regarding the QSPI example merging with my application with other interruptions

RE: regarding the QSPI example merging with my application with other interruptions

Vidar Berg — Mon, 01 Aug 2022 12:29:55 GMT

Thanks for the update. Yes, we should have used __ALIGN(4) or changed the type to a uint32_t in the example code. Now there is always the risk that the linker may decide to place the array at an unaligned start address like what happened in your case.

Also, as you experienced, it doesn't help that the array size is a multiple of 4.

RE: regarding the QSPI example merging with my application with other interruptions

liuluan1984 — Mon, 01 Aug 2022 12:03:28 GMT

Thank you, Now it works fine with your suggested modification.

Strangely that the official example did not use the __ALIGN(4), and itself still run to be fine . Also that QSPI_TEST_DATA_SIZE equals to 256, which equals the 64*4 bytes= 64 words, should theoretically fine. Also in the QSPI official example(below D:\nRF5_SDK_17.0.2_d674dde\examples\peripheral\qspi\pca10056\blank\ses ) not using the __ALIGN(4).

RE: regarding the QSPI example merging with my application with other interruptions

Vidar Berg — Mon, 01 Aug 2022 09:58:55 GMT

Hello,

Please use '__ALIGN(4)' when defining the your tx buffer array to ensure it always starts at a word-aligned address as required by the QSPI peripheral. The write call will return with NRF_ERROR_INVALID_ADDR if the source address is not aligned to a word boundary.

Register description stating that source pointer must be word aligned

Defining the buffers with the ALIGN keyword

__ALIGN(4) static uint8_t m_buffer_tx[QSPI_TEST_DATA_SIZE];
__ALIGN(4) static uint8_t m_buffer_rx[QSPI_TEST_DATA_SIZE];

Best regards,

Vidar