Skip to content

Instantly share code, notes, and snippets.

@tuket

tuket/calc.comp.glsl

Created May 27, 2021 18:04
Show Gist options
  • Save tuket/53c6172e58613b963dc66a27e51dc1c3 to your computer and use it in GitHub Desktop.
Save tuket/53c6172e58613b963dc66a27e51dc1c3 to your computer and use it in GitHub Desktop.
Download ZIP
Vulkan validation error when I try to reset a commandPool after vkQueueWaitIddle
Raw
calc.comp.glsl
#version 450
#extension GL_EXT_shader_explicit_arithmetic_types_int64 : enable
layout (local_size_x = 128) in;
layout(set = 0, binding = 0) uniform Unifs {
uniform int64_t start;
uniform int64_t n;
uniform int64_t N;
} unifs;
layout(set = 1, binding = 0) buffer OutBuffer {
uint64_t outBuf[];
};
void main()
{
uint myId = gl_GlobalInvocationID.x;
outBuf[myId] = 0;
uint64_t start = unifs.start + myId * unifs.n;
uint64_t end = start + unifs.n;
for(uint64_t i = start; i < end; i++) {
if((11 * i) % unifs.N == 1) {
outBuf[myId] = i;
break;
}
}
}
Raw
main.cpp
#include <stdio.h>
#include <string.h>
#include <ctype.h>
#include <stdint.h>
#include <inttypes.h>
#include <time.h>
#include <assert.h>
#include <vulkan/vulkan.h>
typedef int32_t i32;
typedef uint32_t u32;
typedef int64_t i64;
typedef uint64_t u64;
#define MIN(a, b) ((a) < (b) ? (a) : (b))
constexpr i64 NANO = 1'000'000'000;
constexpr i64 NT = 60 * 60 * 12;
constexpr i64 N = NANO * NT;
constexpr i64 perThread = 1'000'000; // how many iterations we will do in each work-item
constexpr i64 numThreads = 10'000;
constexpr i64 perIteration = perThread * numThreads;
constexpr i64 numIterations = N / perIteration;
static const char *strstri(const char *container, const char *contained)
{
const int containerLen = strlen(container);
const int containedLen = strlen(contained);
for (int i = 0; i + containedLen - 1 < containerLen; i++)
{
int j;
for (j = 0; j < containedLen; j++)
{
const char a = tolower(container[i + j]);
const char b = tolower(contained[j]);
if (a != b)
break;
}
if (j == containedLen)
return contained + i;
}
return nullptr;
}
#ifdef _WIN32
#include <windows.h>
#endif
static double getTime()
{
#ifdef _WIN32
LARGE_INTEGER t, f;
bool success = QueryPerformanceCounter(&t);
success &= QueryPerformanceFrequency(&f);
assert(success);
return double(t.QuadPart) / double(f.QuadPart);
#else
timespec t;
clock_gettime(CLOCK_REALTIME, &t);
return t.tv_sec + 1e-9 * t.tv_nsec;
#endif
}
static void printETA(double t)
{
i64 seconds = t;
i64 minutes = seconds / 60;
i64 hours = minutes / 60;
minutes %= 60;
seconds %= 60;
printf("ETA: %" PRId64 "hours, %" PRId64 " minutes, %" PRId64 " seconds\n", hours, minutes, seconds);
}
static void calcElapsedAndPrintETA(double startT, i64 done, i64 total)
{
const double nowT = getTime();
const double elapsedT = nowT - startT;
const double eta = elapsedT * (total - done) / done;
printETA(eta);
}
// --- CPU implementation single threaded ---
static i64 calcWithCpu()
{
constexpr i64 batch = 1'000'000'000;
const double startT = getTime();
i64 percent = 0;
for (i64 i = 0; i < N;)
{
const i64 newPercent = (i64(10000) * i) / N;
if (newPercent != percent)
{
printf("%" PRId64 ".%" PRId64 "%%\n", newPercent / 100, newPercent % 100);
percent = newPercent;
}
if (i != 0)
{
calcElapsedAndPrintETA(startT, i, N);
}
i64 n = MIN(i + batch, N);
for (; i < n; i++)
if ((i64(11) * i) % N == 1)
return i;
}
assert(false);
return 0;
}
// --- Vulkan implementation ---
static void printPhysicalDeviceProps(VkPhysicalDeviceProperties props)
{
printf("device name: %s\n", props.deviceName);
}
static void printPhysicalDeviceProps(VkPhysicalDevice gpu)
{
VkPhysicalDeviceProperties props;
vkGetPhysicalDeviceProperties(gpu, &props);
printPhysicalDeviceProps(props);
}
static void printMemoryProps(const VkPhysicalDeviceMemoryProperties &memProps)
{
printf("Memory Types:\n");
for (int i = 0; i < memProps.memoryTypeCount; i++)
{
auto &t = memProps.memoryTypes[i];
printf("%d)\n", i);
printf(" Heap index: %d\n", t.heapIndex);
printf(" Flags:");
if (t.propertyFlags & VK_MEMORY_PROPERTY_DEVICE_LOCAL_BIT)
printf(" VK_MEMORY_PROPERTY_DEVICE_LOCAL_BIT |");
if (t.propertyFlags & VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT)
printf(" VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT |");
if (t.propertyFlags & VK_MEMORY_PROPERTY_HOST_COHERENT_BIT)
printf(" VK_MEMORY_PROPERTY_HOST_COHERENT_BIT |");
if (t.propertyFlags & VK_MEMORY_PROPERTY_HOST_CACHED_BIT)
printf(" VK_MEMORY_PROPERTY_HOST_CACHED_BIT |");
if (t.propertyFlags & VK_MEMORY_PROPERTY_LAZILY_ALLOCATED_BIT)
printf(" VK_MEMORY_PROPERTY_LAZILY_ALLOCATED_BIT |");
if (t.propertyFlags & VK_MEMORY_PROPERTY_PROTECTED_BIT)
printf(" VK_MEMORY_PROPERTY_PROTECTED_BIT |");
printf("\n");
}
printf("\nMemory Heaps:\n");
for (int i = 0; i < memProps.memoryHeapCount; i++)
{
auto &heap = memProps.memoryHeaps[i];
printf("%d)\n", i);
printf(" Size: %" PRIu64 "MB\n", heap.size / (1024 * 1024));
printf(" Flags:");
if (heap.flags & VK_MEMORY_HEAP_DEVICE_LOCAL_BIT)
printf(" VK_MEMORY_HEAP_DEVICE_LOCAL_BIT");
printf("\n");
}
}
inline size_t aligned(size_t offset, size_t alignment)
{
offset = (offset + alignment - 1) / alignment * alignment;
return offset;
}
static i64 calcWithVulkan()
{
constexpr u32 MAX_LAYERS = 32;
VkLayerProperties* layersProps = new VkLayerProperties[MAX_LAYERS];
// TODO: delete[]
u32 numLayers = MAX_LAYERS;
vkEnumerateInstanceLayerProperties(&numLayers, layersProps);
printf("Available Layers\n");
for (u32 i = 0; i < numLayers; i++)
{
printf("%s: %s\n", layersProps[i].layerName, layersProps[i].description);
}
VkApplicationInfo appInfo = {};
appInfo.sType = VK_STRUCTURE_TYPE_APPLICATION_INFO;
appInfo.pApplicationName = "example";
appInfo.applicationVersion = VK_MAKE_VERSION(1, 0, 0);
appInfo.pEngineName = "none";
appInfo.engineVersion = VK_MAKE_VERSION(1, 0, 0);
appInfo.apiVersion = VK_API_VERSION_1_0;
VkInstanceCreateInfo instanceCreateInfo = {};
instanceCreateInfo.sType = VK_STRUCTURE_TYPE_INSTANCE_CREATE_INFO;
instanceCreateInfo.pApplicationInfo = &appInfo;
instanceCreateInfo.enabledExtensionCount = 1;
static const char* const EXTENSION_NAMES[] = {"VK_EXT_debug_utils"};
instanceCreateInfo.ppEnabledExtensionNames = EXTENSION_NAMES;
instanceCreateInfo.enabledLayerCount = 1;
static const char *const LAYER_NAMES[] = {"VK_LAYER_KHRONOS_validation"};
instanceCreateInfo.ppEnabledLayerNames = LAYER_NAMES;
VkInstance inst;
VkResult result = vkCreateInstance(&instanceCreateInfo, nullptr, &inst);
if (result != VK_SUCCESS)
{
printf("Error creating Vulkan instance\n");
return 0;
}
constexpr u32 MAX_PHYSICAL_DEVICES = 4;
VkPhysicalDevice* physicalDevices = new VkPhysicalDevice[MAX_PHYSICAL_DEVICES];
// TODO: delete[]
u32 numPhysicalDevices = MAX_PHYSICAL_DEVICES;
vkEnumeratePhysicalDevices(inst, &numPhysicalDevices, physicalDevices);
assert(numPhysicalDevices > 0);
VkPhysicalDevice bestGpu = nullptr;
for (u32 i = 0; i < numPhysicalDevices; i++)
{
VkPhysicalDeviceProperties props;
vkGetPhysicalDeviceProperties(physicalDevices[i], &props);
//printPhyiscalDeviceProps(props);
if (!bestGpu || strstri(props.deviceName, "geforce"))
bestGpu = physicalDevices[i];
}
printf("best GPU: ");
printPhysicalDeviceProps(bestGpu);
constexpr u32 MAX_QUEUE_FAMILIES = 8;
VkQueueFamilyProperties* queueFamiles = new VkQueueFamilyProperties[MAX_QUEUE_FAMILIES];
// TODO: delete[]
u32 numQueueFamilies = MAX_QUEUE_FAMILIES;
vkGetPhysicalDeviceQueueFamilyProperties(bestGpu, &numQueueFamilies, queueFamiles);
printf("numQueueFamilies: %d\n", numQueueFamilies);
for (int i = 0; i < numQueueFamilies; i++)
{
printf("%d)\n", i);
printf("flags: ");
if (queueFamiles[i].queueFlags | VK_QUEUE_COMPUTE_BIT)
printf("COMPUTE | ");
if (queueFamiles[i].queueFlags | VK_QUEUE_GRAPHICS_BIT)
printf("GRAPHICS | ");
if (queueFamiles[i].queueFlags | VK_QUEUE_TRANSFER_BIT)
printf("TRANSFER | ");
if (queueFamiles[i].queueFlags | VK_QUEUE_SPARSE_BINDING_BIT)
printf("SPARSE_BINDING | ");
printf("\n");
printf("queueCount: %d\n", queueFamiles[i].queueCount);
}
u32 bestQueueFamilyInd = -1;
for (u32 i = 0; i < numQueueFamilies; i++)
{
if ((queueFamiles[i].queueFlags | VK_QUEUE_COMPUTE_BIT) &&
(bestQueueFamilyInd == -1 ||
queueFamiles[i].queueCount > queueFamiles[bestQueueFamilyInd].queueCount))
{
bestQueueFamilyInd = i;
}
}
const float queuePriorities[] = {0};
VkDeviceQueueCreateInfo queueCreateInfo = {};
queueCreateInfo.sType = VK_STRUCTURE_TYPE_DEVICE_QUEUE_CREATE_INFO;
queueCreateInfo.queueCount = 1;
queueCreateInfo.queueFamilyIndex = bestQueueFamilyInd;
queueCreateInfo.pQueuePriorities = queuePriorities;
VkPhysicalDeviceFeatures gpuFeatures;
vkGetPhysicalDeviceFeatures(bestGpu, &gpuFeatures);
if (!gpuFeatures.shaderInt64)
{
printf("Error: 64 bit integer not supported\n");
return 0;
}
memset(&gpuFeatures, 0, sizeof(gpuFeatures));
gpuFeatures.shaderInt64 = VK_TRUE;
