This is the mail archive of the
gcc-regression@gcc.gnu.org
mailing list for the GCC project.
A recent patch increased GCC's memory consumption in some cases!
- From: gcctest at suse dot de
- To: jh at suse dot cz, gcc-regression at gcc dot gnu dot org
- Date: Tue, 09 Sep 2008 06:25:52 +0000
- Subject: A recent patch increased GCC's memory consumption in some cases!
Hi,
I am a friendly script caring about memory consumption in GCC. Please
contact jh@suse.cz if something is going wrong.
Comparing memory consumption on compilation of combine.i, insn-attrtab.i,
and generate-3.4.ii I got:
comparing empty function compilation at -O0 level:
Ovarall memory allocated via mmap and sbrk decreased from 8475k to 8184k, overall -3.56%
Peak amount of GGC memory allocated before garbage collecting run decreased from 1408k to 1291k, overall -9.06%
Peak amount of GGC memory still allocated after garbage collecting decreased from 1338k to 1217k, overall -9.94%
Amount of produced GGC garbage decreased from 249k to 218k, overall -14.19%
Amount of memory still referenced at the end of compilation decreased from 1331k to 1222k, overall -8.97%
Overall memory needed: 8475k -> 8184k
Peak memory use before GGC: 1408k -> 1291k
Peak memory use after GGC: 1338k -> 1217k
Maximum of released memory in single GGC run: 173k -> 134k
Garbage: 249k -> 218k
Leak: 1331k -> 1222k
Overhead: 159k -> 136k
GGC runs: 3 -> 4
Amount of produced pre-ipa-GGC garbage decreased from 238k to 207k, overall -14.71%
Amount of memory referenced pre-ipa decreased from 1333k to 1224k, overall -8.94%
Pre-IPA-Garbage: 238k -> 207k
Pre-IPA-Leak: 1333k -> 1224k
Pre-IPA-Overhead: 157k -> 135k
Amount of produced post-ipa-GGC garbage decreased from 238k to 207k, overall -14.71%
Amount of memory referenced post-ipa decreased from 1333k to 1224k, overall -8.94%
Post-IPA-Garbage: 238k -> 207k
Post-IPA-Leak: 1333k -> 1224k
Post-IPA-Overhead: 157k -> 135k
comparing empty function compilation at -O0 -g level:
Peak amount of GGC memory allocated before garbage collecting run decreased from 1436k to 1319k, overall -8.87%
Peak amount of GGC memory still allocated after garbage collecting decreased from 1365k to 1245k, overall -9.64%
Amount of produced GGC garbage decreased from 252k to 220k, overall -14.55%
Amount of memory still referenced at the end of compilation decreased from 1364k to 1254k, overall -8.75%
Overall memory needed: 8503k -> 8428k
Peak memory use before GGC: 1436k -> 1319k
Peak memory use after GGC: 1365k -> 1245k
Maximum of released memory in single GGC run: 167k -> 133k
Garbage: 252k -> 220k
Leak: 1364k -> 1254k
Overhead: 163k -> 141k
GGC runs: 4
Amount of produced pre-ipa-GGC garbage decreased from 238k to 207k, overall -14.71%
Amount of memory referenced pre-ipa decreased from 1333k to 1224k, overall -8.94%
Pre-IPA-Garbage: 238k -> 207k
Pre-IPA-Leak: 1333k -> 1224k
Pre-IPA-Overhead: 157k -> 135k
Amount of produced post-ipa-GGC garbage decreased from 238k to 207k, overall -14.71%
Amount of memory referenced post-ipa decreased from 1333k to 1224k, overall -8.94%
Post-IPA-Garbage: 238k -> 207k
Post-IPA-Leak: 1333k -> 1224k
Post-IPA-Overhead: 157k -> 135k
comparing empty function compilation at -O1 level:
Ovarall memory allocated via mmap and sbrk decreased from 8547k to 8240k, overall -3.73%
Peak amount of GGC memory allocated before garbage collecting run decreased from 1408k to 1291k, overall -9.06%
Peak amount of GGC memory still allocated after garbage collecting decreased from 1338k to 1217k, overall -9.94%
Amount of produced GGC garbage decreased from 252k to 221k, overall -13.94%
Amount of memory still referenced at the end of compilation decreased from 1331k to 1222k, overall -8.96%
Overall memory needed: 8547k -> 8240k
Peak memory use before GGC: 1408k -> 1291k
Peak memory use after GGC: 1338k -> 1217k
Maximum of released memory in single GGC run: 168k -> 134k
Garbage: 252k -> 221k
Leak: 1331k -> 1222k
Overhead: 160k -> 137k
GGC runs: 4
Amount of produced pre-ipa-GGC garbage decreased from 238k to 207k, overall -14.71%
Amount of memory referenced pre-ipa decreased from 1334k to 1224k, overall -8.94%
Pre-IPA-Garbage: 238k -> 207k
Pre-IPA-Leak: 1334k -> 1224k
Pre-IPA-Overhead: 157k -> 135k
Amount of produced post-ipa-GGC garbage decreased from 238k to 207k, overall -14.71%
Amount of memory referenced post-ipa decreased from 1334k to 1224k, overall -8.94%
Post-IPA-Garbage: 238k -> 207k
Post-IPA-Leak: 1334k -> 1224k
Post-IPA-Overhead: 157k -> 135k
comparing empty function compilation at -O2 level:
Peak amount of GGC memory allocated before garbage collecting run decreased from 1409k to 1291k, overall -9.14%
Peak amount of GGC memory still allocated after garbage collecting decreased from 1338k to 1218k, overall -9.85%
Amount of produced GGC garbage decreased from 255k to 226k, overall -12.67%
Amount of memory still referenced at the end of compilation decreased from 1332k to 1222k, overall -8.96%
Overall memory needed: 8575k -> 8456k
Peak memory use before GGC: 1409k -> 1291k
Peak memory use after GGC: 1338k -> 1218k
Maximum of released memory in single GGC run: 168k -> 135k
Garbage: 255k -> 226k
Leak: 1332k -> 1222k
Overhead: 160k -> 138k
GGC runs: 4
Amount of produced pre-ipa-GGC garbage decreased from 238k to 207k, overall -14.71%
Amount of memory referenced pre-ipa decreased from 1334k to 1225k, overall -8.94%
Pre-IPA-Garbage: 238k -> 207k
Pre-IPA-Leak: 1334k -> 1225k
Pre-IPA-Overhead: 157k -> 135k
Amount of produced post-ipa-GGC garbage decreased from 238k to 207k, overall -14.71%
Amount of memory referenced post-ipa decreased from 1334k to 1225k, overall -8.94%
Post-IPA-Garbage: 238k -> 207k
Post-IPA-Leak: 1334k -> 1225k
Post-IPA-Overhead: 157k -> 135k
comparing empty function compilation at -O3 level:
Peak amount of GGC memory allocated before garbage collecting run decreased from 1409k to 1291k, overall -9.14%
Peak amount of GGC memory still allocated after garbage collecting decreased from 1338k to 1218k, overall -9.85%
Amount of produced GGC garbage decreased from 255k to 226k, overall -12.67%
Amount of memory still referenced at the end of compilation decreased from 1332k to 1222k, overall -8.96%
Overall memory needed: 8559k -> 8460k
Peak memory use before GGC: 1409k -> 1291k
Peak memory use after GGC: 1338k -> 1218k
Maximum of released memory in single GGC run: 168k -> 135k
Garbage: 255k -> 226k
Leak: 1332k -> 1222k
Overhead: 160k -> 138k
GGC runs: 4
Amount of produced pre-ipa-GGC garbage decreased from 238k to 207k, overall -14.71%
Amount of memory referenced pre-ipa decreased from 1334k to 1225k, overall -8.94%
Pre-IPA-Garbage: 238k -> 207k
Pre-IPA-Leak: 1334k -> 1225k
Pre-IPA-Overhead: 157k -> 135k
Amount of produced post-ipa-GGC garbage decreased from 238k to 207k, overall -14.71%
Amount of memory referenced post-ipa decreased from 1334k to 1225k, overall -8.94%
Post-IPA-Garbage: 238k -> 207k
Post-IPA-Leak: 1334k -> 1225k
Post-IPA-Overhead: 157k -> 135k
comparing combine.c compilation at -O0 level:
Overall memory allocated via mmap and sbrk increased from 31131k to 32044k, overall 2.93%
Overall memory needed: 31131k -> 32044k
Peak memory use before GGC: 18036k -> 18032k
Peak memory use after GGC: 17772k -> 17774k
Maximum of released memory in single GGC run: 1849k -> 1866k
Garbage: 39998k -> 39944k
Leak: 5752k -> 5639k
Overhead: 5321k -> 5267k
GGC runs: 338 -> 339
Amount of produced pre-ipa-GGC garbage increased from 12564k to 12582k, overall 0.14%
Pre-IPA-Garbage: 12564k -> 12582k
Pre-IPA-Leak: 19394k -> 19385k
Pre-IPA-Overhead: 2632k -> 2624k
Amount of produced post-ipa-GGC garbage increased from 12564k to 12582k, overall 0.14%
Post-IPA-Garbage: 12564k -> 12582k
Post-IPA-Leak: 19394k -> 19385k
Post-IPA-Overhead: 2632k -> 2624k
comparing combine.c compilation at -O0 -g level:
Overall memory allocated via mmap and sbrk increased from 32995k to 34056k, overall 3.22%
Overall memory needed: 32995k -> 34056k
Peak memory use before GGC: 19963k -> 19957k
Peak memory use after GGC: 19627k -> 19629k
Maximum of released memory in single GGC run: 1859k -> 1875k
Garbage: 40344k -> 40247k
Leak: 9067k -> 8928k
Overhead: 6141k -> 6084k
GGC runs: 320 -> 322
Amount of produced pre-ipa-GGC garbage increased from 12663k to 12681k, overall 0.14%
Pre-IPA-Garbage: 12663k -> 12681k
Pre-IPA-Leak: 21667k -> 21658k
Pre-IPA-Overhead: 3122k -> 3114k
Amount of produced post-ipa-GGC garbage increased from 12663k to 12681k, overall 0.14%
Post-IPA-Garbage: 12663k -> 12681k
Post-IPA-Leak: 21667k -> 21658k
Post-IPA-Overhead: 3122k -> 3114k
comparing combine.c compilation at -O1 level:
Overall memory allocated via mmap and sbrk increased from 30059k to 31020k, overall 3.20%
Peak amount of GGC memory allocated before garbage collecting increased from 15662k to 15689k, overall 0.17%
Peak amount of GGC memory still allocated after garbage collecting increased from 15480k to 15518k, overall 0.25%
Overall memory needed: 30059k -> 31020k
Peak memory use before GGC: 15662k -> 15689k
Peak memory use after GGC: 15480k -> 15518k
Maximum of released memory in single GGC run: 1355k -> 1366k
Garbage: 47221k -> 47216k
Leak: 5740k -> 5624k
Overhead: 6015k -> 6045k
GGC runs: 392 -> 404
Pre-IPA-Garbage: 13456k -> 13457k
Pre-IPA-Leak: 16881k -> 16877k
Pre-IPA-Overhead: 2538k -> 2530k
Post-IPA-Garbage: 13456k -> 13457k
Post-IPA-Leak: 16881k -> 16877k
Post-IPA-Overhead: 2538k -> 2530k
comparing combine.c compilation at -O2 level:
Overall memory allocated via mmap and sbrk increased from 31115k to 31880k, overall 2.46%
Peak amount of GGC memory still allocated after garbage collecting increased from 15594k to 15621k, overall 0.17%
Amount of produced GGC garbage increased from 59591k to 61217k, overall 2.73%
Amount of memory still referenced at the end of compilation increased from 5776k to 5800k, overall 0.41%
Overall memory needed: 31115k -> 31880k
