aboutsummaryrefslogtreecommitdiff
diff options
context:
space:
mode:
Diffstat
-rw-r--r--layers/core_validation.cpp322
-rw-r--r--layers/core_validation.h4
2 files changed, 167 insertions, 159 deletions
diff --git a/layers/core_validation.cpp b/layers/core_validation.cpp
index 2b1e975e..dd9b8fe9 100644
--- a/layers/core_validation.cpp
+++ b/layers/core_validation.cpp
@@ -109,7 +109,6 @@ struct layer_data {
// MTMERGESOURCE - stuff pulled directly from MT
uint64_t currentFenceId;
// Maps for tracking key structs related to mem_tracker state
- unordered_map<VkDescriptorSet, MT_DESCRIPTOR_SET_INFO> descriptorSetMap;
// Images and Buffers are 2 objects that can have memory bound to them so they get special treatment
unordered_map<uint64_t, MT_OBJ_BINDING_INFO> imageBindingMap;
unordered_map<uint64_t, MT_OBJ_BINDING_INFO> bufferBindingMap;
@@ -606,7 +605,6 @@ static void clear_cmd_buf_and_mem_references(layer_data *dev_data, const VkComma
}
pCBNode->pMemObjList.clear();
}
- pCBNode->activeDescriptorSets.clear();
pCBNode->validate_functions.clear();
}
}
@@ -2827,12 +2825,16 @@ static uint32_t getBindingEndIndex(const LAYOUT_NODE *pLayout, const uint32_t bi
return offsetIndex - 1;
}
-// For the given command buffer, verify that for each set in activeSetBindingsPairs
-// that any dynamic descriptor in that set has a valid dynamic offset bound.
-// To be valid, the dynamic offset combined with the offset and range from its
-// descriptor update must not overflow the size of its buffer being updated
-static VkBool32 validate_dynamic_offsets(layer_data *my_data, const GLOBAL_CB_NODE *pCB,
- const vector<std::pair<SET_NODE *, unordered_set<uint32_t>>> &activeSetBindingsPairs) {
+// For the given command buffer, verify and update the state for activeSetBindingsPairs
+// This includes:
+// 1. Verifying that any dynamic descriptor in that set has a valid dynamic offset bound.
+// To be valid, the dynamic offset combined with the offset and range from its
+// descriptor update must not overflow the size of its buffer being updated
+// 2. Grow updateImages for given pCB to include any bound STORAGE_IMAGE descriptor images
+// 3. Grow updateBuffers for pCB to include buffers from STORAGE*_BUFFER descriptor buffers
+static VkBool32 validate_and_update_drawtime_descriptor_state(
+ layer_data *dev_data, GLOBAL_CB_NODE *pCB,
+ const vector<std::pair<SET_NODE *, unordered_set<uint32_t>>> &activeSetBindingsPairs) {
VkBool32 result = VK_FALSE;
VkWriteDescriptorSet *pWDS = NULL;
@@ -2852,11 +2854,11 @@ static VkBool32 validate_dynamic_offsets(layer_data *my_data, const GLOBAL_CB_NO
if ((pWDS->descriptorType == VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER_DYNAMIC) ||
(pWDS->descriptorType == VK_DESCRIPTOR_TYPE_STORAGE_BUFFER_DYNAMIC)) {
for (uint32_t j = 0; j < pWDS->descriptorCount; ++j) {
- bufferSize = my_data->bufferMap[pWDS->pBufferInfo[j].buffer].create_info->size;
+ bufferSize = dev_data->bufferMap[pWDS->pBufferInfo[j].buffer].create_info->size;
uint32_t dynOffset = pCB->lastBound[VK_PIPELINE_BIND_POINT_GRAPHICS].dynamicOffsets[dynOffsetIndex];
if (pWDS->pBufferInfo[j].range == VK_WHOLE_SIZE) {
if ((dynOffset + pWDS->pBufferInfo[j].offset) > bufferSize) {
- result |= log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT,
+ result |= log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT,
VK_DEBUG_REPORT_OBJECT_TYPE_DESCRIPTOR_SET_EXT,
reinterpret_cast<const uint64_t &>(set_node->set), __LINE__,
DRAWSTATE_DYNAMIC_OFFSET_OVERFLOW, "DS",
@@ -2870,7 +2872,7 @@ static VkBool32 validate_dynamic_offsets(layer_data *my_data, const GLOBAL_CB_NO
}
