diff options
| author | Chris Lattner <sabre@nondot.org> | 2002-11-08 22:29:30 +0000 |
|---|---|---|
| committer | Chris Lattner <sabre@nondot.org> | 2002-11-08 22:29:30 +0000 |
| commit | 2779eff8b00b7ca2976465b900c14b703381dea0 (patch) | |
| tree | 4181a223abeec661a6bc5b306b26399f0bc0f450 /docs | |
| parent | 02fd6870c1e375bc1f1a51f1aad8f04c1dc27a2c (diff) | |
| download | external_llvm-2779eff8b00b7ca2976465b900c14b703381dea0.zip external_llvm-2779eff8b00b7ca2976465b900c14b703381dea0.tar.gz external_llvm-2779eff8b00b7ca2976465b900c14b703381dea0.tar.bz2 | |
Add a bunch more numbers
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@4642 91177308-0d34-0410-b5e6-96231b3b80d8
Diffstat (limited to 'docs')
| -rw-r--r-- | docs/DSGraphStatus.html | 126 |
1 files changed, 122 insertions, 4 deletions
diff --git a/docs/DSGraphStatus.html b/docs/DSGraphStatus.html index 48c40cd..6e46924 100644 --- a/docs/DSGraphStatus.html +++ b/docs/DSGraphStatus.html @@ -6,7 +6,125 @@ This page records information and the progress the data structure analysis passes have been making. Times are in seconds, sizes are #bytes allocated for a particular pass. Runs that are all '*'d out were not completed because they -were taking too long.<p> +were taking too long (time listed show how long I waited).<p> +<hr> +<a name="20021108-2"><h2>Nov 8, 2002 #2:</h2> + +Same as the <a href="#20021108">previous run</a>, but with a huge bug fix: +before, all of the call nodes (from the local pass) were being passed up in the +BU pass, causing graph exposions. Now these call sites are not propogated. The +times and sized of graphs are all much smaller now.<p> + +<pre> +Name: Anlyz: LocTime: BUTime: TDTime: TotTime: LocSize: BUSize: TDSize: TotSize: NumFold NumNodes main __main +254.gap.lib 130.47 * * * * | * * * * | * * * * +255.vortex.lib 86.17 * * * * | * * * * | * * * * +197.parser.lib 26.50 0.1601 1.4931 22.7485 24.7647 | 2187656 4037056 57704672 64165544 | 4550 196932+1184 601+0 601+0 +300.twolf.lib 14.33 0.2311 1.0000 10.1733 12.1591 | 629008 1085616 15564712 17279336 | 4655 161528+1101 854+0 854+0 +burg.llvm.lib 6.71 0.0904 0.3799 5.1871 5.8809 | 745152 1254280 6822216 8821648 | 3253 72533+902 436+0 436+0 +164.gzip.llvm.lib 6.07 0.1158 0.9992 4.1925 5.3966 | 2113120 2772872 9043760 14026344 | 233 16632+232 171+0 171+0 +256.bzip2.lib 2.16 0.0239 0.1987 1.6047 1.8899 | 372776 1059888 7503640 9009560 | 132 11456+201 148+0 148+0 +sim.llvm.lib 1.54 0.0163 0.2356 1.0723 1.3738 | 106096 124280 400760 726152 | 41 2739+67 99+0 99+0 +optimizer-eval.llvm.lib 0.58 0.0126 0.0890 0.3282 0.4696 | 110464 188584 1055832 1373072 | 287 3693+101 63+0 63+0 +181.mcf.lib 0.33 0.0155 0.0465 0.1437 0.2308 | 156864 225184 805584 1204520 | 209 1914+72 44+0 44+0 +voronoi.llvm.lib 0.19 0.0111 0.0253 0.0644 0.1178 | 83248 93616 238648 429504 | 31 1733+106 30+0 30+0 +sgefa.llvm.lib 0.18 0.0069 0.0398 0.0492 0.1156 | 104712 108472 178920 416872 | 67 634+55 31+0 31+0 +bh.llvm.lib 0.16 0.0100 0.0177 0.0355 0.0838 | 81376 93656 190104 428952 | 