Peak memory use before GGC: 15770k -> 15780k
Peak memory use after GGC: 15594k -> 15621k
Maximum of released memory in single GGC run: 1329k -> 1326k
Garbage: 59591k -> 61217k
Leak: 5776k -> 5800k
Overhead: 7679k -> 7951k
GGC runs: 460 -> 473
Pre-IPA-Garbage: 13627k -> 13632k
Pre-IPA-Leak: 16951k -> 16953k
Pre-IPA-Overhead: 2558k -> 2551k
Post-IPA-Garbage: 13627k -> 13632k
Post-IPA-Leak: 16951k -> 16953k
Post-IPA-Overhead: 2558k -> 2551k
comparing combine.c compilation at -O3 level:
Peak amount of GGC memory allocated before garbage collecting increased from 15794k to 15924k, overall 0.82%
Peak amount of GGC memory still allocated after garbage collecting increased from 15637k to 15720k, overall 0.53%
Amount of produced GGC garbage increased from 74240k to 76604k, overall 3.18%
Overall memory needed: 36191k -> 36832k
Peak memory use before GGC: 15794k -> 15924k
Peak memory use after GGC: 15637k -> 15720k
Maximum of released memory in single GGC run: 1860k -> 1820k
Garbage: 74240k -> 76604k
Leak: 5791k -> 5674k
Overhead: 9388k -> 9633k
GGC runs: 498 -> 508
Pre-IPA-Garbage: 13627k -> 13633k
Pre-IPA-Leak: 16962k -> 16964k
Pre-IPA-Overhead: 2559k -> 2552k
Post-IPA-Garbage: 13627k -> 13633k
Post-IPA-Leak: 16962k -> 16964k
Post-IPA-Overhead: 2559k -> 2552k
comparing insn-attrtab.c compilation at -O0 level:
Overall memory allocated via mmap and sbrk increased from 140167k to 158232k, overall 12.89%
Peak amount of GGC memory allocated before garbage collecting increased from 64730k to 65873k, overall 1.77%
Amount of produced GGC garbage increased from 131074k to 134336k, overall 2.49%
Overall memory needed: 140167k -> 158232k
Peak memory use before GGC: 64730k -> 65873k
Peak memory use after GGC: 54607k -> 53242k
Maximum of released memory in single GGC run: 27749k -> 28569k
Garbage: 131074k -> 134336k
Leak: 7601k -> 7489k
Overhead: 15792k -> 16158k
GGC runs: 260 -> 265
Amount of produced pre-ipa-GGC garbage increased from 39184k to 39229k, overall 0.11%
Amount of memory referenced pre-ipa increased from 55268k to 55908k, overall 1.16%
Pre-IPA-Garbage: 39184k -> 39229k
Pre-IPA-Leak: 55268k -> 55908k
Pre-IPA-Overhead: 8585k -> 8653k
Amount of produced post-ipa-GGC garbage increased from 39184k to 39229k, overall 0.11%
Amount of memory referenced post-ipa increased from 55268k to 55908k, overall 1.16%
Post-IPA-Garbage: 39184k -> 39229k
Post-IPA-Leak: 55268k -> 55908k
Post-IPA-Overhead: 8585k -> 8653k
comparing insn-attrtab.c compilation at -O0 -g level:
Overall memory allocated via mmap and sbrk increased from 143075k to 159532k, overall 11.50%
Peak amount of GGC memory allocated before garbage collecting increased from 66005k to 67147k, overall 1.73%
Amount of produced GGC garbage increased from 131590k to 134815k, overall 2.45%
Overall memory needed: 143075k -> 159532k
Peak memory use before GGC: 66005k -> 67147k
Peak memory use after GGC: 55880k -> 54514k
Maximum of released memory in single GGC run: 27750k -> 28569k
Garbage: 131590k -> 134815k
Leak: 9252k -> 9140k
Overhead: 16249k -> 16613k
GGC runs: 253 -> 258
Amount of produced pre-ipa-GGC garbage increased from 39241k to 39285k, overall 0.11%
Amount of memory referenced pre-ipa increased from 56810k to 57449k, overall 1.13%
Pre-IPA-Garbage: 39241k -> 39285k
Pre-IPA-Leak: 56810k -> 57449k
Pre-IPA-Overhead: 8921k -> 8989k
Amount of produced post-ipa-GGC garbage increased from 39241k to 39285k, overall 0.11%
Amount of memory referenced post-ipa increased from 56810k to 57449k, overall 1.13%
Post-IPA-Garbage: 39241k -> 39285k
Post-IPA-Leak: 56810k -> 57449k
Post-IPA-Overhead: 8921k -> 8989k
comparing insn-attrtab.c compilation at -O1 level:
Overall memory allocated via mmap and sbrk increased from 128159k to 134892k, overall 5.25%
Peak amount of GGC memory allocated before garbage collecting increased from 49645k to 50247k, overall 1.21%
Peak amount of GGC memory still allocated after garbage collecting increased from 42480k to 43080k, overall 1.41%
Overall memory needed: 128159k -> 134892k
Peak memory use before GGC: 49645k -> 50247k
Peak memory use after GGC: 42480k -> 43080k
Maximum of released memory in single GGC run: 22930k -> 23393k
Garbage: 185556k -> 184506k
Leak: 6983k -> 6871k
Overhead: 25308k -> 25327k
GGC runs: 299 -> 305
Amount of memory referenced pre-ipa increased from 42797k to 43436k, overall 1.49%
Pre-IPA-Garbage: 45165k -> 45206k
Pre-IPA-Leak: 42797k -> 43436k
Pre-IPA-Overhead: 7984k -> 8052k
Amount of memory referenced post-ipa increased from 42797k to 43436k, overall 1.49%
Post-IPA-Garbage: 45165k -> 45206k
Post-IPA-Leak: 42797k -> 43436k
Post-IPA-Overhead: 7984k -> 8052k
comparing insn-attrtab.c compilation at -O2 level:
Ovarall memory allocated via mmap and sbrk decreased from 170751k to 150264k, overall -13.63%
Peak amount of GGC memory allocated before garbage collecting increased from 49693k to 50279k, overall 1.18%
Peak amount of GGC memory still allocated after garbage collecting increased from 43507k to 43994k, overall 1.12%
Overall memory needed: 170751k -> 150264k
Peak memory use before GGC: 49693k -> 50279k
Peak memory use after GGC: 43507k -> 43994k
Maximum of released memory in single GGC run: 21651k -> 22010k
Garbage: 218681k -> 216414k
Leak: 6992k -> 6879k
Overhead: 30687k -> 30822k
GGC runs: 333 -> 340
Amount of memory referenced pre-ipa increased from 42802k to 43442k, overall 1.49%
Pre-IPA-Garbage: 45243k -> 45283k
Pre-IPA-Leak: 42802k -> 43442k
Pre-IPA-Overhead: 7993k -> 8061k
Amount of memory referenced post-ipa increased from 42802k to 43442k, overall 1.49%
Post-IPA-Garbage: 45243k -> 45283k
Post-IPA-Leak: 42802k -> 43442k
Post-IPA-Overhead: 7993k -> 8061k
comparing insn-attrtab.c compilation at -O3 level:
Ovarall memory allocated via mmap and sbrk decreased from 179091k to 163980k, overall -9.22%
Peak amount of GGC memory still allocated after garbage collecting increased from 58153k to 58454k, overall 0.52%
Overall memory needed: 179091k -> 163980k
Peak memory use before GGC: 61709k -> 61578k
Peak memory use after GGC: 58153k -> 58454k
Maximum of released memory in single GGC run: 23742k -> 24307k
Garbage: 250117k -> 246737k
Leak: 7000k -> 6886k
Overhead: 34258k -> 34267k
GGC runs: 338 -> 343
Amount of memory referenced pre-ipa increased from 42802k to 43442k, overall 1.49%
Pre-IPA-Garbage: 45243k -> 45283k
Pre-IPA-Leak: 42802k -> 43442k
Pre-IPA-Overhead: 7993k -> 8061k
Amount of memory referenced post-ipa increased from 42802k to 43442k, overall 1.49%
Post-IPA-Garbage: 45243k -> 45283k
Post-IPA-Leak: 42802k -> 43442k
Post-IPA-Overhead: 7993k -> 8061k
comparing Gerald's testcase PR8361 compilation at -O0 level:
Peak amount of GGC memory allocated before garbage collecting increased from 81234k to 82283k, overall 1.29%
Peak amount of GGC memory still allocated after garbage collecting increased from 80428k to 81468k, overall 1.29%
Amount of produced GGC garbage increased from 211214k to 212659k, overall 0.68%
Amount of memory still referenced at the end of compilation increased from 49337k to 49808k, overall 0.95%
Overall memory needed: 147891k -> 150422k
Peak memory use before GGC: 81234k -> 82283k
Peak memory use after GGC: 80428k -> 81468k
Maximum of released memory in single GGC run: 15883k -> 16450k
Garbage: 211214k -> 212659k
Leak: 49337k -> 49808k
Overhead: 27074k -> 27261k
GGC runs: 420 -> 424
Amount of produced pre-ipa-GGC garbage increased from 112585k to 114097k, overall 1.34%
Amount of memory referenced pre-ipa increased from 86451k to 87309k, overall 0.99%
Pre-IPA-Garbage: 112585k -> 114097k
Pre-IPA-Leak: 86451k -> 87309k
Pre-IPA-Overhead: 14761k -> 14978k
Amount of produced post-ipa-GGC garbage increased from 112585k to 114097k, overall 1.34%
Amount of memory referenced post-ipa increased from 86451k to 87309k, overall 0.99%
Post-IPA-Garbage: 112585k -> 114097k
Post-IPA-Leak: 86451k -> 87309k
Post-IPA-Overhead: 14761k -> 14978k
comparing Gerald's testcase PR8361 compilation at -O0 -g level:
Peak amount of GGC memory allocated before garbage collecting increased from 94894k to 95772k, overall 0.93%
Peak amount of GGC memory still allocated after garbage collecting increased from 93955k to 94820k, overall 0.92%
Amount of produced GGC garbage increased from 216832k to 218309k, overall 0.68%
Amount of memory still referenced at the end of compilation increased from 75899k to 76370k, overall 0.62%
Overall memory needed: 165367k -> 167982k
Peak memory use before GGC: 94894k -> 95772k
Peak memory use after GGC: 93955k -> 94820k
Maximum of released memory in single GGC run: 16224k -> 16877k
Garbage: 216832k -> 218309k
Leak: 75899k -> 76370k
Overhead: 33731k -> 33925k
GGC runs: 392 -> 394
Amount of produced pre-ipa-GGC garbage increased from 113207k to 114719k, overall 1.34%
Amount of memory referenced pre-ipa increased from 102970k to 103824k, overall 0.83%
Pre-IPA-Garbage: 113207k -> 114719k
Pre-IPA-Leak: 102970k -> 103824k
Pre-IPA-Overhead: 18265k -> 18481k
Amount of produced post-ipa-GGC garbage increased from 113207k to 114719k, overall 1.34%
Amount of memory referenced post-ipa increased from 102970k to 103824k, overall 0.83%
Post-IPA-Garbage: 113207k -> 114719k
Post-IPA-Leak: 102970k -> 103824k
Post-IPA-Overhead: 18265k -> 18481k
comparing Gerald's testcase PR8361 compilation at -O1 level:
Overall memory allocated via mmap and sbrk increased from 108273k to 112082k, overall 3.52%
Peak amount of GGC memory allocated before garbage collecting increased from 80810k to 82325k, overall 1.87%
Peak amount of GGC memory still allocated after garbage collecting increased from 80003k to 81502k, overall 1.87%
Amount of produced GGC garbage increased from 291136k to 291497k, overall 0.12%
Amount of memory still referenced at the end of compilation increased from 46444k to 46593k, overall 0.32%
Overall memory needed: 108273k -> 112082k
Peak memory use before GGC: 80810k -> 82325k
