From eb7ccf38bf2e9208b593d021c50e9ad2ec0781ea Mon Sep 17 00:00:00 2001
From: Jens Axboe <jens.axboe@oracle.com>
Date: Wed, 29 Apr 2009 09:48:04 +0200
Subject: [PATCH] Align io units to processor cache line size

Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
---
 fio.c         | 14 ++++++++++++--
 os/os-linux.h | 20 ++++++++++++++++++++
 os/os.h       | 16 ++++++++++++++++
 3 files changed, 48 insertions(+), 2 deletions(-)

diff --git a/fio.c b/fio.c
index ad2282d7..1aba82de 100644
--- a/fio.c
+++ b/fio.c
@@ -733,7 +733,7 @@ static int init_io_u(struct thread_data *td)
 {
 	struct io_u *io_u;
 	unsigned int max_bs;
-	int i, max_units;
+	int cl_align, i, max_units;
 	char *p;
 
 	max_units = td->o.iodepth;
@@ -761,10 +761,20 @@ static int init_io_u(struct thread_data *td)
 	else
 		p = td->orig_buffer;
 
+	cl_align = os_cache_line_size();
+
 	for (i = 0; i < max_units; i++) {
+		void *ptr;
+
 		if (td->terminate)
 			return 1;
-		io_u = malloc(sizeof(*io_u));
+
+		if (posix_memalign(&ptr, cl_align, sizeof(*io_u))) {
+			log_err("fio: posix_memalign=%s\n", strerror(errno));
+			break;
+		}
+
+		io_u = ptr;
 		memset(io_u, 0, sizeof(*io_u));
 		INIT_FLIST_HEAD(&io_u->list);
 
diff --git a/os/os-linux.h b/os/os-linux.h
index 4460653f..1cd8272d 100644
--- a/os/os-linux.h
+++ b/os/os-linux.h
@@ -30,6 +30,7 @@
 #define FIO_HAVE_FALLOCATE
 #define FIO_HAVE_POSIXAIO_FSYNC
 #define FIO_HAVE_PSHARED_MUTEX
+#define FIO_HAVE_CL_SIZE
 
 #define OS_MAP_ANON		MAP_ANONYMOUS
 
@@ -249,4 +250,23 @@ static inline int fio_lookup_raw(dev_t dev, int *majdev, int *mindev)
 #define FIO_O_NOATIME	0
 #endif
 
+#define CACHE_LINE_FILE	\
+	"/sys/devices/system/cpu/cpu0/cache/index0/coherency_line_size"
+
+static inline int arch_cache_line_size(void)
+{
+	char size[32];
+	int fd, ret;
+
+	fd = open(CACHE_LINE_FILE, O_RDONLY);
+	if (fd < 0)
+		return -1;
+
+	ret = read(fd, size, sizeof(size));
+	if (ret <= 0)
+		return -1;
+
+	return atoi(size);
+}
+
 #endif
diff --git a/os/os.h b/os/os.h
index 5a3bc559..dbf09571 100644
--- a/os/os.h
+++ b/os/os.h
@@ -90,4 +90,20 @@ static inline int load_blktrace(struct thread_data *td, const char *fname)
 }
 #endif
 
+#define FIO_DEF_CL_SIZE		128
+
+static inline int os_cache_line_size(void)
+{
+#ifdef FIO_HAVE_CL_SIZE
+	int ret = arch_cache_line_size();
+
+	if (ret <= 0)
+		return FIO_DEF_CL_SIZE;
+
+	return ret;
+#else
+	return FIO_DEF_CL_SIZE;
+#endif
+}
+
 #endif
-- 
2.25.1