Microbenchmark for small and large allocations of device

and host memory.

microbench/mallocSpeed.cu

@@ -0,0 +1,113 @@
+/*
+ *
+ * mallocSpeed.cu
+ *
+ * Microbenchmark for overhead and per-page speed of host and
+ * device memory allocation.
+ *
+ * Build with: nvcc -I ../chLib <options> mallocSpeed.cu
+ * Requires: No minimum SM requirement.
+ *
+ * Copyright (c) 2013, Archaea Software, LLC.
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions 
+ * are met: 
+ *
+ * 1. Redistributions of source code must retain the above copyright 
+ *    notice, this list of conditions and the following disclaimer. 
+ * 2. Redistributions in binary form must reproduce the above copyright 
+ *    notice, this list of conditions and the following disclaimer in 
+ *    the documentation and/or other materials provided with the 
+ *    distribution. 
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS 
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT 
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS 
+ * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE 
+ * COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, 
+ * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, 
+ * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
+ * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER 
+ * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 
+ * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN 
+ * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 
+ * POSSIBILITY OF SUCH DAMAGE.
+ *
+ */
+
+#include <stdio.h>
+
+#include "chTimer.h"
+
+template <bool bDevice>
+float
+mallocSpeed( int cIterations, size_t N )
+{
+    chTimerTimestamp start, stop;
+    float ret = 0.0f;
+    void **p = (void **) malloc( cIterations*sizeof(void *));
+    if ( ! p )
+        goto Error;
+    memset( p, 0, cIterations*sizeof(void *) );
+
+    chTimerGetTime( &start );
+    for ( size_t i = 0; i < cIterations; i++ ) {
+        if ( bDevice ) {
+            if ( cudaSuccess != cudaMalloc( &p[i], N ) ) {
+                goto Error;
+            }
+        }
+        else {
+            if ( cudaSuccess != cudaMallocHost( &p[i], N ) )
+                goto Error;
+        }
+    }
+    chTimerGetTime( &stop );
+    ret = chTimerElapsedTime( &start, &stop );
+Error:
+    if ( p ) {
+        for ( size_t i = 0; i < cIterations; i++ ) {
+            if ( bDevice ) {
+                cudaFree( p[i] );
+            }
+            else {
+                cudaFreeHost( p[i] );
+            }
+        }
+        free( p );
+    }
+    return ret;
+}
+
+template<bool bDevice>
+float
+mallocSpeed( size_t N )
+{
+    float ret = 0.0f;
+    int cIterations;
+    for ( cIterations = 1; ret < 0.5f; cIterations *= 2 ) {
+        ret = mallocSpeed<bDevice>( cIterations, N );
+        if ( 0.0f == ret ) {
+            return mallocSpeed<bDevice>( cIterations/2, N) / (cIterations/2);
+        }
+    }
+    return ret / cIterations;
+}
+
+
+int
+main( int argc, char *argv[] )
+{
+    if ( cudaSuccess != cudaFree(0) ) {
+        printf( "Initialization failed\n" );
+        exit(1);
+    }
+    printf( "mallocSpeed (4K device): %.2f us/iteration\n", mallocSpeed<true>(4096)*1e6 );
+    printf( "mallocSpeed (1M device): %.2f us/iteration\n", mallocSpeed<true>(1048576)*1e6 );
+    printf( "mallocSpeed (4K host): %.2f us/iteration\n", mallocSpeed<false>(4096)*1e6 );
+    printf( "mallocSpeed (1M host): %.2f us/iteration\n", mallocSpeed<false>(1048576)*1e6 );
+    return 0;
+
+}