dev-amdgpu: Implement GPU clock MMIOs

The ROCr runtime uses a combination of HSA signal timestamps and
hardware MMIOs to calculate profiling times. At the beginning of an
application a timestamp is read from the GPU using MMIOs. The clock
MMIOs reside in the GFX MMIO region, so a new AMDGPUGfx class is added
to handle these MMIOs.

The timestamp value is expected to be in nanoseconds, so we simply use
the gem5 tick converted to ns.

Change-Id: I7d1cba40d5042a7f7a81fd4d132402dc11b71bd4

src/dev/amdgpu/amdgpu_device.cc

@@ -379,6 +379,9 @@ AMDGPUDevice::readMMIO(PacketPtr pkt, Addr offset)
       case GRBM_BASE:
         gpuvm.readMMIO(pkt, aperture_offset >> GRBM_OFFSET_SHIFT);
         break;
+      case GFX_BASE:
+        gfx.readMMIO(pkt, aperture_offset);
+        break;
       case MMHUB_BASE:
         gpuvm.readMMIO(pkt, aperture_offset >> MMHUB_OFFSET_SHIFT);
         break;
@@ -507,6 +510,9 @@ AMDGPUDevice::writeMMIO(PacketPtr pkt, Addr offset)
       case NBIO_BASE:
         nbio.writeMMIO(pkt, aperture_offset);
         break;
+      case GFX_BASE:
+        gfx.writeMMIO(pkt, aperture_offset);
+        break;
       default:
         DPRINTF(AMDGPUDevice, "Unknown MMIO aperture for %#x\n", offset);
         break;