- change the delay from 10 * HZ to 1 * HZ, though (1 * HZ) might not be the
best, just as a base.
1. Since the request has been moved out of interrupt context, there will be no
more calling enable_clk in interrupt context. So it's not necessary to keep
such a long delay on disabling clock in order to save power.
2. Still keeping the 1*HZ of delay is to avoid frequently enabling/disabling
clock.
eMMC card performance test result with bonnie++:
(512M RAM, 1GB data, 1K buffer)
------------------------------------------------------------------------------
| | 1*HZ | 10 * HZ |
------------------------------------------------------------------------------
| Read | ~24.1MB/s | ~23.9MB/s |
------------------------------------------------------------------------------
| Write | ~10.5MB/s | ~10.5MB/s |
------------------------------------------------------------------------------
WiFi card performance test result with iperf is quite same: ~21Mbps
(AR6003@2.4G, TCP, TCP window size option 1MB both for server and client)
Acked-by: Robby CAI <r63905@freescale.com>
Signed-off-by: Ryan QIAN <b32804@freescale.com>
#include "sdhci.h"
#define DRIVER_NAME "sdhci"
-#define CLK_TIMEOUT (10 * HZ)
+#define CLK_TIMEOUT (1 * HZ)
#define DBG(f, x...) \
pr_debug(DRIVER_NAME " [%s()]: " f, __func__,## x)