From mboxrd@z Thu Jan 1 00:00:00 1970 From: Eric Anholt Subject: Re: [PATCH] drm/i915: set IDICOS to medium uncore resources Date: Wed, 13 Jun 2012 23:36:25 -0700 Message-ID: <87y5nqbdsm.fsf@eliezer.anholt.net> References: <1336183139-10286-1-git-send-email-ben@bwidawsk.net> <877gwoo55d.fsf@eliezer.anholt.net> Mime-Version: 1.0 Content-Type: multipart/mixed; boundary="===============1281393954==" Return-path: In-Reply-To: <877gwoo55d.fsf@eliezer.anholt.net> List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Sender: intel-gfx-bounces+gcfxdi-intel-gfx=m.gmane.org@lists.freedesktop.org Errors-To: intel-gfx-bounces+gcfxdi-intel-gfx=m.gmane.org@lists.freedesktop.org To: intel-gfx@lists.freedesktop.org Cc: Ben Widawsky , "Kilarski, Bernard R" List-Id: intel-gfx@lists.freedesktop.org --===============1281393954== Content-Type: multipart/signed; boundary="=-=-="; micalg=pgp-sha1; protocol="application/pgp-signature" --=-=-= Content-Transfer-Encoding: quoted-printable On Sun, 06 May 2012 23:51:26 -0700, Eric Anholt wrote: > On Fri, 4 May 2012 18:58:59 -0700, Ben Widawsky wrote: > > I'm seeing about a 5% FPS improvement across various benchmarks on my > > IVB i3. Rumor has it that the higher end parts show even more benefit. > >=20 > > This derives from a patch originally given to me by Bernard. The docs > > are confusing about the definition names (ie. medium really seems like > > max), but it would seem it gives more cache to the GT at the expense of > > uncore. This configuration makes the split most in favor of the GT. I've > > not tried the other IDICOS values. >=20 > My IVB system started with an SNBPCR value of 0x1b40. To test the effect= of > this patch, I just smashed the value of the reg at runtime: >=20 > Results: > x intel_reg_write 0x900c 0x001b40 > + intel_reg_write 0x900c 0x201b40 >=20 > citybench: > +------------------------------------------------------------------------= ------+ > | xxx + = | > | x xxx + + + = | > | x xxx + +++ = | > | x xxx + + ++++ = | > | xx xxxx + + ++++ = | > |x xxx x xxxx + + ++++++ ++ = | > |xxxxx xxxxxxxx + + + + +*++++ ++ = | > |xxxxxxxxx xxxxxxxx x +++++ ++++ +*+++++++ = | > |xxxxxxxxx xxxxxxxxxx x ++++++++++ ++**+++++++ + + = | > |xxxxxxxxx xxxxxxxxxxxxxx *++++++++++++ ****+*++++++++ + ++++++= +| > | |___________AM_________| |_________A_M_______| = | > +------------------------------------------------------------------------= ------+ > N Min Max Median Avg Stdd= ev > x 118 335.779 387.456 354.306 352.63131 12.4627= 96 > + 118 367.324 419.02 386.14 384.47139 10.9709= 18 > Difference at 95.0% confidence > 31.8401 +/- 2.99585 > 9.02928% +/- 0.849569% > (Student's t, pooled s =3D 11.7406) >=20 > openarena: > +------------------------------------------------------------------------= ------+ > | + ++ xx xx x x x= x | > | + + ++ + + xxxxxxx xxxx= x | > | +++ ++++++++++ ++++++ ++ + x xxxxxxxxxxxxxx= x xx| > |++ + ++++++++++++++++++++++++++++ + + + x x xxxxxxxxxxxxxxx= xxx xx| > | |_________A________| |_____A____|= | > +------------------------------------------------------------------------= ------+ > N Min Max Median Avg Stdd= ev > x 65 273.5 292.2 284 283.73385 4.25618= 22 > + 65 234.8 263.2 249.9 249.6 6.98733= 23 > Difference at 95.0% confidence > -34.1338 +/- 1.989 > -12.0302% +/- 0.70101% >=20 > nexuiz: > +------------------------------------------------------------------------= ------+ > | + = | > | x + + = | > | + x x xx *x+ +x = | > |+ + + x *x+x xxx + x++ + x+*xxxx+x***x+++*x**+x = ++ *+| > | |____|____________AAM____M_____|_____= _| | > +------------------------------------------------------------------------= ------+ > N Min Max Median Avg Stdd= ev > x 33 74.314888 76.177197 75.424764 75.332161 0.482058= 91 > + 33 73.397254 76.219868 75.584749 75.363209 0.694085= 23 >=20 > Using these 3 tests, this patch does not look like a win. Note that OA > is the most CPU-bound of these, while nexuiz is usually on the edge > between being CPU and GPU-bound, and citybench is definitely GPU > limited. Re-testing, openarena (1024x768) currently is about a 2.5% +/- 0.5% loss (lost the numbers). Also, new tropics numbers: N Min Max Median Avg Stddev x 70 21.9718 22.2166 22.1556 22.154501 0.029139432 + 69 22.7246 22.9308 22.8914 22.888735 0.027248939 Difference at 95.0% confidence 0.734233 +/- 0.00938209 3.31415% +/- 0.0423485% (Student's t, pooled s =3D 0.0282169) That swings things in favor of the patch for me. --=-=-= Content-Type: application/pgp-signature -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.12 (GNU/Linux) iEYEARECAAYFAk/ZhmkACgkQHUdvYGzw6vdAXQCeKDbWwNcgX1obEC3Ohlgp073+ HDsAnjv9FDl3SZm4OYtvF5yCAtscbg4G =Q57H -----END PGP SIGNATURE----- --=-=-=-- --===============1281393954== Content-Type: text/plain; charset="us-ascii" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit Content-Disposition: inline _______________________________________________ Intel-gfx mailing list Intel-gfx@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/intel-gfx --===============1281393954==--