Baseline threaded code (computed goto) #495
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
Sweet! Am I reading this correctly? From the first two columns it looks like nothing got worse, and some got as much as 40% or more faster. But I don't really know what any of the columns mean. |
This is output from google/benchmark compare.py script. Only the first number matters which is the relative time change. (Negative values means the performance has increased). The wallclock and CPU times are almost the same. The other columns are the absolute timings, but are rather meaningless because the script rounds them and skips the units. The numbers for GCC are not so great so I'm trying to figure out why. Maybe I will use static table of label addresses for start. When using dynamic table (selected by EVM revision) this occupies one register in the function. |
Codecov Report
@@ Coverage Diff @@
## master #495 +/- ##
=======================================
Coverage 99.40% 99.41%
=======================================
Files 53 53
Lines 5249 5272 +23
=======================================
+ Hits 5218 5241 +23
Misses 31 31
Flags with carried forward coverage won't be shown. Click here to find out more.
|
chfast
force-pushed
the
baseline_cgoto
branch
2 times, most recently
from
b55187a to
0f01455
Compare
|
What are the benchmark results? |
Benchmark resultsClang 15, Haswell: 7%[bnocgoto vs. baseline]/execute/main/blake2b_huff/empty_mean -0.1422 -0.1422 13 11 13 11 [bnocgoto vs. baseline]/execute/main/blake2b_huff/2805nulls_mean -0.1425 -0.1425 289 247 289 247 [bnocgoto vs. baseline]/execute/main/blake2b_huff/5610nulls_mean -0.1489 -0.1489 567 482 567 482 [bnocgoto vs. baseline]/execute/main/blake2b_huff/8415nulls_mean -0.1413 -0.1413 824 707 824 707 [bnocgoto vs. baseline]/execute/main/blake2b_huff/65536nulls_mean -0.1409 -0.1409 6383 5484 6383 5484 [bnocgoto vs. baseline]/execute/main/blake2b_shifts/2805nulls_mean -0.1608 -0.1608 2814 2362 2814 2362 [bnocgoto vs. baseline]/execute/main/blake2b_shifts/5610nulls_mean -0.1706 -0.1706 5709 4735 5709 4735 [bnocgoto vs. baseline]/execute/main/blake2b_shifts/8415nulls_mean -0.1525 -0.1525 8381 7103 8381 7103 [bnocgoto vs. baseline]/execute/main/blake2b_shifts/65536nulls_mean -0.1520 -0.1520 64886 55024 64887 55025 [bnocgoto vs. baseline]/execute/main/sha1_divs/empty_mean -0.0491 -0.0491 53 50 53 50 [bnocgoto vs. baseline]/execute/main/sha1_divs/1351_mean -0.0454 -0.0454 1103 1053 1103 1053 [bnocgoto vs. baseline]/execute/main/sha1_divs/2737_mean -0.0447 -0.0447 2150 2054 2150 2054 [bnocgoto vs. baseline]/execute/main/sha1_divs/5311_mean -0.0468 -0.0468 4202 4006 4202 4006 [bnocgoto vs. baseline]/execute/main/sha1_divs/65536_mean -0.0463 -0.0463 51231 48858 51231 48859 [bnocgoto vs. baseline]/execute/main/sha1_shifts/empty_mean -0.0758 -0.0758 30 27 30 27 [bnocgoto vs. baseline]/execute/main/sha1_shifts/1351_mean -0.0666 -0.0666 631 589 631 589 [bnocgoto vs. baseline]/execute/main/sha1_shifts/2737_mean -0.0680 -0.0680 1233 1149 1233 1149 [bnocgoto