66 715+64 13+0 13+0 +health.llvm.lib 0.10 0.0230 0.0084 0.0138 0.0551 | 31232 31552 56984 134736 | 75 361+40 14+0 14+0 +em3d.llvm.lib 0.09 0.0045 0.0117 0.0255 0.0519 | 37336 67808 126960 248728 | 50 1031+59 28+0 28+0 +power.llvm.lib 0.07 0.0035 0.0053 0.0105 0.0287 | 23432 31680 42680 113616 | 3 277+35 9+0 9+0 +mst.llvm.lib 0.07 0.0033 0.0060 0.0145 0.0333 | 29080 31400 56720 131872 | 35 470+36 21+0 21+0 +lists.llvm.lib 0.07 0.0079 0.0129 0.0170 0.0453 | 27616 39648 75280 154864 | 54 246+34 10+0 10+0 +indvars.llvm.lib 0.07 0.0020 0.0257 0.0269 0.0569 | 85496 87560 161216 334272 | * 18+3 1+0 1+0 +tsp.llvm.lib 0.06 0.0041 0.0049 0.0065 0.0250 | 19360 26216 26712 86912 | 23 247+31 10+0 10+0 +perimeter.llvm.lib 0.05 0.0024 0.0045 0.0039 0.0194 | 15336 22904 16288 67184 | * 105+24 4+0 4+0 +hash.llvm.lib 0.05 0.0029 0.0043 0.0066 0.0211 | 25880 18176 28168 85624 | 2 241+23 7+0 7+0 +bisort.llvm.lib 0.05 0.0022 0.0036 0.0064 0.0192 | 14376 28616 22416 76480 | 3 210+27 10+0 10+0 +llubenchmark.llvm.lib 0.04 0.0017 0.0026 0.0045 0.0169 | 11824 24304 18168 65304 | 8 144+20 15+0 15+0 +treeadd.llvm.lib 0.03 0.0013 0.0016 0.0023 0.0101 | 9080 8408 11224 40704 | 2 76+14 5+0 5+0 +sumarray2d.llvm.lib 0.03 0.0014 0.0097 0.0106 0.0238 | 42056 46440 83192 174008 | * 15+3 1+0 1+0 +sieve.llvm.lib 0.03 0.0008 0.0021 0.0032 0.0103 | 16064 14264 35312 76008 | * 30+4 2+0 2+0 +objinst.llvm.lib 0.03 0.0015 0.0024 0.0037 0.0134 | 11032 15488 11440 51616 | 39 73+17 3+0 3+0 +methcall.llvm.lib 0.03 0.0014 0.0018 0.0030 0.0115 | 10824 9704 11128 45312 | 31 58+14 2+0 2+0 +matrix.llvm.lib 0.03 0.0011 0.0017 0.0018 0.0109 | 8584 17680 4208 46232 | * 46+11 1+0 1+0 +ary3.llvm.lib 0.03 0.0010 0.0014 0.0014 0.0085 | 8752 17136 4168 45720 | 6 50+7 1+0 1+0 +sumarraymalloc.llvm.lib 0.02 0.0007 0.0009 0.0013 0.0070 | 2720 12824 3376 33352 | * 38+8 2+0 2+0 +random.llvm.lib 0.02 0.0006 0.0009 0.0009 0.0060 | 6184 1808 2752 23600 | * 31+5 3+0 3+0 +pi.llvm.lib 0.02 0.0006 0.0007 0.0011 0.0049 | 5696 5272 6544 17512 | * 34+6 3+0 3+0 +matrixTranspose.llvm.lib 0.02 0.0007 0.0033 0.0034 0.0095 | 14088 9424 29480 52992 | * 17+4 3+0 3+0 +heapsort.llvm.lib 0.02 0.0008 0.0009 0.0013 0.0076 | 2936 7224 3912 29240 | * 45+6 3+0 3+0 +fib2.llvm.lib 0.02 0.0014 0.0008 0.0009 0.0067 | 6176 1808 2400 23008 | * 25+6 1+0 1+0 +sumarray.llvm.lib 0.01 0.0003 0.0003 0.0003 0.0022 | 1280 3552 584 7528 | 3 7+2 0+0 0+0 +printargs.llvm.lib 0.01 0.0005 0.0004 0.0006 0.0030 | 1360 816 4760 6936 | * 13+4 1+0 1+0 +hello.llvm.lib 0.01 0.0003 0.0003 0.0003 0.0022 | 880 592 2944 6672 | * 8+3 1+0 1+0 +ackermann.llvm.lib 0.01 0.0000 0.0000 0.0001 0.0004 | 0 0 832 832 | * 0+0 * * +</pre><p> + +<hr> +<a name="20021108"><h2>Nov 8, 2002:</h2> + +Same as the <a href="#20021107-2">previous run</a>, but now the Top-Down pass +has been rewritten to:<p> + +<ul> +<li>Actually work on all of the benchmarks +<li>Require less book-keeping +<li>Only clone a caller graph into a particular callee graph once +<li>Not prune nodes as aggressively, this is probably the reason for many + slowdowns below. +</ul><p> + +Looking deeper into the results, gzip got slower by a factor of 4, optimizer +eval got slower by a factor of 3, sim slowed down just a little bit.