Handle high levels of concurrency in rocm_agent_enumerator.

rocm_agent_enumerator may invoke rocminfo.  Rocminfo opens the
GPU device which allocates limited resource.  Beyond 254
concurrent processes this resource will be exhausted and rocminfo
will return an error.

This patch loops rocm_agent_enumerator when recieving a failure
message from rocminfo indicating KFD is out of memory.

Change-Id: I8637e214f5fa012642975c28578ae6bf9200eda8


[ROCm/rocminfo commit: b57c02d131]

projects/rocminfo/rocm_agent_enumerator

@@ -4,6 +4,7 @@ import os
 import re
 import subprocess
 import sys
+import time
 
 # get current working directory
 CWD = os.path.dirname(os.path.realpath(__file__))
@@ -125,8 +126,24 @@ def readFromROCMINFO(search_arch_name = False):
   rocminfo_executable = os.path.join(CWD, "rocminfo")
 
   try:
-    # run rocminfo
-    rocminfo_output = subprocess.Popen(rocminfo_executable, stdout=subprocess.PIPE).communicate()[0].decode("utf-8").split('\n')
+    t0 = time.time()
+    while 1:
+      t1 = time.time()
+      # quit after retrying rocminfo for a minute.
+      if t1 - t0 > 60.0:
+        print("Timeout querying rocminfo.  Are you compiling with more than 254 threads?")
+        break
+      # run rocminfo
+      rocminfo_output = subprocess.Popen(rocminfo_executable, stdout=subprocess.PIPE).communicate()[0].decode("utf-8").split('\n')
+      term1 = re.compile("Cannot allocate memory")
+      term2 = re.compile("HSA_STATUS_ERROR_OUT_OF_RESOURCES")
+      done = 1
+      for line in rocminfo_output:
+        if term1.search(line) is not None or term2.search(line) is not None:
+          done = 0
+          break
+      if done:
+        break
   except:
     rocminfo_output = []