vs. baseline]/execute/main/sha1_shifts/5311_mean -0.0694 -0.0694 2410 2243 2410 2243 [bnocgoto vs. baseline]/execute/main/sha1_shifts/65536_mean -0.0807 -0.0807 29799 27393 29799 27393 [bnocgoto vs. baseline]/execute/main/weierstrudel/0_mean -0.0212 -0.0212 164 160 164 160 [bnocgoto vs. baseline]/execute/main/weierstrudel/1_mean -0.0345 -0.0345 361 349 361 349 [bnocgoto vs. baseline]/execute/main/weierstrudel/3_mean -0.0324 -0.0324 562 543 562 543 [bnocgoto vs. baseline]/execute/main/weierstrudel/9_mean -0.0357 -0.0357 1164 1123 1164 1123 [bnocgoto vs. baseline]/execute/main/weierstrudel/14_mean -0.0361 -0.0361 1666 1606 1666 1606 [bnocgoto vs. baseline]/execute/micro/jump_around/empty_mean -0.1220 -0.1220 26 23 26 23 [bnocgoto vs. baseline]/execute/micro/jumpdests_0xffff/empty_mean +0.0109 +0.0109 111 112 111 112 [bnocgoto vs. baseline]/execute/micro/loop_with_many_jumpdests/empty_mean -0.0271 -0.0271 14759 14360 14759 14360 [bnocgoto vs. baseline]/execute/micro/memory_grow_mload/nogrow_mean -0.0575 -0.0575 49 46 49 46 [bnocgoto vs. baseline]/execute/micro/memory_grow_mload/by1_mean -0.0596 -0.0596 52 49 52 49 [bnocgoto vs. baseline]/execute/micro/memory_grow_mload/by16_mean -0.0681 -0.0681 58 54 58 54 [bnocgoto vs. baseline]/execute/micro/memory_grow_mload/by32_mean -0.0675 -0.0675 66 62 66 62 [bnocgoto vs. baseline]/execute/micro/memory_grow_mstore/nogrow_mean -0.0253 -0.0253 51 49 51 49 [bnocgoto vs. baseline]/execute/micro/memory_grow_mstore/by1_mean -0.0287 -0.0287 55 53 55 53 [bnocgoto vs. baseline]/execute/micro/memory_grow_mstore/by16_mean -0.0032 -0.0032 61 61 61 61 [bnocgoto vs. baseline]/execute/micro/memory_grow_mstore/by32_mean -0.0130 -0.0130 73 72 73 72 [bnocgoto vs. baseline]/execute/micro/signextend/zero_mean -0.1160 -0.1160 66 59 66 59 [bnocgoto vs. baseline]/execute/micro/signextend/one_mean -0.1150 -0.1150 66 59 66 59 OVERALL_GEOMEAN -0.0769 -0.0769 0 0 0 0 GCC 12, Haswell: 1%[bnocgoto vs. baseline]/execute/main/blake2b_huff/empty_mean -0.0069 -0.0069 14 14 14 14 [bnocgoto vs. baseline]/execute/main/blake2b_huff/2805nulls_mean -0.0044 -0.0044 317 315 317 315 [bnocgoto vs. baseline]/execute/main/blake2b_huff/5610nulls_mean -0.0046 -0.0046 619 616 619 616 [bnocgoto vs. baseline]/execute/main/blake2b_huff/8415nulls_mean -0.0043 -0.0043 907 903 907 903 [bnocgoto vs. baseline]/execute/main/blake2b_huff/65536nulls_mean -0.0049 -0.0049 7049 7015 7049 7015 [bnocgoto vs. baseline]/execute/main/blake2b_shifts/2805nulls_mean +0.0037 +0.0037 2744 2754 2744 2754 [bnocgoto vs. baseline]/execute/main/blake2b_shifts/5610nulls_mean +0.0053 +0.0053 5465 5494 5465 5494 [bnocgoto vs. baseline]/execute/main/blake2b_shifts/8415nulls_mean +0.0069 +0.0069 8190 8246 8190 8246 [bnocgoto vs. baseline]/execute/main/blake2b_shifts/65536nulls_mean +0.0055 +0.0055 63504 63851 63504 63852 [bnocgoto vs. baseline]/execute/main/sha1_divs/empty_mean -0.0159 -0.0159 51 50 51 50 [bnocgoto vs. baseline]/execute/main/sha1_divs/1351_mean +0.0048 +0.0048 1066 1071 1066 1071 [bnocgoto