On Thu, 2015-02-26 at 08:04 +0000, Chris Wilson wrote: > On Thu, Feb 26, 2015 at 01:29:09PM +0530, akash.goel@xxxxxxxxx wrote: > > From: Akash Goel <akash.goel@xxxxxxxxx> > > > > The frequency values(Rp0, Rp1, Rpn) reported by RP_STATE_CAP register > > are stored, initially by the Driver, inside the dev_priv->rps structure. > > Since these values are expected to remain same throughout, there is no real > > need to read this register, on dynamic basis, from debugfs/sysfs functions > > and the values can be instead retrieved from the dev_priv->rps structure > > when needed. > > Inside debugfs, you want both as we want to compare what the hardware > thinks and what we believe it should be. Everywhere else we should > indeed be trusting our bookkeeping. Agree, ideally we should throw an error, if the stored values do not match the ones reported by Hw and report the values read from the registers to User. Actually currently there is a bit of discrepancy. For VLV/CHV, directly the stored values only are being returned to the User. Even for non VLV also, there is inconsistency in the way RP_STATE_CAP register is being used inside i915_max_freq_set/i915_min_freq_set pair of debugfs functions. So tried to bring uniformity by using the stored values everywhere. Best regards Akash > -Chris > _______________________________________________ Intel-gfx mailing list Intel-gfx@xxxxxxxxxxxxxxxxxxxxx http://lists.freedesktop.org/mailman/listinfo/intel-gfx