} else if ((dynOffset + pWDS->pBufferInfo[j].offset + pWDS->pBufferInfo[j].range) > bufferSize) {
result |= log_msg(
- my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT,
+ dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT,
VK_DEBUG_REPORT_OBJECT_TYPE_DESCRIPTOR_SET_EXT,
reinterpret_cast<const uint64_t &>(set_node->set), __LINE__, DRAWSTATE_DYNAMIC_OFFSET_OVERFLOW,
"DS",
@@ -2883,7 +2885,7 @@ static VkBool32 validate_dynamic_offsets(layer_data *my_data, const GLOBAL_CB_NO
reinterpret_cast<const uint64_t &>(pWDS->pBufferInfo[j].buffer), bufferSize);
} else if ((dynOffset + pWDS->pBufferInfo[j].offset + pWDS->pBufferInfo[j].range) > bufferSize) {
result |= log_msg(
- my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT,
+ dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT,
VK_DEBUG_REPORT_OBJECT_TYPE_DESCRIPTOR_SET_EXT,
reinterpret_cast<const uint64_t &>(set_node->set), __LINE__, DRAWSTATE_DYNAMIC_OFFSET_OVERFLOW,
"DS",
@@ -2896,10 +2898,24 @@ static VkBool32 validate_dynamic_offsets(layer_data *my_data, const GLOBAL_CB_NO
reinterpret_cast<const uint64_t &>(pWDS->pBufferInfo[j].buffer), bufferSize);
}
dynOffsetIndex++;
- i += j; // Advance i to end of this set of descriptors (++i at end of for loop will move 1 index past
- // last of these descriptors)
+ }
+ } else if (pWDS->descriptorType == VK_DESCRIPTOR_TYPE_STORAGE_IMAGE) {
+ for (uint32_t j = 0; j < pWDS->descriptorCount; ++j) {
+ pCB->updateImages.insert(pWDS->pImageInfo[j].imageView);
+ }
+ } else if (pWDS->descriptorType == VK_DESCRIPTOR_TYPE_STORAGE_TEXEL_BUFFER) {
+ for (uint32_t j = 0; j < pWDS->descriptorCount; ++j) {
+ assert(dev_data->bufferViewMap.find(pWDS->pTexelBufferView[j]) != dev_data->bufferViewMap.end());
+ pCB->updateBuffers.insert(dev_data->bufferViewMap[pWDS->pTexelBufferView[j]].buffer);
+ }
+ } else if (pWDS->descriptorType == VK_DESCRIPTOR_TYPE_STORAGE_BUFFER ||
+ pWDS->descriptorType == VK_DESCRIPTOR_TYPE_STORAGE_BUFFER_DYNAMIC) {
+ for (uint32_t j = 0; j < pWDS->descriptorCount; ++j) {
+ pCB->updateBuffers.insert(pWDS->pBufferInfo[j].buffer);
}
}
+ i += pWDS->descriptorCount; // Advance i to end of this set of descriptors (++i at end of for loop will move 1
+ // index past last of these descriptors)
break;
default: // Currently only shadowing Write update nodes so shouldn't get here
assert(0);
@@ -2910,18 +2926,57 @@ static VkBool32 validate_dynamic_offsets(layer_data *my_data, const GLOBAL_CB_NO
}
return result;
}
+// TODO : This is a temp function that naively updates bound storage images and buffers based on which descriptor sets are bound.
+// When validate_and_update_draw_state() handles computer shaders so that active_slots is correct for compute pipelines, this
+// function can be killed and validate_and_update_draw_state() used instead
+static void update_shader_storage_images_and_buffers(layer_data *dev_data, GLOBAL_CB_NODE *pCB) {
+ VkWriteDescriptorSet *pWDS = nullptr;
+ SET_NODE *pSet = nullptr;
+ // For the bound descriptor sets, pull off any storage images and buffers
+ // This may be more than are actually updated depending on which are active, but for now this is a stop-gap for compute
+ // pipelines
+ for (auto set : pCB->lastBound[VK_PIPELINE_BIND_POINT_COMPUTE].uniqueBoundSets) {
+ // Get the set node
+ pSet = getSetNode(dev_data, set);
+ // For each update in the set
+ for (auto pUpdate : pSet->pDescriptorUpdates) {
+ // If it's a write update to STORAGE type capture image/buffer being updated
+ if (pUpdate && (pUpdate->sType == VK_STRUCTURE_TYPE_WRITE_DESCRIPTOR_SET)) {
+ pWDS = reinterpret_cast<VkWriteDescriptorSet *>(pUpdate);