Peak memory use after GGC: 80003k -> 81502k
Maximum of released memory in single GGC run: 15998k -> 16685k
Garbage: 291136k -> 291497k
Leak: 46444k -> 46593k
Overhead: 32201k -> 32300k
GGC runs: 510 -> 515
Amount of produced pre-ipa-GGC garbage increased from 160850k to 167344k, overall 4.04%
Amount of memory referenced pre-ipa increased from 83941k to 86076k, overall 2.54%
Pre-IPA-Garbage: 160850k -> 167344k
Pre-IPA-Leak: 83941k -> 86076k
Pre-IPA-Overhead: 19695k -> 20471k
Amount of produced post-ipa-GGC garbage increased from 160850k to 167344k, overall 4.04%
Amount of memory referenced post-ipa increased from 83941k to 86076k, overall 2.54%
Post-IPA-Garbage: 160850k -> 167344k
Post-IPA-Leak: 83941k -> 86076k
Post-IPA-Overhead: 19695k -> 20471k
comparing Gerald's testcase PR8361 compilation at -O2 level:
Overall memory allocated via mmap and sbrk increased from 110293k to 115374k, overall 4.61%
Peak amount of GGC memory allocated before garbage collecting increased from 80908k to 83778k, overall 3.55%
Peak amount of GGC memory still allocated after garbage collecting increased from 80103k to 82799k, overall 3.37%
Amount of produced GGC garbage increased from 338356k to 345533k, overall 2.12%
Overall memory needed: 110293k -> 115374k
Peak memory use before GGC: 80908k -> 83778k
Peak memory use after GGC: 80103k -> 82799k
Maximum of released memory in single GGC run: 16001k -> 16656k
Garbage: 338356k -> 345533k
Leak: 46609k -> 46539k
Overhead: 37447k -> 38274k
GGC runs: 569 -> 580
Amount of produced pre-ipa-GGC garbage increased from 162586k to 171551k, overall 5.51%
Amount of memory referenced pre-ipa increased from 83990k to 86381k, overall 2.85%
Pre-IPA-Garbage: 162586k -> 171551k
Pre-IPA-Leak: 83990k -> 86381k
Pre-IPA-Overhead: 19913k -> 20956k
Amount of produced post-ipa-GGC garbage increased from 162586k to 171551k, overall 5.51%
Amount of memory referenced post-ipa increased from 83990k to 86381k, overall 2.85%
Post-IPA-Garbage: 162586k -> 171551k
Post-IPA-Leak: 83990k -> 86381k
Post-IPA-Overhead: 19913k -> 20956k
comparing Gerald's testcase PR8361 compilation at -O3 level:
Overall memory allocated via mmap and sbrk increased from 116865k to 122734k, overall 5.02%
Peak amount of GGC memory allocated before garbage collecting increased from 80908k to 84478k, overall 4.41%
Peak amount of GGC memory still allocated after garbage collecting increased from 80103k to 83419k, overall 4.14%
Amount of produced GGC garbage increased from 370356k to 376951k, overall 1.78%
Amount of memory still referenced at the end of compilation increased from 46455k to 46523k, overall 0.15%
Overall memory needed: 116865k -> 122734k
Peak memory use before GGC: 80908k -> 84478k
Peak memory use after GGC: 80103k -> 83419k
Maximum of released memory in single GGC run: 16001k -> 16656k
Garbage: 370356k -> 376951k
Leak: 46455k -> 46523k
Overhead: 40663k -> 41481k
GGC runs: 596 -> 615
Amount of produced pre-ipa-GGC garbage increased from 162586k to 171637k, overall 5.57%
Amount of memory referenced pre-ipa increased from 84004k to 87063k, overall 3.64%
Pre-IPA-Garbage: 162586k -> 171637k
Pre-IPA-Leak: 84004k -> 87063k
Pre-IPA-Overhead: 19914k -> 21011k
Amount of produced post-ipa-GGC garbage increased from 162586k to 171637k, overall 5.57%
Amount of memory referenced post-ipa increased from 84004k to 87063k, overall 3.64%
Post-IPA-Garbage: 162586k -> 171637k
Post-IPA-Leak: 84004k -> 87063k
Post-IPA-Overhead: 19914k -> 21011k
comparing PR rtl-optimization/28071 testcase compilation at -O0 level:
Peak amount of GGC memory allocated before garbage collecting run decreased from 100315k to 78451k, overall -27.87%
Peak amount of GGC memory still allocated after garbage collecting decreased from 55968k to 49385k, overall -13.33%
Amount of produced GGC garbage decreased from 181152k to 147530k, overall -22.79%
Overall memory needed: 375460k -> 368861k
Peak memory use before GGC: 100315k -> 78451k
Peak memory use after GGC: 55968k -> 49385k
Maximum of released memory in single GGC run: 50582k -> 38763k
Garbage: 181152k -> 147530k
Leak: 6622k -> 6511k
Overhead: 31238k -> 24890k
GGC runs: 90 -> 87
Amount of memory referenced pre-ipa increased from 19499k to 20190k, overall 3.54%
Pre-IPA-Garbage: 13660k -> 13630k
Pre-IPA-Leak: 19499k -> 20190k
Pre-IPA-Overhead: 2175k -> 2242k
Amount of memory referenced post-ipa increased from 19499k to 20190k, overall 3.54%
Post-IPA-Garbage: 13660k -> 13630k
Post-IPA-Leak: 19499k -> 20190k
Post-IPA-Overhead: 2175k -> 2242k
comparing PR rtl-optimization/28071 testcase compilation at -O0 -g level:
Peak amount of GGC memory allocated before garbage collecting run decreased from 101014k to 79148k, overall -27.63%
Peak amount of GGC memory still allocated after garbage collecting decreased from 56667k to 50082k, overall -13.15%
Amount of produced GGC garbage decreased from 181269k to 147632k, overall -22.78%
Overall memory needed: 376280k -> 369637k
Peak memory use before GGC: 101014k -> 79148k
Peak memory use after GGC: 56667k -> 50082k
Maximum of released memory in single GGC run: 50583k -> 38748k
Garbage: 181269k -> 147632k
Leak: 8666k -> 8553k
Overhead: 31823k -> 25474k
GGC runs: 97 -> 93
Amount of memory referenced pre-ipa increased from 19748k to 20439k, overall 3.50%
Pre-IPA-Garbage: 13668k -> 13638k
Pre-IPA-Leak: 19748k -> 20439k
Pre-IPA-Overhead: 2230k -> 2296k
Amount of memory referenced post-ipa increased from 19748k to 20439k, overall 3.50%
Post-IPA-Garbage: 13668k -> 13638k
Post-IPA-Leak: 19748k -> 20439k
Post-IPA-Overhead: 2230k -> 2296k
comparing PR rtl-optimization/28071 testcase compilation at -O1 level:
Ovarall memory allocated via mmap and sbrk decreased from 363863k to 227772k, overall -59.75%
Peak amount of GGC memory still allocated after garbage collecting increased from 63619k to 63951k, overall 0.52%
Amount of produced GGC garbage decreased from 242138k to 227182k, overall -6.58%
Overall memory needed: 363863k -> 227772k
Peak memory use before GGC: 71726k -> 71433k
Peak memory use after GGC: 63619k -> 63951k
Maximum of released memory in single GGC run: 34866k -> 35392k
Garbage: 242138k -> 227182k
Leak: 7266k -> 7088k
Overhead: 33901k -> 30868k
GGC runs: 96 -> 98
Amount of produced pre-ipa-GGC garbage increased from 51866k to 52164k, overall 0.58%
Amount of memory referenced pre-ipa increased from 61789k to 62151k, overall 0.59%
Pre-IPA-Garbage: 51866k -> 52164k
Pre-IPA-Leak: 61789k -> 62151k
Pre-IPA-Overhead: 8972k -> 9038k
Amount of produced post-ipa-GGC garbage increased from 51866k to 52164k, overall 0.58%
Amount of memory referenced post-ipa increased from 61789k to 62151k, overall 0.59%
Post-IPA-Garbage: 51866k -> 52164k
Post-IPA-Leak: 61789k -> 62151k
Post-IPA-Overhead: 8972k -> 9038k
comparing PR rtl-optimization/28071 testcase compilation at -O2 level:
Overall memory allocated via mmap and sbrk increased from 316167k to 375840k, overall 18.87%
Peak amount of GGC memory still allocated after garbage collecting increased from 63620k to 63952k, overall 0.52%
Amount of produced GGC garbage increased from 245858k to 255525k, overall 3.93%
Amount of memory still referenced at the end of compilation decreased from 7429k to 6910k, overall -7.51%
Overall memory needed: 316167k -> 375840k
Peak memory use before GGC: 71496k -> 71433k
Peak memory use after GGC: 63620k -> 63952k
Maximum of released memory in single GGC run: 36143k -> 36706k
Garbage: 245858k -> 255525k
Leak: 7429k -> 6910k
Overhead: 35288k -> 36963k
GGC runs: 102 -> 109
Amount of produced pre-ipa-GGC garbage increased from 111746k to 112377k, overall 0.56%
Amount of memory referenced pre-ipa increased from 72529k to 73071k, overall 0.75%
Pre-IPA-Garbage: 111746k -> 112377k
Pre-IPA-Leak: 72529k -> 73071k
Pre-IPA-Overhead: 15598k -> 15517k
Amount of produced post-ipa-GGC garbage increased from 111746k to 112377k, overall 0.56%
Amount of memory referenced post-ipa increased from 72529k to 73071k, overall 0.75%
Post-IPA-Garbage: 111746k -> 112377k
Post-IPA-Leak: 72529k -> 73071k
Post-IPA-Overhead: 15598k -> 15517k
comparing PR rtl-optimization/28071 testcase compilation at -O3 -fno-tree-pre -fno-tree-fre level:
Ovarall memory allocated via mmap and sbrk decreased from 1118559k to 1025124k, overall -9.11%
Peak amount of GGC memory allocated before garbage collecting increased from 137913k to 138535k, overall 0.45%
Peak amount of GGC memory still allocated after garbage collecting increased from 125190k to 125812k, overall 0.50%
Amount of memory still referenced at the end of compilation decreased from 9929k to 9422k, overall -5.38%
Overall memory needed: 1118559k -> 1025124k
Peak memory use before GGC: 137913k -> 138535k
Peak memory use after GGC: 125190k -> 125812k
Maximum of released memory in single GGC run: 61836k -> 62503k
Garbage: 366793k -> 366892k
Leak: 9929k -> 9422k
Overhead: 45001k -> 45306k
GGC runs: 97 -> 106
Amount of produced pre-ipa-GGC garbage increased from 111746k to 112377k, overall 0.56%
Amount of memory referenced pre-ipa increased from 72529k to 73071k, overall 0.75%
Pre-IPA-Garbage: 111746k -> 112377k
Pre-IPA-Leak: 72529k -> 73071k
Pre-IPA-Overhead: 15598k -> 15517k
Amount of produced post-ipa-GGC garbage increased from 111746k to 112377k, overall 0.56%
Amount of memory referenced post-ipa increased from 72529k to 73071k, overall 0.75%
Post-IPA-Garbage: 111746k -> 112377k
Post-IPA-Leak: 72529k -> 73071k
Post-IPA-Overhead: 15598k -> 15517k
Head of the ChangeLog is:
--- /usr/src/SpecTests/sandbox-britten-memory/x86_64/mem-result/ChangeLog 2008-08-01 17:44:36.000000000 +0000
+++ /usr/src/SpecTests/sandbox-britten-memory/gcc/gcc/ChangeLog 2008-09-09 02:54:09.000000000 +0000
@@ -1,45 +1,6228 @@
+2008-08-09 Andy Hutchinson <hutchinsonandy@aim.com>
+
+ PR target/36609
+ * config/avr/avr.c (avr_reorg): Create RTL for reversed compare with
+ zero.
