Skip to contentSkip to conversation

mlx4: confiugre cache line size

PATCH Superseded Eli Cohen 4 participants 4 messages Sep 16, 2009 – Sep 22, 2009

Subsystems (2)

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore

Eli Cohen

author
#0 lore raw

PATCH 1/1 mlx4: confiugre cache line size

ConnectX can work more efficiently if the CPU cache line size is confiugred to it at INIT_HCA. This patch configures cache line size for systems that report it.

---
 drivers/net/mlx4/fw.c |    7 +++++++
 1 files changed, 7 insertions(+), 0 deletions(-)
drivers/net/mlx4/fw.c +7 -0
index 20526ce..aa38c06 100644
@@ -699,6 +699,7 @@ int mlx4_INIT_HCA(struct mlx4_dev *dev, struct mlx4_init_hca_param *param)
 #define INIT_HCA_IN_SIZE		 0x200
 #define INIT_HCA_VERSION_OFFSET		 0x000
 #define	 INIT_HCA_VERSION		 2
+#define INIT_HCA_CACHELINE_SZ_OFFSET	 0x0e
 #define INIT_HCA_FLAGS_OFFSET		 0x014
 #define INIT_HCA_QPC_OFFSET		 0x020
 #define	 INIT_HCA_QPC_BASE_OFFSET	 (INIT_HCA_QPC_OFFSET + 0x10)
@@ -736,6 +737,12 @@ int mlx4_INIT_HCA(struct mlx4_dev *dev, struct mlx4_init_hca_param *param)
 
 	*((u8 *) mailbox->buf + INIT_HCA_VERSION_OFFSET) = INIT_HCA_VERSION;
 
+#if defined(cache_line_size)
+	*((u8 *) mailbox->buf + INIT_HCA_CACHELINE_SZ_OFFSET) =
+		order_base_2(cache_line_size() / 16) << 5;
+#endif
+
+
 #if defined(__LITTLE_ENDIAN)
 	*(inbox + INIT_HCA_FLAGS_OFFSET / 4) &= ~cpu_to_be32(1 << 1);
 #elif defined(__BIG_ENDIAN)
Signature · 1 line
1.6.4.3
#1 lore raw
Quoted 2 lines

+#if defined(cache_line_size)

Why the #if here? Do we just need to include <linux/cache.h> explicitly to make sure we get the define?

drivers/net/mlx4/fw.c int mlx4_INIT_HCA(struct mlx4_dev *dev, struct mlx4_init_hca_param *param) · line 742

Quoted 2 lines
drivers/net/mlx4/fw.c +2 -0
+	*((u8 *) mailbox->buf + INIT_HCA_CACHELINE_SZ_OFFSET) =
+		order_base_2(cache_line_size() / 16) << 5;

Trivial but I think it's safe to assume a cacheline is always a power of
2. And I think it's clearer (and avoids generating a divide) to use subtraction rather than division... so this could all become:

		(ilog2(cache_line_size()) - 4) << 5;

- R.
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html

#2 lore raw

On Tue, Sep 22, 2009 at 10:51:10AM -0700, Roland Dreier wrote:
I agree with you on both comments. Would you like me to resend or will you make the necessary changes?

Quoted 20 lines
Quoted 2 lines

+#if defined(cache_line_size)

Why the #if here? Do we just need to include <linux/cache.h> explicitly to make sure we get the define?

drivers/net/mlx4/fw.c int mlx4_INIT_HCA(struct mlx4_dev *dev, struct mlx4_init_hca_param *param) · line 742

Quoted 2 lines
drivers/net/mlx4/fw.c +2 -0
+	*((u8 *) mailbox->buf + INIT_HCA_CACHELINE_SZ_OFFSET) =
+		order_base_2(cache_line_size() / 16) << 5;

Trivial but I think it's safe to assume a cacheline is always a power of
2. And I think it's clearer (and avoids generating a divide) to use subtraction rather than division... so this could all become:

		(ilog2(cache_line_size()) - 4) << 5;

- R.
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

#3 lore raw
Quoted 3 lines

I agree with you on both comments. Would you like me to resend or will you make the necessary changes?

please resend, thanks.