Hello.
I have a similar question.
We are using RT1024 with Keil MDK and MIMXRT1024_QuadSPI_4KB_SEC.FLM
The problem is programming speed too low (via JLink + SWO).
I suppose it because .FLM uses 4kB sectors despite on-chip "Winbond W25Q32JV" supports also 32 and 64kB sectors.
According to Flash datasheet erase time for 4kB sector - 45ms, and 150ms for 64kB.
If we using 4kB it will take 45*16=720ms to erase 64kB instead of 150ms.
And I think overhead with 4kB sectors on extra SWO-transactions is more compare to 64kB option.
For source codes of "MIMXRT105x_HYPER_256KB_SEC.FLM" one can apply to keil support.
As far as I know Keil do not implements *.FLM but MCU manufacturer does it (i.e. NXP in this case).
Can you confirm or refute it?
Of course we can implement our own *.FLM, but it would be much easier to just correct a few lines of code in the existing algorithm.
Thank you.
UPD:
I forgot to mention that "erase" is not the only problem - "programming" has the same issue.
I can see in Keil's progress bar size of chunks - 256 bytes.
It is equals to "page size" of Winbond memory, but FLM allows to set "szPage" field of "struct FlashDevice" to multiple of the real minimum programming size (to increase speed).
For example piece of code for the STM32L0:
struct FlashDevice const FlashDevice = {
FLASH_DRV_VERS, // Driver Version, do not modify!
"STM32L0 8KB Flash", // Device Name
ONCHIP, // Device Type
0x08000000, // Device Start Address
0x00002000, // Device Size in Bytes (8kB)
1024, // Programming Page Size (program several half pages at once)
0, // Reserved, must be 0
0x00, // Initial Content of Erased Memory
500, // Program Page Timeout 500 mSec
500, // Erase Sector Timeout 500 mSec
// Specify Size and Address of Sectors
0x080, 0x000000, // Sector Size 128B (64 Sectors)
SECTOR_END
};
