Adding busy cycles.

4 files changed, 55 insertions, 11 deletions

diff --git a/CMakeLists.txt b/CMakeLists.txt
index 3da4838..47ab5c7 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -37,9 +37,9 @@ add_library(Timer src/Timer.h src/Timer.cpp)
 
 add_library(Types src/Types.h src/Types.cpp)
 
-add_executable (pChase src/Main.h src/Main.cpp)
-target_link_libraries(pChase Run Timer Output Experiment SpinBarrier)
-target_link_libraries(pChase ${CMAKE_THREAD_LIBS_INIT})
+add_executable (chase src/Main.h src/Main.cpp)
+target_link_libraries(chase Run Timer Output Experiment SpinBarrier)
+target_link_libraries(chase ${CMAKE_THREAD_LIBS_INIT})
 if (USE_LIBNUMA)
 	if(LIBNUMA)
 		target_link_libraries(pChase ${LIBNUMA})
diff --git a/src/Experiment.cpp b/src/Experiment.cpp
index 75b1cab..04fb9ef 100644
--- a/src/Experiment.cpp
+++ b/src/Experiment.cpp
@@ -37,6 +37,7 @@ Experiment::Experiment() :
     bytes_per_thread (DEFAULT_BYTES_PER_THREAD),
     num_threads      (DEFAULT_THREADS), 
     bytes_per_test   (DEFAULT_BYTES_PER_TEST),
+    busy_cycles      (DEFAULT_BUSY_CYCLES),
     seconds          (DEFAULT_SECONDS),
     iterations       (DEFAULT_ITERATIONS),
     experiments      (DEFAULT_EXPERIMENTS), 
@@ -66,6 +67,7 @@ Experiment::~Experiment()
 				// -t or --threads          number of threads (concurrency and contention)
 				// -i or --iters            iterations
 				// -e or --experiments      experiments
+				// -b or --busy				amount of cycles processor should remain busy
 				// -a or --access           memory access pattern
 				//         random           random access pattern
 				//         forward <stride> exclusive OR and mask
@@ -130,6 +132,11 @@ Experiment::parse_args(int argc, char* argv[])
 	    if (i == argc) { error = 1; break; }
 	    this->experiments = Experiment::parse_number(argv[i]);
 	    if (this->experiments == 0) { error = 1; break; }
+	} else if (strcasecmp(argv[i], "-b") == 0 || strcasecmp(argv[i], "--busy") == 0) {
+	    i++;
+	    if (i == argc) { error = 1; break; }
+	    this->busy_cycles = Experiment::parse_number(argv[i]);
+	    if (this->experiments == 0) { error = 1; break; }
 	} else if (strcasecmp(argv[i], "-a") == 0 || strcasecmp(argv[i], "--access") == 0) {
 	    i++;
 	    if (i == argc) { error = 1; break; }
@@ -221,6 +228,7 @@ Experiment::parse_args(int argc, char* argv[])
 	printf("    [-o|--output]      <format>    # output format\n");
 	printf("    [-n|--numa]        <placement> # numa placement\n");
 	printf("    [-s|--seconds]     <number>    # run each experiment for <number> seconds\n");
+	printf("    [-b|--busy]        <number>    # how much processing cycles each loop should count\n");
 	printf("    [-x|--strict]                  # fail rather than adjust options to sensible values\n");
 	printf("\n");
 	printf("<pattern> is selected from the following:\n");
diff --git a/src/Experiment.h b/src/Experiment.h
index 2c749d3..4c0ae90 100644
--- a/src/Experiment.h
+++ b/src/Experiment.h
@@ -40,9 +40,10 @@ public:
     int64 links_per_chain;	// working set chain size (links)
     int64 pages_per_chain;	// working set chain size (pages)
     int64 bytes_per_thread;	// thread working set size (bytes)
-    int64 chains_per_thread;	// memory loading per thread
+    int64 chains_per_thread;// memory loading per thread
     int64 num_threads;		// number of threads in the experiment
     int64 bytes_per_test;	// test working set size (bytes)
+    int64 busy_cycles;		// processing cycles
 
