mirror of
https://github.com/golang/go
synced 2024-09-29 12:24:31 -06:00
ed7a8332c4
227 Commits
| Author | SHA1 | Message | Date | |
|---|---|---|---|---|
Keith Randall
|
bba88467f8 |
cmd/compile: add indexed-load CMP instructions
Things like CMPQ 4(AX)(BX*8), CX Fixes #37955 Change-Id: Icbed430f65c91a0e3f38a633d8321d79433ad8b3 Reviewed-on: https://go-review.googlesource.com/c/go/+/224219 Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: David Chase <drchase@google.com> |
||
Josh Bleecher Snyder
|
8114242359 |
cmd/compile, runtime: use more registers for amd64 write barrier calls
The compiler-inserted write barrier calls use a special ABI for speed and to minimize the binary size impact. runtime.gcWriteBarrier takes its args in DI and AX. This change adds gcWriteBarrier wrapper functions, varying only in the register used for the second argument. (Allowing variation in the first argument doesn't offer improvements, which is convenient, as it avoids quadratic API growth.) This reduces the number of register copies. The goals are reduced binary size via reduced register pressure/copies. One downside to this change is that when the write barrier is on, we may bounce through several different write barrier wrappers, which is bad for the instruction cache. Package runtime write barrier benchmarks for this change: name old time/op new time/op delta WriteBarrier-8 16.6ns ± 6% 15.6ns ± 6% -5.73% (p=0.000 n=97+99) BulkWriteBarrier-8 4.37ns ± 7% 4.22ns ± 8% -3.45% (p=0.000 n=96+99) However, I don't particularly trust these numbers. I ran runtime.BenchmarkWriteBarrier multiple times as I rebased this change, and noticed that the results have high variance depending on the parent change, perhaps due to aligment. This change was stress tested with GOGC=1 GODEBUG=gccheckmark=1 go test std. This change reduces binary sizes: file before after Δ % addr2line 4308720 4296688 -12032 -0.279% api 5965592 5945368 -20224 -0.339% asm 5148088 5025464 -122624 -2.382% buildid 2848760 2844904 -3856 -0.135% cgo 4828968 4812840 -16128 -0.334% compile 19754720 19529744 -224976 -1.139% cover 5256840 5236600 -20240 -0.385% dist 3670312 3658264 -12048 -0.328% doc 4669608 4657576 -12032 -0.258% fix 3377976 3365944 -12032 -0.356% link 6614888 6586472 -28416 -0.430% nm 4258368 4254528 -3840 -0.090% objdump 4656336 4644304 -12032 -0.258% pack 2295176 2295432 +256 +0.011% pprof 14762356 14709364 -52992 -0.359% test2json 2824456 2820600 -3856 -0.137% trace 11684404 11643700 -40704 -0.348% vet 8284760 8252248 -32512 -0.392% total 115210328 114580040 -630288 -0.547% This change improves compiler performance: name old time/op new time/op delta Template 208ms ± 3% 207ms ± 3% -0.40% (p=0.030 n=43+44) Unicode 80.2ms ± 3% 81.3ms ± 3% +1.25% (p=0.000 n=41+44) GoTypes 699ms ± 3% 694ms ± 2% -0.71% (p=0.016 n=42+37) Compiler 3.26s ± 2% 3.23s ± 2% -0.86% (p=0.000 n=43+45) SSA 6.97s ± 1% 6.93s ± 1% -0.63% (p=0.000 n=43+45) Flate 134ms ± 3% 133ms ± 2% ~ (p=0.139 n=45+42) GoParser 165ms ± 2% 164ms ± 1% -0.79% (p=0.000 n=45+40) Reflect 434ms ± 4% 435ms ± 4% ~ (p=0.937 n=44+44) Tar 181ms ± 2% 181ms ± 2% ~ (p=0.702 n=43+45) XML 244ms ± 2% 244ms ± 2% ~ (p=0.237 n=45+44) [Geo mean] 403ms 402ms -0.29% name old user-time/op new user-time/op delta Template 271ms ± 2% 268ms ± 1% -1.40% (p=0.000 n=42+42) Unicode 117ms ± 3% 116ms ± 5% ~ (p=0.066 n=45+45) GoTypes 948ms ± 2% 936ms ± 2% -1.30% (p=0.000 n=41+40) Compiler 4.26s ± 1% 4.21s ± 2% -1.25% (p=0.000 n=37+45) SSA 9.52s ± 2% 9.41s ± 1% -1.18% (p=0.000 n=44+45) Flate 167ms ± 2% 165ms ± 2% -1.15% (p=0.000 n=44+41) GoParser 201ms ± 2% 198ms ± 1% -1.40% (p=0.000 n=43+43) Reflect 563ms ± 8% 560ms ± 7% ~ (p=0.206 n=45+44) Tar 224ms ± 2% 222ms ± 2% -0.81% (p=0.000 n=45+45) XML 308ms ± 2% 304ms ± 1% -1.17% (p=0.000 n=42+43) [Geo mean] 525ms 519ms -1.08% name old alloc/op new alloc/op delta Template 36.3MB ± 0% 36.3MB ± 0% ~ (p=0.421 n=5+5) Unicode 28.4MB ± 0% 28.3MB ± 0% ~ (p=0.056 n=5+5) GoTypes 121MB ± 0% 121MB ± 0% -0.14% (p=0.008 n=5+5) Compiler 567MB ± 0% 567MB ± 0% -0.06% (p=0.016 n=4+5) SSA 1.26GB ± 0% 1.26GB ± 0% -0.07% (p=0.008 n=5+5) Flate 22.9MB ± 0% 22.8MB ± 0% ~ (p=0.310 n=5+5) GoParser 28.0MB ± 0% 27.9MB ± 0% -0.09% (p=0.008 n=5+5) Reflect 78.4MB ± 0% 78.4MB ± 0% -0.03% (p=0.008 n=5+5) Tar 34.2MB ± 0% 34.2MB ± 0% -0.05% (p=0.008 n=5+5) XML 44.4MB ± 0% 44.4MB ± 0% -0.04% (p=0.016 n=5+5) [Geo mean] 76.4MB 76.3MB -0.05% name old allocs/op new allocs/op delta Template 356k ± 0% 356k ± 0% -0.13% (p=0.008 n=5+5) Unicode 326k ± 0% 326k ± 0% -0.07% (p=0.008 n=5+5) GoTypes 1.24M ± 0% 1.24M ± 0% -0.24% (p=0.008 n=5+5) Compiler 5.30M ± 0% 5.28M ± 0% -0.34% (p=0.008 n=5+5) SSA 11.9M ± 0% 11.9M ± 0% -0.16% (p=0.008 n=5+5) Flate 226k ± 0% 225k ± 0% -0.12% (p=0.008 n=5+5) GoParser 287k ± 0% 286k ± 0% -0.29% (p=0.008 n=5+5) Reflect 930k ± 0% 929k ± 0% -0.05% (p=0.008 n=5+5) Tar 332k ± 0% 331k ± 0% -0.12% (p=0.008 n=5+5) XML 411k ± 0% 411k ± 0% -0.12% (p=0.008 n=5+5) [Geo mean] 771k 770k -0.16% For some packages, this change significantly reduces the size of executable text. Examples: file before after Δ % cmd/internal/obj/arm.s 68658 66855 -1803 -2.626% cmd/internal/obj/mips.s 57486 56272 -1214 -2.112% cmd/internal/obj/arm64.s 152107 147163 -4944 -3.250% cmd/internal/obj/ppc64.s 125544 120456 -5088 -4.053% cmd/vendor/golang.org/x/tools/go/cfg.s 31699 30742 -957 -3.019% Full listing: file before after Δ % container/ring.s 1890 1870 -20 -1.058% container/list.s 5366 5390 +24 +0.447% internal/cpu.s 3298 3295 -3 -0.091% internal/testlog.s 1507 1501 -6 -0.398% image/color.s 8281 8248 -33 -0.399% runtime.s 480970 480075 -895 -0.186% sync.s 16497 16408 -89 -0.539% internal/singleflight.s 2591 2577 -14 -0.540% math/rand.s 10456 10438 -18 -0.172% cmd/go/internal/par.s 2801 2790 -11 -0.393% internal/reflectlite.s 28477 28417 -60 -0.211% errors.s 2750 2736 -14 -0.509% internal/oserror.s 446 434 -12 -2.691% sort.s 17061 17046 -15 -0.088% io.s 17063 16999 -64 -0.375% vendor/golang.org/x/crypto/hkdf.s 1962 1936 -26 -1.325% text/tabwriter.s 9617 9574 -43 -0.447% hash/crc64.s 3414 3408 -6 -0.176% hash/crc32.s 6657 6651 -6 -0.090% bytes.s 31932 31863 -69 -0.216% strconv.s 53158 52799 -359 -0.675% strings.s 42829 42665 -164 -0.383% encoding/ascii85.s 4833 4791 -42 -0.869% vendor/golang.org/x/text/transform.s 16810 16724 -86 -0.512% path.s 6848 6845 -3 -0.044% encoding/base32.s 9658 9592 -66 -0.683% bufio.s 23051 22908 -143 -0.620% compress/bzip2.s 11773 11764 -9 -0.076% image.s 37565 37502 -63 -0.168% syscall.s 82359 82279 -80 -0.097% regexp/syntax.s 83573 82930 -643 -0.769% image/jpeg.s 36535 36490 -45 -0.123% regexp.s 64396 64214 -182 -0.283% time.s 82724 82622 -102 -0.123% plugin.s 6539 6536 -3 -0.046% context.s 10959 10865 -94 -0.858% internal/poll.s 24286 24270 -16 -0.066% reflect.s 168304 167927 -377 -0.224% internal/fmtsort.s 7416 7376 -40 -0.539% os.s 52465 51787 -678 -1.292% cmd/go/internal/lockedfile/internal/filelock.s 2326 2317 -9 -0.387% os/signal.s 4657 4648 -9 -0.193% runtime/debug.s 6040 5998 -42 -0.695% encoding/binary.s 30838 30801 -37 -0.120% vendor/golang.org/x/net/route.s 23694 23491 -203 -0.857% path/filepath.s 17895 17889 -6 -0.034% cmd/vendor/golang.org/x/sys/unix.s 78125 78109 -16 -0.020% io/ioutil.s 6999 6996 -3 -0.043% encoding/base64.s 12094 12007 -87 -0.719% crypto/cipher.s 20466 20372 -94 -0.459% cmd/go/internal/robustio.s 2672 2669 -3 -0.112% encoding/pem.s 9302 9286 -16 -0.172% internal/obscuretestdata.s 1719 1695 -24 -1.396% crypto/aes.s 11014 11002 -12 -0.109% os/exec.s 29388 29231 -157 -0.534% cmd/internal/browser.s 2266 2260 -6 -0.265% internal/goroot.s 4601 4592 -9 -0.196% vendor/golang.org/x/crypto/chacha20poly1305.s 8945 8942 -3 -0.034% cmd/vendor/golang.org/x/crypto/ssh/terminal.s 27226 27195 -31 -0.114% index/suffixarray.s 36431 36411 -20 -0.055% fmt.s 77017 76709 -308 -0.400% encoding/hex.s 6241 6154 -87 -1.394% compress/lzw.s 7133 7069 -64 -0.897% database/sql/driver.s 18888 18877 -11 -0.058% net/url.s 29838 29739 -99 -0.332% debug/plan9obj.s 8329 8279 -50 -0.600% encoding/csv.s 12986 12902 -84 -0.647% debug/gosym.s 25403 25330 -73 -0.287% compress/flate.s 51192 50970 -222 -0.434% vendor/golang.org/x/net/dns/dnsmessage.s 86769 86208 -561 -0.647% compress/gzip.s 9791 9758 -33 -0.337% compress/zlib.s 7310 7277 -33 -0.451% archive/zip.s 42356 42166 -190 -0.449% debug/dwarf.s 108259 107730 -529 -0.489% encoding/json.s 106378 105910 -468 -0.440% os/user.s 14751 14724 -27 -0.183% database/sql.s 99011 98404 -607 -0.613% log.s 9466 9423 -43 -0.454% debug/pe.s 31272 31182 -90 -0.288% debug/macho.s 32764 32608 -156 -0.476% encoding/gob.s 136976 136517 -459 -0.335% vendor/golang.org/x/text/unicode/bidi.s 27318 27276 -42 -0.154% archive/tar.s 71416 70975 -441 -0.618% vendor/golang.org/x/net/http2/hpack.s 23892 23848 -44 -0.184% vendor/golang.org/x/text/secure/bidirule.s 3354 3351 -3 -0.089% mime/quotedprintable.s 5960 5925 -35 -0.587% net/http/internal.s 5874 5853 -21 -0.358% math/big.s 184147 183692 -455 -0.247% debug/elf.s 63775 63567 -208 -0.326% mime.s 39802 39709 -93 -0.234% encoding/xml.s 111038 110713 -325 -0.293% crypto/dsa.s 6044 6029 -15 -0.248% go/token.s 12139 12077 -62 -0.511% crypto/rand.s 6889 6866 -23 -0.334% go/scanner.s 19030 19008 -22 -0.116% flag.s 22320 22236 -84 -0.376% vendor/golang.org/x/text/unicode/norm.s 66652 66391 -261 -0.392% crypto/rsa.s 31671 31650 -21 -0.066% crypto/elliptic.s 51553 51403 -150 -0.291% internal/xcoff.s 22950 22822 -128 -0.558% go/constant.s 43750 43689 -61 -0.139% encoding/asn1.s 57086 57035 -51 -0.089% runtime/trace.s 2609 2603 -6 -0.230% crypto/x509/pkix.s 10458 10471 +13 +0.124% image/gif.s 27544 27385 -159 -0.577% vendor/golang.org/x/net/idna.s 24558 24502 -56 -0.228% image/png.s 42775 42685 -90 -0.210% vendor/golang.org/x/crypto/cryptobyte.s 33616 33493 -123 -0.366% go/ast.s 80684 80449 -235 -0.291% net/internal/socktest.s 16571 16535 -36 -0.217% crypto/ecdsa.s 11948 11936 -12 -0.100% text/template/parse.s 95138 94002 -1136 -1.194% runtime/pprof.s 59702 59639 -63 -0.106% testing.s 68427 68088 -339 -0.495% internal/testenv.s 5620 5596 -24 -0.427% testing/internal/testdeps.s 3312 3294 -18 -0.543% internal/trace.s 78473 78239 -234 -0.298% testing/iotest.s 4968 4908 -60 -1.208% os/signal/internal/pty.s 3011 2990 -21 -0.697% testing/quick.s 12179 12125 -54 -0.443% cmd/internal/bio.s 9286 9274 -12 -0.129% cmd/internal/src.s 17684 17663 -21 -0.119% cmd/internal/goobj2.s 12588 12558 -30 -0.238% cmd/internal/objabi.s 16408 16390 -18 -0.110% go/printer.s 77417 77308 -109 -0.141% go/parser.s 80045 79113 -932 -1.164% go/format.s 5434 5419 -15 -0.276% cmd/internal/goobj.s 26146 25954 -192 -0.734% runtime/pprof/internal/profile.s 102518 102178 -340 -0.332% text/template.s 95343 94935 -408 -0.428% cmd/internal/dwarf.s 31718 31572 -146 -0.460% cmd/vendor/golang.org/x/arch/arm/armasm.s 45240 45151 -89 -0.197% internal/lazytemplate.s 1470 1457 -13 -0.884% cmd/vendor/golang.org/x/arch/ppc64/ppc64asm.s 37253 37220 -33 -0.089% cmd/asm/internal/flags.s 2593 2590 -3 -0.116% cmd/asm/internal/lex.s 25068 24921 -147 -0.586% cmd/internal/buildid.s 18536 18263 -273 -1.473% cmd/vendor/golang.org/x/arch/x86/x86asm.s 80209 80105 -104 -0.130% go/doc.s 75140 74585 -555 -0.739% cmd/internal/edit.s 3893 3899 +6 +0.154% html/template.s 89377 88809 -568 -0.636% cmd/vendor/golang.org/x/arch/arm64/arm64asm.s 117998 117824 -174 -0.147% cmd/internal/obj.s 115015 114290 -725 -0.630% go/build.s 69379 68862 -517 -0.745% cmd/internal/objfile.s 48106 47982 -124 -0.258% cmd/cover.s 46239 46113 -126 -0.272% cmd/addr2line.s 2845 2833 -12 -0.422% cmd/internal/obj/arm.s 68658 66855 -1803 -2.626% cmd/internal/obj/mips.s 57486 56272 -1214 -2.112% cmd/internal/obj/riscv.s 63834 63006 -828 -1.297% cmd/compile/internal/syntax.s 146582 145456 -1126 -0.768% cmd/internal/obj/wasm.s 44117 44066 -51 -0.116% cmd/cgo.s 242645 241653 -992 -0.409% cmd/internal/obj/arm64.s 152107 147163 -4944 -3.250% net.s 295972 292010 -3962 -1.339% go/types.s 321371 319432 -1939 -0.603% vendor/golang.org/x/net/http/httpproxy.s 9450 9423 -27 -0.286% net/textproto.s 19455 19406 -49 -0.252% cmd/internal/obj/ppc64.s 125544 120456 -5088 -4.053% go/internal/srcimporter.s 6475 6409 -66 -1.019% log/syslog.s 8017 7929 -88 -1.098% cmd/compile/internal/logopt.s 10183 10162 -21 -0.206% net/mail.s 24085 23948 -137 -0.569% mime/multipart.s 21527 21420 -107 -0.497% cmd/internal/obj/s390x.s 127610 127757 +147 +0.115% go/internal/gcimporter.s 34913 34548 -365 -1.045% vendor/golang.org/x/net/nettest.s 28103 28016 -87 -0.310% cmd/go/internal/cfg.s 9967 9916 -51 -0.512% cmd/api.s 39703 39603 -100 -0.252% go/internal/gccgoimporter.s 56470 56120 -350 -0.620% go/importer.s 2077 2056 -21 -1.011% cmd/compile/internal/types.s 48202 47282 -920 -1.909% cmd/go/internal/str.s 4341 4320 -21 -0.484% cmd/internal/obj/x86.s 89440 88625 -815 -0.911% cmd/go/internal/base.s 12667 12580 -87 -0.687% cmd/go/internal/cache.s 30754 30571 -183 -0.595% cmd/doc.s 62976 62755 -221 -0.351% cmd/go/internal/search.s 20114 19993 -121 -0.602% cmd/vendor/golang.org/x/xerrors.s 17923 17855 -68 -0.379% cmd/go/internal/lockedfile.s 16451 16415 -36 -0.219% cmd/vendor/golang.org/x/mod/sumdb/note.s 18200 18150 -50 -0.275% cmd/vendor/golang.org/x/mod/module.s 17869 17851 -18 -0.101% cmd/asm/internal/arch.s 37533 37482 -51 -0.136% cmd/fix.s 87728 87492 -236 -0.269% cmd/vendor/golang.org/x/mod/sumdb/tlog.s 36394 36367 -27 -0.074% cmd/vendor/golang.org/x/mod/sumdb/dirhash.s 4990 4963 -27 -0.541% cmd/go/internal/imports.s 16499 16469 -30 -0.182% cmd/vendor/golang.org/x/mod/zip.s 18816 18745 -71 -0.377% cmd/go/internal/cmdflag.s 5126 5123 -3 -0.059% cmd/internal/test2json.s 9540 9452 -88 -0.922% cmd/go/internal/tool.s 3629 3623 -6 -0.165% cmd/go/internal/version.s 11232 11220 -12 -0.107% cmd/go/internal/mvs.s 25383 25179 -204 -0.804% cmd/nm.s 5815 5803 -12 -0.206% cmd/dist.s 210146 209140 -1006 -0.479% cmd/asm/internal/asm.s 68655 68549 -106 -0.154% cmd/vendor/golang.org/x/mod/modfile.s 72974 72510 -464 -0.636% cmd/go/internal/load.s 107548 106861 -687 -0.639% cmd/link/internal/sym.s 18708 18581 -127 -0.679% cmd/asm.s 3367 3343 -24 -0.713% cmd/gofmt.s 30795 30698 -97 -0.315% cmd/link/internal/objfile.s 21828 21630 -198 -0.907% cmd/pack.s 14878 14869 -9 -0.060% cmd/vendor/github.com/google/pprof/internal/elfexec.s 6788 6782 -6 -0.088% cmd/test2json.s 1647 1641 -6 -0.364% cmd/link/internal/loader.s 48677 48483 -194 -0.399% cmd/vendor/golang.org/x/tools/go/analysis/internal/analysisflags.s 16783 16773 -10 -0.060% cmd/link/internal/loadelf.s 35464 35126 -338 -0.953% cmd/link/internal/loadmacho.s 29438 29180 -258 -0.876% cmd/link/internal/loadpe.s 16440 16371 -69 -0.420% cmd/vendor/golang.org/x/tools/go/analysis/passes/internal/analysisutil.s 2106 2100 -6 -0.285% cmd/link/internal/loadxcoff.s 11711 11615 -96 -0.820% cmd/vendor/golang.org/x/tools/go/analysis/internal/facts.s 14954 14883 -71 -0.475% cmd/vendor/golang.org/x/tools/go/ast/inspector.s 5394 5374 -20 -0.371% cmd/vendor/golang.org/x/tools/go/analysis/passes/asmdecl.s 37029 36822 -207 -0.559% cmd/vendor/golang.org/x/tools/go/analysis/passes/inspect.s 340 337 -3 -0.882% cmd/vendor/golang.org/x/tools/go/analysis/passes/cgocall.s 9919 9858 -61 -0.615% cmd/vendor/golang.org/x/tools/go/analysis/passes/bools.s 6705 