Changes between Version 21 and Version 22 of OpenMPTransformation

Timestamp:

04/21/14 08:13:35 (12 years ago)

Author:

siegel

Comment:

--

OpenMPTransformation

@@ -48 +48 @@
 }}}
 
-== Worksharing state ==
+== Modeling shared variables ==
+
+For each shared variable `v` introduce a second variable `v_state`.  The type of `v_state` is obtained from the type of `v` by replacing all primitive types (leaf nodes in the type tree) by `int`.  Initially all these ints are -1.
+
+A write to (some part of) the shared variable by thread tid:
+* if the state value is -1, set it to tid, then do the write
+* if the state value is tid, do the write
+* else report a data race.
+
+A read from (some part of) the shared variable by thread tid:
+* if the state value is -1 or tid, do the read
+* else report a data race.
+
+Translating `flush`of (some part of) the shared variable by thread tid:
+* if the state value is -1: no-op
+* if the state value is tid: set it to -1
+* else: some other thread has some write to the variable which it hasn't flushed.  But this thread has not done any writes to that variable (since it last flushed).  This is a no-op.
+
+Function
+{{{
+  barrier_and_flush();
+}}}
+does a barrier on `_barrier` and a flush on all shared variables.
+
+
+== Modeling worksharing state ==
 
 The worksharing state will be stored in another handle type object.  The situation here is analogous to the `$gcomm` and `$comm` use for MPI.  Those objects store the shared state for message-passing.    We need similar object for shared state the coordinates work-sharing and barrier constructs:
@@ -99 +124 @@
 
 
-== Translation Strategies ==
-
-
-=== Translating shared variables ===
-
-For each shared variable `v` introduce a second variable `v_state`.  The type of `v_state` is obtained from the type of `v` by replacing all primitive types (leaf nodes in the type tree) by `int`.  Initially all these ints are -1.
-
-A write to (some part of) the shared variable by thread tid:
-* if the state value is -1, set it to tid, then do the write
-* if the state value is tid, do the write
-* else report a data race.
-
-A read from (some part of) the shared variable by thread tid:
-* if the state value is -1 or tid, do the read
-* else report a data race.
-
-Translating `flush`of (some part of) the shared variable by thread tid:
-* if the state value is -1: no-op
-* if the state value is tid: set it to -1
-* else: some other thread has some write to the variable which it hasn't flushed.  But this thread has not done any writes to that variable (since it last flushed).  This is a no-op.
-
-Function
-{{{
-  barrier_and_flush();
-}}}
-does a barrier on `_barrier` and a flush on all shared variables.
+== Translations of specific directives ==
 
 === Translating `parallel` ===
@@ -187 +187 @@
 }
 }}}
+
+We vary the way the way iterators are chosen to explore different tradeoffs and strategies.  On one extreme, every kind of partition can be explored; on the other, some fixed strategy like round-robin with chunksize 1 can be used.  This only changes the definition of `$omp_ws_arrive_loop`, not the translation above.
 
 === Translating `sections` ===
@@ -350 +352 @@
 * `omp_get_num_threads()` => `_nthreads`
 * `omp_get_thread_num()` => `_tid`
-