+ * config/avr/avr.md (QISI) : Define mode iterator.
+ (negated_tst<mode>) : Redefine as split using mode macro.
+ (reversed_tstqi): Define insn as reversed compare with zero.
+ (reversed_tsthi): Ditto.
+ (reversed_tstsi): Ditto.
+
+2008-09-08 Jakub Jelinek <jakub@redhat.com>
+
+ PR middle-end/37393
+ * tree-inline.c (copy_bb): When replacing a gimple_call_va_arg_pack_p
+ call stmt by new_call, clear gimple_bb on stmt after gsi_replace.
+
+ PR middle-end/37414
+ * predict.c (optimize_function_for_size_p): Don't segfault if
+ FUN is NULL.
+ * fold-const.c (LOGICAL_OP_NON_SHORT_CIRCUIT, fold_truthop,
+ tree_swap_operands_p): Don't test cfun != NULL before calling
+ optimize_function_for_s*_p.
+
+2008-09-08 Eric Botcazou <ebotcazou@adacore.com>
+
+ * ira-color.c (ira_reuse_stack_slot): Set slot_num on success at the
+ end of the search.
+
+2008-09-08 M R Swami Reddy <MR.Swami.Reddy@nsc.com>
+
+ * crx/crx.h (IRA_COVER_CLASSES): Define.
+
+2008-09-08 Jakub Jelinek <jakub@redhat.com>
+
+ PR middle-end/37337
+ * tree-ssa-dom.c (optimize_stmt): Call maybe_clean_or_replace_eh_stmt
+ even when a stmt has been gimple_modified_p, but after fold_stmt is
+ not any longer. Remove unneeded may_have_exposed_new_symbols
+ initializations.
+
+2008-09-08 Richard Guenther <rguenther@suse.de>
+
+ PR tree-optimization/37421
+ * tree-ssa-sccvn.c (visit_copy): Make sure to fully
+ valueize the RHS.
+
+2008-09-08 Jakub Jelinek <jakub@redhat.com>
+
+ PR middle-end/37415
+ * opts.c (common_handle_option): Handle OPT_ftree_store_ccp.
+
+2008-09-08 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
+
+ PR driver/37409
+ * pa-hpux.h (LINK_SPEC): Strip -fwhole-program.
+ * pa-hpux10.h (LINK_SPEC): Likewise.
+ * pa-hpux11.h (LINK_SPEC): Likewise.
+
+2008-09-07 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
+
+ * pa32-regs.h (IRA_COVER_CLASSES): Define.
+ * pa64-regs.h (IRA_COVER_CLASSES): Define.
+
+2008-09-07 Helge Deller <deller@gmx.de>
+
+ * pa/linux-atomic.c: New file.
+ * pa/t-linux (LIB2FUNCS_STATIC_EXTRA): Define.
+ * pa/t-linux64 (LIB2FUNCS_STATIC_EXTRA): Define.
+
+2008-09-07 Richard Guenther <rguenther@suse.de>
+ Ira Rosen <irar@il.ibm.com>
+
+ PR tree-optimization/36630
+ * tree-vect-transform.c (vect_update_ivs_after_vectorizer):
+ Call STRIP_NOPS before calling evolution_part_in_loop_num.
+
+2008-09-07 Dorit Nuzman <dorit@il.ibm.com>
+ Ira Rosen <irar@il.ibm.com>
+
+ PR tree-optimization/35642
+ * config/rs6000/altivec.md (mulv8hi3): Implement.
+
+2008-09-06 Jeff Law <law@redhat.com>
+
+ * h8300/h8300.h (IRA_COVER_CLASSES): Define.
+
+2008-09-06 Jan Hubicka <jh@suse.cz>
+
+ PR tree-optimization/14703
+ * tree-ssa-live.c (remove_unused_scope_block_p): Remove ignored declarations.
+ * passes.c (init_optimization_passes): Recompute inline parameters.
+
+2008-09-06 Richard Sandiford <rdsandiford@googlemail.com>
+
+ * config/mips/mips.c (mips_function_ok_for_sibcall): Check for
+ DECL being null.
+
+2008-09-06 Richard Sandiford <rdsandiford@goolemail.com>
+ Peter Fuerst <post@pfrst.de>
+
+ * doc/invoke.texi: Document -mr10k-cache-barrier=.
+ * doc/extend.texi: Document __builtin_mips_cache.
+ * config/mips/mips-ftypes.def: Add a (VOID, SI, CVPOINTER) entry.
+ * config/mips/mips.opt (mr10k-cache-barrier=): New option.
+ * config/mips/mips.h (TARGET_CPU_CPP_BUILTINS): Define
+ __GCC_HAVE_BUILTIN_MIPS_CACHE.
+ (TARGET_CACHE_BUILTIN, ISA_HAS_CACHE): New macros.
+ * config/mips/mips.c (mips_r10k_cache_barrier_setting): New enum.
+ (set_push_mips_isas): New variable.
+ (mips_r10k_cache_barrier): New variable.
+ (cache): New availability predicate.
+ (mips_builtins): Add an entry for __builtin_mips_cache.
+ (mips_build_cvpointer_type): New function.
+ (MIPS_ATYPE_CVPOINTER): New macro.
+ (mips_prepare_builtin_arg): Only use the insn's mode if the rtx's
+ mode is VOIDmode.
+ (r10k_simplified_address_p, r10k_simplify_address)
+ (r10k_uncached_address_p, r10k_safe_address_p)
+ (r10k_needs_protection_p_1, r10k_needs_protection_p_store)
+ (r10k_needs_protection_p_call, r10k_needs_protection_p)
+ (r10k_insert_cache_barriers): New functions.
+ (mips_reorg_process_insns): Delete cache barriers after a
+ branch-likely instruction.
+ (mips_reorg): Call r10k_insert_cache_barriers.
+ (mips_handle_option): Handle OPT_mr10k_cache_barrier_.
+ * config/mips/mips.md (UNSPEC_MIPS_CACHE): New constant.
+ (UNSPEC_R10K_CACHE_BARRIER): Likewise.
+ (mips_cache, r10k_cache_barrier): New define_insns.
+
+2008-09-06 Richard Sandiford <rdsandiford@googlemail.com>
+
+ * ira-int.h (ira_zero_hard_reg_set, ira_one_hard_reg_set): Delete.
+ * ira-color.c (setup_allocno_left_conflicts_num): Use
+ hard_reg_set_equal_p instead of a comparison with
+ ira_zero_hard_reg_set.
+ * ira.c (setup_reg_subclasses): Likewise.
+ (setup_cover_and_important_classes): Likewise.
+ (setup_class_translate): Likewise.
+ (setup_reg_class_intersect_union): Likewise.
+ (ira_zero_hard_reg_set, ira_one_hard_reg_set): Delete.
+ (ira_init_once): Don't initialize them.
+
+2008-09-05 Vladimir Makarov <vmakarov@redhat.com>
+
+ * doc/tm.texi (IRA_COVER_CLASSES): Fix a typo.
+
+2008-09-05 Vladimir Makarov <vmakarov@redhat.com>
+
+ * ira-color.c (ira_fast_allocation): Permit global allocno
+ allocation.
+
+2008-09-05 Trevor Smigiel <Trevor_Smigiel@playstation.sony.com>
+
+ Improved branch hints, safe hints, and scheduling.
+
+ * haifa-sched.c (sched_emit_insn) : Define.
+ * sched-int.h (sched_emit_insn) : Add prototype.
+ * doc/invoke.texi (-mdual-nops, -mhint-max-nops,
+ -mhint-max-distance -msafe-hints) : Document.
+ * config/spu/spu.c (spu_flag_var_tracking): New.
+ (TARGET_SCHED_INIT_GLOBAL, TARGET_SCHED_INIT,
+ TARGET_SCHED_REORDER, TARGET_SCHED_REORDER2,
+ TARGET_ASM_FILE_START): Define.
