TMS570LS0432: Understanding the initialization routine

_MM_

Expert 1995 points

Part Number: TMS570LS0432
Other Parts Discussed in Thread: HALCOGEN

Hello,

I am trying to understand the code for the initialization of the TMS570 out of Halcogen. Currently I am wondering about the scope of the following code in sys_core.asm, and belonging to _coreInitRegisters_

       bl    next1
next1
        bl    next2
next2
        bl    next3
next3
        bl    next4
next4
        bx    r0

Can anyone explain me what the above code actually does for the MCU ? It looks to me not really useful and then removable.

over 7 years ago

0 Sunil Oak over 7 years ago

TI__Mastermind 49120 points

Hi Marco,

The Cortex-R4's prefetch unit (PFU) includes a return stack that is used by the branch prediction unit. This return stack consists of a 4-entry circular buffer. When a taken branch is detected, the PFU pushes the return address onto this return stack. When the return stack detects a taken "return" instruction, the PFU fetches the instruction from the location at the top of the return stack, and pops the return stack.

The sequence of assembly instructions in the CPU registers initialization routine are used to initialize this 4-entry return stack. It must not be removed.

Regards,
Sunil

0 _MM_ over 7 years ago in reply to Sunil Oak

Expert 1995 points

Hello Sunil,

Thanks for the reply.

Do you think there is also a way to watch the behavior of the PFU you described?

0 Sunil Oak over 7 years ago in reply to _MM_

TI__Mastermind 49120 points

Hi Marco,

This 4-entry circular return stack inside the PFU is not accessible via any CPU registers, or memory mapped. You can see the effect of the PFU return stack y enabling / disabling branch prediction within the CPU for a given application.

Regards,
Sunil

0 _MM_ over 7 years ago in reply to Sunil Oak

Expert 1995 points

Ok. And then, how can I observe the taken branches are not predicted anymore ?

0 Sunil Oak over 7 years ago in reply to _MM_

TI__Mastermind 49120 points

Using an application with a few taken branches (e.g. function calls, returns), you would notice an increase in the total number of cycles taken to complete the application, if you can define a consistent measurement point. You can use the CPU's built-in Performance Monitoring Unit (PMU) for the cycle count.

0 Sunil Oak over 7 years ago in reply to Sunil Oak

TI__Mastermind 49120 points

Marco,

Are your questions resolved?

Arm-based microcontrollers

Arm-based microcontrollers forum

TMS570LS0432: Understanding the initialization routine