6690 -15 -0.224% cmd/vendor/golang.org/x/tools/go/analysis/passes/copylock.s 9783 9741 -42 -0.429% cmd/vendor/golang.org/x/tools/go/cfg.s 31699 30742 -957 -3.019% cmd/vendor/golang.org/x/tools/go/analysis/passes/ifaceassert.s 2768 2762 -6 -0.217% cmd/vendor/golang.org/x/tools/go/analysis/passes/loopclosure.s 3031 2998 -33 -1.089% cmd/vendor/golang.org/x/tools/go/analysis/passes/shift.s 4382 4376 -6 -0.137% cmd/vendor/golang.org/x/tools/go/analysis/passes/stdmethods.s 8654 8642 -12 -0.139% cmd/vendor/golang.org/x/tools/go/analysis/passes/stringintconv.s 3458 3446 -12 -0.347% cmd/vendor/golang.org/x/tools/go/analysis/passes/structtag.s 8011 7995 -16 -0.200% cmd/vendor/golang.org/x/tools/go/analysis/passes/tests.s 6205 6193 -12 -0.193% cmd/vendor/golang.org/x/tools/go/ast/astutil.s 66183 65861 -322 -0.487% cmd/vendor/github.com/google/pprof/profile.s 150844 150261 -583 -0.386% cmd/vendor/golang.org/x/tools/go/analysis/passes/unreachable.s 8057 8054 -3 -0.037% cmd/vendor/golang.org/x/tools/go/analysis/passes/unusedresult.s 3670 3667 -3 -0.082% cmd/vendor/github.com/google/pprof/internal/measurement.s 10464 10440 -24 -0.229% cmd/vendor/golang.org/x/tools/go/types/typeutil.s 12319 12274 -45 -0.365% cmd/vendor/golang.org/x/tools/go/analysis/unitchecker.s 13503 13342 -161 -1.192% cmd/vendor/golang.org/x/tools/go/analysis/passes/ctrlflow.s 5261 5218 -43 -0.817% cmd/vendor/golang.org/x/tools/go/analysis/passes/errorsas.s 1462 1459 -3 -0.205% cmd/vendor/golang.org/x/tools/go/analysis/passes/lostcancel.s 9594 9582 -12 -0.125% cmd/vendor/golang.org/x/tools/go/analysis/passes/printf.s 34397 34338 -59 -0.172% cmd/vendor/github.com/google/pprof/internal/graph.s 53225 52936 -289 -0.543% cmd/vendor/github.com/ianlancetaylor/demangle.s 177450 175329 -2121 -1.195% crypto/x509.s 147892 147388 -504 -0.341% cmd/go/internal/work.s 306465 304950 -1515 -0.494% cmd/go/internal/run.s 4664 4657 -7 -0.150% crypto/tls.s 313130 311833 -1297 -0.414% net/http/httptrace.s 3979 3905 -74 -1.860% net/smtp.s 14413 14344 -69 -0.479% cmd/link/internal/ld.s 545343 542279 -3064 -0.562% cmd/link/internal/mips.s 6218 6215 -3 -0.048% cmd/link/internal/mips64.s 6108 6103 -5 -0.082% cmd/link/internal/amd64.s 18154 18112 -42 -0.231% cmd/link/internal/arm64.s 22527 22494 -33 -0.146% cmd/link/internal/arm.s 22574 22494 -80 -0.354% cmd/link/internal/s390x.s 20779 20746 -33 -0.159% cmd/link/internal/wasm.s 16531 16493 -38 -0.230% cmd/link/internal/x86.s 18906 18849 -57 -0.301% cmd/link/internal/ppc64.s 26856 26778 -78 -0.290% net/http.s 559101 556513 -2588 -0.463% net/http/cookiejar.s 15912 15885 -27 -0.170% expvar.s 9531 9525 -6 -0.063% net/http/httptest.s 16616 16475 -141 -0.849% net/http/cgi.s 23624 23458 -166 -0.703% cmd/go/internal/web.s 16546 16489 -57 -0.344% cmd/vendor/golang.org/x/mod/sumdb.s 33197 33117 -80 -0.241% net/http/fcgi.s 19266 19169 -97 -0.503% net/http/httputil.s 39875 39728 -147 -0.369% cmd/vendor/github.com/google/pprof/internal/symbolz.s 5888 5867 -21 -0.357% net/rpc.s 34154 34003 -151 -0.442% cmd/vendor/github.com/google/pprof/internal/transport.s 2746 2716 -30 -1.092% cmd/vendor/github.com/google/pprof/internal/binutils.s 35999 35875 -124 -0.344% net/rpc/jsonrpc.s 6637 6598 -39 -0.588% cmd/vendor/github.com/google/pprof/internal/symbolizer.s 11533 11458 -75 -0.650% cmd/go/internal/get.s 62921 62803 -118 -0.188% cmd/vendor/github.com/google/pprof/internal/report.s 80364 80058 -306 -0.381% cmd/go/internal/modfetch/codehost.s 89680 89066 -614 -0.685% cmd/trace.s 117171 116701 -470 -0.401% cmd/vendor/github.com/google/pprof/internal/driver.s 144268 143297 -971 -0.673% cmd/go/internal/modfetch.s 126299 125860 -439 -0.348% cmd/vendor/github.com/google/pprof/driver.s 9042 9000 -42 -0.464% cmd/go/internal/modconv.s 17947 17889 -58 -0.323% cmd/pprof.s 12399 12326 -73 -0.589% cmd/go/internal/modload.s 151182 150389 -793 -0.525% cmd/go/internal/generate.s 11738 11636 -102 -0.869% cmd/go/internal/help.s 6571 6531 -40 -0.609% cmd/go/internal/clean.s 11174 11142 -32 -0.286% cmd/go/internal/vet.s 7897 7867 -30 -0.380% cmd/go/internal/envcmd.s 22176 22095 -81 -0.365% cmd/go/internal/list.s 15216 15067 -149 -0.979% cmd/go/internal/modget.s 38698 38519 -179 -0.463% cmd/go/internal/modcmd.s 46674 46441 -233 -0.499% cmd/go/internal/test.s 64664 64456 -208 -0.322% cmd/go.s 6730 6703 -27 -0.401% cmd/compile/internal/ssa.s 3592565 3582500 -10065 -0.280% cmd/compile/internal/gc.s 1549123 1537123 -12000 -0.775% cmd/compile/internal/riscv64.s 14579 14483 -96 -0.658% cmd/compile/internal/mips.s 20578 20419 -159 -0.773% cmd/compile/internal/ppc64.s 25524 25359 -165 -0.646% cmd/compile/internal/mips64.s 19795 19636 -159 -0.803% cmd/compile/internal/wasm.s 13329 13290 -39 -0.293% cmd/compile/internal/s390x.s 28097 27892 -205 -0.730% cmd/compile/internal/arm.s 31489 31321 -168 -0.534% cmd/compile/internal/arm64.s 29803 29590 -213 -0.715% cmd/compile/internal/amd64.s 32961 33221 +260 +0.789% cmd/compile/internal/x86.s 31029 30878 -151 -0.487% total 18534966 18440341 -94625 -0.511% Change-Id: I830d37364f14f0297800adc42c99f60a74c51aca Reviewed-on: https://go-review.googlesource.com/c/go/+/226367 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org> |
||
|
Keith Randall
|
33b648c0e9 |
cmd/compile: fix ephemeral pointer problem on amd64
Make sure we don't use the rewrite ptr + (c + x) -> c + (ptr + x), as that may create an ephemeral out-of-bounds pointer. I have not seen an actual bug caused by this yet, but we've seen them in the 386 port so I'm fixing this issue for amd64 as well. The load-combining rules needed to be reworked somewhat to still work without the above broken rule. Update #37881 Change-Id: I8046d170e89e2035195f261535e34ca7d8aca68a Reviewed-on: https://go-review.googlesource.com/c/go/+/226437 Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Cherry Zhang <cherryyz@google.com> |
||
|
Keith Randall
|
af7eafd150 |
cmd/compile: convert 386 port to use addressing modes pass (take 2)
Retrying CL 222782, with a fix that will hopefully stop the random crashing. The issue with the previous CL is that it does pointer arithmetic in a way that may briefly generate an out-of-bounds pointer. If an interrupt happens to occur in that state, the referenced object may be collected incorrectly. Suppose there was code that did s[x+c]. The previous CL had a rule to the effect of ptr + (x + c) -> c + (ptr + x). But ptr+x is not guaranteed to point to the same object as ptr. In contrast, ptr+(x+c) is guaranteed to point to the same object as ptr, because we would have already checked that x+c is in bounds. For example, strconv.trim used to have this code: MOVZX -0x1(BX)(DX*1), BP CMPL $0x30, AL After CL 222782, it had this code: LEAL 0(BX)(DX*1), BP CMPB $0x30, -0x1(BP) An interrupt between those last two instructions could see BP pointing outside the backing store of the slice involved. It's really hard to actually demonstrate a bug. First, you need to have an interrupt occur at exactly the right time. Then, there must be no other pointers to the object in question. Since the interrupted frame will be scanned conservatively, there can't even be a dead pointer in another register or on the stack. (In the example above, a bug can't happen because BX still holds the original pointer.) Then, the object in question needs to be collected (or at least scanned?) before the interrupted code continues. This CL needs to handle load combining somewhat differently than CL 222782 because of the new restriction on arithmetic. That's the only real difference (other than removing the bad rules) from that old CL. This bug is also present in the amd64 rewrite rules, and we haven't seen any crashing as a result. I will fix up that code similarly to this one in a separate CL. Update #37881 Change-Id: I5f0d584d9bef4696bfe89a61ef0a27c8d507329f Reviewed-on: https://go-review.googlesource.com/c/go/+/225798 Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Cherry Zhang <cherryyz@google.com> |