vs. baseline]/execute/main/sha1_divs/2737_mean +0.0025 +0.0025 2083 2088 2083 2088 [bnocgoto vs. baseline]/execute/main/sha1_divs/5311_mean +0.0038 +0.0038 4063 4078 4063 4078 [bnocgoto vs. baseline]/execute/main/sha1_divs/65536_mean +0.0038 +0.0038 49465 49651 49465 49651 [bnocgoto vs. baseline]/execute/main/sha1_shifts/empty_mean -0.0257 -0.0258 25 24 25 24 [bnocgoto vs. baseline]/execute/main/sha1_shifts/1351_mean -0.0245 -0.0245 533 520 533 520 [bnocgoto vs. baseline]/execute/main/sha1_shifts/2737_mean -0.0243 -0.0243 1042 1016 1042 1016 [bnocgoto vs. baseline]/execute/main/sha1_shifts/5311_mean -0.0239 -0.0239 2032 1984 2032 1984 [bnocgoto vs. baseline]/execute/main/sha1_shifts/65536_mean -0.0248 -0.0248 24788 24173 24789 24173 [bnocgoto vs. baseline]/execute/main/weierstrudel/0_mean -0.0113 -0.0113 244 242 244 242 [bnocgoto vs. baseline]/execute/main/weierstrudel/1_mean -0.0174 -0.0174 535 525 535 525 [bnocgoto vs. baseline]/execute/main/weierstrudel/3_mean -0.0150 -0.0150 839 826 839 826 [bnocgoto vs. baseline]/execute/main/weierstrudel/9_mean -0.0140 -0.0140 1740 1716 1740 1716 [bnocgoto vs. baseline]/execute/main/weierstrudel/14_mean -0.0124 -0.0124 2493 2462 2493 2462 [bnocgoto vs. baseline]/execute/micro/jump_around/empty_mean -0.0112 -0.0112 23 23 23 23 [bnocgoto vs. baseline]/execute/micro/jumpdests_0xffff/empty_mean -0.0218 -0.0218 108 105 108 105 [bnocgoto vs. baseline]/execute/micro/loop_with_many_jumpdests/empty_mean -0.0182 -0.0182 14031 13776 14031 13776 [bnocgoto vs. baseline]/execute/micro/memory_grow_mload/nogrow_mean -0.0000 -0.0000 63 63 63 63 [bnocgoto vs. baseline]/execute/micro/memory_grow_mload/by1_mean -0.0008 -0.0008 64 64 64 64 [bnocgoto vs. baseline]/execute/micro/memory_grow_mload/by16_mean -0.0081 -0.0081 72 71 72 71 [bnocgoto vs. baseline]/execute/micro/memory_grow_mload/by32_mean -0.0111 -0.0111 81 80 81 80 [bnocgoto vs. baseline]/execute/micro/memory_grow_mstore/nogrow_mean -0.0591 -0.0591 51 48 51 48 [bnocgoto vs. baseline]/execute/micro/memory_grow_mstore/by1_mean -0.0302 -0.0302 52 50 52 50 [bnocgoto vs. baseline]/execute/micro/memory_grow_mstore/by16_mean -0.0347 -0.0347 62 60 62 60 [bnocgoto vs. baseline]/execute/micro/memory_grow_mstore/by32_mean -0.0197 -0.0197 73 71 73 71 [bnocgoto vs. baseline]/execute/micro/signextend/zero_mean -0.0441 -0.0441 55 52 55 52 [bnocgoto vs. baseline]/execute/micro/signextend/one_mean -0.0433 -0.0433 55 52 55 52 OVERALL_GEOMEAN -0.0137 -0.0137 0 0 0 0 Clang 13, Skylake: 10%[bnocgoto vs. baseline]/execute/main/blake2b_huff/empty_mean -0.1647 -0.1647 13 11 13 11 [bnocgoto vs. baseline]/execute/main/blake2b_huff/2805nulls_mean -0.1671 -0.1671 292 243 292 243 [bnocgoto vs. baseline]/execute/main/blake2b_huff/5610nulls_mean -0.1681 -0.1681 567 472 567 472 [bnocgoto vs. baseline]/execute/main/blake2b_huff/8415nulls_mean -0.1650 -0.1650 834 696 834 696 [bnocgoto vs. baseline]/execute/main/blake2b_huff/65536nulls_mean -0.1686 -0.1687 6463 5373 6463 5373 [bnocgoto vs. baseline]/execute/main/blake2b_shifts/2805nulls_mean -0.1755 -0.1755 2692 2220 2692 2219 [bnocgoto vs. baseline]/execute/main/blake2b_shifts/5610nulls_mean -0.1730 -0.1730 5362 4434 5362 4434 [bnocgoto vs. baseline]/execute/main/blake2b_shifts/8415nulls_mean -0.1703 -0.1703 8005 6642 8005 6642 [bnocgoto vs. baseline]/execute/main/blake2b_shifts/65536nulls_mean -0.1721 -0.1721 62371 51637 62369 51635 [bnocgoto vs. baseline]/execute/main/sha1_divs/empty_mean -0.0567 -0.0567 48 45 48 45 [bnocgoto vs. baseline]/execute/main/sha1_divs/1351_mean -0.0589 -0.0589 998 939 998 939 [bnocgoto vs. baseline]/execute/main/sha1_divs/2737_mean -0.0588 -0.0588 1931 1817 1931 1817 [bnocgoto vs. baseline]/execute/main/sha1_divs/5311_mean -0.0502 -0.0502 3755 3567 3755 3567 [bnocgoto vs. baseline]/execute/main/sha1_divs/65536_mean -0.0505 -0.0505 45650 43344 45649 43342 [bnocgoto vs. baseline]/execute/main/sha1_shifts/empty_mean -0.1035 -0.1035 29 26 29 26 [bnocgoto vs. baseline]/execute/main/sha1_shifts/1351_mean -0.1004 -0.1004 610 548 610 548 [bnocgoto vs. baseline]/execute/main/sha1_shifts/2737_mean -0.0966 -0.0966 1185 1070 1185 1070 [bnocgoto vs. baseline]/execute/main/sha1_shifts/5311_mean -0.1038 -0.1037 2308 2068 2308 2068 [bnocgoto vs. baseline]/execute/main/sha1_shifts/65536_mean -0.1067 -0.1067 28173 25167 28172 25166 [bnocgoto vs. baseline]/execute/main/weierstrudel/0_mean -0.0091 -0.0091 148 147 148 147 [bnocgoto vs. baseline]/execute/main/weierstrudel/1_mean -0.0223 -0.0222 326 318 326 318 [bnocgoto vs. baseline]/execute/main/weierstrudel/3_mean -0.0209 -0.0209 509 498 509 498 [bnocgoto vs. baseline]/execute/main/weierstrudel/9_mean -0.0186 -0.0186 1048 1029 1048 1029 [bnocgoto vs. baseline]/execute/main/weierstrudel/14_mean -0.0220 -0.0220 1505 1472 1505 1472 [bnocgoto vs. baseline]/execute/micro/jump_around/empty_mean -0.0982 -0.0982 28 25 28 25 [bnocgoto vs. baseline]/execute/micro/jumpdests_0xffff/empty_mean -0.0811 -0.0811 80 73 80 73 [bnocgoto vs. baseline]/execute/micro/loop_with_many_jumpdests/empty_mean -0.1082 -0.1081 9649 8606 9649 8605 [bnocgoto vs. baseline]/execute/micro/memory_grow_mload/nogrow_mean -0.0978 -0.0978 48 43 48 43 [bnocgoto vs. baseline]/execute/micro/memory_grow_mload/by1_mean -0.0870 -0.0870 51 47 51 47 [bnocgoto vs. baseline]/execute/micro/memory_grow_mload/by16_mean -0.0596 -0.0596 56 53 56 53 [bnocgoto vs. baseline]/execute/micro/memory_grow_mload/by32_mean -0.0551 -0.0551 63 60 63 60 [bnocgoto vs. baseline]/execute/micro/memory_grow_mstore/nogrow_mean -0.0795 -0.0795 48 44 48 44 [bnocgoto vs. baseline]/execute/micro/memory_grow_mstore/by1_mean -0.0729 -0.0729 51 47 51 47 [bnocgoto vs. baseline]/execute/micro/memory_grow_mstore/by16_mean -0.0577 -0.0577 57 54 57 54 [bnocgoto vs. baseline]/execute/micro/memory_grow_mstore/by32_mean -0.0538 -0.0538 67 63 67 63 [bnocgoto vs. baseline]/execute/micro/signextend/zero_mean -0.2587 -0.2587 74 55 74 55 [bnocgoto vs. baseline]/execute/micro/signextend/one_mean -0.2549 -0.2549 74 55 74 55 OVERALL_GEOMEAN -0.1041 -0.1041 0 0 0 0 Clang 15, Skylake 11%[bnocgoto