And to get this fully on-topic, is there a specific reason why the dri driver is quite a bit slower (up to 50% in some subtests) in SpecViewPerf compared to the cvs version of july? Could this be related to the changes with GLX_NV_vertex_array_range, GLX_MESA_agp_offset, GLX_MESA_allocate_memory (the only extensions which changed in that timeframe) or must this be something else?
Roland, if you're interested in tracking this down, doing a binary search with cvs to find the point at which things slowed down would be hugely helpful.
Ok, I've tried to narrow it down. It happened between 2003-07-25 and 2003-08-01. (For testing, I've only exchanged r200_dri.so and libGL.so, I didn't install a different 2d driver, drm etc. - mostly because I couldn't get everything to compile - guess cvs update -dP -D date doesn't do the trick). btw why exactly isn't it possible to hot-swap (when xfree86 is running) the dri driver (r200_dri.so)? This kinda works, but kwm, kicker kwhatever insists on crashing shortly afterwards :-(
Oprofile output (see below) shows that r200_emit_state_list together with some check_tcl_xx functions gets called approximately 40 times as much.
Maybe dmatmp2 changes related?
http://marc.theaimsgroup.com/?t=105904425800001&r=1&w=2
Interestingly, the non-wireframe light tests got a small boost at the same time.
Oprofile when running the first subtest of light-06 shows this:
(fast, 2003-07-25):
Counted CPU_CLK_UNHALTED events (Cycles outside of halt state) with a unit mask of 0x00 (No unit mask) count 692898
vma samples % image name symbol name
001563f0 5400 17.6916 r200_dri.so r200_Color3fv_ub
0014bc00 3065 10.0416 r200_dri.so tcl_render_line_loop_verts
080fe6a0 2381 7.8007 viewperf rmqNBVc
000513f0 2109 6.9095 libGL.so.1.2.xf86_glx glVertex3fv
00153e60 1725 5.6515 r200_dri.so r200_Begin
001528e0 1443 4.7276 r200_dri.so note_last_prim
00152580 1368 4.4819 r200_dri.so flush_prims
0804ae24 1324 4.3377 viewperf anonymous symbol from section .plt
0014a0a0 1011 3.3123 r200_dri.so r200_mba_z32
00011fc0 1000 3.2762 libpng12.so.0.1.2.4 png_write_find_filter
000056c0 927 3.0371 libz.so.1.1.4 longest_match
001540c0 891 2.9191 r200_dri.so r200_End
00149de0 813 2.6636 r200_dri.so r200ReadRGBASpan_ARGB8888
0014d340 644 2.1099 r200_dri.so r200TclPrimitive
00152860 577 1.8904 r200_dri.so start_prim
00004ec0 554 1.8150 libz.so.1.1.4 deflate_slow
00050c40 454 1.4874 libGL.so.1.2.xf86_glx glColor3fv
0014d2e0 442 1.4481 r200_dri.so r200EmitPrimitive
0005f110 372 1.2188 libc.so.6 _IO_fread
00143830 315 1.0320 r200_dri.so r200_emit_state_list
000755f0 296 0.9698 libc.so.6 memcpy
00050bd0 284 0.9304 libGL.so.1.2.xf86_glx glBegin
000054d0 212 0.6946 libz.so.1.1.4 fill_window
0804e6f0 182 0.5963 viewperf Swap32
00069df0 157 0.5144 libc.so.6 _IO_file_xsgetn
080546f0 147 0.4816 viewperf triquadinput
00143f30 117 0.3833 r200_dri.so r200EmitAOS
0804fcc0 102 0.3342 viewperf loadFacetColors
0006ac60 99 0.3243 libc.so.6 _IO_sgetn
00006e10 96 0.3145 libpthread.so.0 __pthread_mutex_lock
00009d90 95 0.3112 libpthread.so.0 __pthread_unlock
00004d00 95 0.3112 libpthread.so.0 _pthread_cleanup_pop_restore
00005f10 85 0.2785 libz.so.1.1.4 compress_block
0804e750 85 0.2785 viewperf pcfread
080f8e20 85 0.2785 viewperf rmqBVc
000019b0 83 0.2719 libz.so.1.1.4 adler32
00006fe0 81 0.2654 libpthread.so.0 __pthread_mutex_unlock
00136200 75 0.2457 r200_dri.so check_tcl_light
00136250 75 0.2457 r200_dri.so check_tcl_ucp
00050e10 68 0.2228 libGL.so.1.2.xf86_glx glEnd
00009c00 68 0.2228 libpthread.so.0 __pthread_lock
00009360 66 0.2162 libpthread.so.0 __funlockfile