     float seconds;		// number of seconds per experiment
     int64 iterations;		// number of iterations per experiment
@@ -84,6 +85,7 @@ public:
     const static int32 DEFAULT_BYTES_PER_THREAD  = DEFAULT_BYTES_PER_CHAIN * DEFAULT_CHAINS_PER_THREAD;
     const static int32 DEFAULT_THREADS           = 1;
     const static int32 DEFAULT_BYTES_PER_TEST    = DEFAULT_BYTES_PER_THREAD * DEFAULT_THREADS;
+    const static int32 DEFAULT_BUSY_CYCLES       = 10;
     const static int32 DEFAULT_SECONDS           = 1;
     const static int32 DEFAULT_ITERATIONS        = 0;
     const static int32 DEFAULT_EXPERIMENTS       = 1;
diff --git a/src/Run.cpp b/src/Run.cpp
index 4fb8057..1ea4db3 100644
--- a/src/Run.cpp
+++ b/src/Run.cpp
@@ -27,9 +27,9 @@
 
 static double max( double v1, double v2 );
 static double min( double v1, double v2 );
-static void chase_pointers(int64 chains_per_thread, int64 iterations, Chain** root, int64 bytes_per_line, int64 bytes_per_chain, int64 stride);
-static void follow_streams(int64 chains_per_thread, int64 iterations, Chain** root, int64 bytes_per_line, int64 bytes_per_chain, int64 stride);
-static void (*run_benchmark)(int64 chains_per_thread, int64 iterations, Chain** root, int64 bytes_per_line, int64 bytes_per_chain, int64 stride) = chase_pointers;
+static void chase_pointers(int64 chains_per_thread, int64 iterations, Chain** root, int64 bytes_per_line, int64 bytes_per_chain, int64 stride, int64 busy_cycles);
+static void follow_streams(int64 chains_per_thread, int64 iterations, Chain** root, int64 bytes_per_line, int64 bytes_per_chain, int64 stride, int64 busy_cycles);
+static void (*run_benchmark)(int64 chains_per_thread, int64 iterations, Chain** root, int64 bytes_per_line, int64 bytes_per_chain, int64 stride, int64 busy_cycles) = chase_pointers;
 
 Lock   Run::global_mutex;
 int64  Run::_ops_per_chain = 0;
@@ -120,7 +120,7 @@ Run::run()
 	    this->bp->barrier();
 
 				// chase pointers
-	    run_benchmark(this->exp->chains_per_thread, iters, root, this->exp->bytes_per_line, this->exp->bytes_per_chain, this->exp->stride);
+	    run_benchmark(this->exp->chains_per_thread, iters, root, this->exp->bytes_per_line, this->exp->bytes_per_chain, this->exp->stride, this->exp->busy_cycles);
 
 				// barrier
 	    this->bp->barrier();
@@ -156,7 +156,7 @@ Run::run()
 	this->bp->barrier();
 
 				// chase pointers
-	run_benchmark(this->exp->chains_per_thread, this->exp->iterations, root, this->exp->bytes_per_line, this->exp->bytes_per_chain, this->exp->stride);
+	run_benchmark(this->exp->chains_per_thread, this->exp->iterations, root, this->exp->bytes_per_line, this->exp->bytes_per_chain, this->exp->stride, this->exp->busy_cycles);
 
