diff --git a/arch/x86/include/asm/svm.h b/arch/x86/include/asm/svm.h
index c7acaa59..f9260d58 100644
--- a/arch/x86/include/asm/svm.h
+++ b/arch/x86/include/asm/svm.h
@@ -32,7 +32,7 @@ extern "C" {
 
 #ifdef CONFIG_ROCKCREEK
 
-#define SVM_WB
+//#define SVM_WB
 
 #define SVM_STRONG		(1 << 0)
 #define SVM_LAZYRELEASE		(1 << 1)
@@ -96,7 +96,7 @@ void svm_invalidate(void);
 /* *brief flushs the cache for all SVM regions
  */
 #ifndef SVM_WB
-static inline void svm_flush(void)
+static inline void svm_flush(size_t unused)
 {
 	// need to write to another line to make sure the write combine buffer gets flushed
 	*(int *)RCCE_fool_write_combine_buffer = 1;
diff --git a/arch/x86/mm/svm.c b/arch/x86/mm/svm.c
index 4fdfb713..974fbcbe 100644
--- a/arch/x86/mm/svm.c
+++ b/arch/x86/mm/svm.c
@@ -35,6 +35,8 @@
 #include <asm/icc.h>
 #include <asm/svm.h>
 
+#define USE_PERFCOUNTERS	1
+
 #define SHARED_PAGES		(4*(RCCE_SHM_SIZE_MAX >> PAGE_SHIFT))
 #define OWNER_SIZE		((SHARED_PAGES * sizeof(uint8_t) + PAGE_SIZE - 1) & ~(PAGE_SIZE - 1))
 
@@ -65,6 +67,7 @@ static RCCE_FLAG release;
 #define L2_CAPACITY (256*1024UL)
 #define L2_WBSTRIDE (L2_CAPACITY/L2_WAYS)
 
+#ifdef SVM_WB
 /* Helper function to read data into all 4 ways of L2 cache */ 
 
 __attribute__((always_inline)) static inline void svm_purge_set(const size_t set)
@@ -83,6 +86,7 @@ __attribute__((always_inline)) static inline void svm_purge_set(const size_t set
 
 static size_t dummy_base = OWN_MPB + L2_CAPACITY;
 static size_t dummy_offset = 0;
+#endif
 
 /*
  * This array describes the owner of a specific page.
@@ -97,11 +101,16 @@ static const size_t shmbegin = SHM_ADDR;
 static uint32_t emit[RCCE_MAXNP] = {[0 ... RCCE_MAXNP-1] = 0};
 static uint32_t request[RCCE_MAXNP] = {[0 ... RCCE_MAXNP-1] = 0};
 static uint32_t forward[RCCE_MAXNP] = {[0 ... RCCE_MAXNP-1] = 0};
+static uint32_t alloc_page = 0;
+static uint32_t map_page = 0;
+#if USE_PERFCOUNTERS
+static uint64_t alloc_ticks = 0;
 static uint64_t request_ticks = 0;
 static uint64_t emit_ticks = 0;
 static uint64_t wait_ticks = 0;
 static uint64_t max_wait = 0;
 static uint64_t min_wait = (uint64_t) -1;
+#endif
 
 int svm_init(void)
 {
@@ -185,14 +194,19 @@ static size_t get_shpage(void)
  */
 int svm_alloc_page(size_t addr, page_table_t* pgt)
 {
+#if USE_PERFCOUNTERS
+	uint64_t start = rdtsc();
+#endif
 	uint32_t index2 = (addr >> 12) & 0x3FF;
 	size_t phyaddr;
-	t_vcharp mpb = (t_vcharp) ((size_t)(virt_to_phys(addr) >> PAGE_SHIFT) | (size_t) RCCE_comm_buffer[RCCE_IAM]);
+	t_vcharp mpb = (t_vcharp) ((size_t)(virt_to_phys(addr) >> PAGE_SHIFT) | ((size_t) RCCE_comm_buffer[RCCE_IAM] - RCCE_LINE_SIZE));
 	uint16_t offset = 0xFFFF;
+
+	addr &= 0xFFFFF000; // align address to the page boundary
 	