00004ca0 62 0.2031 libpthread.so.0 _pthread_cleanup_push_defer
000bed90 60 0.1966 r200_dri.so _mesa_read_depth_span_float
00143970 51 0.1671 r200_dri.so r200EmitState
00136090 50 0.1638 r200_dri.so check_tex_any
00136190 44 0.1442 r200_dri.so check_tcl_lighting
00075150 41 0.1343 libc.so.6 memset
00009330 41 0.1343 libpthread.so.0 __flockfile
00143cb0 35 0.1147 r200_dri.so r200AllocEltsOpenEnded
00136150 34 0.1114 r200_dri.so check_tcl_tex
0006c440 33 0.1081 r200_dri.so _mesa_pack_depth_span
00016380 32 0.1048 libc.so.6 anonymous symbol from section .plt
and the same with current cvs (slow, forgot to gather data from 2003-08-01, but the performance was pretty close)
Counted CPU_CLK_UNHALTED events (Cycles outside of halt state) with a unit mask of 0x00 (No unit mask) count 692898
vma samples % image name symbol name
001475d0 13780 20.2838 r200_dri.so r200_emit_state_list
00159c00 6633 9.7636 r200_dri.so r200_Color3fv_ub
00139f20 4695 6.9109 r200_dri.so check_tcl_light
00139f70 4118 6.0616 r200_dri.so check_tcl_ucp
0014fa00 2658 3.9125 r200_dri.so tcl_render_line_loop_verts
00139ef0 2513 3.6991 r200_dri.so check_tcl_lighting
00139e30 2210 3.2531 r200_dri.so check_tex_any
00155d80 2210 3.2531 r200_dri.so flush_prims
080fe6a0 2104 3.0970 viewperf rmqNBVc
00147cd0 1916 2.8203 r200_dri.so r200EmitAOS
00139eb0 1849 2.7217 r200_dri.so check_tcl_tex
00157660 1648 2.4258 r200_dri.so r200_Begin
001560e0 1542 2.2698 r200_dri.so note_last_prim
00147a50 1383 2.0357 r200_dri.so r200AllocEltsOpenEnded
0804ae24 1249 1.8385 viewperf anonymous symbol from section .plt
00147930 1237 1.8208 r200_dri.so r200FlushElts
000512b0 1194 1.7575 libGL.so.1.2.xf86_glx glVertex3fv
00147710 1186 1.7458 r200_dri.so r200EmitState
000056c0 1029 1.5147 libz.so.1.1.4 longest_match
0014df80 1006 1.4808 r200_dri.so r200_mba_z32
00139e50 932 1.3719 r200_dri.so check_tex
00011fc0 918 1.3513 libpng12.so.0.1.2.4 png_write_find_filter
00150cb0 890 1.3101 r200_dri.so r200EmitPrimitive
0014dcc0 812 1.1952 r200_dri.so r200ReadRGBASpan_ARGB8888
00139e70 780 1.1481 r200_dri.so check_fog
00150d10 640 0.9421 r200_dri.so r200TclPrimitive
00156060 621 0.9141 r200_dri.so start_prim
0014f520 596 0.8773 r200_dri.so r200AllocElts
00004ec0 533 0.7846 libz.so.1.1.4 deflate_slow
001578c0 519 0.7640 r200_dri.so r200_End
000755f0 445 0.6550 libc.so.6 memcpy
00050b00 425 0.6256 libGL.so.1.2.xf86_glx glColor3fv
0005f110 352 0.5181 libc.so.6 _IO_fread
080f8e20 340 0.5005 viewperf rmqBVc
00050a90 255 0.3754 libGL.so.1.2.xf86_glx glBegin
000054d0 213 0.3135 libz.so.1.1.4 fill_window
0804e6f0 194 0.2856 viewperf Swap32
080546f0 146 0.2149 viewperf triquadinput
00069df0 132 0.1943 libc.so.6 _IO_file_xsgetn
0804e750 124 0.1825 viewperf pcfread
0804fcc0 111 0.1634 viewperf loadFacetColors
00004d00 101 0.1487 libpthread.so.0 _pthread_cleanup_pop_restore
00006e10 100 0.1472 libpthread.so.0 __pthread_mutex_lock
00009d90 92 0.1354 libpthread.so.0 __pthread_unlock
0006ac60 91 0.1339 libc.so.6 _IO_sgetn
0014fff0 89 0.1310 r200_dri.so tcl_render_quads_verts
00006fe0 86 0.1266 libpthread.so.0 __pthread_mutex_unlock
00005f10 83 0.1222 libz.so.1.1.4 compress_block
00050cd0 79 0.1163 libGL.so.1.2.xf86_glx glEnd
000019b0 79 0.1163 libz.so.1.1.4 adler32
------------------------------------------------------- This SF.net email is sponsored by: SF.net Giveback Program. SourceForge.net hosts over 70,000 Open Source Projects. See the people who have HELPED US provide better services: Click here: http://sourceforge.net/supporters.php _______________________________________________ Dri-devel mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/dri-devel
