[OPENMP][NVPTX]Improve performance + reduce number of used registers.

Summary:
Reduced number of the used register + improved performance propagating
the information about current execution/data sharing mode directly from
the compiler, where it is possible.
In some cases, it requires new/reworked interfaces of the runtime
external functions. Old functions are marked as deprecated.

Reviewers: grokos, gtbercea, kkwli0

Subscribers: guansong, jfb, openmp-commits, caomhin

Differential Revision: https://reviews.llvm.org/D56278

git-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@350405 91177308-0d34-0410-b5e6-96231b3b80d8

1 files changed, 28 insertions, 15 deletions

diff --git a/libomptarget/deviceRTLs/nvptx/src/libcall.cu b/libomptarget/deviceRTLs/nvptx/src/libcall.cu
index 91b270c..9abe599 100644
--- a/libomptarget/deviceRTLs/nvptx/src/libcall.cu
+++ b/libomptarget/deviceRTLs/nvptx/src/libcall.cu
@@ -39,14 +39,17 @@ EXTERN void omp_set_num_threads(int num) {
   if (num <= 0) {
     WARNING0(LW_INPUT, "expected positive num; ignore\n");
   } else {
-    omptarget_nvptx_TaskDescr *currTaskDescr = getMyTopTaskDescriptor();
+    omptarget_nvptx_TaskDescr *currTaskDescr =
+        getMyTopTaskDescriptor(/*isSPMDExecutionMode=*/false);
     currTaskDescr->NThreads() = num;
   }
 }
 
 EXTERN int omp_get_num_threads(void) {
-  int tid = GetLogicalThreadIdInBlock();
-  int rc = GetNumberOfOmpThreads(tid, isSPMDMode(), isRuntimeUninitialized());
+  bool isSPMDExecutionMode = isSPMDMode();
+  int tid = GetLogicalThreadIdInBlock(isSPMDExecutionMode);
+  int rc =
+      GetNumberOfOmpThreads(tid, isSPMDExecutionMode, isRuntimeUninitialized());
   PRINT(LD_IO, "call omp_get_num_threads() return %d\n", rc);
   return rc;
 }
@@ -58,7 +61,8 @@ EXTERN int omp_get_max_threads(void) {
     // We're already in parallel region.
     return 1;  // default is 1 thread avail
   }
-  omptarget_nvptx_TaskDescr *currTaskDescr = getMyTopTaskDescriptor();
+  omptarget_nvptx_TaskDescr *currTaskDescr =
+      getMyTopTaskDescriptor(isSPMDMode());
   int rc = 1; // default is 1 thread avail
   if (!currTaskDescr->InParallelRegion()) {
     // Not currently in a parallel region, return what was set.
@@ -76,21 +80,23 @@ EXTERN int omp_get_thread_limit(void) {
     return 0;  // default is 0
   }
   // per contention group.. meaning threads in current team
-  omptarget_nvptx_TaskDescr *currTaskDescr = getMyTopTaskDescriptor();
+  omptarget_nvptx_TaskDescr *currTaskDescr =
+      getMyTopTaskDescriptor(isSPMDMode());
   int rc = currTaskDescr->ThreadLimit();
   PRINT(LD_IO, "call omp_get_thread_limit() return %d\n", rc);
   return rc;
 }
 
 EXTERN int omp_get_thread_num() {
-  int tid = GetLogicalThreadIdInBlock();
-  int rc = GetOmpThreadId(tid, isSPMDMode(), isRuntimeUninitialized());
+  bool isSPMDExecutionMode = isSPMDMode();
+  int tid = GetLogicalThreadIdInBlock(isSPMDExecutionMode);
+  int rc = GetOmpThreadId(tid, isSPMDExecutionMode, isRuntimeUninitialized());
   PRINT(LD_IO, "call omp_get_thread_num() returns %d\n", rc);
   return rc;
 }
 
