[PATCH 3/3] TI QSPI: optimize transfers for dual and quad read

[Jean Pihet <jean.pihet@xxxxxxxxxxxxxx>]

By reading the 32 bits data register and copy the contents to the
receive buffer, according to the single/dual/quad read mode and
the data length to read.

The speed improvement is 3.5x using quad read.
---
 drivers/spi/spi-ti-qspi.c | 48 ++++++++++++++++++++++++++-------------
 1 file changed, 32 insertions(+), 16 deletions(-)

diff --git a/drivers/spi/spi-ti-qspi.c b/drivers/spi/spi-ti-qspi.c
index 13916232a959..65ec3bcb25ae 100644
--- a/drivers/spi/spi-ti-qspi.c
+++ b/drivers/spi/spi-ti-qspi.c
@@ -313,24 +313,25 @@ static int qspi_write_msg(struct ti_qspi *qspi, struct spi_transfer *t,
 static int qspi_read_msg(struct ti_qspi *qspi, struct spi_transfer *t,
 			 int count)
 {
-	int wlen;
 	unsigned int cmd;
+	u32 rx;
 	u8 *rxbuf;
+	u8 xfer_len;
 
 	rxbuf = t->rx_buf;
 	cmd = qspi->cmd;
+	/* Optimize the transfers for dual and quad read */
 	switch (t->rx_nbits) {
-	case SPI_NBITS_DUAL:
-		cmd |= QSPI_RD_DUAL;
-		break;
 	case SPI_NBITS_QUAD:
-		cmd |= QSPI_RD_QUAD;
+		cmd |= QSPI_RD_QUAD | QSPI_WLEN(32);
+		break;
+	case SPI_NBITS_DUAL:
+		cmd |= QSPI_RD_DUAL | QSPI_WLEN(16);
 		break;
 	default:
-		cmd |= QSPI_RD_SNGL;
+		cmd |= QSPI_RD_SNGL | QSPI_WLEN(8);
 		break;
 	}
-	wlen = t->bits_per_word >> 3;	/* in bytes */
 
 	while (count) {
 		dev_dbg(qspi->dev, "rx cmd %08x dc %08x\n", cmd, qspi->dc);
@@ -342,19 +343,34 @@ static int qspi_read_msg(struct ti_qspi *qspi, struct spi_transfer *t,
 			dev_err(qspi->dev, "read timed out\n");
 			return -ETIMEDOUT;
 		}
-		switch (wlen) {
-		case 1:
-			*rxbuf = readb(qspi->base + QSPI_SPI_DATA_REG);
+
+		/* Optimize the transfers for dual and quad read */
+		rx = readl(qspi->base + QSPI_SPI_DATA_REG);
+		switch (t->rx_nbits) {
+		case SPI_NBITS_QUAD:
+			if (count >= 1)
+				*rxbuf++ = rx >> 24;
+			if (count >= 2)
+				*rxbuf++ = rx >> 16;
+			if (count >= 3)
+				*rxbuf++ = rx >> 8;
+			if (count >= 4)
+				*rxbuf++ = rx;
+			xfer_len = min(count, 4);
 			break;
-		case 2:
-			*((u16 *)rxbuf) = readw(qspi->base + QSPI_SPI_DATA_REG);
+		case SPI_NBITS_DUAL:
+			if (count >= 1)
+				*rxbuf++ = rx >> 8;
+			if (count >= 2)
+				*rxbuf++ = rx;
+			xfer_len = min(count, 2);
 			break;
-		case 4:
-			*((u32 *)rxbuf) = readl(qspi->base + QSPI_SPI_DATA_REG);
+		default:
+			*rxbuf++ = rx;
+			xfer_len = 1;
 			break;
 		}
-		rxbuf += wlen;
-		count -= wlen;
+		count -= xfer_len;
 	}
 
 	return 0;
-- 
2.23.0