: unsupported index type 0x%x", type);
return;
}
const auto& vao = *MG_State::pGLContext->GetBoundVertexArray();
const auto* indexBuffer = vao.GetIndexBufferBindingSlot().GetBoundObject().get();
if (!indexBuffer) {
MGLOG_E("MultiDrawElementsIndirect skipped: no element array buffer is bound");
return;
}
const SizeT commandOffset = reinterpret_cast<SizeT>(indirect);
const SizeT commandBytes = commandOffset +
static_cast<SizeT>(stride) * static_cast<SizeT>(drawcount - 1) + sizeof(DrawIndexedCmdParam);
auto drawBuffer = MG_State::pGLContext->GetBufferBindingSlot(BufferTarget::DrawIndirect).GetBoundObject();
if (!drawBuffer || commandBytes > drawBuffer->GetSize()) {
MGLOG_E("MultiDrawElementsIndirect skipped: invalid GL_DRAW_INDIRECT_BUFFER binding or range");
return;
}
// The command parameters live on the GPU; the CPU-visible range that any single
// command may address is the whole element array buffer.
DrawCmdParam vertexRange{};
vertexRange.vertexCount = static_cast<Uint32>(indexBuffer->GetSize() / indexSize);
vertexRange.instanceCount = 1;
IndexBufferView indexBufferView{};
indexBufferView.indexType = type;
indexBufferView.indexByteOffset = 0;
indexBufferView.indexByteSize = indexBuffer->GetSize();
if (!SetupDraw(frame, mode, DrawSetupAspect::IndexBuffer | DrawSetupAspect::IndirectDrawBuffer,
vertexRange, &indexBufferView)) {
return;
}
BufferSlice drawSlice{};
if (!m_bufferManager.AcquireResidentSlice(BufferKind::Indirect, drawBuffer, drawSlice)) {
MGLOG_E("MultiDrawElementsIndirect skipped: failed to sync draw indirect buffer");
return;
}
MOBILEGL_ASSERT(frame.isCommandRecording, "%s: frame recording was not started", __func__);
MGLOG_D("DirectVulkan: glMultiDrawElementsIndirect(drawcount=%d stride=%d)", drawcount, stride);
if (drawcount == 1 || (m_multiDrawIndirectFeatureEnabled && stride % 4 == 0)) {
vkCmdDrawIndexedIndirect(frame.commandBuffer,
drawSlice.buffer,
drawSlice.offset + static_cast<VkDeviceSize>(commandOffset),
static_cast<Uint32>(drawcount),
static_cast<Uint32>(stride));
return;
}
// multiDrawIndirect device feature unavailable: one indirect draw per command is
// valid without it and still consumes the GPU-written parameters.
for (GLsizei idraw = 0; idraw < drawcount; ++idraw) {
vkCmdDrawIndexedIndirect(frame.commandBuffer,
drawSlice.buffer,
drawSlice.offset + static_cast<VkDeviceSize>(commandOffset) +
static_cast<VkDeviceSize>(idraw) * static_cast<VkDeviceSize>(stride),
1, 0);
}
}
void VulkanRenderer::MultiDrawArraysIndirect(GLenum mode, const void* indirect, GLsizei drawcount,
GLsizei stride) {
auto& frame = m_frameContext.GetCurrent();
if (drawcount <= 0) {
return;
}
if (stride == 0) {
stride = sizeof(DrawCmdParam);
}
if (stride < static_cast<GLsizei>(sizeof(DrawCmdParam))) {
MGLOG_E("MultiDrawArraysIndirect skipped: stride %d is smaller than command size %zu",
stride, sizeof(DrawCmdParam));
return;
}
const SizeT commandOffset = reinterpret_cast<SizeT>(indirect);
const SizeT commandBytes = commandOffset +
static_cast<SizeT>(stride) * static_cast<SizeT>(drawcount - 1) + sizeof(DrawCmdParam);
auto drawBuffer = MG_State::pGLContext->GetBufferBindingSlot(BufferTarget::DrawIndirect).GetBoundObject();
if (!drawBuffer || commandBytes > drawBuffer->GetSize()) {
MGLOG_E("MultiDrawArraysIndirect skipped: invalid GL_DRAW_INDIRECT_BUFFER binding or range");
return;
}
// The command parameters live on the GPU, so the vertex range is unknown here;
// resident vertex buffers are uploaded in full regardless.