<p> + +Note that without the globals graph, some benchmarks are still VERY slow (those +with heavy use of globals).<p> + +<pre> +Name: Anlyz: LocTime: BUTime: TDTime: TotTime: LocSize: BUSize: TDSize: TotSize: NumFold NumNodes main __main +254.gap.lib 330.50 * * * * | * * * * | * * * * +255.vortex.lib * * * * * | * * * * | * * * * +197.parser.lib 120.10 0.1649 11.8198 105.3900 118.0119 | 2188536 4525320 60972272 67921544 | 22575 202604+8604 601+0 601+0 +300.twolf.lib 72.38 0.2431 1.8500 30.5671 72.3845 | 629008 1375352 16515832 18520192 | 18090 163168+3906 854+0 854+0 +burg.llvm.lib 34.18 0.0871 1.0919 15.8467 34.1754 | 749304 1755232 7209416 9779360 | 21707 78464+4387 393+0 393+0 +164.gzip.llvm.lib 21.81 0.1149 1.2083 19.6981 21.1201 | 2055024 3145544 8929008 14227896 | 2387 18236+1588 171+0 171+0 +256.bzip2.lib 7.44 0.0250 0.3032 6.2987 7.4430 | 372776 1205520 7660752 9312304 | 1765 11947+1011 148+0 148+0 +optimizer-eval.llvm.lib 3.24 0.0131 1.0648 1.9117 3.0792 | 110456 343032 1971952 2443624 | 5293 4186+1033 63+0 63+0 +sim.llvm.lib 2.27 0.0170 0.2274 1.7720 2.0708 | 104728 130376 451320 759088 | 62 2755+122 99+0 99+0 +181.mcf.lib 0.61 0.0162 0.0556 0.3631 0.4646 | 156648 234864 866704 1275064 | 243 2023+163 44+0 44+0 +voronoi.llvm.lib 0.45 0.0117 0.0447 0.2557 0.3385 | 83088 150936 337552 585520 | 157 2063+390 30+0 30+0 +bh.llvm.lib 0.38 0.0101 0.0253 0.1778 0.2394 | 79736 129328 245408 502512 | 66 858+198 13+0 13+0 +sgefa.llvm.lib 0.27 0.0071 0.0396 0.0780 0.1465 | 104712 112576 204600 446656 | 130 640+103 31+0 31+0 +power.llvm.lib 0.14 0.0035 0.0080 0.0451 0.0675 | 31272 49768 66968 155680 | 6 359+96 9+0 9+0 +em3d.llvm.lib 0.13 0.0046 0.0140 0.0524 0.0825 | 37096 65320 122624 241744 | 141 920+127 21+0 21+0 +mst.llvm.lib 0.13 0.0033 0.0084 0.0381 0.0605 | 29104 36104 70496 150376 | 100 506+91 21+0 21+0 +lists.llvm.lib 0.12 0.0036 0.0118 0.0328 0.0664 | 27880 33064 98728 177520 | 58 270+63 10+0 10+0 +bisort.llvm.lib 0.11 0.0023 0.0047 0.0125 0.0270 | 14376 29520 33808 88776 | 6 226+42 10+0 10+0 +indvars.llvm.lib 0.11 0.0020 0.0254 0.0310 0.0606 | 85520 88232 161216 334968 | * 18+3 1+0 1+0 +health.llvm.lib 0.10 0.0044 0.0113 0.0255 0.0519 | 31232 44592 53456 144248 | 79 385+69 14+0 14+0 +tsp.llvm.lib 0.10 0.0042 0.0062 0.0176 0.0382 | 27840 19008 51376 112528 | 27 271+55 10+0 10+0 +hash.llvm.lib 0.10 0.0026 0.0056 0.0168 0.0336 | 25880 29936 35312 104528 | 6 271+44 7+0 7+0 +heapsort.llvm.lib 0.09 0.0007 0.0009 0.0015 0.0429 | 2936 7384 4064 29552 | * 47+7 3+0 3+0 +perimeter.llvm.lib 0.08 0.0025 0.0064 0.0090 0.0271 | 15432 25616 18552 72440 | * 121+39 4+0 4+0 +objinst.llvm.lib 0.08 0.0015 0.0027 0.0076 0.0180 | 11032 16088 19848 60624 | 39 78+25 3+0 3+0 +sumarray2d.llvm.lib 0.06 0.0011 0.0098 0.0111 0.0241 | 42056 46440 83192 174008 | * 15+3 1+0 1+0 +ary3.llvm.lib 0.06 0.0010 0.0016 0.0021 