Honestly, we don't have documentation handy for the exact size required when enabling PG (Alex and I) are trying to sort it out. Without the patch both ST and CZ are unstable with PG enabled. With it so far they both seem to be running fine (I have a glmark2 + various things) running. The final version will be probably somewhere between the two revisions but for now the old revision isn't stable. Tom On Mon, Jul 11, 2016 at 9:57 PM zhoucm1 <david1.zhou at amd.com> wrote: > > > On 2016å¹´07æ??12æ?¥ 05:22, Tom St Denis wrote: > > Was seeing GPU/CPU hangs with previous CP table size. > > With it doubled I've had a full day of uptime while > > running a variety of activities with PG/CG enabled. > What is the root cause? Does the CP table size need 4K alignment? > We need to identify it before pushing. > > Regards, > David Zhou > > > > Signed-off-by: Tom St Denis <tom.stdenis at amd.com> > > --- > > drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c | 2 +- > > 1 file changed, 1 insertion(+), 1 deletion(-) > > > > diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c > b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c > > index 24faddbd9a5d..cbbbd5f1da0f 100644 > > --- a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c > > +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c > > @@ -1293,7 +1293,7 @@ static int gfx_v8_0_rlc_init(struct amdgpu_device > *adev) > > > > if ((adev->asic_type == CHIP_CARRIZO) || > > (adev->asic_type == CHIP_STONEY)) { > > - adev->gfx.rlc.cp_table_size = ALIGN(96 * 5 * 4, 2048) + > (64 * 1024); /* JT + GDS */ > > + adev->gfx.rlc.cp_table_size = (ALIGN(96 * 5 * 4, 2048) + > (64 * 1024)) * 2; /* JT + GDS */ > > if (adev->gfx.rlc.cp_table_obj == NULL) { > > r = amdgpu_bo_create(adev, > adev->gfx.rlc.cp_table_size, PAGE_SIZE, true, > > AMDGPU_GEM_DOMAIN_VRAM, > > -------------- next part -------------- An HTML attachment was scrubbed... URL: <https://lists.freedesktop.org/archives/amd-gfx/attachments/20160712/a7adbe01/attachment.html>