VkDeviceCreateInfo deviceCreateInfo = {};
deviceCreateInfo.sType = VK_STRUCTURE_TYPE_DEVICE_CREATE_INFO;
deviceCreateInfo.queueCreateInfoCount = 1;
deviceCreateInfo.pQueueCreateInfos = &queueCreateInfo;
deviceCreateInfo.pEnabledFeatures = &gpuFeatures;
VkDevice device;
vkCreateDevice(bestGpu, &deviceCreateInfo, nullptr, &device);
VkQueue queue;
vkGetDeviceQueue(device, bestQueueFamilyInd, 0, &queue);
VkPhysicalDeviceMemoryProperties memProps;
vkGetPhysicalDeviceMemoryProperties(bestGpu, &memProps);
printMemoryProps(memProps);
const u32 localMemTypeInd = [&memProps]() -> u32 {
for (u32 i = 0; i < memProps.memoryTypeCount; i++)
{
const VkMemoryPropertyFlags typeFlags = memProps.memoryTypes[i].propertyFlags;
if (typeFlags & VK_MEMORY_PROPERTY_DEVICE_LOCAL_BIT)
return i;
}
assert(false);
}();
const u32 hostVisibleMemTypeInd = [&memProps]() -> u32 {
for (u32 i = 0; i < memProps.memoryTypeCount; i++)
{
const VkMemoryPropertyFlags typeFlags = memProps.memoryTypes[i].propertyFlags;
if (typeFlags & VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT)
return i;
}
assert(false);
}();
struct Uniforms { i64 start, perThread, N; };
Uniforms uniforms = { 0, perThread, N };
VkBuffer buffer;
{
VkBufferCreateInfo info = {};
info.sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO;
info.size = sizeof(i64) * numThreads;
info.usage = VK_BUFFER_USAGE_STORAGE_BUFFER_BIT | VK_BUFFER_USAGE_TRANSFER_SRC_BIT;
info.sharingMode = VK_SHARING_MODE_EXCLUSIVE;
vkCreateBuffer(device, &info, nullptr, &buffer);
}
VkBuffer unifsBuffer;
{
VkBufferCreateInfo info = {};
info.sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO;
info.size = sizeof(Uniforms);
info.usage = VK_BUFFER_USAGE_UNIFORM_BUFFER_BIT | VK_BUFFER_USAGE_TRANSFER_DST_BIT;
info.sharingMode = VK_SHARING_MODE_EXCLUSIVE;
vkCreateBuffer(device, &info, nullptr, &unifsBuffer);
}
VkBuffer stagingBuffer;
{
VkBufferCreateInfo info = {};
info.sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO;
info.size = sizeof(i64) * numThreads;
info.usage = VK_BUFFER_USAGE_TRANSFER_DST_BIT | VK_BUFFER_USAGE_TRANSFER_DST_BIT;
info.sharingMode = VK_SHARING_MODE_EXCLUSIVE;
vkCreateBuffer(device, &info, nullptr, &stagingBuffer);
}
VkDeviceMemory bufferMem = nullptr;
{
VkMemoryRequirements bufferMemReqs[2];
vkGetBufferMemoryRequirements(device, buffer, &bufferMemReqs[0]);
vkGetBufferMemoryRequirements(device, unifsBuffer, &bufferMemReqs[1]);
const size_t unifsBufferOffset = aligned(bufferMemReqs[0].size, bufferMemReqs[1].alignment);
const size_t memSize = unifsBufferOffset + bufferMemReqs[1].size;
//VkMemoryPropertyFlagBits memPropFlags;
VkMemoryAllocateInfo allocInfo = {};
allocInfo.sType = VK_STRUCTURE_TYPE_MEMORY_ALLOCATE_INFO;
allocInfo.allocationSize = memSize;
allocInfo.memoryTypeIndex = localMemTypeInd;
vkAllocateMemory(device, &allocInfo, nullptr, &bufferMem);
assert(bufferMem && "Error allocating memory");
vkBindBufferMemory(device, buffer, bufferMem, 0);