0.0096 | 8752 17528 4696 46640 | 6 55+10 1+0 1+0 +llubenchmark.llvm.lib 0.05 0.0018 0.0034 0.0063 0.0199 | 11824 20384 28056 71272 | 32 141+31 15+0 15+0 +treeadd.llvm.lib 0.05 0.0012 0.0023 0.0036 0.0121 | 9120 17640 10960 53880 | 4 82+18 5+0 5+0 +sumarray.llvm.lib 0.04 0.0003 0.0003 0.0003 0.0022 | 1280 3552 584 7528 | 3 7+2 0+0 0+0 +methcall.llvm.lib 0.03 0.0014 0.0021 0.0045 0.0147 | 10680 16000 11784 51720 | 31 63+19 2+0 2+0 +sieve.llvm.lib 0.03 0.0008 0.0022 0.0047 0.0120 | 16064 14528 39672 80632 | * 32+5 2+0 2+0 +fib2.llvm.lib 0.03 0.0007 0.0009 0.0011 0.0063 | 5512 6088 2544 26432 | * 27+7 1+0 1+0 +matrix.llvm.lib 0.02 0.0010 0.0017 0.0021 0.0100 | 8760 8104 8976 42184 | * 48+12 1+0 1+0 +matrixTranspose.llvm.lib 0.02 0.0007 0.0034 0.0037 0.0098 | 14088 9424 29480 52992 | * 17+4 3+0 3+0 +sumarraymalloc.llvm.lib 0.02 0.0007 0.0010 0.0016 0.0075 | 2720 12928 3536 33544 | * 40+9 2+0 2+0 +pi.llvm.lib 0.02 0.0006 0.0007 0.0014 0.0052 | 5696 5368 5680 16744 | * 35+7 3+0 3+0 +random.llvm.lib 0.02 0.0006 0.0007 0.0012 0.0061 | 2136 6040 7248 31176 | * 33+6 3+0 3+0 +printargs.llvm.lib 0.02 0.0003 0.0004 0.0007 0.0030 | 1360 1024 1728 4112 | * 14+5 1+0 1+0 +hello.llvm.lib 0.02 0.0003 0.0003 0.0005 0.0023 | 880 648 768 4552 | * 9+4 1+0 1+0 +ackermann.llvm.lib 0.02 0.0000 0.0000 0.0002 0.0004 | 0 0 832 832 | * 0+0 * * +</pre><p> + <hr> <a name="20021107-2"><h2>Nov 7, 2002 #2:</h2> @@ -18,12 +136,12 @@ gzip significantly.<p> <pre> Name: Anlyz: LocTime: BUTime: TDTime: TotTime: LocSize: BUSize: TDSize: TotSize: NumFold NumNodes main __main +300.twolf.lib 30.88 * * * * | * * * * | * * * * +255.vortex.lib 15.71 * * * * | * * * * | * * * * 254.gap.lib 6.07 analyze: ../../../include/llvm/Analysis/DSNode.h:77: DSNode::~DSNode(): Assertion `Referrers.empty() && "Referrers to dead node exist!"' failed. 164.gzip.llvm.lib 5.32 0.1157 1.0296 3.4025 4.6326 | 2054352 2631576 7804656 12588904 | 398 13372+217 174+3 174+3 -300.twolf.lib 30.88 * * * * | * * * * | * * * * 256.bzip2.lib 2.36 0.0250 0.1913 1.8190 2.0966 | 371928 1045320 7533440 9023600 | 407 11415+149 151+2 151+2 sim.llvm.lib 1.97 0.0174 0.2155 1.4731 1.7539 | 104584 120112 449736 747416 | 46 2745+52 101+2 101+2 -255.vortex.lib 15.71 * * * * | * * * * | * * * * 197.parser.lib 1.27 analyze: ../../../include/llvm/Analysis/DSNode.h:77: DSNode::~DSNode(): Assertion `Referrers.empty() && "Referrers to dead node exist!"' failed. optimizer-eval.llvm.lib 1.09 0.0132 0.0926 0.8104 0.9528 | 109416 186592 1827184 2141376 | 1709 8189+43 63+0 63+0 burg.llvm.lib 0.81 analyze: ../../../include/llvm/Analysis/DSNode.h:77: DSNode::~DSNode(): Assertion `Referrers.empty() && "Referrers to dead node exist!"' failed. @@ -171,7 +289,7 @@ burg.llvm.lib analyze: ../../../include/llvm/Analysis/DSNode.h:7 <address><a href="mailto:sabre@nondot.org">Chris Lattner</a></address> <!-- Created: Wed Nov 6 19:57:57 CST 2002 --> <!-- hhmts start --> -Last modified: Thu Nov 7 23:00:33 CST 2002 +Last modified: Fri Nov 8 16:26:15 CST 2002 <!-- hhmts end --> </body> </html> |