slight API change

Author: devshgraphicsprogramming

examples_tests/49.ComputeFFT/fft_convolve_ifft.comp

@@ -26,7 +26,7 @@ void convolve(in uint item_per_thread_count, in uint ch)
 	for(uint t=0u; t<item_per_thread_count; t++)
 	{
 		uint tid = gl_LocalInvocationIndex + t * _NBL_GLSL_WORKGROUP_SIZE_;
-		// TODO: refactor
+		// TODO: refactor for smem usage
 		uvec3 coords = nbl_glsl_ext_FFT_getCoordinates(tid);
 		const uvec3 log2_size = uvec3(11u, 10u, 0u);
         coords = bitfieldReverse(coords)>>(uvec3(32u)-log2_size);
@@ -42,8 +42,9 @@ void convolve(in uint item_per_thread_count, in uint ch)
 
 void main()
 {
-    const uint dataLength = nbl_glsl_ext_FFT_Parameters_t_getFFTLength();
-    const uint item_per_thread_count = dataLength>>_NBL_GLSL_WORKGROUP_SIZE_LOG2_;
+    const uint log2FFTSize = nbl_glsl_ext_FFT_Parameters_t_getLog2FFTSize();
+    const uint item_per_thread_count = 0x1u<<(log2FFTSize-_NBL_GLSL_WORKGROUP_SIZE_LOG2_);
+
     const uint numChannels = nbl_glsl_ext_FFT_Parameters_t_getNumChannels();
 	for(uint ch = 0u; ch < numChannels; ++ch)
 	{
@@ -53,13 +54,13 @@ void main()
 			const uint tid = (t<<_NBL_GLSL_WORKGROUP_SIZE_LOG2_)|gl_LocalInvocationIndex;
 			nbl_glsl_ext_FFT_impl_values[t] = nbl_glsl_ext_FFT_getPaddedData(nbl_glsl_ext_FFT_getCoordinates(tid),ch);
 		}
-		nbl_glsl_ext_FFT_preloaded(false,dataLength);
+		nbl_glsl_ext_FFT_preloaded(false,log2FFTSize);
 		barrier();
 
 		convolve(item_per_thread_count,ch);
 
 		barrier();
-		nbl_glsl_ext_FFT_preloaded(true,dataLength);
+		nbl_glsl_ext_FFT_preloaded(true,log2FFTSize);
 		// write out to main memory
 		for(uint t=0u; t<item_per_thread_count; t++)
 		{

include/nbl/builtin/glsl/ext/FFT/fft.glsl

@@ -67,9 +67,10 @@ void nbl_glsl_ext_FFT_loop(in bool is_inverse, in uint virtual_thread_count, in
     }
 }
 
-void nbl_glsl_ext_FFT_preloaded(bool is_inverse, in uint dataLength)
+void nbl_glsl_ext_FFT_preloaded(bool is_inverse, in uint log2FFTSize)
 {
     // Virtual Threads Calculation
+    const uint dataLength = 1u<<log2FFTSize;
     const uint halfDataLength = dataLength>>1u;
     const uint virtual_thread_count = halfDataLength>>_NBL_GLSL_WORKGROUP_SIZE_LOG2_;
 
@@ -98,8 +99,8 @@ void nbl_glsl_ext_FFT_preloaded(bool is_inverse, in uint dataLength)
 void nbl_glsl_ext_FFT(bool is_inverse, uint channel)
 {
     // Virtual Threads Calculation
-    const uint dataLength = nbl_glsl_ext_FFT_Parameters_t_getFFTLength();
-    const uint item_per_thread_count = dataLength>>_NBL_GLSL_WORKGROUP_SIZE_LOG2_;
+    const uint log2FFTSize = nbl_glsl_ext_FFT_Parameters_t_getLog2FFTSize();
+    const uint item_per_thread_count = 0x1u<<(log2FFTSize-_NBL_GLSL_WORKGROUP_SIZE_LOG2_);
 
     // Load Values into local memory
     for(uint t=0u; t<item_per_thread_count; t++)
@@ -108,7 +109,7 @@ void nbl_glsl_ext_FFT(bool is_inverse, uint channel)
         nbl_glsl_ext_FFT_impl_values[t] = nbl_glsl_ext_FFT_getPaddedData(nbl_glsl_ext_FFT_getCoordinates(tid),channel);
     }
     // do FFT
-    nbl_glsl_ext_FFT_preloaded(is_inverse,dataLength);
+    nbl_glsl_ext_FFT_preloaded(is_inverse,log2FFTSize);
     // write out to main memory
     for(uint t=0u; t<item_per_thread_count; t++)
     {

include/nbl/builtin/glsl/ext/FFT/parameters.glsl

@@ -38,6 +38,10 @@ uint nbl_glsl_ext_FFT_Parameters_t_getFFTLength() {
     const uint direction = nbl_glsl_ext_FFT_Parameters_t_getDirection();
     return nbl_glsl_ext_FFT_Parameters_t_getPaddedDimensions()[direction];
 }
+uint nbl_glsl_ext_FFT_Parameters_t_getLog2FFTSize()
+{
+    return findMSB(nbl_glsl_ext_FFT_Parameters_t_getFFTLength());
+}
 
 bool nbl_glsl_ext_FFT_Parameters_t_getIsInverse() {
     nbl_glsl_ext_FFT_Parameters_t params = nbl_glsl_ext_FFT_getParameters();