Add basic support to the harness for parallel test execution (#1687)

This change introduces a new command-line parameter to enable
parallel execution by a specified number of worker threads. When
parallel execution is requested, tests are distributed across
the worker threads. This behaviour is disabled by default.

This does not currently work for all suites as some of them are
using global variables to configure tests. For the suites that
do not use global state, this change reduced the execution time
by up to 5x on an 8-core machine.

Signed-off-by: Kévin Petit <kpet@free.fr>

test_common/harness/parseParameters.cpp

@@ -36,6 +36,7 @@ std::string gCompilationCachePath = ".";
 std::string gCompilationProgram = DEFAULT_COMPILATION_PROGRAM;
 bool gDisableSPIRVValidation = false;
 std::string gSPIRVValidator = DEFAULT_SPIRV_VALIDATOR;
+unsigned gNumWorkerThreads;
 
 void helpInfo()
 {
@@ -48,6 +49,8 @@ void helpInfo()
             online     Use online compilation (default)
             binary     Use binary offline compilation
             spir-v     Use SPIR-V offline compilation
+    --num-worker-threads <num>
+        Select parallel execution with the specified number of worker threads.
 
 For offline compilation (binary and spir-v modes) only:
     --compilation-cache-mode <cache-mode>
@@ -137,6 +140,23 @@ int parseCustomParam(int argc, const char *argv[], const char *ignore)
                 return -1;
             }
         }
+        else if (!strcmp(argv[i], "--num-worker-threads"))
+        {
+            delArg++;
+            if ((i + 1) < argc)
+            {
+                delArg++;
+                const char *numthstr = argv[i + 1];
+
+                gNumWorkerThreads = atoi(numthstr);
+            }
+            else
+            {
+                log_error(
+                    "A parameter to --num-worker-threads must be provided!\n");
+                return -1;
+            }
+        }
         else if (!strcmp(argv[i], "--compilation-cache-mode"))
         {
             delArg++;