+ if (pWDS->descriptorType == VK_DESCRIPTOR_TYPE_STORAGE_IMAGE) {
+ for (uint32_t j = 0; j < pWDS->descriptorCount; ++j) {
+ pCB->updateImages.insert(pWDS->pImageInfo[j].imageView);
+ }
+ } else if (pWDS->descriptorType == VK_DESCRIPTOR_TYPE_STORAGE_TEXEL_BUFFER) {
+ for (uint32_t j = 0; j < pWDS->descriptorCount; ++j) {
+ pCB->updateBuffers.insert(dev_data->bufferViewMap[pWDS->pTexelBufferView[j]].buffer);
+ }
+ } else if (pWDS->descriptorType == VK_DESCRIPTOR_TYPE_STORAGE_BUFFER ||
+ pWDS->descriptorType == VK_DESCRIPTOR_TYPE_STORAGE_BUFFER_DYNAMIC) {
+ for (uint32_t j = 0; j < pWDS->descriptorCount; ++j) {
+ pCB->updateBuffers.insert(pWDS->pBufferInfo[j].buffer);
+ }
+ }
+ }
+ }
+ }
+}
// Validate overall state at the time of a draw call
-static VkBool32 validate_draw_state(layer_data *my_data, GLOBAL_CB_NODE *pCB, VkBool32 indexedDraw) {
- PIPELINE_NODE *pPipe = getPipeline(my_data, pCB->lastBound[VK_PIPELINE_BIND_POINT_GRAPHICS].pipeline);
+static VkBool32 validate_and_update_draw_state(layer_data *my_data, GLOBAL_CB_NODE *pCB, const VkBool32 indexedDraw,
+ const VkPipelineBindPoint bindPoint) {
+ VkBool32 result = VK_FALSE;
+ auto const &state = pCB->lastBound[bindPoint];
+ PIPELINE_NODE *pPipe = getPipeline(my_data, state.pipeline);
// First check flag states
- VkBool32 result = validate_draw_state_flags(my_data, pCB, pPipe, indexedDraw);
+ if (VK_PIPELINE_BIND_POINT_GRAPHICS == bindPoint)
+ result = validate_draw_state_flags(my_data, pCB, pPipe, indexedDraw);
+
// Now complete other state checks
// TODO : Currently only performing next check if *something* was bound (non-zero last bound)
// There is probably a better way to gate when this check happens, and to know if something *should* have been bound
// We should have that check separately and then gate this check based on that check
if (pPipe) {
- auto const &state = pCB->lastBound[VK_PIPELINE_BIND_POINT_GRAPHICS];
if (state.pipelineLayout) {
string errorString;
// Need a vector (vs. std::set) of active Sets for dynamicOffset validation in case same set bound w/ different offsets
@@ -2944,7 +2999,7 @@ static VkBool32 validate_draw_state(layer_data *my_data, GLOBAL_CB_NODE *pCB, Vk
"VkDescriptorSet (%#" PRIxLEAST64
") bound as set #%u is not compatible with overlapping VkPipelineLayout %#" PRIxLEAST64 " due to: %s",
(uint64_t)setHandle, setIndex, (uint64_t)pPipe->graphicsPipelineCI.layout, errorString.c_str());
- } else { // Valid set is bound and layout compatible, validate that it's updated and verify any dynamic offsets
+ } else { // Valid set is bound and layout compatible, validate that it's updated
// Pull the set node
SET_NODE *pSet = my_data->setMap[state.boundDescriptorSets[setIndex]];
// Save vector of all active sets to verify dynamicOffsets below
@@ -2961,54 +3016,55 @@ static VkBool32 validate_draw_state(layer_data *my_data, GLOBAL_CB_NODE *pCB, Vk
}
}
}
- // For each dynamic descriptor, make sure dynamic offset doesn't overstep buffer
- if (!state.dynamicOffsets.empty())
- result |= validate_dynamic_offsets(my_data, pCB, activeSetBindingsPairs);
- }
- // Verify Vtx binding
- if (pPipe->vertexBindingDescriptions.size() > 0) {
- for (size_t i = 0; i < pPipe->vertexBindingDescriptions.size(); i++) {
- if ((pCB->currentDrawData.buffers.size() < (i + 1)) || (pCB->currentDrawData.buffers[i] == VK_NULL_HANDLE)) {
- result |= log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, (VkDebugReportObjectTypeEXT)0, 0,
- __LINE__, DRAWSTATE_VTX_INDEX_OUT_OF_BOUNDS, "DS",
- "The Pipeline State Object (%#" PRIxLEAST64
- ") expects that this Command Buffer's vertex binding Index " PRINTF_SIZE_T_SPECIFIER