+ (TARGET_SCHED_ADJUST_PRIORITY): Remove.
+ (STOP_HINT_P, HINTED_P, SCHED_ON_EVEN_P): Define.
+ (spu_emit_branch_hint): Add blocks argument.
+ (insert_branch_hints, insert_nops): Remove.
+ (pad_bb, insert_hbrp_for_ilb_runout, insert_hbrp, in_spu_reorg,
+ uses_ls_unit, spu_sched_init_global, spu_sched_init,
+ spu_sched_reorder, asm_file_start): New functions.
+ (clock_var, spu_sched_length, pipe0_clock,
+ pipe1_clock, prev_clock_var, prev_priority,
+ spu_ls_first, prev_ls_clock): New static variables.
+ * config/spu/spu.h (TARGET_DEFAULT): Add MASK_SAFE_HINTS.
+ * config/spu.md (iprefetch): Add operand, make it clobber MEM.
+ (nopn_nv): Add a non-volatile version of nop.
+ * config/spu/spu.opt (-mdual-nops, -mhint-max-nops,
+ -mhint-max-distance, -msafe-hints): New options.
+
+2008-09-05 Janis Johnson <janis187@us.ibm.com>
+ Samuel Tardieu <sam@rfc1149.net>
+
+ * opts.c (decode_options): Combine nested if statements.
+
+ PR target/37283
+ * opts.c (decode_options): Handle more relationships among
+ unit-at-a-time, toplevel-reorder, and section-anchors.
+
+2008-09-05 David Daney <ddaney@avtrex.com>
+
+ * doc/install.texi (--enable-reduced-reflection): Document new option.
+
+2008-09-05 Bob Wilson <bob.wilson@acm.org>
+
+ * config/xtensa/predicates.md (nonimmed_operand, mem_operand): Use
+ constantpool_mem_p.
+ (constantpool_operand): New.
+ (move_operand): Disallow sub-word modes for the constant pool.
+ * config/xtensa/xtensa.c (TARGET_SECONDARY_RELOAD): Define.
+ (xtensa_secondary_reload_class): Replace with....
+ (xtensa_secondary_reload): this function. Remove SIGN_EXTEND check.
+ Set icode for sub-word reloads from the constant pool.
+ * config/xtensa/xtensa.h (SECONDARY_INPUT_RELOAD_CLASS): Delete.
+ (SECONDARY_OUTPUT_RELOAD_CLASS): Delete.
+ * config/xtensa/xtensa.md (reload<mode>_literal): New.
+ * config/xtensa/xtensa-protos.h: Update prototypes.
+
+2008-09-05 Joseph Myers <joseph@codesourcery.com>
+
+ * config/mips/mips.h (enum reg_class): Add FRAME_REGS.
+ (REG_CLASS_NAMES): Update.
+ (REG_CLASS_CONTENTS): Update.
+ * config/mips/mips.c (mips_regno_to_class): Use FRAME_REGS instead
+ of ALL_REGS for regs 77 and 78.
+ * function.c (instantiate_virtual_regs_in_insn): Assert that
+ return value of simplify_gen_subreg is not NULL.
+
+2008-09-05 Paolo Bonzini <bonzini@gnu.org>
+
+ * emit-rtl.c (gen_rtvec): Rewrite not using gen_rtvec_v.
+ (gen_rtvec_v): Fix coding standards.
+
+2008-09-04 Jan Hubicka <jh@suse.cz>
+
+ * i386.c (decide_alg): Be more conservative about optimizing for size.
+
+2008-09-04 Ian Lance Taylor <iant@google.com>
+
+ * varasm.c (narrowing_initializer_constant_valid_p): New
+ static function.
+ (initializer_constant_valid_p): Call it.
+
+2008-09-04 Jeff Law <law@redhat.com>
+
+ * fold-const.c (native_encode_real): Fix computation of WORDS.
+ (native_interpret_real): Likewise.
+
+2008-09-04 Janis Johnson <janis187@us.ibm.com>
+
+ * config/rs6000/t-rs6000: Remove target gt-rs6000.h.
+
+2008-09-04 Vladimir Makarov <vmakarov@redhat.com>
+
+ * ira-conflicts.c (process_regs_for_copy): Check insn to check
+ that the cost is already taken into account in ira-costs.c
+
+ * ira-int.h (ira_debug_copy, ira_debug_copies): New.
+
+ * ira-build.c (print_copy, print_copies, ira_debug_copy,
+ ira_debug_copies): New.
+ (ira_bulid): Call print_copies.
+
+ * doc/tm.texi (IRA_COVER_CLASSES): Fix the description.
+
+2008-09-04 Samuel Tardieu <sam@rfc1149.net>
+
+ PR target/32783
+ * config/rs6000/rs6000.c (optimization_options): Remove check of
+ flag_toplevel_order.
+
+2008-09-04 Adam Nemet <anemet@caviumnetworks.com>
+
+ * config/mips/mips.h (ISA_HAS_SEQ_SNE): New macro.
+ * config/mips/mips.c (mips_expand_scc): Also expand seq and sne if
+ second operand is a reg_imm10_operand.
+ * config/mips/mips.md (*seq_<GPR:mode><GPR2:mode>_seq,
+ *sne_<GPR:mode><GPR2:mode>_sne): New patterns.
+ (*seq_<GPR:mode><GPR2:mode>): Rename to
+ *seq_zero_<GPR:mode><GPR2:mode>. Don't match if
+ ISA_HAS_SEQ_SNE.
+ (*seq_<GPR:mode><GPR2:mode>_mips16): Rename to
+ *seq_zero_<GPR:mode><GPR2:mode>_mip16. Don't match if
+ ISA_HAS_SEQ_SNE.
+ (*sne_<GPR:mode><GPR2:mode>): Rename to
+ *sne_zero_<GPR:mode><GPR2:mode>. Don't match if
+ ISA_HAS_SEQ_SNE.
+
+2008-09-04 Adam Nemet <anemet@caviumnetworks.com>
+
+ * config/mips/mips.h (ISA_HAS_EXTS): New macro.
+ * config/mips/mips.md (*ashr_trunc<mode>): Name the pattern
+ combining an arithmetic right shift by more than 31 and a
+ trunction. Don't match for out-of-range shift amounts. Set
+ attribute mode to <MODE>.
+ (*lshr32_trunc<mode>): Name the pattern combining a logical right
+ shift by 32 and and a truncation. Set attribute mode to <MODE>.
+ (*<optab>_trunc<mode>_exts): New pattern for truncated right
+ shifts by less than 32.
+ (extv): Change predicate on first operand to accept registers.
+ Change predicate of the other operands from immediate_operand to
+ const_int_operand. Expand exts when source is a register.
+ (extzv): Change predicate of the constant operands from
+ immediate_operand to const_int_operand.
+ (extzv<mode>): Change predicate of the constant operands from
+ immediate_operand to const_int_operand and no constraint. Also
+ remove mode.
+ (*extzv_trunc<mode>_exts): New pattern.
+
+2008-09-04 Adam Nemet <anemet@caviumnetworks.com>
+
+ * config/mips/mips.h (ISA_HAS_CINS): New macro.
+ * config/mips/mips-protos.h (mask_low_and_shift_p,
+ mask_low_and_shift_len): Declare.
+ * config/mips/mips.c (mask_low_and_shift_p,
+ mask_low_and_shift_len): New functions.
+ (mips_print_operand): Handle new operand prefix "m".
+ * config/mips/mips.md (*cins<mode>): New pattern.
+
+2008-09-04 Bernd Schmidt <bernd.schmidt@analog.com>
+
+ * config/bfin/bfin.c (gen_one_bundle): Don't create new nops when
+ optimizing for size.
+
+2008-09-04 Richard Sandiford <rdsandiford@googlemail.com>
+
+ * df-scan.c (df_get_entry_block_def_set): Add STACK_POINTER_REGNUM
+ regardless of epilogue_completed.
+
+2008-09-04 Vladimir Makarov <vmakarov@redhat.com>
+
+ PR middle-end/37359
+ * ira-lives.c (process_bb_node_lives): Check setjmp.
+
+2008-09-04 Richard Sandiford <rdsandiford@googlemail.com>
+
+ PR middle-end/37243
+ * ira-build.c (form_loop_tree): Reverse BB walk.
+ (create_bb_allocnos): Likewise.
+ * ira-lives.c (make_regno_born_and_dead, regs_set): Delete.
+ (mark_reg_store): Rename to...
+ (mark_ref_live): ...this and take a df_ref argument instead of
+ note_stores arguments. Assert that we have a register.
+ (mark_reg_clobber): Delete.
+ (def_conflicts_with_inputs_p): New function.
+ (mark_reg_conflicts): Delete.
+ (mark_reg_death): Rename to...
+ (mark_ref_dead): ...this and take a df_ref argument instead of
+ a register. Assert that we have a register.
+ (process_bb_node_lives): Hoist frequency calculation out of
+ instruction walk. Convert from a forwards scan to a backwards scan.
+ Use DF_REF_USES and DF_REF_DEFS instead of register notes and
+ note_stores. Remove EH_RETURN_DATA_REGNO and regs_set handling.
+ (create_allocno_live_ranges): Don't create regs_set.
+
+2008-09-04 Ian Lance Taylor <iant@google.com>
+
+ * rtl.h (LABEL_REF_NONLOCAL_P): Don't check for REG_LABEL_OPERAND
+ or REG_LABEL_TARGET.
+ * calls.c (emit_library_call_value_1): Use MEM_P rather than
+ comparing MODE with MEM.
+ * gimple.c (gimple_build_predict): Cast END_PREDICTORS before
+ comparing with GF_PREDICT_TAKEN.
+ (gimple_get_lhs): Change code to enum gimple_code.
+ (gimple_set_lhs): Likewise.
+ * ifcvt.c (noce_process_if_block): Correct GET_MODE to GET_CODE.
+ * omp-low.c (find_omp_clause): Change kind parameter to enum
+ omp_clause_code.
+ * tree-flow.h (find_omp_clause): Update declaration.
+ * regrename.c (clear_dead_regs): Change kind parameter to enum
+ reg_note.
+ * reload1.c (eliminate_regs_1): Use REG_NOTE_KIND rather than
+ GET_MODE.
+ * see.c (see_get_extension_data): Change return type to enum
+ entry_type. Change UNKNOWN to NOT_RELEVANT, SIGN_EXTEND to
+ SIGNED_EXTENDED_DEF, ZERO_EXTEND to ZERO_EXTENDED_DEF.
+ (see_gen_normalized_extension): Change extension_code parameter to
+ enum entry_type.
+ (see_seek_pre_extension_expr): Change extension_code to enum
+ entry_type.
+ (see_merge_one_def_extension): Likewise.
+ (see_handle_relevant_defs): Likewise.
+ (see_handle_relevant_uses): Likewise.
+ (see_analyze_one_def): Likewise.
