CoE Cluster April 2012 - Revision history

Pasa: /* CUDA Code Listing */

2012-05-02T07:11:07Z

‎CUDA Code Listing

2012-05-01T19:23:26Z

‎Results

2012-05-01T19:21:41Z

‎Results

2012-05-01T19:20:57Z

‎Results

2012-05-01T19:20:21Z

‎Results

2012-05-01T19:19:09Z

‎Results

2012-05-01T19:17:52Z

‎Results

2012-05-01T19:11:54Z

‎Results

2012-05-01T19:07:10Z

‎Results

2012-05-01T19:06:34Z

‎Results

@@ Line 35: / Line 35: @@
 // mandelbrot.cu
-// Calculate number of iterations required to make each point in Mandelbrot Set diverge and colour the corresponding pixel
+// Calculate number of iterations required to make each point
-// Tends to work faster with floats rather than doubles - but at the expense of "colour blocking" at lower resolutions
+// in Mandelbrot Set diverge and colour the corresponding pixel
 // Paul Saunders

@@ Line 17: / Line 17: @@
 *[http://mars.tekkom.dk/js/cuda_benchmark.htm CUDA Benchmark] (External JavaScript)
-Along with the CUDA C programming Guide (see Literature List), the [http://developer.download.nvidia.com/compute/cuda/CUDA_Occupancy_calculator.xls CUDA Occupancy Calculator] provides a good basis for understanding and explaining the results seen.  In general, it was found that more than 32 blocks and/or 128 threads per block did not provide noticable performance gains, whereas there were significant performance gains for every step increase below these values.  This [Media:cuda_presentation.ppt independent presentation] provides a comprehensive explanation and confirmation of the results we obtained.
+Along with the CUDA C programming Guide (see Literature List), the [http://developer.download.nvidia.com/compute/cuda/CUDA_Occupancy_calculator.xls CUDA Occupancy Calculator] provides a good basis for understanding and explaining the results seen.  In general, it was found that more than 32 blocks and/or 128 threads per block did not provide noticable performance gains, whereas there were significant performance gains for every step increase below these values.  This independent presentation: [[Media:cuda_presentation.ppt]] provides a comprehensive explanation and confirmation of the results we obtained.
 [[File:cuda_mandelbrot.png]]

@@ Line 17: / Line 17: @@
 *[http://mars.tekkom.dk/js/cuda_benchmark.htm CUDA Benchmark] (External JavaScript)
-Along with the CUDA C programming Guide (see Literature List), the [http://developer.download.nvidia.com/compute/cuda/CUDA_Occupancy_calculator.xls CUDA Occupancy Calculator] provides a good basis for understanding and explaining the results seen.  In general, it was found that more than 32 blocks and/or 128 threads per block did not provide noticable performance gains, whereas there were significant performance gains for every step increase below these values.  This [[independent presentation Media:cuda_presentation.ppt]] provides a comprehensive explanation and confirmation of the results we obtained.
+Along with the CUDA C programming Guide (see Literature List), the [http://developer.download.nvidia.com/compute/cuda/CUDA_Occupancy_calculator.xls CUDA Occupancy Calculator] provides a good basis for understanding and explaining the results seen.  In general, it was found that more than 32 blocks and/or 128 threads per block did not provide noticable performance gains, whereas there were significant performance gains for every step increase below these values.  This [Media:cuda_presentation.ppt independent presentation] provides a comprehensive explanation and confirmation of the results we obtained.
 [[File:cuda_mandelbrot.png]]

@@ Line 17: / Line 17: @@
 *[http://mars.tekkom.dk/js/cuda_benchmark.htm CUDA Benchmark] (External JavaScript)
-Along with the CUDA C programming Guide (see Literature List), the [http://developer.download.nvidia.com/compute/cuda/CUDA_Occupancy_calculator.xls CUDA Occupancy Calculator] provides a good basis for understanding and explaining the results seen.  In general, it was found that more than 32 blocks and/or 128 threads per block did not provide noticable performance gains, whereas there were significant performance gains for every step increase below these values.  This [[Media:cuda_presentation.ppt independent presentation]] provides a comprehensive explanation and confirmation of the results we obtained.
+Along with the CUDA C programming Guide (see Literature List), the [http://developer.download.nvidia.com/compute/cuda/CUDA_Occupancy_calculator.xls CUDA Occupancy Calculator] provides a good basis for understanding and explaining the results seen.  In general, it was found that more than 32 blocks and/or 128 threads per block did not provide noticable performance gains, whereas there were significant performance gains for every step increase below these values.  This [[independent presentation Media:cuda_presentation.ppt]] provides a comprehensive explanation and confirmation of the results we obtained.
 [[File:cuda_mandelbrot.png]]

@@ Line 17: / Line 17: @@
 *[http://mars.tekkom.dk/js/cuda_benchmark.htm CUDA Benchmark] (External JavaScript)
-Along with the CUDA C programming Guide (see Literature List), the [http://developer.download.nvidia.com/compute/cuda/CUDA_Occupancy_calculator.xls CUDA Occupancy Calculator] provides a good basis for understanding and explaining the results seen.  In general, it was found that more than 32 blocks and/or 128 threads per block did not provide noticable performance gains, whereas there were significant performance gains for every step increase below these values.  This [[Media:cuda_presentation.ppt]] provides a comprehensive explanation and confirmation of the results we obtained.
+Along with the CUDA C programming Guide (see Literature List), the [http://developer.download.nvidia.com/compute/cuda/CUDA_Occupancy_calculator.xls CUDA Occupancy Calculator] provides a good basis for understanding and explaining the results seen.  In general, it was found that more than 32 blocks and/or 128 threads per block did not provide noticable performance gains, whereas there were significant performance gains for every step increase below these values.  This [[Media:cuda_presentation.ppt independent presentation]] provides a comprehensive explanation and confirmation of the results we obtained.
 [[File:cuda_mandelbrot.png]]