 EXTERN int omp_get_num_procs(void) {
-  int rc = GetNumberOfProcsInDevice();
+  int rc = GetNumberOfProcsInDevice(isSPMDMode());
   PRINT(LD_IO, "call omp_get_num_procs() returns %d\n", rc);
   return rc;
 }
@@ -102,7 +108,8 @@ EXTERN int omp_in_parallel(void) {
             "Expected SPMD mode only with uninitialized runtime.");
     rc = 1;  // SPMD mode is always in parallel.
   } else {
-    omptarget_nvptx_TaskDescr *currTaskDescr = getMyTopTaskDescriptor();
+    omptarget_nvptx_TaskDescr *currTaskDescr =
+        getMyTopTaskDescriptor(isSPMDMode());
     if (currTaskDescr->InParallelRegion()) {
       rc = 1;
     }
@@ -161,7 +168,8 @@ EXTERN int omp_get_level(void) {
     return omptarget_nvptx_simpleThreadPrivateContext->GetParallelLevel();
   }
   int level = 0;
-  omptarget_nvptx_TaskDescr *currTaskDescr = getMyTopTaskDescriptor();
+  omptarget_nvptx_TaskDescr *currTaskDescr =
+      getMyTopTaskDescriptor(isSPMDMode());
   ASSERT0(LT_FUSSY, currTaskDescr,
           "do not expect fct to be called in a non-active thread");
   do {
@@ -181,7 +189,8 @@ EXTERN int omp_get_active_level(void) {
     return 1;
   }
   int level = 0; // no active level parallelism
-  omptarget_nvptx_TaskDescr *currTaskDescr = getMyTopTaskDescriptor();
+  omptarget_nvptx_TaskDescr *currTaskDescr =
+      getMyTopTaskDescriptor(isSPMDMode());
   ASSERT0(LT_FUSSY, currTaskDescr,
           "do not expect fct to be called in a non-active thread");
   do {
@@ -208,7 +217,8 @@ EXTERN int omp_get_ancestor_thread_num(int level) {
   } else if (level > 0) {
     int totLevel = omp_get_level();
     if (level <= totLevel) {
-      omptarget_nvptx_TaskDescr *currTaskDescr = getMyTopTaskDescriptor();
+      omptarget_nvptx_TaskDescr *currTaskDescr =
+          getMyTopTaskDescriptor(isSPMDMode());
       int steps = totLevel - level;
       PRINT(LD_IO, "backtrack %d steps\n", steps);
       ASSERT0(LT_FUSSY, currTaskDescr,
@@ -259,7 +269,8 @@ EXTERN int omp_get_team_size(int level) {
   } else if (level > 0) {
     int totLevel = omp_get_level();
     if (level <= totLevel) {
-      omptarget_nvptx_TaskDescr *currTaskDescr = getMyTopTaskDescriptor();
+      omptarget_nvptx_TaskDescr *currTaskDescr =
+          getMyTopTaskDescriptor(isSPMDMode());
       int steps = totLevel - level;
       ASSERT0(LT_FUSSY, currTaskDescr,
               "do not expect fct to be called in a non-active thread");
@@ -288,7 +299,8 @@ EXTERN void omp_get_schedule(omp_sched_t *kind, int *modifier) {
     *kind = omp_sched_static;
     *modifier = 1;
   } else {
-    omptarget_nvptx_TaskDescr *currTaskDescr = getMyTopTaskDescriptor();
+    omptarget_nvptx_TaskDescr *currTaskDescr =
+        getMyTopTaskDescriptor(isSPMDMode());
     *kind = currTaskDescr->GetRuntimeSched();
     *modifier = currTaskDescr->RuntimeChunkSize();
   }
@@ -305,7 +317,8 @@ EXTERN void omp_set_schedule(omp_sched_t kind, int modifier) {
     return;
   }
   if (kind >= omp_sched_static && kind < omp_sched_auto) {
-    omptarget_nvptx_TaskDescr *currTaskDescr = getMyTopTaskDescriptor();
+    omptarget_nvptx_TaskDescr *currTaskDescr =
+        getMyTopTaskDescriptor(isSPMDMode());
     currTaskDescr->SetRuntimeSched(kind);
     currTaskDescr->RuntimeChunkSize() = modifier;
     PRINT(LD_IOD, "omp_set_schedule did set sched %d & modif %" PRIu64 "\n",