1.c.-OpenMP-directives.md

@@ -23,7 +23,7 @@ With teams, the code runs on GPU using multiple "teams".
 - Must be within "omp target"
 
 **distribute**  
-Iterations of the loop below are partitioned across "teams".
+Iterations of the loop below are partitioned across "teams".  
 Without, only thread 0 of each team would be used.
 
 - Loop partitioning (like omp do/far)
@@ -32,7 +32,8 @@ Without, only thread 0 of each team would be used.
 - Best practice to combine with teams (target teams distribute)
 
 **simd**  
-Optimize internal functions with SIMD instructions if possible. This directive gets ignored if this is not possible, thus it can always be used.  
+Optimize internal functions with SIMD instructions if possible.  
+This directive gets ignored if this is not possible, thus it can always be used.  
 Maps the work below to GPU threads within the "teams" blocks.  
 
 - Used for two-level GPU parallelism
@@ -47,16 +48,45 @@ Maps the work below to GPU threads within the "teams" blocks.
 - This will change for CCE16, where "parallel" will have the function that "simd" has in CCE15.
 
 **parallel do**  
-Maps the kernel to the available threads.
-Most usefull in combination with "teams" blocks.
+Maps the kernel to the available threads.  
+Most useful in combination with "teams" blocks.  
 Causes the work done in a loop inside a parallel region to be divided among threads.  
 
 **loop**  
-Needs to be bounded to the teams with bind(teams) or is done implicitely when called inside a "teams" region.
-Basically does a *parallel do simd* internally? But it is newer.
+Needs to be bounded to the teams with `bind(teams)` or is done implicitly when called inside a **teams** region.  
+Basically does a *parallel do simd* internally with some compiler optimizations.
 
 ## Memory directives.
 
+These directives can be used for invoking GPU memory operations for a single kernel.
+
+**map([\<var\>])**  
+Maps a variable to and from host.
+
+**map(from:[\<var\>])**  
+Maps a variable from host.
+
+**map(to:[\<var\>])**  
+Maps a variable to host after the kernel finsihes.
+
+**map(fromto:[\<var\>])**  
+Maps a variable from host and back after the kernel finishes.
+
+**map(alloc:[\<var\>])**  
+Allocate space for the variable on device.
+
+## Allocative memory operations
+
+There are also directives that you can use to allocate memory on the device and is accessible by the kernel.
+
+One can allocate memory on the device with:  
+```!$OMP target enter data map (\<specifier\>)```  
+Where the specifier is one of the operations listed in the section above.
+
+One can deallocate memory on the device (after the kernel finishes) with:  
+```!$OMP target exit data map (delete:<var>)```  
+where the data is deleted. If you want to copy over the data, you can use:  
+```!$OMP target exit data map (from:<var>)```
 
 ## Environment variables.