+ * tree-cfg.c (need_fake_edge_p): Compare gimple code with
+ GIMPLE_ASM rather than ASM_EXPR.
+ * tree-ssa-alias.c (is_escape_site): Compare gimple code with
+ GIMPLE_RETURN rather than RETURN_EXPR.
+ * tree-ssa-ccp.c (likely_value): Change code to enum gimple_code.
+ (evaluate_stmt): Likewise.
+ * tree-vect-analyze.c (vect_analyze_operations): Change relevance
+ to enum vect_relevant.
+ (vect_mark_stmts_to_be_vectorized): Change assertion to not
+ compare gimple codes with tree codes.
+
+2008-09-04 Paul Brook <paul@codesourcery.com>
+
+ * config/arm/arm.c (arm_size_rtx_costs): Call cost function.
+
+2008-09-04 Bernd Schmidt <bernd.schmidt@analog.com>
+
+ * config/bfin/bfin.c (bfin_function_ok_for_sibcall): Restore a null
+ pointer check lost in the recent no-unit-at-a-time patch.
+
+2008-09-04 Jan Hubicka <jh@suse.cz>
+
+ PR middle-end/37343
+ * tree-switch-conversion.c (check_final_bb): Accept only IP
+ invariants.
+
+2008-09-04 Jan Hubicka <jh@suse.cz>
+
+ * tree-vrp.c (vrp_evaluate_conditional_warnv_with_ops_using_ranges):
+ Break out from ...
+ (vrp_evaluate_conditional_warnv_with_ops): ... this one. Add
+ using_ranges argument.
+ (vrp_evaluate_conditional): Avoid bogus warning for type range.
+ (vrp_visit_cond_stmt): Update call of
+ vrp_evaluate_conditional_warnv_with_ops
+
+2008-09-04 Jan Hubicka <jh@suse.cz>
+
+ PR tree-optimization/37345
+ PR tree-optimization/37358
+ PR tree-optimization/37357
+ * tree.c (build_function_type_skip_args): Build distinct type copy;
+ set TYPE_CONTEXT.
+ (build_function_decl_skip_args): Set type of new decl not orig decl;
+ clear DECL_VINDEX for methods turned into functions.
+
+2008-09-04 Nick Clifton <nickc@redhat.com>
+
+ * configure.ac (HAVE_GAS_LCOMM_WITH_ALIGNMENT): New assembler
+ check.
+ * configure: Regenerate.
+ * config.in: Regenerate.
+ * config/i386/bsd.h (ASM_OUTPUT_ALIGNED_LOCAL): Use .lcomm with an
+ alignment field, if it is supported.
+
+2008-09-04 Kai Tietz <kai.tietz@onevision.com>
+
+ * config/i386/t-mingw32 (SHLIB_LC): Change order of import
+ libraries.
+
+2008-09-03 David Edelsohn <edelsohn@gnu.org>
+
+ * config/rs6000/rs6000.c (processor_target_table): Do not
+ enable VMX for POWER6.
+
+2008-09-03 Anton Blanchard <anton@samba.org>
+
+ * config/rs6000/rs6000.c (rs6000_split_lock_test_and_set): Do not
+ emit memory barrier before operation.
+
+2008-09-03 John David Anglin <dave.anglin@nrc.cnrc.gc.ca>
+
+ PR testsuite/37325
+ * config/pa/pa64-hpux.h (ASM_OUTPUT_EXTERNAL_REAL): Call
+ default_elf_asm_output_external.
+
+2008-09-03 Jeff Law <law@redhat.com>
+
+ * h8300.c (expand_a_shift): Avoid invalid RTL sharing.
+
+2008-09-03 Richard Henderson <rth@redhat.com>
+
+ * config/alpha/alpha.c (alpha_split_lock_test_and_set): Move
+ memory barrier to below the test-and-set.
+ (alpha_split_lock_test_and_set_12): Likewise.
+
+2008-09-03 Vladimir Makarov <vmakarov@redhat.com>
+
+ PR rtl-opt/37243
+
+ * ira-int.h (ira_loop_tree_node): Rename mentioned_allocnos to
+ all_allocnos.
+
+ * ira-color.c (print_loop_title): Use all_allocnos.
+ (color_pass): Ditto. Don't add border_allocnos. Check that
+ subloop allocno in the correspdoning bitmap all_allocnos.
+
+ * ira-emit.c (change_loop): Use all_allocnos.
+
+ * ira-build.c (create_loop_tree_nodes, finish_loop_tree_node):
+ Ditto.
+ (ira_create_allocno): Set up all_allocnos bit for the created
+ allocno.
+ (create_cap_allocno): Remove setting mentioned_allocnos.
+ (create_insn_allocnos): Ditto.
+ (remove_unnecessary_allocnos): Use all_allocnos.
+ (check_allocno_creation): Check that allocnos are in the
+ corresponding bitmap all_allocnos.
+
+2008-09-03 Vladimir Makarov <vmakarov@redhat.com>
+
+ PR rtl-opt/37243
+
+ * ira-conflicts.c (REG_SUBREG_P, go_through_subreg): New.
+ (process_regs_for_copy): Process subregs. Refine check when cost
+ is taken into account in ira-costs.c.
+ (process_reg_shuffles): Use REG_SUBREG_P.
+ (add_insn_allocno_copies): Ditto. Ignore modes.
+
+ * ira-color.c (conflict_allocno_vec): New.
+ (COST_HOP_DIVISOR): New macro.
+ (update_copy_costs_1): Use it.
+ (update_conflict_hard_regno_costs): New function.
+ (assign_hard_reg): Use it.
+ (ira_color): Allocate and free conflict_allocno_vec.
+
+2008-09-03 Vladimir Makarov <vmakarov@redhat.com>
+
+ PR rtl-opt/37296
+
+ * ira-int.h (ira_sort_insn_chain): Remove.
+
+ * ira.c (basic_block_order_nums, chain_insn_order,
+ chain_freq_compare, chain_bb_compare, ira_sort_insn_chain): Remove.
+ (ira): Don't call ira_sort_insn_chain.
+
+ * reload1.c (reload): Don't call ira_sort_insn_chain.
+
+2008-09-03 Andrew Pinski <andrew_pinski@playstation.sony.com>
+
+ PR middle-end/37293
+ * cgraphunit.c (update_call_expr): Remove eh regions from statements
+ which become non throw.
+ (cgraph_function_versioning): Also clear DECL_WEAK. Call
+ update_call_expr after updating the flags on the decl.
+
+2008-09-03 Jan Hubicka <jh@suse.cz>
+
+ PR tree-optimization/37315
+ * cgraph.c (cgraph_create_edge): Use gimple_has_body_p.
+ * cgraphunit.c (verify_cgraph_node): drop gimple_body check.
+ (cgraph_analyze_functions): Use node->analyzed
+ (cgraph_mark_functions_to_output): Likewise.
+ (cgraph_expand_function): All functions can be released after
+ expanding.
+ (cgraph_optimize): Use gimple_has_body_p.
+ * ipa-inline.c (cgraph_clone_inlined_nodes): Use analyzed flag.
+ (cgraph_decide_inlining_incrementally): Likewise.
+ (inline_transform): Inline transform.
+ * tree-inline.c (initialize_cfun): Do now shallow copy structure;
+ copy fields needed.
+ (inlinable_function_p): Drop gimple_body check.
+ (expand_call_inline): Use gimple_has_body_p.
+ * gimple.c (gimple_has_body_p): New.
+ * gimple.h (gimple_has_body_p): Add prototype.
+ * tree-cfg.c (execute_build_cfg): Remove gimple_body.
+ (dump_function_to_file): Use gimple_has_body_p check.
+
+2008-09-03 Jakub Jelinek <jakub@redhat.com>
+
+ PR c++/37346
+ * gimple.h (gimple_has_substatements): GIMPLE_OMP_CRITICAL has
+ substatements.
+ * tree-inline.c (remap_gimple_stmt): Handle GIMPLE_OMP_CRITICAL.
+
+2008-09-03 Bernd Schmidt <bernd.schmidt@analog.com>
+
+ From Michael Frysinger <michael.frysinger@analog.com>
+ * config/bfin/bfin.c (bfin_cpus[]): Add 0.1 for bf522, bf523, bf524,
+ bf525, bf526, bf527, bf542, bf544, bf547, bf548, and bf549. Add 0.2
+ for bf538.
+
+2008-09-03 Hari Sandanagobalane <hariharan@picochip.com>
+
+ Add picoChip port.
+ * doc/extend.texi: Document picoChip builtin functions.
+ * doc/invoke.texi: Document picoChip options.
+ * doc/contrib.texi: Add picoChip contribution.
+ * doc/md.texi: Document picoChip constraints.
+ * config.gcc: Add picochip-*-*.
+ * config/picochip/: Add new port.
+
+2008-09-03 Richard Guenther <rguenther@suse.de>
+
+ PR tree-optimization/37328
+ * tree-sra.c (sra_build_assignment): Gimplify properly.
+ (generate_copy_inout): Take the correct stmt as definition,
+ remove bogus assert.
+
+2008-09-03 Bernd Schmidt <bernd.schmidt@analog.com>
+
+ * config/bfin/bfin.c (bfin_expand_prologue): Honour no_stack_limit
+ attributes.
+
+ From Michael Frysinger <michael.frysinger@analog.com>
+ * config/bfin/bfin.c (bfin_handle_option): Remove BF561 warning.
+
+2008-09-03 Danny Smith <dannysmith@usrs.sourceforge.net>
+
+ * gthr-win32.h (CONST_CAST2): Really make sure CONST_CAST2 is
+ defined.
+ (__gthread_setspecific): Revert 2008-08-31 change to
+ __GTHREAD_HIDE_W32API case. Apply it to !__GTHREAD_HIDE_W32API
+ case.
+
+2008-09-02 Aldy Hernandez <aldyh@redhat.com>
+
+ * diagnostic.c (error_at): New.
+ * toplev.h (error_at): New prototype.
+ * c-typeck.c (build_array_ref): Call error_at instead of error.
+ Pass location to pedwarn.
+
+2008-09-02 Adam Nemet <anemet@caviumnetworks.com>
+
+ * sel-sched.c (sel_hard_regno_rename_ok): Mark arguments unused.
+
+2008-09-02 Jakub Jelinek <jakub@redhat.com>
+
+ * dwarf2out.c (tree_add_const_value_attribute): Only handle
+ VAR_DECL and CONST_DECL.
+
+2008-09-02 H.J. Lu <hongjiu.lu@intel.com>
+ Jakub Jelinek <jakub@redhat.com>
+
+ * config/i386/i386.c (X86_64_VARARGS_SIZE): Removed.
+ (setup_incoming_varargs_64): Assume cum != NULL. Set/check
+ ix86_varargs_gpr_size and ix86_varargs_fpr_size. Use
+ ix86_varargs_gpr_size instead of X86_64_REGPARM_MAX.
