Improve SPI block write #6607
Improve SPI block write #6607
Conversation
| req.rx_data = (uint8_t *)(rx_buffer + tx_length); | ||
| req.len = rx_length - tx_length; | ||
| SPIM_Trans(obj->spi, &req); | ||
| } |
There was a problem hiding this comment.
can you fix alignment for } . this should be 4 spaces right. Line 180 should be 4 spaces right
There was a problem hiding this comment.
Indentation fixed at line 204.
| char in = spi_master_write(obj, out); | ||
| if (i < rx_length) { | ||
| rx_buffer[i] = in; | ||
| core_util_critical_section_enter(); |
There was a problem hiding this comment.
Why is this in critical section?
There was a problem hiding this comment.
Hardware limitation may cause data corruption if the TX fifo is allowed to run empty. The critical section ensures a transaction will complete without interruption.
There was a problem hiding this comment.
Hardware limitation may cause data corruption if the TX fifo is allowed to run empty. The critical section ensures a transaction will complete without interruption.
these type of details are worth in the commit msg. I imagine coming over and removing it like as the upper layer should take care of synchronization (this is however hw protection)
Replace looping construct with actual block writes. Transaction in spi_master_block_write funtion is protected by critical section to ensure it is completed without interruption. A hardware limitation may cause data corruption if the TX fifo is allowed to run empty.
|
Commit message updated for spi block write. |
|
/morph build |
Build : SUCCESSBuild number : 1755 Triggering tests/morph test |
Exporter Build : SUCCESSBuild number : 1392 |
Test : SUCCESSBuild number : 1561 |
|
/morph mbed2-build |
Description
Improve SPI block write by replacing byte write loop with hardware assisted block transaction.
Tested all supported toolchains with MAX32630FTHR.
Pull request type
[X] Fix
[ ] Refactor
[ ] New target
[ ] Feature
[ ] Breaking change