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

14
src/dev/amdgpu/pm4_packet_processor.cc
View File

@@ -441,12 +441,17 @@ void
PM4PacketProcessor::processSDMAMQD(PM4MapQueues *pkt, PM4Queue *q, Addr addr,
SDMAQueueDesc *mqd, uint16_t vmid)
{
uint32_t rlc_size = 4UL << bits(mqd->sdmax_rlcx_rb_cntl, 6, 1);
Addr rptr_wb_addr = mqd->sdmax_rlcx_rb_rptr_addr_hi;
rptr_wb_addr <<= 32;
rptr_wb_addr |= mqd->sdmax_rlcx_rb_rptr_addr_lo;
DPRINTF(PM4PacketProcessor, "SDMAMQD: rb base: %#lx rptr: %#x/%#x wptr: "
"%#x/%#x ib: %#x/%#x size: %d ctrl: %#x\n", mqd->rb_base,
mqd->sdmax_rlcx_rb_rptr, mqd->sdmax_rlcx_rb_rptr_hi,
"%#x/%#x ib: %#x/%#x size: %d ctrl: %#x rptr wb addr: %#lx\n",
mqd->rb_base, mqd->sdmax_rlcx_rb_rptr, mqd->sdmax_rlcx_rb_rptr_hi,
mqd->sdmax_rlcx_rb_wptr, mqd->sdmax_rlcx_rb_wptr_hi,
mqd->sdmax_rlcx_ib_base_lo, mqd->sdmax_rlcx_ib_base_hi,
mqd->sdmax_rlcx_ib_size, mqd->sdmax_rlcx_rb_cntl);
rlc_size, mqd->sdmax_rlcx_rb_cntl, rptr_wb_addr);
// Engine 2 points to SDMA0 while engine 3 points to SDMA1
assert(pkt->engineSel == 2 || pkt->engineSel == 3);
@@ -454,7 +459,8 @@ PM4PacketProcessor::processSDMAMQD(PM4MapQueues *pkt, PM4Queue *q, Addr addr,
// Register RLC queue with SDMA
sdma_eng->registerRLCQueue(pkt->doorbellOffset << 2,
mqd->rb_base << 8);
mqd->rb_base << 8, rlc_size,
rptr_wb_addr);
// Register doorbell with GPU device
gpuDevice->setSDMAEngine(pkt->doorbellOffset << 2, sdma_eng);