On 06/27/2012 07:55 PM, Ramana Radhakrishnan wrote:
> I must admit that I had been suggesting to Zhenqiang about turning > this off by tightening the movsi_insn predicates rather than adding a > split, but given that it appears to produce enough benefit in this > case I don't have any reasons to object ... > > However it's interesting that this doesn't seem to help vpr .... We retested vpr, but it just seems to be unstable: base peak time time 175.vpr 1400 502 X 1400 526 X 175.vpr 1400 500 X 1400 524 X 175.vpr 1400 516 X 1400 526 X 175.vpr 1400 492 X 1400 481 X 175.vpr 1400 496 X 1400 485 X median 500 524 However, the minimum time is still better with the patch. And here are all 3 runs for previously reported data: test base ratio peak ratio median base median peak improvement 164.gzip 284 281 284 282 -0.70% 164.gzip 284 282 164.gzip 285 283 175.vpr 329 306 323 306 -5.26% 175.vpr 323 305 175.vpr 306 307 176.gcc 542 554 542 557 2.77% 176.gcc 541 558 176.gcc 544 557 181.mcf 343 340 340 341 0.29% 181.mcf 339 342 181.mcf 340 341 186.crafty 383 399 383 391 2.09% 186.crafty 390 391 186.crafty 380 386 197.parser 254 257 254 257 1.18% 197.parser 254 257 197.parser 254 257 252.eon 591 644 594 644 8.42% 252.eon 598 644 252.eon 594 643 253.perlbmk 462 490 463 490 5.83% 253.perlbmk 463 490 253.perlbmk 463 490 254.gap 415 473 425 467 9.88% 254.gap 430 467 254.gap 425 464 255.vortex 384 430 382 430 12.57% 255.vortex 382 430 255.vortex 381 430 256.bzip2 331 354 332 354 6.63% 256.bzip2 332 354 256.bzip2 335 349 300.twolf 323 356 328 352 7.32% 300.twolf 347 337 300.twolf 328 352 -- Best regards, Dmitry
