v2 release. Large devices are bigger than >64MB in size. - Remove unused macro for fclk rate. - Since the TI QSPI IP block only maps 64MB of MMIO, use MMIO below the 64MB boundary and software generated transfers above. - Optimize the software generated byte-transfers for dual and quad I/O read operations. The speed-up is 4.9x for quad I/O reads. Note: depends on Tony's patches for hwmod cleanup, in order to get the desired QSPI clk rate: - [PATCH 1/2] ARM: dts: Configure interconnect target module for am4 qspi - [PATCH 2/2] ARM: OMAP2+: Drop legacy platform data for am4 qspi Tested using raw accesses (mtd_debug) and JFFS2 FS read/write/erase; in single, dual and quad modes. All accesses have been checked on the logic analyzer. Jean Pihet (3): spi: spi-ti-qspi: Remove unused macro for fclk frequency spi: spi-ti-qspi: support large flash devices spi: spi-ti-qspi: optimize byte-transfers drivers/spi/spi-ti-qspi.c | 84 ++++++++++++++++++++++++++++++++++++--- 1 file changed, 79 insertions(+), 5 deletions(-) -- 2.23.0
