derek/gem5
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

dev-amdgpu: Fix SDMA ring buffer wrap around

Browse Source 
The current SDMA wrap around handling only considers the ring buffer
location as seen by the GPU. Eventually when the end of the SDMA ring
buffer is reached, the driver waits until the rptr written back to the
host catches up to what the driver sees before wrapping around back to
the beginning of the buffer. This writeback currently does not happen at
all, causing hangs for applications with a lot of SDMA commands.

This changeset first fixes the sizes of the queues, especially RLC
queues, so that the wrap around occurs in the correct place. Second, we
now store the rptr writeback address and the absoluate (unwrapped) rptr
value in each SDMA queue. The absolulte rptr is what the driver sends to
the device and what it expects to be written back.

This was tested with an application which basically does a few hundred
thousand hipMemcpy() calls in a loop. It should also fix the issue with
pannotia BC in fullsystem mode.

Change-Id: I53ebdcc6b02fb4eb4da435c9a509544066a97069
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/65351
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Reviewed-by: Matt Sinclair <mattdsinclair@gmail.com>
Maintainer: Matt Sinclair <mattdsinclair@gmail.com>
This commit is contained in:
Matthew Poremba
2022-11-07 16:28:15 -08:00
parent 8d63c9fc06
commit c8d687b05c
3 changed files with 47 additions and 15 deletions
Download Patch File Download Diff File Expand all files Collapse all files

18
src/dev/amdgpu/sdma_engine.hh
View File

@@ -58,6 +58,8 @@ class SDMAEngine : public DmaVirtDevice
Addr _rptr;
Addr _wptr;
Addr _size;
Addr _rptr_wb_addr = 0;
Addr _global_rptr = 0;
bool _valid;
bool _processing;
SDMAQueue *_parent;
@@ -72,6 +74,8 @@ class SDMAEngine : public DmaVirtDevice
Addr wptr() { return _base + _wptr; }
Addr getWptr() { return _wptr; }
Addr size() { return _size; }
Addr rptrWbAddr() { return _rptr_wb_addr; }
Addr globalRptr() { return _global_rptr; }
bool valid() { return _valid; }
bool processing() { return _processing; }
SDMAQueue* parent() { return _parent; }
@@ -82,22 +86,27 @@ class SDMAEngine : public DmaVirtDevice
void
incRptr(uint32_t value)
{
//assert((_rptr + value) <= (_size << 1));
_rptr = (_rptr + value) % _size;
_global_rptr += value;
}
void rptr(Addr value) { _rptr = value; }
void
rptr(Addr value)
{
_rptr = value;
_global_rptr = value;
}
void
setWptr(Addr value)
{
//assert(value <= (_size << 1));
_wptr = value % _size;
}
void wptr(Addr value) { _wptr = value; }
void size(Addr value) { _size = value; }
void rptrWbAddr(Addr value) { _rptr_wb_addr = value; }
void valid(bool v) { _valid = v; }
void processing(bool value) { _processing = value; }
void parent(SDMAQueue* q) { _parent = q; }
@@ -268,7 +277,8 @@ class SDMAEngine : public DmaVirtDevice
/**
* Methods for RLC queues
*/
void registerRLCQueue(Addr doorbell, Addr rb_base);
void registerRLCQueue(Addr doorbell, Addr rb_base, uint32_t size,
Addr rptr_wb_addr);
void unregisterRLCQueue(Addr doorbell);
void deallocateRLCQueues();