DrawCmdParam vertexRange{};
vertexRange.vertexCount = 0;
vertexRange.instanceCount = 1;
if (!SetupDraw(frame, mode, DrawSetupAspect::IndirectDrawBuffer, vertexRange)) {
return;
}
BufferSlice drawSlice{};
if (!m_bufferManager.AcquireResidentSlice(BufferKind::Indirect, drawBuffer, drawSlice)) {
MGLOG_E("MultiDrawArraysIndirect skipped: failed to sync draw indirect buffer");
return;
}
MOBILEGL_ASSERT(frame.isCommandRecording, "%s: frame recording was not started", __func__);
MGLOG_D("DirectVulkan: glMultiDrawArraysIndirect(drawcount=%d stride=%d)", drawcount, stride);
if (drawcount == 1 || (m_multiDrawIndirectFeatureEnabled && stride % 4 == 0)) {
vkCmdDrawIndirect(frame.commandBuffer,
drawSlice.buffer,
drawSlice.offset + static_cast<VkDeviceSize>(commandOffset),
static_cast<Uint32>(drawcount),
static_cast<Uint32>(stride));
return;
}
for (GLsizei idraw = 0; idraw < drawcount; ++idraw) {
vkCmdDrawIndirect(frame.commandBuffer,
drawSlice.buffer,
drawSlice.offset + static_cast<VkDeviceSize>(commandOffset) +
static_cast<VkDeviceSize>(idraw) * static_cast<VkDeviceSize>(stride),
1, 0);
}
}
VkCommandBuffer VulkanRenderer::AcquireBufferCopyCommandBuffer() {
if (m_device == VK_NULL_HANDLE || m_frameContext.GetFrameCount() == 0) {
return VK_NULL_HANDLE;
@@ -5755,6 +5901,8 @@ void main() {
m_logicOpFeatureEnabled = deviceFeatures.logicOp == VK_TRUE;
deviceFeatures.shaderInt64 = supportedDeviceFeatures.shaderInt64;
deviceFeatures.drawIndirectFirstInstance = supportedDeviceFeatures.drawIndirectFirstInstance;
deviceFeatures.multiDrawIndirect = supportedDeviceFeatures.multiDrawIndirect;
m_multiDrawIndirectFeatureEnabled = deviceFeatures.multiDrawIndirect == VK_TRUE;
m_logicOpFeatureEnabled = deviceFeatures.logicOp == VK_TRUE;
VkDeviceCreateInfo deviceCreateInfo{};
@@ -5787,18 +5935,19 @@ void main() {
indexTypeUint8ExtensionName = VK_EXT_INDEX_TYPE_UINT8_EXTENSION_NAME;
}
auto getPhysicalDeviceFeatures2 = reinterpret_cast<PFN_vkGetPhysicalDeviceFeatures2>(
vkGetInstanceProcAddr(m_instance, "vkGetPhysicalDeviceFeatures2"));
if (getPhysicalDeviceFeatures2 == nullptr) {
getPhysicalDeviceFeatures2 = reinterpret_cast<PFN_vkGetPhysicalDeviceFeatures2>(
vkGetInstanceProcAddr(m_instance, "vkGetPhysicalDeviceFeatures2KHR"));
}
VkPhysicalDeviceIndexTypeUint8Features indexTypeUint8Features{};
indexTypeUint8Features.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_INDEX_TYPE_UINT8_FEATURES;
if (indexTypeUint8ExtensionName != nullptr) {
VkPhysicalDeviceFeatures2 featureQuery{};
featureQuery.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_FEATURES_2;
featureQuery.pNext = &indexTypeUint8Features;
auto getPhysicalDeviceFeatures2 = reinterpret_cast<PFN_vkGetPhysicalDeviceFeatures2>(
vkGetInstanceProcAddr(m_instance, "vkGetPhysicalDeviceFeatures2"));
if (getPhysicalDeviceFeatures2 == nullptr) {
getPhysicalDeviceFeatures2 = reinterpret_cast<PFN_vkGetPhysicalDeviceFeatures2>(
