This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [patch] for PR 18316 (updated)
Hello,
> The rest of the patch looks reasonable. Do you have any benchmark
> figures on the impact of this patch? We've had problems in the past
> with the number of induction variables we find and their effect on
> register pressure. If there's no (large) performance degradation on
> SPEC with this change, I'll approve it for mainline.
here are the spec results from i686 (at -O2, peak with the patch).
Zdenek
164.gzip 1400 198 706 * 1400 196 713 *
175.vpr 1400 392 358 * 1400 392 357 *
176.gcc X X
181.mcf 1800 733 246 * 1800 733 246 *
186.crafty 1000 118 848 * 1000 118 847 *
197.parser 1800 394 457 * 1800 394 457 *
252.eon 1300 142 918 * 1300 142 918 *
253.perlbmk 1800 203 887 * 1800 203 889 *
254.gap 1100 176 624 * 1100 175 629 *
255.vortex 1900 234 811 * 1900 234 811 *
256.bzip2 1500 336 447 * 1500 337 445 *
300.twolf 3000 784 383 * 3000 784 383 *
Est. SPECint_base2000 560
Est. SPECint2000 560
168.wupwise 1600 288 555 * 1600 279 574 *
171.swim 3100 653 475 * 3100 710 437 *
172.mgrid 1800 508 355 * 1800 709 254 *
173.applu 2100 573 367 * 2100 572 367 *
177.mesa 1400 200 698 * 1400 202 694 *
178.galgel X X
179.art 2600 1463 178 * 2600 1466 177 *
183.equake 1300 292 446 * 1300 291 447 *
187.facerec 1900 474 401 * 1900 469 405 *
188.ammp 2200 607 362 * 2200 608 362 *
189.lucas 2000 430 465 * 2000 431 464 *
191.fma3d 2100 395 532 * 2100 397 529 *
200.sixtrack 1100 235 468 * 1100 235 468 *
301.apsi 2600 689 377 * 2600 659 394 *
Est. SPECfp_base2000 418
Est. SPECfp2000 407