 	RCCE_acquire_lock(RC_COREID[LOCK_ID]);
 
-	RCCE_get((t_vcharp) &offset, mpb, sizeof(uint16_t), RCCE_IAM);
+	iRCCE_get((t_vcharp) &offset, mpb, sizeof(uint16_t), RCCE_IAM);
 
 	if (!offset) {
 		int i;
@@ -200,7 +214,7 @@ int svm_alloc_page(size_t addr, page_table_t* pgt)
 		phyaddr = get_shpage();
 		offset = (uint16_t) ((phyaddr - shmbegin) >> PAGE_SHIFT);
 		for(i=0; i<RCCE_NP; i++)
-			RCCE_put(mpb, (t_vcharp) &offset, sizeof(uint16_t), i);
+			iRCCE_put(mpb, (t_vcharp) &offset, sizeof(uint16_t), i);
 
 		RCCE_release_lock(RC_COREID[LOCK_ID]);
 
@@ -210,8 +224,12 @@ int svm_alloc_page(size_t addr, page_table_t* pgt)
 		phys2virt[(phyaddr - shmbegin) >> PAGE_SHIFT] = addr;
 		tlb_flush_one_page(addr);
 
-		kprintf("map new page frame 0x%x at 0x%x, flags0x%x, offset 0x%x, mpb 0x%x\n", phyaddr, addr, pgt->entries[index2] & 0xFFF, (int) offset, mpb);
+		alloc_page++;
+		//kprintf("map new page frame 0x%x at 0x%x, flags0x%x, offset 0x%x, mpb 0x%x\n", phyaddr, addr, pgt->entries[index2] & 0xFFF, (int) offset, mpb);
 
+#if USE_PERFCOUNTERS
+		alloc_ticks += rdtsc() - start;
+#endif
 		return 0;
 	} else {
 		RCCE_release_lock(RC_COREID[LOCK_ID]);
@@ -226,11 +244,14 @@ int svm_alloc_page(size_t addr, page_table_t* pgt)
 		phys2virt[(phyaddr - shmbegin) >> PAGE_SHIFT] = addr;
 		tlb_flush_one_page(addr);
 
-		kprintf("map existing page frame 0x%x at 0x%x, offset 0x%x, mpb 0x%x\n", phyaddr, addr, offset, mpb);
+		map_page++;
+		//kprintf("map existing page frame 0x%x at 0x%x, offset 0x%x, mpb 0x%x\n", phyaddr, addr, offset, mpb);
 
+#if USE_PERFCOUNTERS
+		alloc_ticks += rdtsc() - start;
+#endif
 		if (pgt->entries[index2] & PG_SVM_LAZYRELEASE)
 			return 0;
-		kprintf("send request to %d, 0x%x\n", (int) page_owner[(phyaddr - shmbegin) >> PAGE_SHIFT], (phyaddr - shmbegin) >> PAGE_SHIFT);
 		return svm_access_request(addr);
 	}
 }
@@ -241,7 +262,9 @@ int svm_alloc_page(size_t addr, page_table_t* pgt)
  */
 int svm_access_request(size_t addr)
 {
+#if USE_PERFCOUNTERS
 	uint64_t start = rdtsc();
+#endif
 	size_t phyaddr = virt_to_phys(addr);
 	uint32_t pageid;
 	int remote_rank;
@@ -261,24 +284,31 @@ int svm_access_request(size_t addr)
 	((size_t*) payload)[0] = RCCE_IAM;
 	((size_t*) payload)[1] = phyaddr;
 
+	//kprintf("send request (0x%x) to %d\n", addr, remote_rank);
 	/* send ping request */
 	iRCCE_mail_send(2*sizeof(size_t), SVM_REQ, 0, (char*) payload, remote_rank);
 	icc_send_gic_irq(remote_rank);	
 	request[remote_rank]++;
 