||
Lynn Boger
|
e4a1cf8a56 |
cmd/compile: add rules to eliminate unnecessary signed shifts
This change to the rules removes some unnecessary signed shifts that appear in the math/rand functions. Existing rules did not cover some of the signed cases. A little improvement seen in math/rand due to removing 1 of 2 instructions generated for Int31n, which is inlined quite a bit. Intn1000 46.9ns ± 0% 45.5ns ± 0% -2.99% (p=1.000 n=1+1) Int63n1000 33.5ns ± 0% 32.8ns ± 0% -2.09% (p=1.000 n=1+1) Int31n1000 32.7ns ± 0% 32.6ns ± 0% -0.31% (p=1.000 n=1+1) Float32 32.7ns ± 0% 30.3ns ± 0% -7.34% (p=1.000 n=1+1) Float64 21.7ns ± 0% 20.9ns ± 0% -3.69% (p=1.000 n=1+1) Perm3 205ns ± 0% 202ns ± 0% -1.46% (p=1.000 n=1+1) Perm30 1.71µs ± 0% 1.68µs ± 0% -1.35% (p=1.000 n=1+1) Perm30ViaShuffle 1.65µs ± 0% 1.65µs ± 0% -0.30% (p=1.000 n=1+1) ShuffleOverhead 2.83µs ± 0% 2.83µs ± 0% -0.07% (p=1.000 n=1+1) Read3 18.7ns ± 0% 16.1ns ± 0% -13.90% (p=1.000 n=1+1) Read64 126ns ± 0% 124ns ± 0% -1.59% (p=1.000 n=1+1) Read1000 1.75µs ± 0% 1.63µs ± 0% -7.08% (p=1.000 n=1+1) Change-Id: I11502dfca7d65aafc76749a8d713e9e50c24a858 Reviewed-on: https://go-review.googlesource.com/c/go/+/225917 Run-TryBot: Lynn Boger <laboger@linux.vnet.ibm.com> Reviewed-by: Cherry Zhang <cherryyz@google.com> |
||
Ruixin(Peter) Bao
|
16cfab8d89 |
cmd/compile: use load and test instructions on s390x
The load and test instructions compare the given value against zero and will produce a condition code indicating one of the following scenarios: 0: Result is zero 1: Result is less than zero 2: Result is greater than zero 3: Result is not a number (NaN) The instruction can be used to simplify floating point comparisons against zero, which can enable further optimizations. This CL also reduces the size of .text section of math.test binary by around 0.7 KB (in hexadecimal, from 1358f0 to 135620). Change-Id: I33cb714f0c6feebac7a1c46dfcc735e7daceff9c Reviewed-on: https://go-review.googlesource.com/c/go/+/209159 Reviewed-by: Michael Munday <mike.munday@ibm.com> Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> |
||
|
Keith Randall
|
c785633941 |
Revert "cmd/compile: convert 386 port to use addressing modes pass"
This reverts commit CL 222782. Reason for revert: Reverting to see if 386 errors go away Update #37881 Change-Id: I74f287404c52414db1b6ff1649effa4ed9e5cc0c Reviewed-on: https://go-review.googlesource.com/c/go/+/225218 Reviewed-by: Bryan C. Mills <bcmills@google.com> |
||
|
Keith Randall
|
e0deacd1c0 |
Revert "cmd/compile: disable mem+op operations on 386"
This reverts commit CL 224837. Reason for revert: Reverting partial reverts of 222782. Update #37881 Change-Id: Ie9bf84d6e17ed214abe538965e5ff03936886826 Reviewed-on: https://go-review.googlesource.com/c/go/+/225217 Reviewed-by: Bryan C. Mills <bcmills@google.com> |
||
|
Keith Randall
|
f975485ad1 |
Revert "cmd/compile: disable addressingmodes pass for 386"
This reverts commit CL 225057. Reason for revert: Undoing partial reverts of CL 222782 Update #37881 Change-Id: Iee024cab2a580a37a0fc355e0e3c5ad3d8fdaf7d Reviewed-on: https://go-review.googlesource.com/c/go/+/225197 Reviewed-by: Bryan C. Mills <bcmills@google.com> |
||
|
Keith Randall
|
5b897ec017 |
cmd/compile: disable addressingmodes pass for 386
Update #37881 Change-Id: I1f9a3f57f6215a19c31765c257ee78715eab36b7 Reviewed-on: https://go-review.googlesource.com/c/go/+/225057 Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: Bryan C. Mills <bcmills@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> |
||
|
Keith Randall
|
3adbdb6d99 |
cmd/compile: disable mem+op operations on 386
Rolling back portions of CL 222782 to see if that helps issue #37881 any. Update #37881 Change-Id: I9cc3ff8c469fa5e4b22daec715d04148033f46f7 Reviewed-on: https://go-review.googlesource.com/c/go/+/224837 Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Bryan C. Mills <bcmills@google.com> |
||
Russ Cox
|
fc8a6336d1 |
cmd/asm, cmd/compile, runtime: add -spectre=ret mode
This commit extends the -spectre flag to cmd/asm and adds a new Spectre mitigation mode "ret", which enables the use of retpolines. Retpolines prevent speculation about the target of an indirect jump or call and are described in more detail here: https://support.google.com/faqs/answer/7625886 Change-Id: I4f2cb982fa94e44d91e49bd98974fd125619c93a Reviewed-on: https://go-review.googlesource.com/c/go/+/222661 Reviewed-by: Keith Randall <khr@golang.org> |
||
|
Russ Cox
|
877ef86bec |
cmd/compile: add spectre mitigation mode enabled by -spectre
This commit adds a new cmd/compile flag -spectre, which accepts a comma-separated list of possible Spectre mitigations to apply, or the empty string (none), or "all". The only known mitigation right now is "index", which uses conditional moves to ensure that x86-64 CPUs do not speculate past index bounds checks. Speculating past index bounds checks may be problematic on systems running privileged servers that accept requests from untrusted users who can execute their own programs on the same machine. (And some more constraints that make it even more unlikely in practice.) The cases this protects against are analogous to the ones Microsoft explains in the "Array out of bounds load/store feeding ..." sections here: https://docs.microsoft.com/en-us/cpp/security/developer-guidance-speculative-execution?view=vs-2019#array-out-of-bounds-load-feeding-an-indirect-branch Change-Id: Ib7532d7e12466b17e04c4e2075c2a456dc98f610 Reviewed-on: https://go-review.googlesource.com/c/go/+/222660 Reviewed-by: Keith Randall <khr@golang.org> |
||
|
Keith Randall
|
d84cbec890 |
cmd/compile: convert 386 port to use addressing modes pass
Update #36468 Change-Id: Idfdb845d097994689be450d6e8a57fa9adb57166 Reviewed-on: https://go-review.googlesource.com/c/go/+/222782 Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com> |
||
|
Russ Cox
|
801a9d9a0c |
test/codegen: mention in README that tests only run on Linux without -all_codegen
This took me a while to figure out. The relevant code is in
test/run.go (note the "linux" hard-coded strings):
var arch, subarch, os string
switch {
case archspec[2] != "": // 3 components: "linux/386/sse2"
os, arch, subarch = archspec[0], archspec[1][1:], archspec[2][1:]
case archspec[1] != "": // 2 components: "386/sse2"
os, arch, subarch = "linux", archspec[0], archspec[1][1:]
default: // 1 component: "386"
os, arch, subarch = "linux", archspec[0], ""
if arch == "wasm" {
os = "js"
}
}
Change-Id: I92ba280025d2072e17532a5e43cf1d676789c167
Reviewed-on: https://go-review.googlesource.com/c/go/+/222819
Reviewed-by: Keith Randall <khr@golang.org>
|
||
|
Keith Randall
|
98cb76799c |
cmd/compile: insert complicated x86 addressing modes as a separate pass
Use a separate compiler pass to introduce complicated x86 addressing modes. Loads in the normal architecture rules (for x86 and all other platforms) can have constant offsets (AuxInt values) and symbols (Aux values), but no more. The complex addressing modes (x+y, x+2*y, etc.) are introduced in a separate pass that combines loads with LEAQx ops. Organizing rewrites this way simplifies the number of rewrites required, as there are lots of different rule orderings that have to be specified to ensure these complex addressing modes are always found if they are possible. Update #36468 Change-Id: I5b4bf7b03a1e731d6dfeb9ef19b376175f3b4b44 Reviewed-on: https://go-review.googlesource.com/c/go/+/217097 Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com> |
||
Diogo Pinela
|
19ed0d993c |
cmd/compile: use staticuint64s instead of staticbytes
There are still two places in src/runtime/string.go that use staticbytes, so we cannot delete it just yet. There is a new codegen test to verify that the index calculation is constant-folded, at least on amd64. ppc64, mips[64] and s390x cannot currently do that. There is also a new runtime benchmark to ensure that this does not slow down performance (tested against parent commit): name old time/op new time/op delta ConvT2EByteSized/bool-4 1.07ns ± 1% 1.07ns ± 1% ~ (p=0.060 n=14+15) ConvT2EByteSized/uint8-4 1.06ns ± 1% 1.07ns ± 1% ~ (p=0.095 n=14+15) Updates #37612 Change-Id: I5ec30738edaa48cda78dfab4a78e24a32fa7fd6a Reviewed-on: https://go-review.googlesource.com/c/go/+/221957 Run-TryBot: Ian Lance Taylor <iant@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com> |
||
|
Keith Randall
|
cd9fd640db |
cmd/compile: don't allow NaNs in floating-point constant ops
Trying this CL again, with a fixed test that allows platforms to disagree on the exact behavior of converting NaNs. We store 32-bit floating point constants in a 64-bit field, by converting that 32-bit float to 64-bit float to store it, and convert it back to use it. That works for *almost* all floating-point constants. The exception is signaling NaNs. The round trip described above means we can't represent a 32-bit signaling NaN, because conversions strip the signaling bit. To fix this issue, just forbid NaNs as floating-point constants in SSA form. This shouldn't affect any real-world code, as people seldom constant-propagate NaNs (except in test code). Additionally, NaNs are somewhat underspecified (which of the many NaNs do you get when dividing 0/0?), so when cross-compiling there's a danger of using the compiler machine's NaN regime for some math, and the target machine's NaN regime for other math. Better to use the target machine's NaN regime always. Update #36400 Change-Id: Idf203b688a15abceabbd66ba290d4e9f63619ecb Reviewed-on: https://go-review.googlesource.com/c/go/+/221790 Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com> |
||
|
Josh Bleecher Snyder
|
b49d8ce2fa |
all: fix two minor typos in comments
Change-Id: Iec6cd81c9787d3419850aa97e75052956ad139bc Reviewed-on: https://go-review.googlesource.com/c/go/+/221789 Reviewed-by: Emmanuel Odeke <emm.odeke@gmail.com> |
||
Michael Munday
|
e37cc29863 |
cmd/compile: optimize integer-in-range checks
This CL incorporates code from CL 201206 by Josh Bleecher Snyder (thanks Josh). This CL restores the integer-in-range optimizations in the SSA backend. The fuse pass is enhanced to detect inequalities that could be merged and fuse their associated blocks while the generic rules optimize them into a single unsigned comparison. For example, the inequality `x >= 0 && x < 10` will now be optimized to `unsigned(x) < 10`. Overall has a fairly positive impact on binary sizes. name old time/op new time/op delta Template 192ms ± 1% 192ms ± 1% ~ (p=0.757 n=17+18) Unicode 76.6ms ± 2% 76.5ms ± 2% ~ (p=0.603 n=19+19) GoTypes 694ms ± 1% 693ms ± 1% ~ (p=0.569 n=19+20) Compiler 3.26s ± 0% 3.27s ± 0% +0.25% (p=0.000 n=20+20) SSA 7.41s ± 0% 7.49s ± 0% +1.10% (p=0.000 n=17+19) Flate 120ms ± 1% 120ms ± 1% +0.38% (p=0.003 n=19+19) GoParser 152ms ± 1% 152ms ± 1% ~ (p=0.061 n=17+19) Reflect 422ms ± 1% 425ms ± 2% +0.76% (p=0.001 n=18+20) Tar 167ms ± 1% 167ms ± 0% ~ (p=0.730 n=18+19) XML 233ms ± 4% 231ms ± 1% ~ (p=0.752 n=20+17) LinkCompiler 927ms ± 8% 928ms ± 8% ~ (p=0.857 n=19+20) ExternalLinkCompiler 1.81s ± 2% 1.81s ± 2% ~ (p=0.513 n=19+20) LinkWithoutDebugCompiler 556ms ±10% 583ms ±13% +4.95% (p=0.007 n=20+20) [Geo mean] 478ms 481ms +0.52% name old user-time/op new user-time/op delta Template 270ms ± 5% 269ms ± 7% ~ (p=0.925 n=20+20) Unicode 134ms ± 7% 131ms ±14% ~ (p=0.593 n=18+20) GoTypes 981ms ± 3% 987ms ± 2% +0.63% (p=0.049 n=19+18) Compiler 4.50s ± 2% 4.50s ± 1% ~ (p=0.588 n=19+20) SSA 10.6s ± 2% 10.6s ± 1% ~ (p=0.141 n=20+19) Flate 164ms ± 8% 165ms ±10% ~ (p=0.738 n=20+20) GoParser 202ms ± 5% 203ms ± 6% ~ (p=0.820 n=20+20) Reflect 587ms ± 6% 597ms ± 3% ~ (p=0.087 n=20+18) Tar 230ms ± 6% 228ms ± 8% ~ (p=0.569 n=19+20) XML 311ms ± 6% 314ms ± 5% ~ (p=0.369 n=20+20) LinkCompiler 878ms ± 8% 887ms ± 7% ~ (p=0.289 n=20+20) ExternalLinkCompiler 1.60s ± 7% 1.60s ± 7% ~ (p=0.820 n=20+20) LinkWithoutDebugCompiler 498ms ±12% 489ms ±11% ~ (p=0.398 n=20+20) [Geo mean] 611ms 611ms +0.05% name old alloc/op new alloc/op delta Template 36.1MB ± 0% 36.0MB ± 0% -0.32% (p=0.000 n=20+20) Unicode 28.3MB ± 0% 28.3MB ± 0% -0.03% (p=0.000 n=19+20) GoTypes 121MB ± 0% 121MB ± 0% ~ (p=0.226 n=16+20) Compiler 563MB ± 0% 563MB ± 0% ~ (p=0.166 n=20+19) SSA 1.32GB ± 0% 1.33GB ± 0% +0.88% (p=0.000 n=20+19) Flate 22.7MB ± 0% 22.7MB ± 0% -0.02% (p=0.033 n=19+20) GoParser 27.9MB ± 0% 27.9MB ± 0% -0.02% (p=0.001 n=20+20) Reflect 78.3MB ± 0% 78.2MB ± 0% -0.01% (p=0.019 n=20+20) Tar 34.0MB ± 0% 34.0MB ± 0% -0.04% (p=0.000 n=20+20) XML 43.9MB ± 0% 43.9MB ± 0% -0.07% (p=0.000 n=20+19) LinkCompiler 205MB ± 0% 205MB ± 0% +0.44% (p=0.000 n=20+18) ExternalLinkCompiler 223MB ± 0% 223MB ± 0% +0.03% (p=0.000 n=20+20) LinkWithoutDebugCompiler 139MB ± 0% 142MB ± 0% +1.75% (p=0.000 n=20+20) [Geo mean] 93.7MB 93.9MB +0.20% name old allocs/op new allocs/op delta Template 363k ± 0% 361k ± 0% -0.58% (p=0.000 n=20+19) Unicode 329k ± 0% 329k ± 0% -0.06% (p=0.000 n=19+20) GoTypes 1.28M ± 0% 1.28M ± 0% -0.01% (p=0.000 n=20+20) Compiler 5.40M ± 0% 5.40M ± 0% -0.01% (p=0.000 n=20+20) SSA 12.7M ± 0% 12.8M ± 0% +0.80% (p=0.000 n=20+20) Flate 228k ± 0% 228k ± 0% ~ (p=0.194 n=20+20) GoParser 295k ± 0% 295k ± 0% -0.04% (p=0.000 n=20+20) Reflect 949k ± 0% 949k ± 0% -0.01% (p=0.000 n=20+20) Tar 337k ± 0% 337k ± 0% -0.06% (p=0.000 n=20+20) XML 418k ± 0% 417k ± 0% -0.17% (p=0.000 n=20+20) LinkCompiler 553k ± 0% 554k ± 0% +0.22% (p=0.000 n=20+19) ExternalLinkCompiler 1.52M ± 0% 1.52M ± 0% +0.27% (p=0.000 n=20+20) LinkWithoutDebugCompiler 186k ± 0% 186k ± 0% +0.06% (p=0.000 n=20+20) [Geo mean] 723k 723k +0.03% name old text-bytes new text-bytes delta HelloSize 828kB ± 0% 828kB ± 0% -0.01% (p=0.000 n=20+20) name old data-bytes new data-bytes delta HelloSize 13.4kB ± 0% 13.4kB ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 180kB ± 0% 180kB ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 1.23MB ± 0% 1.23MB ± 0% -0.33% (p=0.000 n=20+20) file before after Δ % addr2line 4320075 4311883 -8192 -0.190% asm 5191932 5187836 -4096 -0.079% buildid 2835338 2831242 -4096 -0.144% compile 20531717 20569099 +37382 +0.182% cover 5322511 5318415 -4096 -0.077% dist 3723749 3719653 -4096 -0.110% doc 4743515 4739419 -4096 -0.086% fix 3413960 3409864 -4096 -0.120% link 6690119 6686023 -4096 -0.061% nm 4269616 4265520 -4096 -0.096% pprof 14942189 14929901 -12288 -0.082% trace 11807164 11790780 -16384 -0.139% vet 8384104 8388200 +4096 +0.049% go 15339076 15334980 -4096 -0.027% total 132258257 132226007 -32250 -0.024% Fixes #30645. Change-Id: If551ac5996097f3685870d083151b5843170aab0 Reviewed-on: https://go-review.googlesource.com/c/go/+/165998 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org> |
||
|
Michael Munday
|
44fe355694 |
cmd/compile: canonicalize comparison argument order
Ensure that any comparison between two values has the same argument
order. This helps ensure that they can be eliminated during the
lowered CSE pass which will be particularly important if we eliminate
the Greater and Geq ops (see #37316).