vkBindBufferMemory(device, unifsBuffer, bufferMem, unifsBufferOffset);
}
VkDeviceMemory stagingBufferMem = nullptr;
{
VkMemoryRequirements bufferMemReqs;
vkGetBufferMemoryRequirements(device, stagingBuffer, &bufferMemReqs);
VkMemoryAllocateInfo allocInfo = {};
allocInfo.sType = VK_STRUCTURE_TYPE_MEMORY_ALLOCATE_INFO;
allocInfo.allocationSize = bufferMemReqs.size;
allocInfo.memoryTypeIndex = hostVisibleMemTypeInd;
vkAllocateMemory(device, &allocInfo, nullptr, &stagingBufferMem);
vkBindBufferMemory(device, stagingBuffer, stagingBufferMem, 0);
}
VkShaderModule shaderModule;
{
VkShaderModuleCreateInfo shaderModuleCreateInfo = {};
FILE *file = fopen("../calc.spv", "rb");
const int MAX_CODE_SIZE = 1 << 20;
u32 *code = new u32[MAX_CODE_SIZE];
const int codeSize = fread(code, sizeof(u32), MAX_CODE_SIZE - 1, file);
shaderModuleCreateInfo.sType = VK_STRUCTURE_TYPE_SHADER_MODULE_CREATE_INFO;
shaderModuleCreateInfo.pCode = code;
shaderModuleCreateInfo.codeSize = 4 * codeSize;
vkCreateShaderModule(device, &shaderModuleCreateInfo, nullptr, &shaderModule);
delete[] code;
fclose(file);
}
VkPipelineShaderStageCreateInfo stageCreateInfo = {};
stageCreateInfo.sType = VK_STRUCTURE_TYPE_PIPELINE_SHADER_STAGE_CREATE_INFO;
stageCreateInfo.stage = VK_SHADER_STAGE_COMPUTE_BIT;
stageCreateInfo.module = shaderModule;
stageCreateInfo.pName = "main";
//info.pSpecializationInfo = nullptr;
VkDescriptorSetLayout* descriptorSetLayout = new VkDescriptorSetLayout[2];
// TODO: delete[]
{
VkDescriptorSetLayoutBinding descriptorSetLayoutBinding;
descriptorSetLayoutBinding.binding = 0;
descriptorSetLayoutBinding.descriptorType = VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER;
descriptorSetLayoutBinding.descriptorCount = 1;
descriptorSetLayoutBinding.stageFlags = VK_SHADER_STAGE_COMPUTE_BIT;
VkDescriptorSetLayoutCreateInfo descriptorSetLayoutCreateInfo = {};
descriptorSetLayoutCreateInfo.sType = VK_STRUCTURE_TYPE_DESCRIPTOR_SET_LAYOUT_CREATE_INFO;
descriptorSetLayoutCreateInfo.bindingCount = 1;
descriptorSetLayoutCreateInfo.pBindings = &descriptorSetLayoutBinding;
vkCreateDescriptorSetLayout(device, &descriptorSetLayoutCreateInfo, nullptr, &descriptorSetLayout[0]);
descriptorSetLayoutBinding.descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER;
vkCreateDescriptorSetLayout(device, &descriptorSetLayoutCreateInfo, nullptr, &descriptorSetLayout[1]);
}
VkPipelineLayoutCreateInfo pipelineLayoutCreateInfo = {};
pipelineLayoutCreateInfo.sType = VK_STRUCTURE_TYPE_PIPELINE_LAYOUT_CREATE_INFO;
pipelineLayoutCreateInfo.setLayoutCount = 2;
pipelineLayoutCreateInfo.pSetLayouts = descriptorSetLayout;
VkPipelineLayout pipelineLayout;
vkCreatePipelineLayout(device, &pipelineLayoutCreateInfo, nullptr, &pipelineLayout);
VkComputePipelineCreateInfo pipelineCreateInfo = {};
pipelineCreateInfo.sType = VK_STRUCTURE_TYPE_COMPUTE_PIPELINE_CREATE_INFO;
pipelineCreateInfo.stage = stageCreateInfo;
pipelineCreateInfo.layout = pipelineLayout;