vs. baseline]/execute/main/blake2b_huff/empty_mean -0.1757 -0.1757 16 13 16 13 [bnocgoto vs. baseline]/execute/main/blake2b_huff/2805nulls_mean -0.1703 -0.1703 340 282 340 282 [bnocgoto vs. baseline]/execute/main/blake2b_huff/5610nulls_mean -0.1767 -0.1767 665 548 665 548 [bnocgoto vs. baseline]/execute/main/blake2b_huff/8415nulls_mean -0.1723 -0.1723 974 806 974 806 [bnocgoto vs. baseline]/execute/main/blake2b_huff/65536nulls_mean -0.1826 -0.1826 7615 6225 7615 6225 [bnocgoto vs. baseline]/execute/main/blake2b_shifts/2805nulls_mean -0.2145 -0.2145 3174 2493 3174 2493 [bnocgoto vs. baseline]/execute/main/blake2b_shifts/5610nulls_mean -0.2057 -0.2057 6322 5021 6322 5021 [bnocgoto vs. baseline]/execute/main/blake2b_shifts/8415nulls_mean -0.2004 -0.2004 9365 7488 9365 7488 [bnocgoto vs. baseline]/execute/main/blake2b_shifts/65536nulls_mean -0.2130 -0.2130 73398 57763 73394 57760 [bnocgoto vs. baseline]/execute/main/sha1_divs/empty_mean -0.0626 -0.0626 53 50 53 50 [bnocgoto vs. baseline]/execute/main/sha1_divs/1351_mean -0.0716 -0.0716 1119 1039 1119 1039 [bnocgoto vs. baseline]/execute/main/sha1_divs/2737_mean -0.0809 -0.0809 2174 1998 2173 1998 [bnocgoto vs. baseline]/execute/main/sha1_divs/5311_mean -0.0698 -0.0698 4205 3911 4204 3911 [bnocgoto vs. baseline]/execute/main/sha1_divs/65536_mean -0.0846 -0.0846 51568 47204 51567 47202 [bnocgoto vs. baseline]/execute/main/sha1_shifts/empty_mean -0.0927 -0.0927 32 29 32 29 [bnocgoto vs. baseline]/execute/main/sha1_shifts/1351_mean -0.0851 -0.0851 684 625 684 625 [bnocgoto vs. baseline]/execute/main/sha1_shifts/2737_mean -0.0847 -0.0847 1317 1205 1317 1205 [bnocgoto vs. baseline]/execute/main/sha1_shifts/5311_mean -0.0806 -0.0806 2561 2354 2561 2354 [bnocgoto vs. baseline]/execute/main/sha1_shifts/65536_mean -0.0793 -0.0793 31022 28563 31021 28562 [bnocgoto vs. baseline]/execute/main/weierstrudel/0_mean -0.0173 -0.0173 165 162 165 162 [bnocgoto vs. baseline]/execute/main/weierstrudel/1_mean -0.0277 -0.0277 363 353 363 353 [bnocgoto vs. baseline]/execute/main/weierstrudel/3_mean -0.0322 -0.0322 569 551 569 551 [bnocgoto vs. baseline]/execute/main/weierstrudel/9_mean -0.0389 -0.0389 1171 1126 1171 1125 [bnocgoto vs. baseline]/execute/main/weierstrudel/14_mean -0.0316 -0.0316 1660 1607 1660 1607 [bnocgoto vs. baseline]/execute/micro/jump_around/empty_mean -0.0109 -0.0109 27 27 27 27 [bnocgoto vs. baseline]/execute/micro/jumpdests_0xffff/empty_mean -0.0139 -0.0139 78 77 78 77 [bnocgoto vs. baseline]/execute/micro/loop_with_many_jumpdests/empty_mean -0.0005 -0.0005 9400 9395 9399 9394 [bnocgoto vs. baseline]/execute/micro/memory_grow_mload/nogrow_mean -0.1116 -0.1116 54 48 54 48 [bnocgoto vs. baseline]/execute/micro/memory_grow_mload/by1_mean -0.1117 -0.1117 59 53 59 52 [bnocgoto vs. baseline]/execute/micro/memory_grow_mload/by16_mean -0.4076 -0.4076 100 59 100 59 [bnocgoto vs. baseline]/execute/micro/memory_grow_mload/by32_mean -0.1171 -0.1171 76 67 76 67 [bnocgoto vs. baseline]/execute/micro/memory_grow_mstore/nogrow_mean -0.1462 -0.1462 59 51 59 51 [bnocgoto