Example:
CMP R0, R1
BLT L1
CMP R1, R0 // different order, cannot eliminate
BEQ L2
CMP R0, R1
BLT L1
CMP R0, R1 // same order, can eliminate
BEQ L2
This does have some drawbacks. Notably comparisons might 'flip'
direction in the assembly output after even small changes to the
code or compiler. It should help make optimizations more reliable
however.
compilecmp master -> HEAD
master (
|
||
Bryan C. Mills
|
a9f1ea4a83 |
Revert "cmd/compile: don't allow NaNs in floating-point constant ops"
This reverts CL 213477. Reason for revert: tests are failing on linux-mips*-rtrk builders. Change-Id: I8168f7450890233f1bd7e53930b73693c26d4dc0 Reviewed-on: https://go-review.googlesource.com/c/go/+/220897 Run-TryBot: Bryan C. Mills <bcmills@google.com> Reviewed-by: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> |
||
|
Keith Randall
|
2aa7c6c548 |
cmd/compile: don't allow NaNs in floating-point constant ops
We store 32-bit floating point constants in a 64-bit field, by converting that 32-bit float to 64-bit float to store it, and convert it back to use it. That works for *almost* all floating-point constants. The exception is signaling NaNs. The round trip described above means we can't represent a 32-bit signaling NaN, because conversions strip the signaling bit. To fix this issue, just forbid NaNs as floating-point constants in SSA form. This shouldn't affect any real-world code, as people seldom constant-propagate NaNs (except in test code). Additionally, NaNs are somewhat underspecified (which of the many NaNs do you get when dividing 0/0?), so when cross-compiling there's a danger of using the compiler machine's NaN regime for some math, and the target machine's NaN regime for other math. Better to use the target machine's NaN regime always. This has been a bug since 1.10, and there's an easy workaround (declare a global varaible containing the signaling NaN pattern, and use that as the argument to math.Float32frombits) so we'll fix it in 1.15. Fixes #36400 Update #36399 Change-Id: Icf155e743281560eda2eed953d19a829552ccfda Reviewed-on: https://go-review.googlesource.com/c/go/+/213477 Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com> |
||
|
Keith Randall
|
1cfe8e91b6 |
cmd/compile: use ADDQ instead of LEAQ when we can
The address calculations in the example end up doing x << 4 + y + 0. Before this CL we use a SHLQ+LEAQ. Since the constant offset is 0, we can use SHLQ+ADDQ instead. Change-Id: Ia048c4fdbb3a42121c7e1ab707961062e8247fca Reviewed-on: https://go-review.googlesource.com/c/go/+/209959 Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Cherry Zhang <cherryyz@google.com> |
||
Brian Kessler
|
6b1d5471b9 |
cmd/compile: add signed indivisibility by power of 2 rules
Commit
|
||
|
Russ Cox
|
543c6d2e0d |
math, cmd/compile: rename Fma to FMA
This API was added for #25819, where it was discussed as math.FMA. The commit adding it used math.Fma, presumably for consistency with the rest of the unusual names in package math (Sincos, Acosh, Erfcinv, Float32bits, etc). I believe that using an idiomatic Go name is more important here than consistency with these other names, most of which are historical baggage from C's standard library. Early additions like Float32frombits happened before "uppercase for export" (so they were originally like "float32frombits") and they were not properly reconsidered when we uppercased the symbols to export them. That's a mistake we live with. The names of functions we have added since then, and even a few that were legacy, are more properly Go-cased, such as IsNaN, IsInf, and RoundToEven, rather than Isnan, Isinf, and Roundtoeven. And also constants like MaxFloat32. For new API, we should keep using proper Go-cased symbols instead of minimally-upper-cased-C symbols. So math.FMA, not math.Fma. This API has not yet been released, so this change does not break the compatibility promise. This CL also modifies cmd/compile, since the compiler knows the name of the function. I could have stopped at changing the string constants, but it seemed to make more sense to use a consistent casing everywhere. Change-Id: I0f6f3407f41e99bfa8239467345c33945088896e Reviewed-on: https://go-review.googlesource.com/c/go/+/205317 Run-TryBot: Russ Cox <rsc@golang.org> Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org> |
||
smasher164
|
58b031949b |
cmd/compile: add fma intrinsic for arm
This change introduces an arm intrinsic that generates the FMULAD instruction for the fused-multiply-add operation on systems that support it. System support is detected via cpu.ARM.HasVFPv4. A rewrite rule translates the generic intrinsic to FMULAD. Updates #25819. Change-Id: I8459e5dd1cdbdca35f88a78dbeb7d387f1e20efa Reviewed-on: https://go-review.googlesource.com/c/go/+/142117 Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org> |
||
|
smasher164
|
7a6da218b1 |
cmd/compile: add fma intrinsic for amd64
To permit ssa-level optimization, this change introduces an amd64 intrinsic
that generates the VFMADD231SD instruction for the fused-multiply-add
operation on systems that support it. System support is detected via
cpu.X86.HasFMA. A rewrite rule can then translate the generic ssa intrinsic
("Fma") to VFMADD231SD.
The benchmark compares the software implementation (old) with the intrinsic
(new).
name old time/op new time/op delta
Fma-4 27.2ns ± 1% 1.0ns ± 9% -96.48% (p=0.008 n=5+5)
Updates #25819.
Change-Id: I966655e5f96817a5d06dff5942418a3915b09584
Reviewed-on: https://go-review.googlesource.com/c/go/+/137156
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
|
||
|
smasher164
|
33425ab8db |
cmd/compile: introduce generic ssa intrinsic for fused-multiply-add
In order to make math.FMA a compiler intrinsic for ISAs like ARM64,
PPC64[le], and S390X, a generic 3-argument opcode "Fma" is provided and
rewritten as
ARM64: (Fma x y z) -> (FMADDD z x y)
PPC64: (Fma x y z) -> (FMADD x y z)
S390X: (Fma x y z) -> (FMADD z x y)
Updates #25819.
Change-Id: Ie5bc628311e6feeb28ddf9adaa6e702c8c291efa
Reviewed-on: https://go-review.googlesource.com/c/go/+/131959
Run-TryBot: Akhil Indurti <aindurti@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
|
||
David Chase
|
6adaf17eaa |
cmd/compile: preserve statements in late nilcheckelim optimization
When a subsequent load/store of a ptr makes the nil check of that pointer unnecessary, if their lines differ, change the line of the load/store to that of the nilcheck, and attempt to rehome the load/store position instead. This fix makes profiling less accurate in order to make panics more informative. Fixes #33724 Change-Id: Ib9afaac12fe0d0320aea1bf493617facc34034b3 Reviewed-on: https://go-review.googlesource.com/c/go/+/200197 Run-TryBot: David Chase <drchase@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org> |
||
Meng Zhuo
|
50f1157760 |
cmd/compile: add math/bits.Mul64 intrinsic on mips64x
Benchmark: name old time/op new time/op delta Mul 36.0ns ± 1% 2.8ns ± 0% -92.31% (p=0.000 n=10+10) Mul32 4.37ns ± 0% 4.37ns ± 0% ~ (p=0.429 n=6+10) Mul64 36.4ns ± 0% 2.8ns ± 0% -92.37% (p=0.000 n=10+9) Change-Id: Ic4f4e5958adbf24999abcee721d0180b5413fca7 Reviewed-on: https://go-review.googlesource.com/c/go/+/200582 Reviewed-by: Keith Randall <khr@golang.org> Reviewed-by: Cherry Zhang <cherryyz@google.com> Run-TryBot: Cherry Zhang <cherryyz@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> |
||
|
Michael Munday
|
6ec4c71eef |
cmd/compile: add SSA rules for s390x compare-and-branch instructions
This commit adds SSA rules for the s390x combined compare-and-branch instructions. These have a shorter encoding than separate compare and branch instructions and they also don't clobber the condition code (a.k.a. flag register) reducing pressure on the flag allocator. I have deleted the 'loop_test.go' file and replaced it with a new codegen test which performs a wider range of checks. Object sizes from compilebench: name old object-bytes new object-bytes delta Template 562kB ± 0% 561kB ± 0% -0.28% (p=0.000 n=10+10) Unicode 217kB ± 0% 217kB ± 0% -0.17% (p=0.000 n=10+10) GoTypes 2.03MB ± 0% 2.02MB ± 0% -0.59% (p=0.000 n=10+10) Compiler 8.16MB ± 0% 8.11MB ± 0% -0.62% (p=0.000 n=10+10) SSA 27.4MB ± 0% 27.0MB ± 0% -1.45% (p=0.000 n=10+10) Flate 356kB ± 0% 356kB ± 0% -0.12% (p=0.000 n=10+10) GoParser 438kB ± 0% 436kB ± 0% -0.51% (p=0.000 n=10+10) Reflect 1.37MB ± 0% 1.37MB ± 0% -0.42% (p=0.000 n=10+10) Tar 485kB ± 0% 483kB ± 0% -0.39% (p=0.000 n=10+10) XML 630kB ± 0% 621kB ± 0% -1.45% (p=0.000 n=10+10) [Geo mean] 1.14MB 1.13MB -0.60% name old text-bytes new text-bytes delta HelloSize 763kB ± 0% 754kB ± 0% -1.30% (p=0.000 n=10+10) CmdGoSize 10.7MB ± 0% 10.6MB ± 0% -0.91% (p=0.000 n=10+10) [Geo mean] 2.86MB 2.82MB -1.10% Change-Id: Ibca55d9c0aa1254aee69433731ab5d26a43a7c18 Reviewed-on: https://go-review.googlesource.com/c/go/+/198037 Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org> |
||
Cuong Manh Le
|
77f5adba55 |
cmd/compile: don't use statictmps for small object in slice literal
Fixes #21561 Change-Id: I89c59752060dd9570d17d73acbbaceaefce5d8ce Reviewed-on: https://go-review.googlesource.com/c/go/+/197560 Run-TryBot: Cuong Manh Le <cuong.manhle.vn@gmail.com> Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> |
||
Keith Randall
|
72dc9ab191 |
cmd/compile: reuse dead register before reusing register holding constant
For commuting ops, check whether the second argument is dead before checking if the first argument is rematerializeable. Reusing the register holding a dead value is always best. Fixes #33580 Change-Id: I7372cfc03d514e6774d2d9cc727a3e6bf6ce2657 Reviewed-on: https://go-review.googlesource.com/c/go/+/199559 Run-TryBot: Keith Randall <khr@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: David Chase <drchase@google.com> |
||
Dan Scales
|
225f484c88 |
misc, runtime, test: extra tests and benchmarks for defer
Add a bunch of extra tests and benchmarks for defer, in preparation for new low-cost (open-coded) implementation of defers (see #34481), - New file defer_test.go that tests a bunch more unusual defer scenarios, including things that might have problems for open-coded defers. - Additions to callers_test.go actually verifying what the stack trace looks like for various panic or panic-recover scenarios. - Additions to crash_test.go testing several more crash scenarios involving recursive panics. - New benchmark in runtime_test.go measuring speed of panic-recover - New CGo benchmark in cgo_test.go calling from Go to C back to Go that shows defer overhead Updates #34481 Change-Id: I423523f3e05fc0229d4277dd00073289a5526188 Reviewed-on: https://go-review.googlesource.com/c/go/+/197017 Run-TryBot: Dan Scales <danscales@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org> Reviewed-by: Austin Clements <austin@google.com> |
||
Martin Möhrmann
|
f41451e7eb |
compile: prefer an AND instead of SHR+SHL instructions
On modern 64bit CPUs a SHR, SHL or AND instruction take 1 cycle to execute.