 				// barrier
 	this->bp->barrier();
@@ -348,7 +348,8 @@ chase_pointers(
     Chain** root,			// root(s) of the chain(s) to follow
     int64 bytes_per_line,		// ignored
     int64 bytes_per_chain,		// ignored
-    int64 stride			// ignored
+    int64 stride,			// ignored
+    int64 busy_cycles	// processing cycles
 )
 {
 				// chase pointers
@@ -359,6 +360,8 @@ chase_pointers(
 	    Chain* a = root[0];
 	    while (a != NULL) {
 		a = a->next;
+		for (int64 j=0; j < busy_cycles; j++)
+			asm("nop");
 	    }
 	    mem_chk( a );
 	}
@@ -370,6 +373,8 @@ chase_pointers(
 	    while (a != NULL) {
 		a = a->next;
 		b = b->next;
+		for (int64 j=0; j < busy_cycles; j++)
+			asm("nop");
 	    }
 	    mem_chk( a );
 	    mem_chk( b );
@@ -384,6 +389,8 @@ chase_pointers(
 		a = a->next;
 		b = b->next;
 		c = c->next;
+		for (int64 j=0; j < busy_cycles; j++)
+			asm("nop");
 	    }
 	    mem_chk( a );
 	    mem_chk( b );
@@ -401,6 +408,8 @@ chase_pointers(
 		b = b->next;
 		c = c->next;
 		d = d->next;
+		for (int64 j=0; j < busy_cycles; j++)
+			asm("nop");
 	    }
 	    mem_chk( a );
 	    mem_chk( b );
@@ -421,6 +430,8 @@ chase_pointers(
 		c = c->next;
 		d = d->next;
 		e = e->next;
+		for (int64 j=0; j < busy_cycles; j++)
+			asm("nop");
 	    }
 	    mem_chk( a );
 	    mem_chk( b );
@@ -444,6 +455,8 @@ chase_pointers(
 		d = d->next;
 		e = e->next;
 		f = f->next;
+		for (int64 j=0; j < busy_cycles; j++)
+			asm("nop");
 	    }
 	    mem_chk( a );
 	    mem_chk( b );
@@ -470,6 +483,8 @@ chase_pointers(
 		e = e->next;
 		f = f->next;
 		g = g->next;
+		for (int64 j=0; j < busy_cycles; j++)
+			asm("nop");
 	    }
 	    mem_chk( a );
 	    mem_chk( b );
@@ -499,6 +514,8 @@ chase_pointers(
 		f = f->next;
 		g = g->next;
 		h = h->next;
+		for (int64 j=0; j < busy_cycles; j++)
+			asm("nop");
 	    }
 	    mem_chk( a );
 	    mem_chk( b );
@@ -531,6 +548,8 @@ chase_pointers(
 		g = g->next;
 		h = h->next;
 		j = j->next;
+		for (int64 j=0; j < busy_cycles; j++)
+			asm("nop");
 	    }
 	    mem_chk( a );
 	    mem_chk( b );
@@ -566,6 +585,8 @@ chase_pointers(
 		h = h->next;
 		j = j->next;
 		k = k->next;
+		for (int64 j=0; j < busy_cycles; j++)
+			asm("nop");
 	    }
 	    mem_chk( a );
 	    mem_chk( b );
@@ -604,6 +625,8 @@ chase_pointers(
 		j = j->next;
 		k = k->next;
 		l = l->next;
+		for (int64 j=0; j < busy_cycles; j++)
+			asm("nop");
 	    }
 	    mem_chk( a );
 	    mem_chk( b );
@@ -645,6 +668,8 @@ chase_pointers(
 		k = k->next;
 		l = l->next;
 		m = m->next;
+		for (int64 j=0; j < busy_cycles; j++)
+			asm("nop");
 	    }
 	    mem_chk( a );
 	    mem_chk( b );
@@ -689,6 +714,8 @@ chase_pointers(
 		l = l->next;
 		m = m->next;
 		n = n->next;
+		for (int64 j=0; j < busy_cycles; j++)
+			asm("nop");
 	    }
 	    mem_chk( a );
 	    mem_chk( b );
@@ -736,6 +763,8 @@ chase_pointers(
 		m = m->next;
 		n = n->next;
 		o = o->next;
+		for (int64 j=0; j < busy_cycles; j++)
+			asm("nop");
 	    }
 	    mem_chk( a );
 	    mem_chk( b );
@@ -786,6 +815,8 @@ chase_pointers(
 		n = n->next;
 		o = o->next;
 		p = p->next;
+		for (int64 j=0; j < busy_cycles; j++)
+			asm("nop");
 	    }
 	    mem_chk( a );
 	    mem_chk( b );
@@ -839,6 +870,8 @@ chase_pointers(
 		o = o->next;
 		p = p->next;
 		q = q->next;
+		for (int64 j=0; j < busy_cycles; j++)
+			asm("nop");
 	    }
 	    mem_chk( a );
 	    mem_chk( b );
@@ -905,7 +938,8 @@ follow_streams(
     Chain** root,			// root(s) of the chain(s) to follow
     int64 bytes_per_line,		// ignored
     int64 bytes_per_chain,		// ignored
-    int64 stride			// ignored
+    int64 stride,			// ignored
+    int64 busy_cycles	// ignored
 )
 {
     int64 refs_per_line  = bytes_per_line  / sizeof(double);