Move totaliterations calculation to CPU. This frees 4 VGPRs.

Now the code is back to the limit of 32 VGPRs without the need to directly operate on the state buffer.
This means that now it's bound by memory access.

2 files changed, 5 insertions, 5 deletions

diff --git a/BuddhaTest/Shaders/BuddhaCompute.glsl b/BuddhaTest/Shaders/BuddhaCompute.glsl
index 10dfe23..7cd4efe 100644
--- a/BuddhaTest/Shaders/BuddhaCompute.glsl
+++ b/BuddhaTest/Shaders/BuddhaCompute.glsl
@@ -34,6 +34,7 @@ uniform uint height;
 uniform uvec4 orbitLength;
 
 uniform uint iterationsPerDispatch;
+uniform uint totalIterations;
 
 void addToColorOfCell(uvec2 cell, uvec3 toAdd)
 {
@@ -202,15 +203,11 @@ vec2 getCurrentOrbitOffset(const uint orbitNumber, const uint totalWorkers, cons
 
 void main() {
     //we need to know how many total work groups are running this iteration
-
     const uvec3 totalWorkersPerDimension = gl_WorkGroupSize * gl_NumWorkGroups;
     const uint totalWorkers = totalWorkersPerDimension.x*totalWorkersPerDimension.y*totalWorkersPerDimension.z;
 
     const uint uniqueWorkerID = gl_GlobalInvocationID.x + gl_GlobalInvocationID.y*totalWorkersPerDimension.x + gl_GlobalInvocationID.z*(totalWorkersPerDimension.x * totalWorkersPerDimension.y);
 
-    const uint _totalIterations = orbitLength.x > orbitLength.y ? orbitLength.x : orbitLength.y;
-    const uint totalIterations = _totalIterations > orbitLength.z ? _totalIterations : orbitLength.z;
-
     individualData state = stateArray[uniqueWorkerID];
 
     //getIndividualState(in uint CellID, out vec2 offset, out vec2 coordinates, out uint phase, out uint orbitNumber, out uint doneIterations)
diff --git a/BuddhaTest/src/BuddhaTest.cpp b/BuddhaTest/src/BuddhaTest.cpp
index 3b99169..0a99fee 100644
--- a/BuddhaTest/src/BuddhaTest.cpp
+++ b/BuddhaTest/src/BuddhaTest.cpp
@@ -148,6 +148,7 @@ int main(int argc, char * argv[])
 
     glUseProgram(ComputeShader);
     GLint orbitLengthUniformHandle = glGetUniformLocation(ComputeShader, "orbitLength");
+    GLint totalIterationsUniformHandle = glGetUniformLocation(ComputeShader, "totalIterations");
     GLint widthUniformComputeHandle = glGetUniformLocation(ComputeShader, "width");
     GLint heightUniformComputeHandle = glGetUniformLocation(ComputeShader, "height");
     GLint iterationsPerDispatchHandle = glGetUniformLocation(ComputeShader, "iterationsPerDispatch");
@@ -155,6 +156,9 @@ int main(int argc, char * argv[])
     glUniform1ui(widthUniformComputeHandle, settings.imageWidth);
     glUniform1ui(heightUniformComputeHandle, bufferHeight);
 
+    const uint32_t maxOrbitlength = std::max(std::max(settings.orbitLengthBlue,settings.orbitLengthGreen),settings.orbitLengthRed);
+    glUniform1ui(totalIterationsUniformHandle, maxOrbitlength);
+
     glUseProgram(VertexAndFragmentShaders);
     GLint widthUniformFragmentHandle = glGetUniformLocation(VertexAndFragmentShaders, "width");
     GLint heightUniformFragmentHandle = glGetUniformLocation(VertexAndFragmentShaders, "height");
@@ -169,7 +173,6 @@ int main(int argc, char * argv[])
 
     uint64_t totalIterationCount{0};
     uint64_t lastMessage{0};
-    const uint64_t maxOrbitlength = std::max(std::max(settings.orbitLengthBlue,settings.orbitLengthGreen),settings.orbitLengthRed);
 
     /* Loop until the user closes the window */
     while (!glfwWindowShouldClose(window))