A pair of shifts that operate on the same register will take 2 cycles
and needs to wait for the input register value to be available.
Large constants used to mask the high bits of a register with an AND
instruction can not be encoded as an immediate in the AND instruction
on amd64 and therefore need to be loaded into a register with a MOV
instruction.
However that MOV instruction is not dependent on the output register and
on many CPUs does not compete with the AND or shift instructions for
execution ports.
Using a pair of shifts to mask high bits instead of an AND to mask high
bits of a register has a shorter encoding and uses one less general
purpose register but is slower due to taking one clock cycle longer
if there is no register pressure that would make the AND variant need to
generate a spill.
For example the instructions emitted for (x & 1 << 63) before this CL are:
48c1ea3f SHRQ $0x3f, DX
48c1e23f SHLQ $0x3f, DX
after this CL the instructions are the same as GCC and LLVM use:
48b80000000000000080 MOVQ $0x8000000000000000, AX
4821d0 ANDQ DX, AX
Some platforms such as arm64 already have SSA optimization rules to fuse
two shift instructions back into an AND.
Removing the general rule to rewrite AND to SHR+SHL speeds up this benchmark:
var GlobalU uint
func BenchmarkAndHighBits(b *testing.B) {
x := uint(0)
for i := 0; i < b.N; i++ {
x &= 1 << 63
}
GlobalU = x
}
amd64/darwin on Intel(R) Core(TM) i7-3520M CPU @ 2.90GHz:
name old time/op new time/op delta
AndHighBits-4 0.61ns ± 6% 0.42ns ± 6% -31.42% (p=0.000 n=25+25):
'go run run.go -all_codegen -v codegen' passes with following adjustments:
ARM64: The BFXIL pattern ((x << lc) >> rc | y & ac) needed adjustment
since ORshiftRL generation fusing '>> rc' and '|' interferes
with matching ((x << lc) >> rc) to generate UBFX. Previously
ORshiftLL was created first using the shifts generated for (y & ac).
S390X: Add rules for abs and copysign to match use of AND instead of SHIFTs.
Updates #33826
Updates #32781
Change-Id: I5a59f6239660d53c029cd22dfb44ddf39f93a56c
Reviewed-on: https://go-review.googlesource.com/c/go/+/196810
Run-TryBot: Martin Möhrmann <moehrmann@google.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
|
||
|
Bryan C. Mills
|
34fe8295c5 |
Revert "compile: prefer an AND instead of SHR+SHL instructions"
This reverts CL 194297. Reason for revert: introduced register allocation failures on PPC64LE builders. Updates #33826 Updates #32781 Updates #34468 Change-Id: I7d0b55df8cdf8e7d2277f1814299b083c2692e48 Reviewed-on: https://go-review.googlesource.com/c/go/+/196957 Run-TryBot: Bryan C. Mills <bcmills@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Dmitri Shuralyov <dmitshur@golang.org> Reviewed-by: Cherry Zhang <cherryyz@google.com> Reviewed-by: Martin Möhrmann <moehrmann@google.com> |
||
|
Martin Möhrmann
|
4e2b84ffc5 |
compile: prefer an AND instead of SHR+SHL instructions
On modern 64bit CPUs a SHR, SHL or AND instruction take 1 cycle to execute.
A pair of shifts that operate on the same register will take 2 cycles
and needs to wait for the input register value to be available.
Large constants used to mask the high bits of a register with an AND
instruction can not be encoded as an immediate in the AND instruction
on amd64 and therefore need to be loaded into a register with a MOV
instruction.
However that MOV instruction is not dependent on the output register and
on many CPUs does not compete with the AND or shift instructions for
execution ports.
Using a pair of shifts to mask high bits instead of an AND to mask high
bits of a register has a shorter encoding and uses one less general
purpose register but is slower due to taking one clock cycle longer
if there is no register pressure that would make the AND variant need to
generate a spill.
For example the instructions emitted for (x & 1 << 63) before this CL are:
48c1ea3f SHRQ $0x3f, DX
48c1e23f SHLQ $0x3f, DX
after this CL the instructions are the same as GCC and LLVM use:
48b80000000000000080 MOVQ $0x8000000000000000, AX
4821d0 ANDQ DX, AX
Some platforms such as arm64 already have SSA optimization rules to fuse
two shift instructions back into an AND.
Removing the general rule to rewrite AND to SHR+SHL speeds up this benchmark:
var GlobalU uint
func BenchmarkAndHighBits(b *testing.B) {
x := uint(0)
for i := 0; i < b.N; i++ {
x &= 1 << 63
}
GlobalU = x
}
amd64/darwin on Intel(R) Core(TM) i7-3520M CPU @ 2.90GHz:
name old time/op new time/op delta
AndHighBits-4 0.61ns ± 6% 0.42ns ± 6% -31.42% (p=0.000 n=25+25):
'go run run.go -all_codegen -v codegen' passes with following adjustments:
ARM64: The BFXIL pattern ((x << lc) >> rc | y & ac) needed adjustment
since ORshiftRL generation fusing '>> rc' and '|' interferes
with matching ((x << lc) >> rc) to generate UBFX. Previously
ORshiftLL was created first using the shifts generated for (y & ac).
S390X: Add rules for abs and copysign to match use of AND instead of SHIFTs.
Updates #33826
Updates #32781
Change-Id: I43227da76b625de03fbc51117162b23b9c678cdb
Reviewed-on: https://go-review.googlesource.com/c/go/+/194297
Run-TryBot: Martin Möhrmann <martisch@uos.de>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
|
||
Agniva De Sarker
|
ecc7dd5469 |
test/codegen: fix wasm codegen breakage
i32.eqz instructions don't appear unless needed in if conditions anymore after CL 195204. I forgot to run the codegen tests while submitting the CL. Thanks to @martisch for catching it. Fixes #34442 Change-Id: I177b064b389be48e39d564849714d7a8839be13e Reviewed-on: https://go-review.googlesource.com/c/go/+/196580 Run-TryBot: Agniva De Sarker <agniva.quicksilver@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Martin Möhrmann <moehrmann@google.com> |
||
Matthew Dempsky
|
85fc765341 |
cmd/compile: optimize switch on strings
When compiling expression switches, we try to optimize runs of
constants into binary searches. The ordering used isn't visible to the
application, so it's unimportant as long as we're consistent between
sorting and searching.
For strings, it's much cheaper to compare string lengths than strings
themselves, so instead of ordering strings by "si <= sj", we currently
order them by "len(si) < len(sj) || len(si) == len(sj) && si <= sj"
(i.e., the lexicographical ordering on the 2-tuple (len(s), s)).
However, it's also somewhat cheaper to compare strings for equality
(i.e., ==) than for ordering (i.e., <=). And if there were two or
three string constants of the same length in a switch statement, we
might unnecessarily emit ordering comparisons.
For example, given:
switch s {
case "", "1", "2", "3": // ordered by length then content
goto L
}
we currently compile this as:
if len(s) < 1 || len(s) == 1 && s <= "1" {
if s == "" { goto L }
else if s == "1" { goto L }
} else {
if s == "2" { goto L }
else if s == "3" { goto L }
}
This CL switches to using a 2-level binary search---first on len(s),
then on s itself---so that string ordering comparisons are only needed
when there are 4 or more strings of the same length. (4 being the
cut-off for when using binary search is actually worthwhile.)
