[V2,5/8] spi: spi_amd: Optimize IO operations

Commit Message

Rangoju, Raju Sept. 25, 2024, 1:36 p.m. UTC

Read and write the maximum number of data bytes at once, rather than byte
by byte. This improves AMD SPI controller driver performance by reducing
the time required to access FIFO registers. For example, with the new
changes, 64 bytes of data from the FIFO queue can be read in 8 read calls
(8 bytes per call) instead of 64 read calls(1 byte per call).

Co-developed-by: Krishnamoorthi M <[e-mail]>
Signed-off-by: Krishnamoorthi M <[e-mail]>
Co-developed-by: Akshata MukundShetty <[e-mail]>
Signed-off-by: Akshata MukundShetty <[e-mail]>
Signed-off-by: Raju Rangoju <[e-mail]>
---
 drivers/spi/spi-amd.c | 53 +++++++++++++++++++++++++++++++++----------
 1 file changed, 41 insertions(+), 12 deletions(-)

Patch

diff --git a/drivers/spi/spi-amd.c b/drivers/spi/spi-amd.c
index 7841f3292a62..00fcec903d91 100644
--- a/drivers/spi/spi-amd.c
+++ b/drivers/spi/spi-amd.c
@@ -9,6 +9,7 @@ 
 #include <linux/acpi.h>
 #include <linux/delay.h>
 #include <linux/init.h>
+#include <linux/io-64-nonatomic-lo-hi.h>
 #include <linux/iopoll.h>
 #include <linux/module.h>
 #include <linux/platform_device.h>
@@ -140,6 +141,16 @@  static inline void amd_spi_writereg32(struct amd_spi *amd_spi, int idx, u32 val)
 	writel(val, ((u8 __iomem *)amd_spi->io_remap_addr + idx));
 }
 
+static inline u64 amd_spi_readreg64(struct amd_spi *amd_spi, int idx)
+{
+	return readq((u8 __iomem *)amd_spi->io_remap_addr + idx);
+}
+
+static inline void amd_spi_writereg64(struct amd_spi *amd_spi, int idx, u64 val)
+{
+	writeq(val, ((u8 __iomem *)amd_spi->io_remap_addr + idx));
+}
+
 static inline void amd_spi_setclear_reg32(struct amd_spi *amd_spi, int idx, u32 set, u32 clear)
 {
 	u32 tmp = amd_spi_readreg32(amd_spi, idx);
@@ -448,15 +459,23 @@  static void amd_spi_mem_data_out(struct amd_spi *amd_spi,
 				 const struct spi_mem_op *op)
 {
 	int base_addr = AMD_SPI_FIFO_BASE + op->addr.nbytes;
-	u8 *buf = (u8 *)op->data.buf.out;
+	u64 *buf_64 = (u64 *)op->data.buf.out;
 	u32 nbytes = op->data.nbytes;
+	u32 left_data = nbytes;
+	u8 *buf;
 	int i;
 
 	amd_spi_set_opcode(amd_spi, op->cmd.opcode);
 	amd_spi_set_addr(amd_spi, op);
 
-	for (i = 0; i < nbytes; i++)
-		amd_spi_writereg8(amd_spi, (base_addr + i), buf[i]);
+	for (i = 0; left_data >= 8; i++, left_data -= 8)
+		amd_spi_writereg64(amd_spi, base_addr + op->dummy.nbytes + (i * 8), *buf_64++);
+
+	buf = (u8 *)buf_64;
+	for (i = 0; i < left_data; i++) {
+		amd_spi_writereg8(amd_spi, base_addr + op->dummy.nbytes + nbytes + i - left_data,
+				  buf[i]);
+	}
 
 	amd_spi_set_tx_count(amd_spi, op->addr.nbytes + op->data.nbytes);
 	amd_spi_set_rx_count(amd_spi, 0);
@@ -467,23 +486,33 @@  static void amd_spi_mem_data_out(struct amd_spi *amd_spi,
 static void amd_spi_mem_data_in(struct amd_spi *amd_spi,
 				const struct spi_mem_op *op)
 {
-	int offset = (op->addr.nbytes == 0) ? 0 : 1;
-	u8 *buf = (u8 *)op->data.buf.in;
+	int base_addr = AMD_SPI_FIFO_BASE + op->addr.nbytes;
+	u64 *buf_64 = (u64 *)op->data.buf.in;
 	u32 nbytes = op->data.nbytes;
-	int base_addr, i;
-
-	base_addr = AMD_SPI_FIFO_BASE + op->addr.nbytes + offset;
+	u32 left_data = nbytes;
+	u8 *buf;
+	int i;
 
 	amd_spi_set_opcode(amd_spi, op->cmd.opcode);
 	amd_spi_set_addr(amd_spi, op);
-	amd_spi_set_tx_count(amd_spi, op->addr.nbytes);
-	amd_spi_set_rx_count(amd_spi, op->data.nbytes + 1);
+	amd_spi_set_tx_count(amd_spi, op->addr.nbytes + op->dummy.nbytes);
+
+	for (i = 0; i < op->dummy.nbytes; i++)
+		amd_spi_writereg8(amd_spi, (base_addr + i), 0xff);
+
+	amd_spi_set_rx_count(amd_spi, op->data.nbytes);
 	amd_spi_clear_fifo_ptr(amd_spi);
 	amd_spi_execute_opcode(amd_spi);
 	amd_spi_busy_wait(amd_spi);
 
-	for (i = 0; i < nbytes; i++)
-		buf[i] = amd_spi_readreg8(amd_spi, base_addr + i);
+	for (i = 0; left_data >= 8; i++, left_data -= 8)
+		*buf_64++ = amd_spi_readreg64(amd_spi, base_addr + op->dummy.nbytes +
+					      (i * 8));
+
+	buf = (u8 *)buf_64;
+	for (i = 0; i < left_data; i++)
+		buf[i] = amd_spi_readreg8(amd_spi, base_addr + op->dummy.nbytes +
+					  nbytes + i - left_data);
 }
 
 static int amd_spi_exec_mem_op(struct spi_mem *mem,