20:52 <aka_[m]> lumag: im scrolling around my old code and noticed Vladmir once noted nb_stages might be off in mdp5_cfg.c... (full message at <https://matrix.org/_matrix/media/v3/download/matrix.org/YcCLpNmqrDumaLCecfGcFjPh>)

20:59 <lumag> aka_[m], well. According to the docs I have here, 8916 has base, stage0-stage3. which gives us 5 for nb_stages

20:59 <aka_[m]> also i believe 8953 sets wrong perf factors

21:01 <aka_[m]> 8916 has these... (full message at <https://matrix.org/_matrix/media/v3/download/matrix.org/VJGLcakbipVMSkMihQcnuGFB>)

21:01 <aka_[m]> .ib_inefficiency should be set to 100 for 8953

21:03 <lumag> correct for ib_inefficiency

21:05 <aka_[m]> lumag: current is correct or 100?

21:05 <lumag> 100

21:05 <aka_[m]> so 8x96 is also wrong probably

21:06 <aka_[m]> https://android.googlesource.com/kernel/msm.git/+/android-7.1.0_r0.3/arch/arm/boot/dts/qcom/msm8996-mdss.dtsi#39

21:06 <aka_[m]> unless they hardcode it in driver

21:08 <aka_[m]> which they do in mdss_mdp_parse_dt_misc

21:09 <aka_[m]> its all crap because there is no single shared mdss_mdp and quite often they hardcode only specific verisons in techpacks

21:19 <lumag> I believe msm8996 also should have ib_inefficiency 100

21:19 <lumag> Maybe 120 to account for small inefficiencies

21:31 <lumag> aka_[m], I think 8x36 also has 5 stages (and not 8)

21:31 <lumag> And 8917

21:32 <lumag> Let's move that to freedreno

21:34 <lumag> Hmm, you are not on #freedreno

21:34 <lumag> robclark, ^^

21:35 <lumag> Do you remember, why 8916 got 8 stages?