So the above switch instead now compiles to:
if len(s) == 0 {
if s == "" { goto L }
} else if len(s) == 1 {
if s == "1" { goto L }
else if s == "2" { goto L }
else if s == "3" { goto L }
}
which is better optimized by walk and SSA. (Notably, because there are
only two distinct lengths and no more than three strings of any
particular length, this example ends up falling back to simply using
linear search.)
Test case by khr@ from CL 195138.
Fixes #33934.
Change-Id: I8eeebcaf7e26343223be5f443d6a97a0daf84f07
Reviewed-on: https://go-review.googlesource.com/c/go/+/195340
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
|
||
|
LE Manh Cuong
|
ec4e8517cd |
cmd/compile: support more length types for slice extension optimization
golang.org/cl/109517 optimized the compiler to avoid the allocation for make in append(x, make([]T, y)...). This was only implemented for the case that y has type int. This change extends the optimization to trigger for all integer types where the value is known at compile time to fit into an int. name old time/op new time/op delta ExtendInt-12 106ns ± 4% 106ns ± 0% ~ (p=0.351 n=10+6) ExtendUint64-12 1.03µs ± 5% 0.10µs ± 4% -90.01% (p=0.000 n=9+10) name old alloc/op new alloc/op delta ExtendInt-12 0.00B 0.00B ~ (all equal) ExtendUint64-12 13.6kB ± 0% 0.0kB -100.00% (p=0.000 n=10+10) name old allocs/op new allocs/op delta ExtendInt-12 0.00 0.00 ~ (all equal) ExtendUint64-12 1.00 ± 0% 0.00 -100.00% (p=0.000 n=10+10) Updates #29785 Change-Id: Ief7760097c285abd591712da98c5b02bc3961fcd Reviewed-on: https://go-review.googlesource.com/c/go/+/182559 Run-TryBot: Cuong Manh Le <cuong.manhle.vn@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org> |
||
Alberto Donizetti
|
c2facbe937 |
Revert "test/codegen: document -all_codegen option in README"
This reverts CL 192101. Reason for revert: The same paragraph was added 2 weeks ago (look a few lines above) Change-Id: I05efb2631d7b4966f66493f178f2a649c715a3cc Reviewed-on: https://go-review.googlesource.com/c/go/+/195637 Reviewed-by: Cherry Zhang <cherryyz@google.com> |
||
Cherry Zhang
|
d9b8ffa51c |
test/codegen: document -all_codegen option in README
It is useful to know about the -all_codegen option for running codegen tests for all platforms. I was puzzling that some codegen test was not failing on my local machine or on trybot, until I found this option. Change-Id: I062cf4d73f6a6c9ebc2258195779d2dab21bc36d Reviewed-on: https://go-review.googlesource.com/c/go/+/192101 Reviewed-by: Daniel Martí <mvdan@mvdan.cc> Run-TryBot: Daniel Martí <mvdan@mvdan.cc> TryBot-Result: Gobot Gobot <gobot@golang.org> |
||
|
Ruixin Bao
|
98aa97806b |
cmd/compile: add math/bits.Mul64 intrinsic on s390x
This change adds an intrinsic for Mul64 on s390x. To achieve that, a new assembly instruction, MLGR, is introduced in s390x/asmz.go. This assembly instruction directly uses an existing instruction on Z and supports multiplication of two 64 bit unsigned integer and stores the result in two separate registers. In this case, we require the multiplcand to be stored in register R3 and the output result (the high and low 64 bit of the product) to be stored in R2 and R3 respectively. A test case is also added. Benchmark: name old time/op new time/op delta Mul-18 11.1ns ± 0% 1.4ns ± 0% -87.39% (p=0.002 n=8+10) Mul32-18 2.07ns ± 0% 2.07ns ± 0% ~ (all equal) Mul64-18 11.1ns ± 1% 1.4ns ± 0% -87.42% (p=0.000 n=10+10) Change-Id: Ieca6ad1f61fff9a48a31d50bbd3f3c6d9e6675c1 Reviewed-on: https://go-review.googlesource.com/c/go/+/194572 Reviewed-by: Michael Munday <mike.munday@ibm.com> Run-TryBot: Michael Munday <mike.munday@ibm.com> TryBot-Result: Gobot Gobot <gobot@golang.org> |
||
|
Michael Munday
|
5c5f217b63 |
cmd/compile: improve s390x sign/zero extension removal
This CL gets rid of the MOVDreg and MOVDnop SSA operations on
s390x. They were originally inserted to help avoid situations
where a sign/zero extension was elided but a spill invalidated
the optimization. It's not really clear we need to do this though
(amd64 doesn't have these ops for example) so long as we are
careful when removing sign/zero extensions. Also, the MOVDreg
technique doesn't work if the register is spilled before the
MOVDreg op (I haven't seen that in practice).
Removing these ops reduces the complexity of the rules and also
allows us to unblock optimizations. For example, the compiler can
now merge the loads in binary.{Big,Little}Endian.PutUint16 which
it wasn't able to do before. This CL reduces the size of the .text
section in the go tool by about 4.7KB (0.09%).
Change-Id: Icaddae7f2e4f9b2debb6fabae845adb3f73b41db
Reviewed-on: https://go-review.googlesource.com/c/go/+/173897
Run-TryBot: Michael Munday <mike.munday@ibm.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
|
||
|
Martin Möhrmann
|
5bb59b6d16 |
Revert "compile: prefer an AND instead of SHR+SHL instructions"
This reverts commit
|
||
|
Martin Möhrmann
|
9ec7074a94 |
compile: prefer an AND instead of SHR+SHL instructions
On modern 64bit CPUs a SHR, SHL or AND instruction take 1 cycle to execute.
A pair of shifts that operate on the same register will take 2 cycles
and needs to wait for the input register value to be available.
Large constants used to mask the high bits of a register with an AND
instruction can not be encoded as an immediate in the AND instruction
on amd64 and therefore need to be loaded into a register with a MOV
instruction.
However that MOV instruction is not dependent on the output register and
on many CPUs does not compete with the AND or shift instructions for
execution ports.
Using a pair of shifts to mask high bits instead of an AND to mask high
bits of a register has a shorter encoding and uses one less general
purpose register but is slower due to taking one clock cycle longer
if there is no register pressure that would make the AND variant need to
generate a spill.
For example the instructions emitted for (x & 1 << 63) before this CL are:
48c1ea3f SHRQ $0x3f, DX
48c1e23f SHLQ $0x3f, DX
after this CL the instructions are the same as GCC and LLVM use:
48b80000000000000080 MOVQ $0x8000000000000000, AX
4821d0 ANDQ DX, AX
Some platforms such as arm64 already have SSA optimization rules to fuse
two shift instructions back into an AND.
Removing the general rule to rewrite AND to SHR+SHL speeds up this benchmark:
var GlobalU uint
func BenchmarkAndHighBits(b *testing.B) {
x := uint(0)
for i := 0; i < b.N; i++ {
x &= 1 << 63
}
GlobalU = x
}
amd64/darwin on Intel(R) Core(TM) i7-3520M CPU @ 2.90GHz:
name old time/op new time/op delta
AndHighBits-4 0.61ns ± 6% 0.42ns ± 6% -31.42% (p=0.000 n=25+25):
Updates #33826
Updates #32781
Change-Id: I862d3587446410c447b9a7265196b57f85358633
Reviewed-on: https://go-review.googlesource.com/c/go/+/191780
Run-TryBot: Martin Möhrmann <moehrmann@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
|
||
|
Alberto Donizetti
|
e6d2544d20 |
test/codegen: mention -all_codegen in the README
For performance reasons (avoiding costly cross-compilations) CL 177577 changed the codegen test harness to only run the tests for the machine's GOARCH by default. This change updates the codegen README accordingly, explaining what all.bash does run by default and how to perform the tests for all architectures. Fixes #33924 Change-Id: I43328d878c3e449ebfda46f7e69963a44a511d40 Reviewed-on: https://go-review.googlesource.com/c/go/+/192619 Reviewed-by: Daniel Martí <mvdan@mvdan.cc> |
||
|
Brian Kessler
|
b003afe4fe |
cmd/compile: intrinsify RotateLeft32 on wasm
wasm has 32-bit versions of all integer operations. This change lowers RotateLeft32 to i32.rotl on wasm and intrinsifies the math/bits call. Benchmarking on amd64 under node.js this is ~25% faster. node v10.15.3/amd64 name old time/op new time/op delta RotateLeft 8.37ns ± 1% 8.28ns ± 0% -1.05% (p=0.029 n=4+4) RotateLeft8 11.9ns ± 1% 11.8ns ± 0% ~ (p=0.167 n=5+5) RotateLeft16 11.8ns ± 0% 11.8ns ± 0% ~ (all equal) RotateLeft32 11.9ns ± 1% 8.7ns ± 0% -26.32% (p=0.008 n=5+5) RotateLeft64 8.31ns ± 1% 8.43ns ± 2% ~ (p=0.063 n=5+5) Updates #31265 Change-Id: I5b8e155978faeea536c4f6427ac9564d2f096a46 Reviewed-on: https://go-review.googlesource.com/c/go/+/182359 Run-TryBot: Brian Kessler <brian.m.kessler@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Richard Musiol <neelance@gmail.com> |
||
Ben Shi
|
1786ecd502 |
cmd/compile: eliminate WASM's redundant extension & wrapping
This CL eliminates unnecessary pairs of I32WrapI64 and I64ExtendI32U generated by the WASM backend for IF statements. And it makes the total size of pkg/js_wasm/ decreases about 490KB. Change-Id: I16b0abb686c4e30d5624323166ec2d0ec57dbe2d Reviewed-on: https://go-review.googlesource.com/c/go/+/191758 Run-TryBot: Ben Shi <powerman1st@163.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Richard Musiol <neelance@gmail.com> |