vs. baseline]/execute/micro/memory_grow_mstore/by1_mean -0.1332 -0.1332 62 54 62 54 [bnocgoto vs. baseline]/execute/micro/memory_grow_mstore/by16_mean -0.1517 -0.1517 72 61 72 61 [bnocgoto vs. baseline]/execute/micro/memory_grow_mstore/by32_mean -0.1351 -0.1351 82 71 82 71 [bnocgoto vs. baseline]/execute/micro/signextend/zero_mean +0.0159 +0.0160 70 71 70 71 [bnocgoto vs. baseline]/execute/micro/signextend/one_mean +0.0190 +0.0189 70 71 70 71 OVERALL_GEOMEAN -0.1112 -0.1112 0 0 0 0 GCC 12, Skylake: 2%[bnocgoto vs. baseline]/execute/main/blake2b_huff/empty_mean -0.0199 -0.0200 14 14 14 14 [bnocgoto vs. baseline]/execute/main/blake2b_huff/2805nulls_mean -0.0085 -0.0085 310 308 310 308 [bnocgoto vs. baseline]/execute/main/blake2b_huff/5610nulls_mean -0.0111 -0.0111 608 601 608 601 [bnocgoto vs. baseline]/execute/main/blake2b_huff/8415nulls_mean -0.0084 -0.0084 887 880 887 880 [bnocgoto vs. baseline]/execute/main/blake2b_huff/65536nulls_mean -0.0091 -0.0091 6919 6856 6919 6856 [bnocgoto vs. baseline]/execute/main/blake2b_shifts/2805nulls_mean -0.0327 -0.0327 2695 2607 2695 2607 [bnocgoto vs. baseline]/execute/main/blake2b_shifts/5610nulls_mean -0.0312 -0.0312 5380 5212 5380 5212 [bnocgoto vs. baseline]/execute/main/blake2b_shifts/8415nulls_mean -0.0275 -0.0275 8047 7826 8047 7825 [bnocgoto vs. baseline]/execute/main/blake2b_shifts/65536nulls_mean -0.0276 -0.0276 62323 60603 62320 60600 [bnocgoto vs. baseline]/execute/main/sha1_divs/empty_mean -0.0368 -0.0368 49 47 49 47 [bnocgoto vs. baseline]/execute/main/sha1_divs/1351_mean -0.0175 -0.0175 1028 1010 1028 1010 [bnocgoto vs. baseline]/execute/main/sha1_divs/2737_mean -0.0242 -0.0242 1996 1948 1996 1948 [bnocgoto vs. baseline]/execute/main/sha1_divs/5311_mean -0.0187 -0.0187 3870 3797 3870 3797 [bnocgoto vs. baseline]/execute/main/sha1_divs/65536_mean -0.0186 -0.0186 46916 46043 46914 46042 [bnocgoto vs. baseline]/execute/main/sha1_shifts/empty_mean -0.0332 -0.0332 25 24 25 24 [bnocgoto vs. baseline]/execute/main/sha1_shifts/1351_mean -0.0288 -0.0288 530 515 530 515 [bnocgoto vs. baseline]/execute/main/sha1_shifts/2737_mean -0.0303 -0.0303 1031 999 1031 999 [bnocgoto vs. baseline]/execute/main/sha1_shifts/5311_mean -0.0373 -0.0373 2002 1927 2002 1927 [bnocgoto vs. baseline]/execute/main/sha1_shifts/65536_mean -0.0349 -0.0349 24114 23273 24113 23272 [bnocgoto vs. baseline]/execute/main/weierstrudel/0_mean -0.0021 -0.0021 212 211 212 211 [bnocgoto vs. baseline]/execute/main/weierstrudel/1_mean -0.0109 -0.0109 465 460 465 460 [bnocgoto vs. baseline]/execute/main/weierstrudel/3_mean -0.0097 -0.0097 724 717 724 717 [bnocgoto vs. baseline]/execute/main/weierstrudel/9_mean -0.0024 -0.0024 1482 1479 1482 1479 [bnocgoto vs. baseline]/execute/main/weierstrudel/14_mean +0.0015 +0.0015 2113 2116 2113 2116 [bnocgoto vs. baseline]/execute/micro/jump_around/empty_mean -0.0827 -0.0827 22 21 22 21 [bnocgoto vs. baseline]/execute/micro/jumpdests_0xffff/empty_mean +0.0261 +0.0261 72 74 72 74 [bnocgoto vs. baseline]/execute/micro/loop_with_many_jumpdests/empty_mean -0.0005 -0.0005 8690 8686 8690 8685 [bnocgoto vs. baseline]/execute/micro/memory_grow_mload/nogrow_mean -0.0033 -0.0033 59 59 59 59 [bnocgoto vs. baseline]/execute/micro/memory_grow_mload/by1_mean -0.0052 -0.0052 60 60 60 60 [bnocgoto vs. baseline]/execute/micro/memory_grow_mload/by16_mean -0.0129 -0.0129 70 69 70 69 [bnocgoto vs. baseline]/execute/micro/memory_grow_mload/by32_mean -0.0193 -0.0193 77 76 77 76 [bnocgoto vs. baseline]/execute/micro/memory_grow_mstore/nogrow_mean -0.0302 -0.0302 48 47 48 47 [bnocgoto vs. baseline]/execute/micro/memory_grow_mstore/by1_mean -0.0244 -0.0244 49 48 49 48 [bnocgoto vs. baseline]/execute/micro/memory_grow_mstore/by16_mean -0.0211 -0.0212 58 56 58 56 [bnocgoto vs. baseline]/execute/micro/memory_grow_mstore/by32_mean -0.0184 -0.0184 67 66 67 66 [bnocgoto vs. baseline]/execute/micro/signextend/zero_mean -0.1905 -0.1905 62 50 62 50 [bnocgoto vs. baseline]/execute/micro/signextend/one_mean -0.1989 -0.1989 63 50 63 50 OVERALL_GEOMEAN -0.0297 -0.0297 0 0 0 0 SummaryThe improvements for GCC are minimal, but at least it is not a regression. Clang benefits more. Run benchmarksRun benchmarks yourself and attach the bin/evmone-bench ../../test/benchmarks --benchmark_display_aggregates_only --benchmark_filter='/execute/' --benchmark_repetitions=23 --benchmark_min_time=0.1 --benchmark_out=results.json |
axic
reviewed
Aug 23, 2022
yperbasis
approved these changes
Aug 24, 2022
axic
reviewed
Aug 24, 2022
| dispatch<false>(cost_table, state); | ||
| { | ||
| #if EVMONE_CGOTO_SUPPORTED | ||
| if (vm.cgoto) |
There was a problem hiding this comment.
I wonder if making this else if (INTX_LIKELY(vm.cgoto)) can trigger any better layout? Since the default case is cgoto unless disabled.
There was a problem hiding this comment.
We don't want to use likely/unlikely here, because there is a chance compiler will compile the unlikely branch for code size. I want them to have equal chances because they are competing against each other in benchmarks.
chfast
commented
Aug 24, 2022
| @@ -276,7 +323,14 @@ evmc_result execute(const VM& vm, ExecutionState& state, const CodeAnalysis& ana | |||
| if (INTX_UNLIKELY(tracer != nullptr)) | |||
| dispatch<true>(cost_table, state, tracer); | |||
| else | |||
| dispatch<false>(cost_table, state); | |||
| { | |||
| #if EVMONE_CGOTO_SUPPORTED | |||
There was a problem hiding this comment.
We can eliminate the #if by providing a stub dispatch_cgoto() implementation for unsupported compilers and making sure vm.cgoto is always false. However, the last condition can only be checked at runtime. With the current version we will get compilation error if something is messed up.
Add option to handle undefined opcodes, do not pass identifier by default.
Add EVMC option cgoto=no to disable the "computed goto" dispatch in Baseline.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This finally adds "computed goto" dispatch loop implementation. This is enabled by default for Baseline if a compiler supports the Labels as Values extension.