1.c.-OpenMP-directives.md

@@ -9,7 +9,6 @@ It will only cover the GPU related options.
 NOTE: This wiki does not contain all information, please also read the [specification](https://www.openmp.org/specifications/).
 
 ## Parallel directives
-
 These directives can be used for invoking GPU parallelization.
 
 | Directive | Description |
@@ -28,20 +27,22 @@ Within the group of memory operations, some operate on a single kernel while oth
 ### Single kernel memory operations
 
 These directives can be used for invoking GPU memory operations for a single kernel.
+It is generally placed before a loop which is offloaded to the GPU.
 
 | Directive | Description |
 |:---------:|:------------|
 | map([\<var\>]) | Maps a variable from host and back after the kernel finishes. |
-| map(from:[\<var\>]) | Maps a variable from host. |
+| map(from:[\<var\>]) | Maps a variable from host, it's initial value is undefined. |
+| map(to:[\<var\>]) | Maps a variable to device, but not back. |
 | map(fromto:[\<var\>]) | Maps a variable from host and back after the kernel finishes. |
 | map(alloc:[\<var\>] | Allocate space for the variable on device, but do not copy anything. |
 
-### Allocative memory operations
+Note that the `from`, `to`, `fromto`, and `alloc` options are called the *specifiers*.
 
+### Allocative memory operations
 There are also directives that you can use to allocate memory on the device and is accessible by the kernel.
 You have the `target enter data map` and `target exit data map` structure, or the `target data map` and `end target data map` structure. The difference is that the latter is bounded to a lexical scope, and the first is not. Both will be described below.
 
-
 #### Data Map
 One can allocate memory on the device with:  
 `!$omp target data map (<specifier>:<var>)`