vkGetInstanceProcAddr(m_instance, "vkGetPhysicalDeviceFeatures2KHR"));
}
MOBILEGL_ASSERT(getPhysicalDeviceFeatures2 != nullptr,
"CreateLogicalDeviceAndQueues: vkGetPhysicalDeviceFeatures2 is unavailable");
getPhysicalDeviceFeatures2(m_physicalDevice.handle, &featureQuery);
@@ -5818,10 +5967,32 @@ void main() {
MGLOG_W("VK_KHR_index_type_uint8 / VK_EXT_index_type_uint8 not supported; uint8 index buffers will stay disabled");
}
m_shaderDrawParametersFeatureEnabled = false;
VkPhysicalDeviceShaderDrawParametersFeatures shaderDrawParametersFeatures{};
shaderDrawParametersFeatures.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_SHADER_DRAW_PARAMETERS_FEATURES;
if (m_physicalDevice.properties.apiVersion >= VK_API_VERSION_1_1 && getPhysicalDeviceFeatures2 != nullptr) {
VkPhysicalDeviceFeatures2 featureQuery{};
featureQuery.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_FEATURES_2;
featureQuery.pNext = &shaderDrawParametersFeatures;
getPhysicalDeviceFeatures2(m_physicalDevice.handle, &featureQuery);
if (shaderDrawParametersFeatures.shaderDrawParameters == VK_TRUE) {
shaderDrawParametersFeatures.pNext = const_cast<void*>(deviceCreateInfo.pNext);
deviceCreateInfo.pNext = &shaderDrawParametersFeatures;
m_shaderDrawParametersFeatureEnabled = true;
}
} else if (m_shaderDrawParametersExtensionEnabled) {
// Vulkan 1.0 device: enabling VK_KHR_shader_draw_parameters alone exposes the SPIR-V
// DrawParameters capability; the shaderDrawParameters feature struct only exists from 1.1.
m_shaderDrawParametersFeatureEnabled = true;
}
if (!m_shaderDrawParametersFeatureEnabled) {
MGLOG_W("shaderDrawParameters is unavailable; shaders using gl_DrawID/gl_BaseInstance will not work");
}
deviceCreateInfo.enabledExtensionCount = static_cast<Uint32>(enabledDeviceExtensions.size());
deviceCreateInfo.ppEnabledExtensionNames = enabledDeviceExtensions.data();
MGLOG_I("Device feature support: geometryShader=%s independentBlend=%s logicOp=%s shaderClipDistance=%s "
"shaderCullDistance=%s wideLines=%s shaderInt64=%s drawIndirectFirstInstance=%s",
"shaderCullDistance=%s wideLines=%s shaderInt64=%s drawIndirectFirstInstance=%s multiDrawIndirect=%s",
supportedDeviceFeatures.geometryShader ? "true" : "false",
supportedDeviceFeatures.independentBlend ? "true" : "false",
supportedDeviceFeatures.logicOp ? "true" : "false",
@@ -5829,9 +6000,11 @@ void main() {
supportedDeviceFeatures.shaderCullDistance ? "true" : "false",
supportedDeviceFeatures.wideLines ? "true" : "false",
supportedDeviceFeatures.shaderInt64 ? "true" : "false",
supportedDeviceFeatures.drawIndirectFirstInstance ? "true" : "false");
supportedDeviceFeatures.drawIndirectFirstInstance ? "true" : "false",
supportedDeviceFeatures.multiDrawIndirect ? "true" : "false");
MGLOG_I("Device feature enabled: geometryShader=%s independentBlend=%s logicOp=%s shaderClipDistance=%s "