+ Don't set ix86_save_varrargs_registers.
+ (ix86_setup_incoming_varargs): Assume cum != NULL.
+ (ix86_va_start): Check ix86_varargs_gpr_size and
+ ix86_varargs_fpr_size instead of cfun->va_list_gpr_size and
+ cfun->va_list_fpr_size, respectively. Subtract 8*X86_64_REGPARM_MAX
+ from frame pointer if ix86_varargs_gpr_size == 0.
+ (ix86_compute_frame_layout): Updated.
+ * config/i386/i386.h (ix86_save_varrargs_registers): Removed.
+ (ix86_varargs_gpr_size): Define.
+ (ix86_varargs_fpr_size): Likewise.
+ (machine_function): Remove save_varrargs_registers.
+ Add varargs_gpr_size and varargs_fpr_size.
+
+2008-09-02 Jakub Jelinek <jakub@redhat.com>
+
+ * config/alpha/alpha.c (va_list_skip_additions,
+ alpha_stdarg_optimize_hook, alpha_gimplify_va_arg_1): Tuplify.
+ (alpha_gimplify_va_arg): Call unshare_expr on second use of
+ offset_field.
+
+ PR tree-optimization/36766
+ * tree-cfg.c (gimple_purge_all_dead_eh_edges): Do nothing
+ for already removed basic blocks.
+
+ PR target/36332
+ * real.c (real_maxval): Clear a lower bit to make real_maxval
+ match get_max_float for IBM long double format.
+
+2008-09-02 Andreas Schwab <schwab@suse.de>
+
+ * configure.ac: Use m4_quote to quote the expansion of m4_do.
+ * configure: Regenerate.
+
+2008-09-02 Richard Sandiford <rdsandiford@googlemail.com>
+
+ * simplify-rtx.c (simplify_binary_operation_1): Check for CONST,
+ SYMBOL_REF and LABEL_REF when applying plus_constant, instead of
+ checking for a non-VOID CONSTANT_P.
+
+2008-09-02 Jan Hubicka <jh@suse.cz>
+
+ * predict.c (pass_strip_predict_hints): Avoid bugs dump file.
+ * expmed.c (expand_divmod): Fix test of smod_pow2_cheap and sdiv_pow2_cheap
+
+2008-09-02 H.J. Lu <hongjiu.lu@intel.com>
+
+ * expr.c (emit_group_store): Don't assert stack temp mode size.
+
+2008-09-02 Sebastian Pop <sebastian.pop@amd.com>
+ Tobias Grosser <grosser@fim.uni-passau.de>
+ Jan Sjodin <jan.sjodin@amd.com>
+ Harsha Jagasia <harsha.jagasia@amd.com>
+ Dwarakanath Rajagopal <dwarak.rajagopal@amd.com>
+ Konrad Trifunovic <konrad.trifunovic@inria.fr>
+ Adrien Eliche <aeliche@isty.uvsq.fr>
+
+ Merge from graphite branch.
+ * graphite.c: New.
+ * graphite.h: New.
+ * tree-loop-linear.c (perfect_loop_nest_depth): Export.
+ * doc/invoke.texi (-floop-block, -floop-interchange,
+ -floop-strip-mine): Document new flags.
+ * tree-into-ssa.c (gimple_vec): Moved...
+ * tree-loop-distribution.c (rdg_component): Moved...
+ * cfgloopmanip.c: Include tree-flow.h.
+ (update_dominators_in_loop): New.
+ (create_empty_if_region_on_edge): New.
+ (create_empty_loop_on_edge): New.
+ (loopify): Use update_dominators_in_loop.
+ * tree-pass.h (pass_graphite_transforms): Declared.
+ * configure: Regenerate.
+ * tree-phinodes.c (make_phi_node): Export.
+ (add_phi_node_to_bb): New, split from create_phi_node.
+ * tree-chrec.c (for_each_scev_op): New.
+ * tree-chrec.h (for_each_scev_op): Declared.
+ * tree-ssa-loop-ivopts.c (get_phi_with_result): New.
+ (remove_statement): Call get_phi_with_result.
+ * config.in (HAVE_cloog): Undef.
+ * gdbinit.in (pgg): New.
+ * timevar.def (TV_GRAPHITE_TRANSFORMS): New.
+ * tree-ssa-loop.c (graphite_transforms): New.
+ (gate_graphite_transforms): New.
+ (pass_graphite_transforms): New.
+ * configure.ac (PPLLIBS, PPLINC, CLOOGLIBS, CLOOGINC,
+ HAVE_cloog): Defined.
+ * tree-vectorizer.c (rename_variables_in_bb): Export.
+ * tree-data-ref.c (dr_may_alias_p): Export.
+ (stmt_simple_memref_p): New.
+ (find_data_references_in_stmt): Export.
+ (find_data_references_in_loop): Export.
+ (create_rdg_edge_for_ddr): Initialize RDGE_RELATION.
+ (create_rdg_edges_for_scalar): Initialize RDGE_RELATION.
+ (create_rdg_vertices): Export.
+ (build_empty_rdg): New.
+ (build_rdg): Call build_empty_rdg. Free dependence_relations.
+ * tree-data-ref.h (rdg_component): ... here.
+ (scop_p): New.
+ (struct data_reference): Add a field scop.
+ (DR_SCOP): New.
+ (find_data_references_in_loop): Declared.
+ (find_data_references_in_stmt): Declared.
+ (create_rdg_vertices): Declared.
+ (dr_may_alias_p): Declared.
+ (stmt_simple_memref_p): Declared.
+ (struct rdg_edge): Add a field ddr_p relation.
+ (build_empty_rdg): Declared.
+ * lambda.h (lambda_matrix): Declare a VEC of.
+ (find_induction_var_from_exit_cond): Declared.
+ (lambda_vector_compare): New.
+ * common.opt (fgraphite, floop-strip-mine,
+ floop-interchange, floop-block): New flags.
+ * lambda-code.c (find_induction_var_from_exit_cond): Export.
+ * cfgloop.c (is_loop_exit): New.
+ * cfgloop.h (is_loop_exit): Declared.
+ (create_empty_if_region_on_edge): Declared.
+ (create_empty_loop_on_edge): Declared.
+ * tree-flow.h (add_phi_node_to_bb): Declared.
+ (make_phi_node): Declared.
+ (rename_variables_in_bb): Declared.
+ (perfect_loop_nest_depth): Declared.
+ (graphite_transform_loops): Declared.
+ * Makefile.in (cfgloopmanip.o): Depend on TREE_FLOW_H.
+ (graphite.o-warn): Add -Wno-error.
+ (PPLLIBS, PPLINC, CLOOGLIBS, CLOOGINC): Declared.
+ (LIBS): Add GMPLIBS, CLOOGLIBS, PPLLIBS.
+ (INCLUDES): Add PPLINC, CLOOGINC.
+ (OBJS-common): Add graphite.o.
+ (graphite.o): Add rule.
+ * gimple.h (gimple_vec): ... here.
+ * tree-cfg.c (print_loops): Start printing at ENTRY_BLOCK_PTR.
+ * passes.c (init_optimization_passes): Schedule
+ pass_graphite_transforms.
+
+2008-09-02 Samuel Tardieu <sam@rfc1149.net>
+
+ PR target/37283
+ * opts.c (decode_options): Move processing of -fno-unit-at-a-time
+ and -fno-toplevel-reorder after handle_options.
+
+2008-09-02 Paul Brook <paul@codesourcery.com>
+
+ * config/arm/arm.c (arm_tune_cortex_a9): Define.
+ (arm_override_options): Set arm_tune_cortex_a9.
+ (arm_rtx_costs_1): Make register shifts more expensive on Cortex-A9.
+ (arm_issue_rate): Handle cortexa9.
+ * config/arm/arm.h (arm_tune_cortex_a9): Declare.
+ * config/arm/arm-cores.def: Add cortex-a9.
+ * config/arm/arm-tune.md: Regenerate.
+ * config/arm/arm.md: Include cortex-a9.md.
+ (generic_sched, generic_vfp): Add cortexa9.
+ * config/arm/cortex-a9.md: New file.
+
+2008-09-02 Richard Guenther <rguenther@suse.de>
+
+ PR tree-optimization/37327
+ * tree-vrp.c (register_new_assert_for): Make sure to not have
+ TREE_OVERFLOW set on the bound.
+
+2008-09-02 Alexander Monakov <amonakov@ispras.ru>
+
+ Fixed typo in date of the 2008-09-01 MOVE_RATIO commit.
+
+ * config/ia64/ia64.opt (msched-ar-data-spec): Default to 1.
+ * config/ia64/ia64.c (ia64_set_sched_flags): Always initialize
+ spec_info->mask.
+
+2008-09-02 Victor Kaplansky <victork@il.ibm.com>
+
+ * gcc/config/spu/spu.md (divdf3): Removed.
+
+2008-09-02 Jakub Jelinek <jakub@redhat.com>
+
+ PR tree-optimization/37095
+ * cgraph.c (cgraph_node): When creating new cgraph node after
+ assembler_name_hash has been populated, record it in the hash
+ table.
+
+2008-09-01 Paul Brook <paul@codesourcery.com>
+
+ * doc/invoke.texi: Document -mword-relocations.
+ * config/arm/uclinux-elf.h (TARGET_DEFAULT_WORD_RELOCATIONS): Define.
+ * config/arm/symbian.h (TARGET_DEFAULT_WORD_RELOCATIONS): Define.
+ * config/arm/vxworks.h (TARGET_DEFAULT_WORD_RELOCATIONS): Define.
+ * config/arm/arm.c (arm_split_constant): Use arm_emit_movpair.
+ (arm_rtx_costs_1, arm_size_rtx_costs): Handle HIGH and LO_SUM.
+ (arm_emit_movpair): New function.
+ (arm_print_operand <c>): Handle sybolic addresses.
+ * config/arm/arm.h (TARGET_USE_MOVT): Define.
+ (TARGET_DEFAULT_WORD_RELOCATIONS): Define.
+ * config/arm/arm-protos.h (arm_emit_movpair): Add prototype.
+ * config/arm/arm.opt: Add -mword-relocations.
+ * config/arm/arm.md (movsi): Use arm_emit_movpair.
+ (arm_movt, arm_movw): New.
+
+2008-09-01 Paul Brook <paul@codesourcery.com>
+
+ * config/arm/arm.c (arm_override_options): Set arm_abi earlier.
+ Allow Interworking on ARMv4 EABI based targets.
+ * config/arm/bpabi.h (TARGET_FIX_V4BX_SPEC): Define.
+ (SUBTARGET_EXTRA_ASM_SPEC, LINK_SPEC): Add TARGET_FIX_V4BX_SPEC.
+
+2008-09-01 Paul Brook <paul@codesourcery.com>
+
+ * config/arm/thumb2.md (thumb2_negscc): Match the correct operand for
+ optimized LT0 test. Remove optimization for GT.
