I cat'n find out how does BIOS heap operates with gates.
I define BIOS's heap in app.cfg like this:
var Defaults = xdc.useModule('xdc.runtime.Defaults');
var BIOS = xdc.useModule('ti.sysbios.BIOS');
var HeapTrack = xdc.useModule('ti.sysbios.heaps.HeapTrack');
BIOS.heapSize = 0x4000;
BIOS.heapTrackEnabled = true;
Do I have to take steps to protect heap corruptness during multi thread usage?
Is it safe to allocate new records in std::map<...> in different thread simultaneous?
By default the system heap is created using a HeapMem instance (see 7.7.2, Specifying the Default System Heap). And by default, HeapMem uses GateMutex for a lock.
But you can change it to a different gate type, for example as shown in 7.8.2:
var GateMutexPri = xdc.useModule('ti.sysbios.gates.GateMutexPri');
var HeapMem = xdc.useModule('ti.sysbios.heaps.HeapMem');
HeapMem.common$.gate = GateMutexPri.create();
If you open up the RTOS Object Viewer (ROV) after starting your program, you can see the gate configuration. Click on HeapMem, then the Raw tab, then Configuration, and then common$, and look at $name. For example (for a CC3200 that I have handy):
You’d asked if it was safe to use with std::map<>. I think that the gate you choose will be used here. But I’m not absolutely positive. I think to verify whether this is the case, you can run your app to a breakpoint just before the call. Then set a breakpoint on the gate enter function (for example ti_sysbios_gates_GateMutexPri_enter__E()), and run. I think you should see the breakpoint stop the CPU. Then look in the callstack to see the gate used for the allocation. For example, I tried this on my board for a simple malloc call and saw the callstack:
Hope this helps.
Regards,
Scott