- " should be set via vkCmdBindVertexBuffers.",
- (uint64_t)pCB->lastBound[VK_PIPELINE_BIND_POINT_GRAPHICS].pipeline, i);
+ // For given active slots, verify any dynamic descriptors and record updated images & buffers
+ result |= validate_and_update_drawtime_descriptor_state(my_data, pCB, activeSetBindingsPairs);
+ }
+ if (VK_PIPELINE_BIND_POINT_GRAPHICS == bindPoint) {
+ // Verify Vtx binding
+ if (pPipe->vertexBindingDescriptions.size() > 0) {
+ for (size_t i = 0; i < pPipe->vertexBindingDescriptions.size(); i++) {
+ if ((pCB->currentDrawData.buffers.size() < (i + 1)) || (pCB->currentDrawData.buffers[i] == VK_NULL_HANDLE)) {
+ result |= log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, (VkDebugReportObjectTypeEXT)0, 0,
+ __LINE__, DRAWSTATE_VTX_INDEX_OUT_OF_BOUNDS, "DS",
+ "The Pipeline State Object (%#" PRIxLEAST64
+ ") expects that this Command Buffer's vertex binding Index " PRINTF_SIZE_T_SPECIFIER
+ " should be set via vkCmdBindVertexBuffers.",
+ (uint64_t)state.pipeline, i);
+ }
}
- }
- } else {
- if (!pCB->currentDrawData.buffers.empty()) {
- result |= log_msg(my_data->report_data, VK_DEBUG_REPORT_PERFORMANCE_WARNING_BIT_EXT, (VkDebugReportObjectTypeEXT)0,
- 0, __LINE__, DRAWSTATE_VTX_INDEX_OUT_OF_BOUNDS, "DS",
- "Vertex buffers are bound to command buffer (%#" PRIxLEAST64
- ") but no vertex buffers are attached to this Pipeline State Object (%#" PRIxLEAST64 ").",
- (uint64_t)pCB->commandBuffer, (uint64_t)pCB->lastBound[VK_PIPELINE_BIND_POINT_GRAPHICS].pipeline);
- }
- }
- // If Viewport or scissors are dynamic, verify that dynamic count matches PSO count.
- // Skip check if rasterization is disabled or there is no viewport.
- if ((!pPipe->graphicsPipelineCI.pRasterizationState ||
- !pPipe->graphicsPipelineCI.pRasterizationState->rasterizerDiscardEnable) &&
- pPipe->graphicsPipelineCI.pViewportState) {
- VkBool32 dynViewport = isDynamic(pPipe, VK_DYNAMIC_STATE_VIEWPORT);
- VkBool32 dynScissor = isDynamic(pPipe, VK_DYNAMIC_STATE_SCISSOR);
- if (dynViewport) {
- if (pCB->viewports.size() != pPipe->graphicsPipelineCI.pViewportState->viewportCount) {
- result |= log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, (VkDebugReportObjectTypeEXT)0, 0,
- __LINE__, DRAWSTATE_VIEWPORT_SCISSOR_MISMATCH, "DS",
- "Dynamic viewportCount from vkCmdSetViewport() is " PRINTF_SIZE_T_SPECIFIER
- ", but PSO viewportCount is %u. These counts must match.",
- pCB->viewports.size(), pPipe->graphicsPipelineCI.pViewportState->viewportCount);
+ } else {
+ if (!pCB->currentDrawData.buffers.empty()) {
+ result |= log_msg(my_data->report_data, VK_DEBUG_REPORT_PERFORMANCE_WARNING_BIT_EXT,
+ (VkDebugReportObjectTypeEXT)0, 0, __LINE__, DRAWSTATE_VTX_INDEX_OUT_OF_BOUNDS, "DS",
+ "Vertex buffers are bound to command buffer (%#" PRIxLEAST64
+ ") but no vertex buffers are attached to this Pipeline State Object (%#" PRIxLEAST64 ").",
+ (uint64_t)pCB->commandBuffer, (uint64_t)state.pipeline);
}
}
- if (dynScissor) {
- if (pCB->scissors.size() != pPipe->graphicsPipelineCI.pViewportState->scissorCount) {
- result |= log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, (VkDebugReportObjectTypeEXT)0, 0,
- __LINE__, DRAWSTATE_VIEWPORT_SCISSOR_MISMATCH, "DS",
- "Dynamic scissorCount from vkCmdSetScissor() is " PRINTF_SIZE_T_SPECIFIER
- ", but PSO scissorCount is %u. These counts must match.",
- pCB->scissors.size(), pPipe->graphicsPipelineCI.pViewportState->scissorCount);
+ // If Viewport or scissors are dynamic, verify that dynamic count matches PSO count.
+ // Skip check if rasterization is disabled or there is no viewport.