+
+2008-09-01 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
+
+ * pa64-hpux.h (LIB_SPEC): Link against librt when building static
+ openmp applications.
+ * pa-hpux11.h (LIB_SPEC): Likewise.
+
+ * ira.c (setup_eliminable_regset): Only declare i if ELIMINABLE_REGS is
+ defined.
+
+2008-09-01 Jan Hubicka <jh@suse.cz>
+
+ * predict.c (maybe_hot_count_p, maybe_hot_edge_p): When profile is
+ present, use only counts; when guessed use only frequencies.
+ (estimate_bb_frequencies): Use function_frequency.
+ * profile.c (compute_branch_probabilities): Exit early when
+ profile reading failed. Set profile_status.
+ (branch_prob): Do not set profile_status.
+
+2008-09-01 Aldy Hernandez <aldyh@redhat.com>
+
+ * tree.c (protected_set_expr_location): New.
+ * tree.h (protected_set_expr_location): New prototype.
+ * c-tree.h (build_array_ref): Add argument.
+ (parser_build_unary_op): Same.
+ * c-typeck.c (build_indirect_ref): Handle new location argument.
+ (build_array_ref): Same.
+ (parser_build_unary_op): Same.
+ * gimplify.c (gimplify_asm_expr): Set input_location before calling
+ error.
+ * c-omp.c (c_finish_omp_atomic): Pass location when calling
+ build_indirect_ref.
+ * c-common.c (finish_label_address_expr): Handle new location
+ argument.
+ * c-common.h (build_indirect_ref): Add argument.
+ (finish_label_address_expr): Same.
+ * c-parser.c (c_parser_unary_expression): Pass location to build
+ functions.
+ (c_parser_postfix_expression): Same.
+
+2008-09-01 Paul Brook <paul@codesourcery.com>
+
+ * config/arm/arm.md: Include cortex-r4f.md.
+ (attr fpu): Update type list.
+ (attr type): Add fcpys, ffariths, ffarithd, fadds, faddd, fconsts,
+ fconstd, fcmps and fcmpd.
+ (attr tune_cortexr4): Define.
+ (attr generic_sched, attr generic_vfp): Use tune_cortexr4.
+ * config/arm/vfp.md: Document fcpys, ffariths, ffarithd, fadds, faddd,
+ fconsts, fconstd, fcmps and fcmpd. Use them in insn patterns.
+ * config/arm/arm.c (arm_issue_rate): Add cortexr4f.
+ * config/arm/arm1020e.md (v10_ffarith, v10_farith): Use new insn
+ types.
+ * config/arm/cortex-a8-neon.md (cortex_a8_vfp_add_sub,
+ cortex_a8_vfp_farith: Ditto.
+ * config/arm/vfp11.md (vfp_ffarith, vfp_farith): Ditto.
+ * config/arm/cortex-r4.md: Use tune_cortexr4.
+ * config/arm/cortex-r4f.md: New file.
+
+2008-09-01 Richard Guenther <rguenther@suse.de>
+
+ PR tree-optimization/37305
+ * tree-ssa-ccp.c (ccp_fold): Do not set TREE_OVERFLOW on
+ the result of constant conversions.
+ (fold_gimple_assign): Likewise.
+
+2008-09-01 Andrey Belevantsev <abel@ispras.ru>
+
+ * sel-sched-ir.c (cmp_v_in_regset_pool): Surround with
+ #ifdef ENABLE_CHECKING.
+
+2008-09-01 Andrey Belevantsev <abel@ispras.ru>
+
+ * sel-sched-dump.c: Surround with #ifdef INSN_SCHEDULING.
+ * sched-int.h (print_insn, print_pattern, print_value): Declare also
+ when !INSN_SCHEDULING.
+
+2008-09-01 Andreas Schwab <schwab@suse.de>
+
+ * config/ia64/ia64.c (TARGET_ADDRESS_COST): Update.
+
+2008-09-01 Paul Brook <paul@codesourcery.com>
+
+ * config/arm/arm.c (arm_no_early_mul_dep): Handle multiply-subtract.
+
+2008-09-01 Jan Hubicka <jh@suse.cz>
+
+ * config/spu/spu.c (spu_expand_block_move) Update call of MOVE_RATIO.
+
+ * expr.c (MOVE_BY_PIECES_P, CLEAR_BY_PIECES_P, SET_BY_PIECES_P): Pass speed operand.
+ * expr.h (MOVE_RATIO, CLEAR_RATIO, SET_RATIO): Update.
+ * gimplify.c (gimplify_init_constructor): Add speed operand.
+ * tree-sra.c (decide_block_copy): Likewise.
+ * tree-inline.c (estimate_move_cost): Likewise.
+ * config/alpha/alpha.h (MOVE_RATIO): Update.
+ * config/frv/frv.c (MOVE_RATIO): Update.
+ * config/spu/spu.h (MOVE_RATIO): Update.
+ * config/sparc/sparc.h (MOVE_RATIO): Update.
+ * config/i386/i386.h (MOVE_RATIO, CLEAR_RATIO): Update.
+ * config/m68hc11/m68hc11.h (MOVE_RATIO): Update.
+ * config/cris/cris.h (MOVE_RATIO): Update.
+ * config/mn10300/mn10300.h (MOVE_RATIO): Update.
+ * config/arm/arm.h (MOVE_RATIO): Update.
+ * config/pa/pa.md: Update uses of MOVE_RATIO
+ * config/pa/pa.h (MOVE_RATIO): Update.
+ * config/mips/mips.h (MOVE_RATIO, MOVE_BY_PIECES, CLEAR_RATIO, SET_RATIO): Update.
+ * config/h8300/h8300.h (MOVE_RATIO): Update.
+ * config/v850/v850.h (MOVE_RATIO): Update.
+ * config/bfin/bfin.h (MOVE_RATIO): Update.
+
+2008-08-31 Andrey Belevantsev <abel@ispras.ru>
+ Dmitry Melnik <dm@ispras.ru>
+ Dmitry Zhurikhin <zhur@ispras.ru>
+ Alexander Monakov <amonakov@ispras.ru>
+ Maxim Kuvyrkov <maxim@codesourcery.com>
+
+ * sel-sched.h, sel-sched-dump.h, sel-sched-ir.h, sel-sched.c,
+ sel-sched-dump.c, sel-sched-ir.c: New files.
+ * Makefile.in (OBJS-common): Add selective scheduling object
+ files.
+ (sel-sched.o, sel-sched-dump.o, sel-sched-ir.o): New entries.
+ (SEL_SCHED_IR_H, SEL_SCHED_DUMP_H): New entries.
+ (sched-vis.o): Add dependency on $(INSN_ATTR_H).
+ * cfghooks.h (get_cfg_hooks, set_cfg_hooks): New prototypes.
+ * cfghooks.c (get_cfg_hooks, set_cfg_hooks): New functions.
+ (make_forwarder_block): Update loop latch if we have redirected
+ the loop latch edge.
+ * cfgloop.c (get_loop_body_in_custom_order): New function.
+ * cfgloop.h (LOOPS_HAVE_FALLTHRU_PREHEADERS): New enum field.
+ (CP_FALLTHRU_PREHEADERS): Likewise.
+ (get_loop_body_in_custom_order): Declare.
+ * cfgloopmanip.c (has_preds_from_loop): New.
+ (create_preheader): Honor CP_FALLTHRU_PREHEADERS.
+ Assert that the preheader edge will be fall thru when it is set.
+ * common.opt (fsel-sched-bookkeeping, fsel-sched-pipelining,
+ fsel-sched-pipelining-outer-loops, fsel-sched-renaming,
+ fsel-sched-substitution, fselective-scheduling): New flags.
+ * cse.c (hash_rtx_cb): New.
+ (hash_rtx): Use it.
+ * dbgcnt.def (sel_sched_cnt, sel_sched_region_cnt,
+ sel_sched_insn_cnt): New counters.
+ * final.c (compute_alignments): Export. Free dominance info after loop_optimizer_finalize.
+ * genattr.c (main): Output maximal_insn_latency prototype.
+ * genautomata.c (output_default_latencies): New. Factor its code from ...
+ (output_internal_insn_latency_func): ... here.
+ (output_internal_maximal_insn_latency_func): New.
+ (output_maximal_insn_latency_func): New.
+ * hard-reg-set.h (UHOST_BITS_PER_WIDE_INT): Define unconditionally.
+ (struct hard_reg_set_iterator): New.
+ (hard_reg_set_iter_init, hard_reg_set_iter_set,
+ hard_reg_set_iter_next): New functions.
+ (EXECUTE_IF_SET_IN_HARD_REG_SET): New macro.
+ * lists.c (remove_free_INSN_LIST_node,
+ remove_free_EXPR_LIST_node): New functions.
+ * loop-init.c (loop_optimizer_init): When LOOPS_HAVE_FALLTHRU_PREHEADERS,
+ set CP_FALLTHRU_PREHEADERS when calling create_preheaders.
+ (loop_optimizer_finalize): Do not verify flow info after reload.
+ * recog.c (validate_replace_rtx_1): New parameter simplify.
+ Default it to true. Update all uses. Factor out simplifying
+ code to ...
+ (simplify_while_replacing): ... this new function.
+ (validate_replace_rtx_part,
+ validate_replace_rtx_part_nosimplify): New.
+ * recog.h (validate_replace_rtx_part,
+ validate_replace_rtx_part_nosimplify): Declare.
+ * rtl.c (rtx_equal_p_cb): New.
+ (rtx_equal_p): Use it.
+ * rtl.h (rtx_equal_p_cb, hash_rtx_cb): Declare.
+ (remove_free_INSN_LIST_NODE, remove_free_EXPR_LIST_node,
+ debug_bb_n_slim, debug_bb_slim, print_rtl_slim): Likewise.
+ * vecprim.h: Add a vector type for unsigned int.
+ * haifa-sched.c: Include vecprim.h and cfgloop.h.
+ (issue_rate, sched_verbose_param, note_list, dfa_state_size,
+ ready_try, cycle_issued_insns, spec_info): Make global.
+ (readyp): Initialize.
+ (dfa_lookahead): New global variable.
+ (old_max_uid, old_last_basic_block): Remove.
+ (h_i_d): Make it a vector.
+ (INSN_TICK, INTER_TICK, QUEUE_INDEX, INSN_COST): Make them work
+ through HID macro.
The results can be reproduced by building a compiler with
--enable-gather-detailed-mem-stats targetting x86-64
and compiling preprocessed combine.c or testcase from PR8632 with:
-fmem-report --param=ggc-min-heapsize=1024 --param=ggc-min-expand=1 -Ox -Q
The memory consumption summary appears in the dump after detailed listing
of the places they are allocated in. Peak memory consumption is actually
computed by looking for maximal value in {GC XXXX -> YYYY} report.
Your testing script.