Avoid function calls in the inner PIO loops. On a Centris 660av this
improves throughput for sequential read transfers by about 40% and
sequential write by about 10%.
Unfortunately it is not possible to have methods like .esp_write8
placed inline so this is always going to be slow, even with LTO.
Tested-by: Stan Johnson <user...@yahoo.com>
Signed-off-by: Finn Thain <fth...@telegraphics.com.au>
Tested-by: Michael Schmitz <schmitz...@gmail.com>
---
Changed since v1:
- Don't touch the scsi_esp_cmd(esp, ESP_CMD_FLUSH) as it's outside the loop.
---
drivers/scsi/esp_scsi.c | 12 ++++++------
1 file changed, 6 insertions(+), 6 deletions(-)
diff --git a/drivers/scsi/esp_scsi.c b/drivers/scsi/esp_scsi.c
index 0f0d486be486..c3077c064522 100644
--- a/drivers/scsi/esp_scsi.c
+++ b/drivers/scsi/esp_scsi.c
@@ -2783,7 +2783,7 @@ static inline unsigned int esp_wait_for_fifo(struct esp
*esp)
if (fbytes)
return fbytes;
- udelay(2);
+ udelay(1);
} while (--i);
shost_printk(KERN_ERR, esp->host, "FIFO is empty. sreg [%02x]\n",
@@ -2800,7 +2800,7 @@ static inline int esp_wait_for_intr(struct esp *esp)
if (esp->sreg & ESP_STAT_INTR)
return 0;
- udelay(2);
+ udelay(1);
} while (--i);
shost_printk(KERN_ERR, esp->host, "IRQ timeout. sreg [%02x]\n",
@@ -2828,7 +2828,7 @@ void esp_send_pio_cmd(struct esp *esp, u32 addr, u32
esp_count,
if (!esp_wait_for_fifo(esp))
break;
- *dst++ = esp_read8(ESP_FDATA);
+ *dst++ = readb(esp->fifo_reg);
--esp_count;
if (!esp_count)
@@ -2849,9 +2849,9 @@ void esp_send_pio_cmd(struct esp *esp, u32 addr, u32
esp_count,
}
if (phase == ESP_MIP)
- scsi_esp_cmd(esp, ESP_CMD_MOK);
+ esp_write8(ESP_CMD_MOK, ESP_CMD);
- scsi_esp_cmd(esp, ESP_CMD_TI);
+ esp_write8(ESP_CMD_TI, ESP_CMD);
}
} else {
unsigned int n = ESP_FIFO_SIZE;
@@ -2891,7 +2891,7 @@ void esp_send_pio_cmd(struct esp *esp, u32 addr, u32
esp_count,
src += n;
esp_count -= n;
- scsi_esp_cmd(esp, ESP_CMD_TI);
+ esp_write8(ESP_CMD_TI, ESP_CMD);
}
}
--
2.18.1