+ if ((!pPipe->graphicsPipelineCI.pRasterizationState ||
+ !pPipe->graphicsPipelineCI.pRasterizationState->rasterizerDiscardEnable) &&
+ pPipe->graphicsPipelineCI.pViewportState) {
+ VkBool32 dynViewport = isDynamic(pPipe, VK_DYNAMIC_STATE_VIEWPORT);
+ VkBool32 dynScissor = isDynamic(pPipe, VK_DYNAMIC_STATE_SCISSOR);
+ if (dynViewport) {
+ if (pCB->viewports.size() != pPipe->graphicsPipelineCI.pViewportState->viewportCount) {
+ result |= log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, (VkDebugReportObjectTypeEXT)0, 0,
+ __LINE__, DRAWSTATE_VIEWPORT_SCISSOR_MISMATCH, "DS",
+ "Dynamic viewportCount from vkCmdSetViewport() is " PRINTF_SIZE_T_SPECIFIER
+ ", but PSO viewportCount is %u. These counts must match.",
+ pCB->viewports.size(), pPipe->graphicsPipelineCI.pViewportState->viewportCount);
+ }
+ }
+ if (dynScissor) {
+ if (pCB->scissors.size() != pPipe->graphicsPipelineCI.pViewportState->scissorCount) {
+ result |= log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, (VkDebugReportObjectTypeEXT)0, 0,
+ __LINE__, DRAWSTATE_VIEWPORT_SCISSOR_MISMATCH, "DS",
+ "Dynamic scissorCount from vkCmdSetScissor() is " PRINTF_SIZE_T_SPECIFIER
+ ", but PSO scissorCount is %u. These counts must match.",
+ pCB->scissors.size(), pPipe->graphicsPipelineCI.pViewportState->scissorCount);
+ }
}
}
}
@@ -4465,7 +4521,8 @@ static void resetCB(layer_data *my_data, const VkCommandBuffer cb) {
pCB->currentDrawData.buffers.clear();
pCB->primaryCommandBuffer = VK_NULL_HANDLE;
pCB->secondaryCommandBuffers.clear();
- pCB->activeDescriptorSets.clear();
+ pCB->updateImages.clear();
+ pCB->updateBuffers.clear();
pCB->validate_functions.clear();
pCB->pMemObjList.clear();
pCB->eventUpdates.clear();
@@ -6864,29 +6921,6 @@ vkUpdateDescriptorSets(VkDevice device, uint32_t descriptorWriteCount, const VkW
// dsUpdate will return VK_TRUE only if a bailout error occurs, so we want to call down tree when update returns VK_FALSE
layer_data *dev_data = get_my_data_ptr(get_dispatch_key(device), layer_data_map);
loader_platform_thread_lock_mutex(&globalLock);
-#if MTMERGESOURCE
- // MTMTODO : Merge this in with existing update code below and handle descriptor copies case
- uint32_t j = 0;
- for (uint32_t i = 0; i < descriptorWriteCount; ++i) {
- if (pDescriptorWrites[i].descriptorType == VK_DESCRIPTOR_TYPE_STORAGE_IMAGE) {
- for (j = 0; j < pDescriptorWrites[i].descriptorCount; ++j) {
- dev_data->descriptorSetMap[pDescriptorWrites[i].dstSet].images.push_back(
- pDescriptorWrites[i].pImageInfo[j].imageView);
- }
- } else if (pDescriptorWrites[i].descriptorType == VK_DESCRIPTOR_TYPE_STORAGE_TEXEL_BUFFER) {
- for (j = 0; j < pDescriptorWrites[i].descriptorCount; ++j) {
- dev_data->descriptorSetMap[pDescriptorWrites[i].dstSet].buffers.push_back(
- dev_data->bufferViewMap[pDescriptorWrites[i].pTexelBufferView[j]].buffer);
- }
- } else if (pDescriptorWrites[i].descriptorType == VK_DESCRIPTOR_TYPE_STORAGE_BUFFER ||
- pDescriptorWrites[i].descriptorType == VK_DESCRIPTOR_TYPE_STORAGE_BUFFER_DYNAMIC) {
- for (j = 0; j < pDescriptorWrites[i].descriptorCount; ++j) {
- dev_data->descriptorSetMap[pDescriptorWrites[i].dstSet].buffers.push_back(
- pDescriptorWrites[i].pBufferInfo[j].buffer);
- }
- }
- }
-#endif
VkBool32 rtn = dsUpdate(dev_data, device, descriptorWriteCount, pDescriptorWrites, descriptorCopyCount, pDescriptorCopies);
loader_platform_thread_unlock_mutex(&globalLock);
if (!rtn) {
@@ -7325,21 +7359,6 @@ vkCmdBindDescriptorSets(VkCommandBuffer commandBuffer, VkPipelineBindPoint pipel
VkBool32 skipCall = VK_FALSE;
layer_data *dev_data = get_my_data_ptr(get_dispatch_key(commandBuffer), layer_data_map);