VkPipeline pipeline;
vkCreateComputePipelines(device, VK_NULL_HANDLE, 1, &pipelineCreateInfo, nullptr, &pipeline);
VkDescriptorPool descriptorPool;
{
VkDescriptorPoolSize poolSizes[2] = {};
poolSizes[0].type = VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER;
poolSizes[0].descriptorCount = 1;
poolSizes[1].type = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER;
poolSizes[1].descriptorCount = 1;
VkDescriptorPoolCreateInfo descriptorPoolCreateInfo = {};
descriptorPoolCreateInfo.sType = VK_STRUCTURE_TYPE_DESCRIPTOR_POOL_CREATE_INFO;
descriptorPoolCreateInfo.maxSets = 2;
descriptorPoolCreateInfo.poolSizeCount = 2;
descriptorPoolCreateInfo.pPoolSizes = poolSizes;
vkCreateDescriptorPool(device, &descriptorPoolCreateInfo, nullptr, &descriptorPool);
}
VkDescriptorSet* descriptorSets = new VkDescriptorSet[2];
{
VkDescriptorSetAllocateInfo allocInfo = {};
allocInfo.sType = VK_STRUCTURE_TYPE_DESCRIPTOR_SET_ALLOCATE_INFO;
allocInfo.descriptorPool = descriptorPool;
allocInfo.descriptorSetCount = 2;
allocInfo.pSetLayouts = descriptorSetLayout;
vkAllocateDescriptorSets(device, &allocInfo, descriptorSets);
VkDescriptorBufferInfo bufferInfos[2];
bufferInfos[0].buffer = unifsBuffer;
bufferInfos[0].offset = 0;
bufferInfos[0].range = VK_WHOLE_SIZE;
bufferInfos[1].buffer = buffer;
bufferInfos[1].offset = 0;
bufferInfos[1].range = VK_WHOLE_SIZE;
VkWriteDescriptorSet writes[2] = {};
writes[0].sType = VK_STRUCTURE_TYPE_WRITE_DESCRIPTOR_SET;
writes[0].dstSet = descriptorSets[0];
writes[0].dstBinding = 0;
writes[0].dstArrayElement = 0;
writes[0].descriptorCount = 1;
writes[0].descriptorType = VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER;
writes[0].pBufferInfo = &bufferInfos[0];
writes[1].sType = VK_STRUCTURE_TYPE_WRITE_DESCRIPTOR_SET;
writes[1].dstSet = descriptorSets[1];
writes[1].dstBinding = 0;
writes[1].dstArrayElement = 0;
writes[1].descriptorCount = 1;
writes[1].descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER;
writes[1].pBufferInfo = &bufferInfos[1];
vkUpdateDescriptorSets(device, 2, writes, 0, nullptr);
}
VkCommandPool commandPool;
{
VkCommandPoolCreateInfo info = {};
info.sType = VK_STRUCTURE_TYPE_COMMAND_POOL_CREATE_INFO;
info.queueFamilyIndex = bestQueueFamilyInd;
vkCreateCommandPool(device, &info, nullptr, &commandPool);
}
VkCommandBuffer commandBuffer;
{
VkCommandBufferAllocateInfo allocInfo = {};
allocInfo.sType = VK_STRUCTURE_TYPE_COMMAND_BUFFER_ALLOCATE_INFO;
allocInfo.commandPool = commandPool;
allocInfo.level = VK_COMMAND_BUFFER_LEVEL_PRIMARY;
allocInfo.commandBufferCount = 1;
vkAllocateCommandBuffers(device, &allocInfo, &commandBuffer);
}
VkBufferMemoryBarrier* memBarriers = new VkBufferMemoryBarrier[2];
memBarriers[0] = {};
memBarriers[0].sType = VK_STRUCTURE_TYPE_BUFFER_MEMORY_BARRIER;
memBarriers[0].srcAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT;
memBarriers[0].dstAccessMask = VK_ACCESS_UNIFORM_READ_BIT;