+#if USE_PERFCOUNTERS
 	uint64_t wait_start = rdtsc();
+#endif
 	// wait for response
 	icc_wait(SVM_RESP);
+#if USE_PERFCOUNTERS
 	uint64_t res = rdtsc() - wait_start;
 	wait_ticks += res;
 	if (min_wait > res)
 		min_wait = res;
 	if (max_wait < res)
 		max_wait = res;
+#endif
 
 	ret = change_page_permissions(addr, addr+PAGE_SIZE, VMA_READ|VMA_WRITE|VMA_CACHEABLE);
 
+#if USE_PERFCOUNTERS
 	request_ticks += rdtsc() - start;
+#endif
 
 	return ret;
 }
@@ -384,6 +414,9 @@ void* svm_malloc(size_t size, uint32_t consistency)
 #endif
 
 	map_flags |= MAP_NO_ACCESS;
+#ifndef SVM_WB
+	map_flags |= MAP_MPE;
+#endif
 
 	viraddr = map_region(0, 0, size >> PAGE_SHIFT, map_flags);
 	kprintf("svmmalloc: viraddr 0x%x, size 0x%x\n", viraddr, size);
@@ -400,7 +433,7 @@ void* svm_malloc(size_t size, uint32_t consistency)
 			}
 
 			//kprintf("mpb_addr 0x%x\n", mpb_addr);
-			RCCE_put(mpb_addr, buffer, RCCE_LINE_SIZE, RCCE_IAM);
+			iRCCE_put(mpb_addr, buffer, RCCE_LINE_SIZE, RCCE_IAM);
 
 			irq_nested_enable(flags);
 		}
@@ -454,7 +487,9 @@ void svm_free(void* addr, size_t size)
  */
 int svm_emit_page(size_t phyaddr, int ue)
 {
+#if USE_PERFCOUNTERS
 	uint64_t start = rdtsc();	
+#endif
 	uint32_t pageid;
 	int remote_rank;
 
@@ -497,7 +532,9 @@ int svm_emit_page(size_t phyaddr, int ue)
 		change_page_permissions(viraddr, viraddr+PAGE_SIZE, VMA_NOACCESS|VMA_READ|VMA_CACHEABLE);
 	}
 
+#if USE_PERFCOUNTERS
 	emit_ticks += rdtsc() - start;
+#endif
 
 	return 0;
 }
@@ -545,7 +582,7 @@ void svm_invalidate(void)
 
 }
 
-void svm_flush( size_t phyaddr )
+void svm_flush(size_t phyaddr)
 {
 	task_t* task = per_core(current_task);
 	page_dir_t* pgd = task->pgd;
@@ -659,8 +696,8 @@ int svm_barrier(uint32_t flags)
 	return 0;
 }
 
-extern uint64_t check_ticks;
-extern uint64_t recv_ticks;
+//extern uint64_t check_ticks;
+//extern uint64_t recv_ticks;
 
 int svm_statistics(void)
 {
@@ -676,13 +713,18 @@ int svm_statistics(void)
 	for(i=0; i<RCCE_MAXNP; i++)
 		kprintf("\t%u", forward[i]);
 	kputs("\n");
+	kprintf("allocate page frame: %u\n", alloc_page);
+	kprintf("map page frame: %d\n", map_page);
+#if USE_PERFCOUNTERS
+	kprintf("alloc ticks: %llu\n", alloc_ticks);
 	kprintf("request ticks: %llu\n", request_ticks);
 	kprintf("wait ticks: %llu\n", wait_ticks);
 	kprintf("emit ticks: %llu\n", emit_ticks);
 	kprintf("max wait: %llu\n", max_wait);
 	kprintf("min wait: %llu\n", min_wait);
-	kprintf("check_ticks: %llu\n", check_ticks);
-	kprintf("recv_tick: %llu\n", recv_ticks);
+	//kprintf("check_ticks: %llu\n", check_ticks);
+	//kprintf("recv_tick: %llu\n", recv_ticks);
+#endif
 
 	return 0;
 }