GPU Tutorial/SAXPY CUDA C - Revision history

Marius-neumann-d848@uni-bielefeld.de at 10:17, 3 January 2022

2022-01-03T10:17:55Z

2021-11-11T12:39:17Z

2021-11-11T10:45:15Z

2021-11-11T10:44:47Z

2021-11-11T10:42:04Z

2021-11-11T10:41:03Z

2021-11-11T10:21:03Z

2021-11-11T10:20:31Z

2021-11-11T09:52:48Z

2021-11-10T15:15:37Z

@@ Line 94: / Line 94: @@
 - You flag a line to be parallelized via keywords, e.g.: __device__ y=a*x+b
 || Wrong. These modifiers are used at function definitions.
 </quiz>
 {{hidden end}}

← Older revision		Revision as of 12:39, 11 November 2021
Line 5:		Line 5:

	This video discusses the SAXPY via NVIDIA CUDA C/C++.		This video discusses the SAXPY via NVIDIA CUDA C/C++.
		+	CUDA is an application programming interface (API) for NVIDIA GPUs. In general, CUDA works with many programming languages, but this tutorial is going to focus on C/C++. CUDA gives access to a GPUs instruction set, which means we have to go through everything step-by-step, since many things do not happen automatically.

	=== Video === <!--T:5-->		=== Video === <!--T:5-->

@@ Line 89: / Line 89: @@
 + Each thread has has an index attached to it, which is addressed via threadIdx.x
 || Correct
-- If you use array-element-wise operations, e.g.: y.=a.*x.+b . this is managed by the NVIDIA preprocessor.
+- If you use array-element-wise operations, e.g.: y.=a.*x.+b . This is managed by the NVIDIA preprocessor.
 || Wrong. There are no element-wise operators in C/C++
 - You flag a line to be parallelized via keywords, e.g.: __device__ y=a*x+b

@@ Line 68: / Line 68: @@
+{
 |type="()"}
-- MyKernel()
+- MyKernel();
 || Wrong. This would just execute an ordinary function.
-- CUDA.run(NoBlocks, NoThreads, MyKernel())
+- CUDA.run(NoBlocks, NoThreads, MyKernel());
 || Wrong. There is no CUDA.run()
-+ <<<NoBlocks, NoThreads>>>MyKernel()
++ <<<NoBlocks, NoThreads>>>MyKernel();
 || Correct
-- __global(NoBlocks, NoThreads)__ MyKernel()
+- __global(NoBlocks, NoThreads)__ MyKernel();
 || Wrong. __global__ and other modifiers cant have arguments and are part of a function definition, not launch.
 </quiz>
@@ Line 89: / Line 89: @@
 + Each thread has has an index attached to it, which is addressed via threadIdx.x
 || Correct
-- If you use array-element-wise operations (like y.=a.*x.+b ), this is managed by the NVIDIA preprocessor.
+- If you use array-element-wise operations, e.g.: y.=a.*x.+b . this is managed by the NVIDIA preprocessor.
 || Wrong. There are no element-wise operators in C/C++
 - You flag a line to be parallelized via keywords, e.g.: __device__ y=a*x+b

@@ Line 81: / Line 81: @@
 {{hidden begin
-|title = 5. Inside your kernel function, how do you distribute your data over the threads?}}
+|title = 5. Inside your kernel function, how do you distribute your data over the GPU threads?}}
 <quiz display=simple>
+{

@@ Line 128: / Line 128: @@
 {{hidden begin
-|title = 3. Why do we expect an onverhead in the GPU timings?}}
+|title = 3. Why do we expect an overhead in the GPU timings?}}
 <quiz display=simple>
+{

@@ Line 12: / Line 12: @@
 ([[Media:GPU_tutorial_saxpy_cuda_c.pdf |Slides as pdf]])
-=== Quiz === <!--T:5-->
+=== Introduction Quiz === <!--T:5-->
 {{hidden begin
-|title = 1. For which kind of programm can we expect improvements with GPUs?}}
+|title = 1. For which kind of program can we expect improvements with GPUs?}}
 <quiz display=simple>
+{
@@ Line 52: / Line 52: @@
 + All of the above
 || Correct!
 </quiz>
 {{hidden end}}

@@ Line 18: / Line 18: @@
+{
 |type="()"}
-+ serial programs
+- serial programs
 || Correct: CPU: optimized for low latency (strong single thread); GPU: optimized for throughput (massive parallelism)
-- parallel programs
++ parallel programs
 || Wrong: CPU: optimized for low latency (strong single thread); GPU: optimized for throughput (massive parallelism)
 </quiz>