drm/radeon: re-enable selective GPUVM flushing
authorAlex Deucher <alexander.deucher@amd.com>
Tue, 19 Aug 2014 15:48:30 +0000 (11:48 -0400)
committerAlex Deucher <alexander.deucher@amd.com>
Tue, 19 Aug 2014 15:48:30 +0000 (11:48 -0400)
Now that the PFP and ME synchronization is fixed, we
can enable this again reliably.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Tested-by: Michel Dänzer <michel.daenzer@amd.com>
drivers/gpu/drm/radeon/radeon_vm.c

index 832ef320c26de5fd92cb9ad858b9c3aaeb061d70..088ffdc2f577c06519e2a31bc90bbd402c940125 100644 (file)
@@ -238,9 +238,7 @@ void radeon_vm_flush(struct radeon_device *rdev,
        uint64_t pd_addr = radeon_bo_gpu_offset(vm->page_directory);
 
        /* if we can't remember our last VM flush then flush now! */
-       /* XXX figure out why we have to flush all the time before CIK */
-       if (rdev->family < CHIP_BONAIRE ||
-           !vm->last_flush || pd_addr != vm->pd_gpu_addr) {
+       if (!vm->last_flush || pd_addr != vm->pd_gpu_addr) {
                trace_radeon_vm_flush(pd_addr, ring, vm->id);
                vm->pd_gpu_addr = pd_addr;
                radeon_ring_vm_flush(rdev, ring, vm);