get slow but correct results for intermediate

.gitignore

@@ -43,3 +43,4 @@ include/divisible/*
 build/
 .cache/
 .vscode/
+assignment/

CMakeLists.txt

@@ -34,6 +34,7 @@ elseif (WIN32)
         set(RENDERDOC_PATH "C:\\Program Files\\RenderDoc")
     endif()
 else ()
+    set(RENDERDOC_PATH "/usr/lib/")
     #LINUX PATH HERE
 endif ()
 

shaders/A2Task1Interleaved.comp

@@ -12,10 +12,12 @@ layout(local_size_x_id = 0, local_size_y = 1, local_size_z = 1) in;
 layout(push_constant) uniform PushStruct {
     uint size;
     uint stride;
-} p;
+}p;
 
 layout(binding = 0) buffer inoutBufer {uint v[];};
 
 void main() {
-    // TODO: Kernel implementation
+    uint idx = gl_GlobalInvocationID.x;
+    
+    v[idx] += v[idx + p.stride];
 }

shaders/A2Task1Sequential.comp

@@ -16,6 +16,11 @@ layout(push_constant) uniform PushStruct {
 
 layout(binding = 0) buffer inoutBufer { uint v[]; };
 
-void main() {
+void main() {    
-    // TODO: Kernel implementation
+    uint gIDx = gl_GlobalInvocationID.x;
+    if (gIDx >= p.size) {
+        return;
+    }
+
+    v[gIDx] = v[gIDx] + v[gIDx + p.offset];
 }

shaders/A2Task2Naive.comp

@@ -19,5 +19,4 @@ layout(binding = 0) buffer inBuffer { uint v[]; };
 layout(binding = 1) buffer outBufer { uint g_v[]; };
 
 void main() {
-    // TODO: Kernel implementation
 }

src/A2Task1Solution/Interleaved.cpp

@@ -49,6 +49,25 @@ void A2Task1SolutionInterleaved::compute()
 
     // TODO: Implement reduction with interleaved addressing
 
+    cb.bindPipeline(vk::PipelineBindPoint::eCompute, pipeline);
+    cb.bindDescriptorSets(vk::PipelineBindPoint::eCompute, pipelineLayout, 0, 1, &descriptorSet, 0, nullptr);
+
+    uint inputSize = mpInput->size();
+    PushConstant pc;
+    pc.size = inputSize;
+
+    for (uint stride = inputSize / 2; stride > 0; stride >>= 1)
+    {
+        pc.stride = stride;
+        cb.pushConstants(pipelineLayout, vk::ShaderStageFlagBits::eCompute, 0, sizeof(PushConstant), &pc);
+
+        cb.dispatch((stride + workGroupSize - 1) / workGroupSize, 1, 1);
+
+        vk::MemoryBarrier memoryBarrier(vk::AccessFlagBits::eShaderWrite, vk::AccessFlagBits::eShaderRead);
+        cb.pipelineBarrier(vk::PipelineStageFlagBits::eComputeShader, vk::PipelineStageFlagBits::eComputeShader,
+                           vk::DependencyFlags(), 1, &memoryBarrier, 0, nullptr, 0, nullptr);
+    }
+
     cb.end();
 
     vk::SubmitInfo submitInfo = vk::SubmitInfo(0, nullptr, nullptr, 1, &cb);

src/A2Task1Solution/Sequential.cpp

@@ -1,6 +1,7 @@
 #include "Sequential.h"
 
 #include "host_timer.h"
+#include <iostream>
 
 A2Task1SolutionSequential::A2Task1SolutionSequential(AppResources &app, uint workGroupSize) :
     app(app), workGroupSize(workGroupSize) {}
@@ -47,7 +48,27 @@ void A2Task1SolutionSequential::compute()
 
     cb.begin(beginInfo);
 
-    // TODO: Implement reduction with sequential addressing
+    cb.resetQueryPool(app.queryPool, 0, 1);
+   
+    cb.bindPipeline(vk::PipelineBindPoint::eCompute, pipeline);
+    cb.bindDescriptorSets(vk::PipelineBindPoint::eCompute, pipelineLayout, 0, 1, &descriptorSet, 0, nullptr);
+
+    uint inputSize = mpInput->size();
+    PushConstant pc;
+    pc.size = inputSize;
+    pc.offset = inputSize / 2;
+
+    for (; pc.offset != 0 ; pc.offset = pc.offset / 2)
+    {
+        cb.pushConstants(pipelineLayout, vk::ShaderStageFlagBits::eCompute, 0, sizeof(PushConstant), &pc);
+
+        uint groupCount = (pc.size + pc.offset + workGroupSize - 1) / workGroupSize;
+        cb.dispatch(groupCount, 1, 1);
+
+        vk::MemoryBarrier memoryBarrier(vk::AccessFlagBits::eShaderWrite, vk::AccessFlagBits::eShaderRead);
+        cb.pipelineBarrier(vk::PipelineStageFlagBits::eComputeShader, vk::PipelineStageFlagBits::eComputeShader,
+                           vk::DependencyFlags(), 1, &memoryBarrier, 0, nullptr, 0, nullptr);
+    }
 
     cb.end();
 

src/main.cpp

@@ -45,14 +45,14 @@ void run_A2_task1(AppResources &app){
     A2Task1SolutionSequential sequentialSolution(app, 128);
     evaluateTask1Solution(&sequentialSolution, "Sequential");
 
-    A2Task1SolutionKernelDecomposition kernelDecompositionSolution(app, 128, workingDir +"build/shaders/A2Task1KernelDecomposition.comp.spv");
+    // A2Task1SolutionKernelDecomposition kernelDecompositionSolution(app, 128, workingDir +"build/shaders/A2Task1KernelDecomposition.comp.spv");
-    evaluateTask1Solution(&kernelDecompositionSolution, "KernelDecomposition");
+    // evaluateTask1Solution(&kernelDecompositionSolution, "KernelDecomposition");
 
-    A2Task1SolutionKernelDecomposition kernelDecompositionUnrollSolution(app, 128, workingDir +"build/shaders/A2Task1KernelDecompositionUnroll.comp.spv");
+    // A2Task1SolutionKernelDecomposition kernelDecompositionUnrollSolution(app, 128, workingDir +"build/shaders/A2Task1KernelDecompositionUnroll.comp.spv");
-    evaluateTask1Solution(&kernelDecompositionUnrollSolution, "KernelDecomposition Unroll");
+    // evaluateTask1Solution(&kernelDecompositionUnrollSolution, "KernelDecomposition Unroll");
 
-    A2Task1SolutionKernelDecomposition kernelDecompositionAtomicSolution(app, 128, workingDir +"build/shaders/A2Task1KernelDecompositionAtomic.comp.spv");
+    // A2Task1SolutionKernelDecomposition kernelDecompositionAtomicSolution(app, 128, workingDir +"build/shaders/A2Task1KernelDecompositionAtomic.comp.spv");
-    evaluateTask1Solution(&kernelDecompositionAtomicSolution, "KernelDecomposition Atomic");
+    // evaluateTask1Solution(&kernelDecompositionAtomicSolution, "KernelDecomposition Atomic");
 }
 void run_A2_task2(AppResources& app){
     
@@ -111,7 +111,7 @@ int main()
         
         run_A2_task1(app);
        
-        run_A2_task2(app);
+        // run_A2_task2(app);
 
         renderdoc::endCapture();