loader_platform_thread_lock_mutex(&globalLock);
-#if MTMERGESOURCE
- // MTMTODO : Merge this with code below
- auto cb_data = dev_data->commandBufferMap.find(commandBuffer);
- if (cb_data != dev_data->commandBufferMap.end()) {
- // MTMTODO : activeDescriptorSets should be merged with lastBound.boundDescriptorSets
- std::vector<VkDescriptorSet> &activeDescriptorSets = cb_data->second->activeDescriptorSets;
- if (activeDescriptorSets.size() < (setCount + firstSet)) {
- activeDescriptorSets.resize(setCount + firstSet);
- }
- for (uint32_t i = 0; i < setCount; ++i) {
- activeDescriptorSets[i + firstSet] = pDescriptorSets[i];
- }
- }
- // TODO : Somewhere need to verify that all textures referenced by shaders in DS are in some type of *SHADER_READ* state
-#endif
GLOBAL_CB_NODE *pCB = getCBNode(dev_data, commandBuffer);
if (pCB) {
if (pCB->state == CB_RECORDING) {
@@ -7598,64 +7617,48 @@ VK_LAYER_EXPORT VKAPI_ATTR void VKAPI_CALL vkCmdBindVertexBuffers(VkCommandBuffe
dev_data->device_dispatch_table->CmdBindVertexBuffers(commandBuffer, firstBinding, bindingCount, pBuffers, pOffsets);
}
-#if MTMERGESOURCE
/* expects globalLock to be held by caller */
-bool markStoreImagesAndBuffersAsWritten(VkCommandBuffer commandBuffer) {
+bool markStoreImagesAndBuffersAsWritten(layer_data *dev_data, GLOBAL_CB_NODE *pCB) {
bool skip_call = false;
- layer_data *my_data = get_my_data_ptr(get_dispatch_key(commandBuffer), layer_data_map);
- auto cb_data = my_data->commandBufferMap.find(commandBuffer);
- if (cb_data == my_data->commandBufferMap.end())
- return skip_call;
- std::vector<VkDescriptorSet> &activeDescriptorSets = cb_data->second->activeDescriptorSets;
- for (auto descriptorSet : activeDescriptorSets) {
- auto ds_data = my_data->descriptorSetMap.find(descriptorSet);
- if (ds_data == my_data->descriptorSetMap.end())
+
+ for (auto imageView : pCB->updateImages) {
+ auto iv_data = dev_data->imageViewMap.find(imageView);
+ if (iv_data == dev_data->imageViewMap.end())
continue;
- std::vector<VkImageView> images = ds_data->second.images;
- std::vector<VkBuffer> buffers = ds_data->second.buffers;
- for (auto imageView : images) {
- auto iv_data = my_data->imageViewMap.find(imageView);
- if (iv_data == my_data->imageViewMap.end())
- continue;
- VkImage image = iv_data->second.image;
- VkDeviceMemory mem;
- skip_call |=
- get_mem_binding_from_object(my_data, commandBuffer, (uint64_t)image, VK_DEBUG_REPORT_OBJECT_TYPE_IMAGE_EXT, &mem);
- std::function<VkBool32()> function = [=]() {
- set_memory_valid(my_data, mem, true, image);
- return VK_FALSE;
- };
- cb_data->second->validate_functions.push_back(function);
- }
- for (auto buffer : buffers) {
- VkDeviceMemory mem;
- skip_call |=
- get_mem_binding_from_object(my_data, commandBuffer, (uint64_t)buffer, VK_DEBUG_REPORT_OBJECT_TYPE_BUFFER_EXT, &mem);
- std::function<VkBool32()> function = [=]() {
- set_memory_valid(my_data, mem, true);
- return VK_FALSE;
- };
- cb_data->second->validate_functions.push_back(function);
- }
+ VkImage image = iv_data->second.image;
+ VkDeviceMemory mem;
+ skip_call |=
+ get_mem_binding_from_object(dev_data, pCB->commandBuffer, (uint64_t)image, VK_DEBUG_REPORT_OBJECT_TYPE_IMAGE_EXT, &mem);
+ std::function<VkBool32()> function = [=]() {
+ set_memory_valid(dev_data, mem, true, image);
+ return VK_FALSE;
+ };
+ pCB->validate_functions.push_back(function);
+ }
+ for (auto buffer : pCB->updateBuffers) {
+ VkDeviceMemory mem;
+ skip_call |= get_mem_binding_from_object(dev_data, pCB->commandBuffer, (uint64_t)buffer,
+ VK_DEBUG_REPORT_OBJECT_TYPE_BUFFER_EXT, &mem);
+ std::function<VkBool32()> function = [=]() {
