In ACM, when the accelerator is processing a set of TCBs, the core can halt the accelerator TCB
processing at any stage using the CTL1.HALT bit and dynamically insert new TCBs for processing by the
accelerator. The advantage of this mechanism is that if a request for processing of a set of TCBs arrives when the accelerator is already processing another set of TCBs, the core can dynamically halt the accelerator, insert the new TCB list, and resume the processing.
Here Resuming refers to that after Halt is off, the current TCB which is being processed will continue the processing until the window size number of samples, are complete as configured for that TCB. Only after completing the current TCB processing, the newly injected TCB processing will start.
So, here the model of the sequence be like
(TCB1 --> Halt the accelerator --> Inject TCB2 --> Resume accelerator(Halt off) --> resuming of TCB1 --> After TCB1, process TCB2 --> Next TCBs.)
Please refer the below Application note to know more about ACM Mode in FIR Accelerator.
https://www.analog.com/media/en/technical-documentation/application-notes/EE408V02.pdf