From 1884daea1f6db7269be074a6dd436227ae4fa370 Mon Sep 17 00:00:00 2001
From: Arif Hasanic <arif@hasanic.com>
Date: Sat, 27 Dec 2025 22:01:42 +0100
Subject: [PATCH] stop being dumb dumb

---
 shaders/A2Task1Interleaved.comp     | 8 ++++++--
 shaders/A2Task1Sequential.comp      | 5 +++--
 src/A2Task1Solution/Interleaved.cpp | 4 ++--
 src/A2Task1Solution/Sequential.cpp  | 2 +-
 4 files changed, 12 insertions(+), 7 deletions(-)

diff --git a/shaders/A2Task1Interleaved.comp b/shaders/A2Task1Interleaved.comp
index cc8be56..fa4c372 100644
--- a/shaders/A2Task1Interleaved.comp
+++ b/shaders/A2Task1Interleaved.comp
@@ -17,7 +17,11 @@ layout(push_constant) uniform PushStruct {
 layout(binding = 0) buffer inoutBufer {uint v[];};
 
 void main() {
-    uint idx = gl_GlobalInvocationID.x;
-    
+    uint idx = gl_GlobalInvocationID.x * 2 * p.stride;
+
+    if (idx + p.stride >= p.size) {
+        return;
+    }
+
     v[idx] += v[idx + p.stride];
 }
diff --git a/shaders/A2Task1Sequential.comp b/shaders/A2Task1Sequential.comp
index 6b33125..afcb285 100644
--- a/shaders/A2Task1Sequential.comp
+++ b/shaders/A2Task1Sequential.comp
@@ -18,9 +18,10 @@ layout(binding = 0) buffer inoutBufer { uint v[]; };
 
 void main() {    
     uint gIDx = gl_GlobalInvocationID.x;
-    if (gIDx >= p.size) {
+    
+    if (gIDx + p.offset >= p.size) {
         return;
     }
 
-    v[gIDx] = v[gIDx] + v[gIDx + p.offset];
+    v[gIDx] += v[gIDx + p.offset];
 }
diff --git a/src/A2Task1Solution/Interleaved.cpp b/src/A2Task1Solution/Interleaved.cpp
index 88d2ad7..7b963a4 100644
--- a/src/A2Task1Solution/Interleaved.cpp
+++ b/src/A2Task1Solution/Interleaved.cpp
@@ -56,12 +56,12 @@ void A2Task1SolutionInterleaved::compute()
     PushConstant pc;
     pc.size = inputSize;
 
-    for (uint stride = inputSize / 2; stride > 0; stride >>= 1)
+    for (uint stride = 1; stride < inputSize; stride <<= 1)
     {
         pc.stride = stride;
         cb.pushConstants(pipelineLayout, vk::ShaderStageFlagBits::eCompute, 0, sizeof(PushConstant), &pc);
 
-        cb.dispatch((stride + workGroupSize - 1) / workGroupSize, 1, 1);
+        cb.dispatch((inputSize / (stride * 2) + workGroupSize - 1) / workGroupSize, 1, 1);
 
         vk::MemoryBarrier memoryBarrier(vk::AccessFlagBits::eShaderWrite, vk::AccessFlagBits::eShaderRead);
         cb.pipelineBarrier(vk::PipelineStageFlagBits::eComputeShader, vk::PipelineStageFlagBits::eComputeShader,
diff --git a/src/A2Task1Solution/Sequential.cpp b/src/A2Task1Solution/Sequential.cpp
index 0016f54..f204433 100644
--- a/src/A2Task1Solution/Sequential.cpp
+++ b/src/A2Task1Solution/Sequential.cpp
@@ -62,7 +62,7 @@ void A2Task1SolutionSequential::compute()
     {
         cb.pushConstants(pipelineLayout, vk::ShaderStageFlagBits::eCompute, 0, sizeof(PushConstant), &pc);
 
-        uint groupCount = (pc.size + pc.offset + workGroupSize - 1) / workGroupSize;
+        uint groupCount = (pc.offset + workGroupSize - 1) / workGroupSize;
         cb.dispatch(groupCount, 1, 1);
 
         vk::MemoryBarrier memoryBarrier(vk::AccessFlagBits::eShaderWrite, vk::AccessFlagBits::eShaderRead);