+ set_memory_valid(dev_data, mem, true);
+ return VK_FALSE;
+ };
+ pCB->validate_functions.push_back(function);
}
return skip_call;
}
-#endif
VK_LAYER_EXPORT VKAPI_ATTR void VKAPI_CALL vkCmdDraw(VkCommandBuffer commandBuffer, uint32_t vertexCount, uint32_t instanceCount,
uint32_t firstVertex, uint32_t firstInstance) {
VkBool32 skipCall = VK_FALSE;
layer_data *dev_data = get_my_data_ptr(get_dispatch_key(commandBuffer), layer_data_map);
loader_platform_thread_lock_mutex(&globalLock);
-#if MTMERGESOURCE
- // MTMTODO : merge with code below
- skipCall = markStoreImagesAndBuffersAsWritten(commandBuffer);
-#endif
GLOBAL_CB_NODE *pCB = getCBNode(dev_data, commandBuffer);
if (pCB) {
skipCall |= addCmd(dev_data, pCB, CMD_DRAW, "vkCmdDraw()");
pCB->drawCount[DRAW]++;
- skipCall |= validate_draw_state(dev_data, pCB, VK_FALSE);
+ skipCall |= validate_and_update_draw_state(dev_data, pCB, VK_FALSE, VK_PIPELINE_BIND_POINT_GRAPHICS);
+ skipCall |= markStoreImagesAndBuffersAsWritten(dev_data, pCB);
// TODO : Need to pass commandBuffer as srcObj here
skipCall |=
log_msg(dev_data->report_data, VK_DEBUG_REPORT_INFORMATION_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_COMMAND_BUFFER_EXT, 0,
@@ -7677,15 +7680,12 @@ VK_LAYER_EXPORT VKAPI_ATTR void VKAPI_CALL vkCmdDrawIndexed(VkCommandBuffer comm
layer_data *dev_data = get_my_data_ptr(get_dispatch_key(commandBuffer), layer_data_map);
VkBool32 skipCall = VK_FALSE;
loader_platform_thread_lock_mutex(&globalLock);
-#if MTMERGESOURCE
- // MTMTODO : merge with code below
- skipCall = markStoreImagesAndBuffersAsWritten(commandBuffer);
-#endif
GLOBAL_CB_NODE *pCB = getCBNode(dev_data, commandBuffer);
if (pCB) {
skipCall |= addCmd(dev_data, pCB, CMD_DRAWINDEXED, "vkCmdDrawIndexed()");
pCB->drawCount[DRAW_INDEXED]++;
- skipCall |= validate_draw_state(dev_data, pCB, VK_TRUE);
+ skipCall |= validate_and_update_draw_state(dev_data, pCB, VK_TRUE, VK_PIPELINE_BIND_POINT_GRAPHICS);
+ skipCall |= markStoreImagesAndBuffersAsWritten(dev_data, pCB);
// TODO : Need to pass commandBuffer as srcObj here
skipCall |= log_msg(dev_data->report_data, VK_DEBUG_REPORT_INFORMATION_BIT_EXT,
VK_DEBUG_REPORT_OBJECT_TYPE_COMMAND_BUFFER_EXT, 0, __LINE__, DRAWSTATE_NONE, "DS",
@@ -7713,13 +7713,13 @@ vkCmdDrawIndirect(VkCommandBuffer commandBuffer, VkBuffer buffer, VkDeviceSize o
skipCall =
get_mem_binding_from_object(dev_data, commandBuffer, (uint64_t)buffer, VK_DEBUG_REPORT_OBJECT_TYPE_BUFFER_EXT, &mem);
skipCall |= update_cmd_buf_and_mem_references(dev_data, commandBuffer, mem, "vkCmdDrawIndirect");
- skipCall |= markStoreImagesAndBuffersAsWritten(commandBuffer);
#endif
GLOBAL_CB_NODE *pCB = getCBNode(dev_data, commandBuffer);
if (pCB) {
skipCall |= addCmd(dev_data, pCB, CMD_DRAWINDIRECT, "vkCmdDrawIndirect()");
pCB->drawCount[DRAW_INDIRECT]++;
- skipCall |= validate_draw_state(dev_data, pCB, VK_FALSE);
+ skipCall |= validate_and_update_draw_state(dev_data, pCB, VK_FALSE, VK_PIPELINE_BIND_POINT_GRAPHICS);
+ skipCall |= markStoreImagesAndBuffersAsWritten(dev_data, pCB);
// TODO : Need to pass commandBuffer as srcObj here
skipCall |= log_msg(dev_data->report_data, VK_DEBUG_REPORT_INFORMATION_BIT_EXT,
VK_DEBUG_REPORT_OBJECT_TYPE_COMMAND_BUFFER_EXT, 0, __LINE__, DRAWSTATE_NONE, "DS",
@@ -7746,13 +7746,13 @@ vkCmdDrawIndexedIndirect(VkCommandBuffer commandBuffer, VkBuffer buffer, VkDevic
skipCall =
get_mem_binding_from_object(dev_data, commandBuffer, (uint64_t)buffer, VK_DEBUG_REPORT_OBJECT_TYPE_BUFFER_EXT, &mem);
skipCall |= update_cmd_buf_and_mem_references(dev_data, commandBuffer, mem, "vkCmdDrawIndexedIndirect");
- skipCall |= markStoreImagesAndBuffersAsWritten(commandBuffer);