memBarriers[0].srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
memBarriers[0].dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
memBarriers[0].buffer = unifsBuffer;
memBarriers[0].offset = 0;
memBarriers[0].size = VK_WHOLE_SIZE;
memBarriers[1] = {};
memBarriers[1].sType = VK_STRUCTURE_TYPE_BUFFER_MEMORY_BARRIER;
memBarriers[1].srcAccessMask = VK_ACCESS_SHADER_WRITE_BIT;
memBarriers[1].dstAccessMask = VK_ACCESS_TRANSFER_READ_BIT;
memBarriers[1].srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
memBarriers[1].dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
memBarriers[1].buffer = buffer;
memBarriers[1].offset = 0;
memBarriers[1].size = VK_WHOLE_SIZE;
VkCommandBufferBeginInfo beginInfo = {};
beginInfo.sType = VK_STRUCTURE_TYPE_COMMAND_BUFFER_BEGIN_INFO;
beginInfo.flags = VK_COMMAND_BUFFER_USAGE_ONE_TIME_SUBMIT_BIT;
beginInfo.pInheritanceInfo = nullptr;
const double startT = getTime();
i64 percent = 0;
for (i64 i = 0; i < numIterations; i++)
{
i64 start = i * perIteration;
i64 newPercent = (10000 * start) / N;
if (newPercent/1 > percent/1)
{
printf("%" PRId64 ".%" PRId64 "%%\n", newPercent / 100, newPercent % 100);
percent = newPercent;
calcElapsedAndPrintETA(startT, i, numIterations);
}
vkBeginCommandBuffer(commandBuffer, &beginInfo);
vkCmdBindPipeline(commandBuffer, VK_PIPELINE_BIND_POINT_COMPUTE, pipeline);
vkCmdBindDescriptorSets(commandBuffer, VK_PIPELINE_BIND_POINT_COMPUTE, pipelineLayout,
0, 2, descriptorSets, 0, nullptr);
uniforms.start = i * numThreads;
vkCmdUpdateBuffer(commandBuffer, unifsBuffer, 0, sizeof(uniforms), &uniforms);
vkCmdPipelineBarrier(commandBuffer,
VK_PIPELINE_STAGE_TRANSFER_BIT, VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT, 0,
0, nullptr,
1, &memBarriers[0],
0, nullptr);
vkCmdDispatch(commandBuffer, numThreads, 1, 1);
vkCmdPipelineBarrier(commandBuffer,
VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT, VK_PIPELINE_STAGE_TRANSFER_BIT, 0,
0, nullptr,
1, &memBarriers[1],
0, nullptr);
VkBufferCopy copyInfo = {};
copyInfo.srcOffset = 0;
copyInfo.dstOffset = 0;
copyInfo.size = sizeof(i64) * numThreads;
vkCmdCopyBuffer(commandBuffer,
buffer, stagingBuffer, 1, &copyInfo);
vkEndCommandBuffer(commandBuffer);
{
VkSubmitInfo info = {};
info.sType = VK_STRUCTURE_TYPE_SUBMIT_INFO;
info.commandBufferCount = 1;
info.pCommandBuffers = &commandBuffer;
vkQueueSubmit(queue, 1, &info, VK_NULL_HANDLE);
}
vkQueueWaitIdle(queue);
vkResetCommandPool(device, commandPool, VK_COMMAND_BUFFER_USAGE_ONE_TIME_SUBMIT_BIT);
i64* result;
vkMapMemory(device, stagingBufferMem, 0, sizeof(i64) * numThreads, 0, (void**)&result);
for (int i = 0; i < numThreads; i++)
{
if (result[i]) {
auto res = result[i];
vkUnmapMemory(device, stagingBufferMem);
return res;
}
}
vkUnmapMemory(device, stagingBufferMem);
}
return 0;
}
int main()
{
i64 res =
//calcWithCpu();
//calcWithOpenCl();
calcWithVulkan();
printf("RESULT: %" PRId64 "\n", res);
}
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment