Patch for 8051 timer

Two issues when I'm tring to use Timer 1 mode 2 ( 8-bit auto reload mode, it's used by uart to set the baud rate)

- TL1's value is not updated UNLESS mcu monitor is displayed with TCON monitored.

It seems that some operation in i51core directly access mem , instead of calling GET_RAM, so the watcher on the TL1 won't be triggered and it's value not get updated. I changed all direct access to m_dataMem to GET_RAM and SET_RAM.

- TL1's value is counted starting from 0, instead from value in TH1.

When TLx get updated, add the value in THx to it.

I don't remember much about 8051 Timers, but for what I see in the code, mode 2 is supposed to be 8 bits??

arcachofo wrote:I don't remember much about 8051 Timers, but for what I see in the code, mode 2 is supposed to be 8 bits??

yes, it's the 8 bits auto reload mode.

Ok, I see.
But I think that is already done:
When you write TH1 McuTimer::countWriteH() is called.
This calls: 51Timer::updtCount(), and here:
case 2: // 8 bits
m_countStart = COUNT_H; // Load this value after overflow

I think this problem is probably solved with GET_RAM()

Patch for GET_RAM() applied at Rev 1959.

arcachofo wrote:Ok, I see.
But I think that is already done:
When you write TH1 McuTimer::countWriteH() is called.
This calls: 51Timer::updtCount(), and here:
   case 2: // 8 bits
       m_countStart = COUNT_H; // Load this value after overflow

Nop. in  I51Timer::updtCount(), countVal and  COUNT_L is calculated as the following:

Code:


        uint64_t timTime = m_ovfCycle-Simulator::self()->circTime(); // Next overflow time - current time
        uint16_t countVal = timTime/m_mcu->psInst()/m_prescaler;
        if( m_mode == 0 )  // 13 bits
        ....
        else if( m_mode == 2 ) // 8 bits
        {
            COUNT_L = ( countVal ) & 0xFF;
            //if( m_countH ) m_countH[0] = (countVal>>8) & 0xFF;
        }


So countVal (COUNT_L) is just the difference between overflow time and current time.

This can be verified by watch TL1's value change in the mcu monitor.

royqh1979 wrote:So countVal (COUNT_L) is just the difference between overflow time and current time.

This can be verified by watch TL1's value change in the mcu monitor.

I rechecked McuTimer::updtCount() and found that I've also made a mistake here:

COUNT_L should be (m_ovfMatch-countVal) & 0xFF

And I'm afraid the calculation in I51Timer::updtCount() for mode 0 is wrong, either.

So countVal (COUNT_L) is just the difference between overflow time and current time, it always start from 0.

But overflow time is not calculated from 0, it is calculated from m_countStart:

Code:

void McuTimer::runEvent()            // Overflow
{
...
    m_countVal = m_countStart;            // Reset count value
...
    sheduleEvents();
}
void McuTimer::sheduleEvents()
{
...
        uint64_t cycles = (ovfPeriod-m_countVal)*m_scale; // cycles in ps
...
}


This can be verified by watch TL1's value change in the mcu monitor.

I will have a look, but this must be done after GET_RAM patch.

arcachofo wrote:
But overflow time is not calculated from 0, it is calculated from m_countStart:

I compared  McuTimer::updtCount() with I51Timer::updtCount() and found that I've also made a mistake:

COUNT_L should be (m_ovfMatch-countVal) & 0xFF

And I'm afraid the calculation in I51Timer::updtCount() for mode 0 is wrong, either.

I rechecked McuTimer::updtCount() and found that I've also made a mistake here:

COUNT_L should be (m_ovfMatch-countVal) & 0xFF

And I'm afraid the calculation in I51Timer::updtCount() for mode 0 is wrong, either.

Yes, it could be some issue here.
If I remember well at some point I changed McuTimer::updtCount() but maybe not updated I51Timer::updtCount()

arcachofo wrote:
Yes, it could be some issue here.
If I remember well at some point I changed McuTimer::updtCount() but maybe not updated I51Timer::updtCount()

Yeah. And I think we could just update m_countVal here as well.

Yeah. And I think we could just update m_countVal here as well.

At Rev 1960 counter calculation is splitted in it's own function and called from McuTimer::updtCount() and I51Timer::updtCount().

arcachofo wrote:
At Rev 1960 counter calculation is splitted in it's own function and called from McuTimer::updtCount() and I51Timer::updtCount().

Great!