"shaderCullDistance=%s wideLines=%s shaderInt64=%s drawIndirectFirstInstance=%s",
"shaderCullDistance=%s wideLines=%s shaderInt64=%s drawIndirectFirstInstance=%s multiDrawIndirect=%s "
"shaderDrawParameters=%s",
deviceFeatures.geometryShader ? "true" : "false",
deviceFeatures.independentBlend ? "true" : "false",
deviceFeatures.logicOp ? "true" : "false",
@@ -5839,7 +6012,9 @@ void main() {
deviceFeatures.shaderCullDistance ? "true" : "false",
deviceFeatures.wideLines ? "true" : "false",
deviceFeatures.shaderInt64 ? "true" : "false",
deviceFeatures.drawIndirectFirstInstance ? "true" : "false");
deviceFeatures.drawIndirectFirstInstance ? "true" : "false",
deviceFeatures.multiDrawIndirect ? "true" : "false",
m_shaderDrawParametersFeatureEnabled ? "true" : "false");
VK_VERIFY(vkCreateDevice(m_physicalDevice.handle, &deviceCreateInfo, nullptr, &m_device), "vkCreateDevice");
s_vkCmdDrawIndexedIndirectCount = reinterpret_cast<PFNDrawIndexedIndirectCountFunc>(
@@ -6067,8 +6242,9 @@ void main() {
Vector<const char*>& inOutEnabledExtensions) {
m_drawIndirectCountExtensionEnabled = EnableOptionalDeviceExtension(availableExtensions, inOutEnabledExtensions,
VK_KHR_DRAW_INDIRECT_COUNT_EXTENSION_NAME);
EnableOptionalDeviceExtension(availableExtensions, inOutEnabledExtensions,
VK_KHR_SHADER_DRAW_PARAMETERS_EXTENSION_NAME);
m_shaderDrawParametersExtensionEnabled =
EnableOptionalDeviceExtension(availableExtensions, inOutEnabledExtensions,
VK_KHR_SHADER_DRAW_PARAMETERS_EXTENSION_NAME);
#ifdef VK_KHR_PORTABILITY_SUBSET_EXTENSION_NAME
EnableOptionalDeviceExtension(availableExtensions, inOutEnabledExtensions,
VK_KHR_PORTABILITY_SUBSET_EXTENSION_NAME);
MobileGL/MG_Backend/DirectVulkan/Renderer/VulkanRenderer.h
+6
View File
@@ -156,6 +156,9 @@ namespace MobileGL::MG_Backend::DirectVulkan {
void DrawArrays(const DrawCmd& payload);
void DrawElements(const DrawIndexedCmd& payload);
void MultiDrawElements(const MultiDrawIndexedCmd& payloads);
void MultiDrawElementsIndirect(GLenum mode, GLenum type, const void* indirect, GLsizei drawcount,
GLsizei stride);
void MultiDrawArraysIndirect(GLenum mode, const void* indirect, GLsizei drawcount, GLsizei stride);
void MultiDrawElementsIndirectCount(GLenum mode, GLenum type, const void* indirect, GLintptr drawcount,
GLsizei maxdrawcount, GLsizei stride);
void Present();
@@ -229,6 +232,9 @@ namespace MobileGL::MG_Backend::DirectVulkan {
Bool m_drawIndirectCountExtensionEnabled = false;
Bool m_indexTypeUint8ExtensionEnabled = false;
Bool m_logicOpFeatureEnabled = false;
Bool m_multiDrawIndirectFeatureEnabled = false;
Bool m_shaderDrawParametersExtensionEnabled = false;
Bool m_shaderDrawParametersFeatureEnabled = false;
using PFNDrawIndexedIndirectCountFunc = void(VKAPI_PTR*)(VkCommandBuffer commandBuffer, VkBuffer buffer,
VkDeviceSize offset, VkBuffer countBuffer,
VkDeviceSize countBufferOffset, Uint32 maxDrawCount,