#endif
GLOBAL_CB_NODE *pCB = getCBNode(dev_data, commandBuffer);
if (pCB) {
skipCall |= addCmd(dev_data, pCB, CMD_DRAWINDEXEDINDIRECT, "vkCmdDrawIndexedIndirect()");
pCB->drawCount[DRAW_INDEXED_INDIRECT]++;
- skipCall |= validate_draw_state(dev_data, pCB, VK_TRUE);
+ skipCall |= validate_and_update_draw_state(dev_data, pCB, VK_TRUE, VK_PIPELINE_BIND_POINT_GRAPHICS);
+ skipCall |= markStoreImagesAndBuffersAsWritten(dev_data, pCB);
// TODO : Need to pass commandBuffer as srcObj here
skipCall |=
log_msg(dev_data->report_data, VK_DEBUG_REPORT_INFORMATION_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_COMMAND_BUFFER_EXT, 0,
@@ -7773,11 +7773,13 @@ VK_LAYER_EXPORT VKAPI_ATTR void VKAPI_CALL vkCmdDispatch(VkCommandBuffer command
VkBool32 skipCall = VK_FALSE;
layer_data *dev_data = get_my_data_ptr(get_dispatch_key(commandBuffer), layer_data_map);
loader_platform_thread_lock_mutex(&globalLock);
-#if MTMERGESOURCE
- skipCall = markStoreImagesAndBuffersAsWritten(commandBuffer);
-#endif
GLOBAL_CB_NODE *pCB = getCBNode(dev_data, commandBuffer);
if (pCB) {
+ // TODO : Re-enable validate_and_update_draw_state() when it supports compute shaders
+ // skipCall |= validate_and_update_draw_state(dev_data, pCB, VK_FALSE, VK_PIPELINE_BIND_POINT_COMPUTE);
+ // TODO : Call below is temporary until call above can be re-enabled
+ update_shader_storage_images_and_buffers(dev_data, pCB);
+ skipCall |= markStoreImagesAndBuffersAsWritten(dev_data, pCB);
skipCall |= addCmd(dev_data, pCB, CMD_DISPATCH, "vkCmdDispatch()");
skipCall |= insideRenderPass(dev_data, pCB, "vkCmdDispatch");
}
@@ -7796,10 +7798,14 @@ vkCmdDispatchIndirect(VkCommandBuffer commandBuffer, VkBuffer buffer, VkDeviceSi
skipCall =
get_mem_binding_from_object(dev_data, commandBuffer, (uint64_t)buffer, VK_DEBUG_REPORT_OBJECT_TYPE_BUFFER_EXT, &mem);
skipCall |= update_cmd_buf_and_mem_references(dev_data, commandBuffer, mem, "vkCmdDispatchIndirect");
- skipCall |= markStoreImagesAndBuffersAsWritten(commandBuffer);
#endif
GLOBAL_CB_NODE *pCB = getCBNode(dev_data, commandBuffer);
if (pCB) {
+ // TODO : Re-enable validate_and_update_draw_state() when it supports compute shaders
+ // skipCall |= validate_and_update_draw_state(dev_data, pCB, VK_FALSE, VK_PIPELINE_BIND_POINT_COMPUTE);
+ // TODO : Call below is temporary until call above can be re-enabled
+ update_shader_storage_images_and_buffers(dev_data, pCB);
+ skipCall |= markStoreImagesAndBuffersAsWritten(dev_data, pCB);
skipCall |= addCmd(dev_data, pCB, CMD_DISPATCHINDIRECT, "vkCmdDispatchIndirect()");
skipCall |= insideRenderPass(dev_data, pCB, "vkCmdDispatchIndirect");
}
diff --git a/layers/core_validation.h b/layers/core_validation.h
index 31132a93..0518da77 100644
--- a/layers/core_validation.h
+++ b/layers/core_validation.h
@@ -868,11 +868,13 @@ struct GLOBAL_CB_NODE {
vector<DRAW_DATA> drawData;
DRAW_DATA currentDrawData;
VkCommandBuffer primaryCommandBuffer;
+ // Track images and buffers that are updated by this CB at the point of a draw
+ unordered_set<VkImageView> updateImages;
+ unordered_set<VkBuffer> updateBuffers;
// If cmd buffer is primary, track secondary command buffers pending
// execution
std::unordered_set<VkCommandBuffer> secondaryCommandBuffers;
// MTMTODO : Scrub these data fields and merge active sets w/ lastBound as appropriate
- vector<VkDescriptorSet> activeDescriptorSets;
vector<std::function<VkBool32()>> validate_functions;
list<VkDeviceMemory> pMemObjList; // List container of Mem objs referenced by this CB
vector<std::function<bool(VkQueue)>> eventUpdates;
generated by cgit v1.2.3 (git 2.39.1) at 2026-08-31 16:18:52 +0000