Skip to contentSkip to conversation

drm/amdgpu/gfx: use a single pipe for vega10 for now

PATCH In review Alex Deucher 1 participant 2 messages Jun 12, 2017 – Jun 13, 2017 1 message not in the archive

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 drm/amdgpu/gfx: use a single pipe for vega10 for now

Spreading queues across pipes regresses performance right now disable until we root cause the issue.

---
 drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c +5 -1
index dfbf027..6540cd8 100644
@@ -125,7 +125,11 @@ void amdgpu_gfx_compute_queue_acquire(struct amdgpu_device *adev)
 		if (mec >= adev->gfx.mec.num_mec)
 			break;
 
-		if (adev->gfx.mec.num_mec > 1) {
+		if (adev->asic_type >= CHIP_VEGA10) {
+			/* policy: amdgpu owns all queues in the first pipe */
+			if (mec == 0 && pipe == 0)
+				set_bit(i, adev->gfx.mec.queue_bitmap);
+		} else if (adev->gfx.mec.num_mec > 1) {
 			/* policy: amdgpu owns the first two queues of the first MEC */
 			if (mec == 0 && queue < 2)
 				set_bit(i, adev->gfx.mec.queue_bitmap);
Signature · 6 lines
2.5.5

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

A message that is not in the archive (sent off-list or not yet received): 1497295071-3717-1-git-send-email-alexander.deucher-5C7GfCeVMHo@public.gmane.org

#2 lore raw
Quoted 4 lines · Alex Deucher

Spreading queues across pipes regresses performance right now disable until we root cause the issue.

Ignore this patch. Evan found the root cause.

Alex

Quoted 22 lines
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c +5 -2
index dfbf027..6540cd8 100644
@@ -125,7 +125,11 @@ void amdgpu_gfx_compute_queue_acquire(struct amdgpu_device *adev)
                if (mec >= adev->gfx.mec.num_mec)
                        break;
 
-               if (adev->gfx.mec.num_mec > 1) {
+               if (adev->asic_type >= CHIP_VEGA10) {
+                       /* policy: amdgpu owns all queues in the first pipe */
+                       if (mec == 0 && pipe == 0)
+                               set_bit(i, adev->gfx.mec.queue_bitmap);
+               } else if (adev->gfx.mec.num_mec > 1) {
                        /* policy: amdgpu owns the first two queues of the first MEC */
                        if (mec == 0 && queue < 2)
                                set_bit(i, adev->gfx.mec.queue_bitmap);
--
Quoted 2 lines

2.5.5

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx