typenative 0.6.2 → 0.7.0
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/CHANGELOG.md +388 -0
- package/README.md +1 -1
- package/bin/cli.js +18 -8
- package/bin/go/overrides.js +21 -1
- package/bin/index.js +48 -25
- package/bin/main.js +2 -1
- package/bin/node/assert.js +54 -15
- package/bin/node/async_hooks.js +69 -10
- package/bin/node/buffer.js +321 -7
- package/bin/node/child_process.js +154 -8
- package/bin/node/cluster.js +436 -32
- package/bin/node/crypto.js +814 -5
- package/bin/node/crypto_curves.js +142 -0
- package/bin/node/crypto_keys.js +1147 -48
- package/bin/node/crypto_pem.js +313 -0
- package/bin/node/crypto_pkcs12.js +17 -0
- package/bin/node/dgram.js +161 -10
- package/bin/node/diagnostics_channel.js +5 -1
- package/bin/node/events.js +358 -8
- package/bin/node/ffi.js +933 -0
- package/bin/node/fs.js +334 -6
- package/bin/node/http.js +630 -55
- package/bin/node/http2.js +1141 -88
- package/bin/node/intl.js +4576 -0
- package/bin/node/intl_data.js +65412 -0
- package/bin/node/module_v8.js +155 -16
- package/bin/node/net.js +59 -10
- package/bin/node/path.js +5 -1
- package/bin/node/process.js +342 -3
- package/bin/node/readline.js +558 -8
- package/bin/node/registry.js +2 -0
- package/bin/node/sqlite.js +51 -2
- package/bin/node/stream.js +370 -32
- package/bin/node/stream_web.js +422 -15
- package/bin/node/system_extra.js +154 -1
- package/bin/node/temporal.js +96 -0
- package/bin/node/temporal_runtime.js +5075 -0
- package/bin/node/temporal_zones.js +460 -0
- package/bin/node/test.js +153 -8
- package/bin/node/test_coverage.js +58 -0
- package/bin/node/test_report.js +3 -3
- package/bin/node/test_run.js +7 -1
- package/bin/node/timers.js +123 -0
- package/bin/node/tls.js +346 -12
- package/bin/node/tls_roots.js +122 -0
- package/bin/node/typed_arrays.js +867 -93
- package/bin/node/url.js +657 -0
- package/bin/node/util.js +441 -2
- package/bin/node/util_extra.js +6 -0
- package/bin/node/wasi.js +214 -0
- package/bin/node/webassembly.js +1122 -0
- package/bin/node/websocket.js +763 -0
- package/bin/node/worker_threads.js +434 -52
- package/bin/node/zlib.js +338 -9
- package/bin/node/zlib_zip.js +2392 -0
- package/bin/transpiler.js +2057 -253
- package/bin/typecheck-node.js +49 -17
- package/bin/version.js +1 -1
- package/conformance.md +160 -113
- package/examples/README.md +21 -20
- package/examples/angular-templates/README.md +12 -12
- package/examples/desktop-notes/README.md +7 -7
- package/examples/dupe-finder/README.md +4 -4
- package/examples/fibonacci/README.md +6 -6
- package/examples/hello-world/README.md +4 -4
- package/examples/json-server/README.md +13 -12
- package/examples/mandelbrot-png/README.md +4 -4
- package/examples/marked-cli/README.md +16 -17
- package/examples/openai-chat/README.md +5 -5
- package/examples/react-compiler/README.md +10 -10
- package/examples/todo-cli/README.md +8 -8
- package/examples/yaml-cli/README.md +10 -10
- package/govendor/github.com/cloudflare/circl/.golangci.yaml +64 -0
- package/govendor/github.com/cloudflare/circl/.semgrepignore +1 -0
- package/govendor/github.com/cloudflare/circl/AGENTS.md +91 -0
- package/govendor/github.com/cloudflare/circl/CITATION.cff +30 -0
- package/govendor/github.com/cloudflare/circl/LICENSE +57 -0
- package/govendor/github.com/cloudflare/circl/Makefile +70 -0
- package/govendor/github.com/cloudflare/circl/README.md +206 -0
- package/govendor/github.com/cloudflare/circl/REVIEW.md +83 -0
- package/govendor/github.com/cloudflare/circl/codecov.yml +21 -0
- package/govendor/github.com/cloudflare/circl/dh/x448/curve.go +104 -0
- package/govendor/github.com/cloudflare/circl/dh/x448/curve_amd64.go +30 -0
- package/govendor/github.com/cloudflare/circl/dh/x448/curve_amd64.h +111 -0
- package/govendor/github.com/cloudflare/circl/dh/x448/curve_amd64.s +194 -0
- package/govendor/github.com/cloudflare/circl/dh/x448/curve_generic.go +100 -0
- package/govendor/github.com/cloudflare/circl/dh/x448/curve_noasm.go +11 -0
- package/govendor/github.com/cloudflare/circl/dh/x448/doc.go +19 -0
- package/govendor/github.com/cloudflare/circl/dh/x448/key.go +46 -0
- package/govendor/github.com/cloudflare/circl/dh/x448/table.go +460 -0
- package/govendor/github.com/cloudflare/circl/doc.go +8 -0
- package/govendor/github.com/cloudflare/circl/ecc/goldilocks/constants.go +71 -0
- package/govendor/github.com/cloudflare/circl/ecc/goldilocks/curve.go +90 -0
- package/govendor/github.com/cloudflare/circl/ecc/goldilocks/isogeny.go +52 -0
- package/govendor/github.com/cloudflare/circl/ecc/goldilocks/point.go +181 -0
- package/govendor/github.com/cloudflare/circl/ecc/goldilocks/scalar.go +203 -0
- package/govendor/github.com/cloudflare/circl/ecc/goldilocks/twist.go +138 -0
- package/govendor/github.com/cloudflare/circl/ecc/goldilocks/twistPoint.go +135 -0
- package/govendor/github.com/cloudflare/circl/ecc/goldilocks/twistTables.go +216 -0
- package/govendor/github.com/cloudflare/circl/ecc/goldilocks/twist_basemult.go +62 -0
- package/govendor/github.com/cloudflare/circl/go.mod +3 -0
- package/govendor/github.com/cloudflare/circl/internal/conv/conv.go +173 -0
- package/govendor/github.com/cloudflare/circl/internal/sha3/doc.go +62 -0
- package/govendor/github.com/cloudflare/circl/internal/sha3/hashes.go +69 -0
- package/govendor/github.com/cloudflare/circl/internal/sha3/keccakf.go +391 -0
- package/govendor/github.com/cloudflare/circl/internal/sha3/rc.go +29 -0
- package/govendor/github.com/cloudflare/circl/internal/sha3/sha3.go +200 -0
- package/govendor/github.com/cloudflare/circl/internal/sha3/sha3_s390x.s +33 -0
- package/govendor/github.com/cloudflare/circl/internal/sha3/shake.go +119 -0
- package/govendor/github.com/cloudflare/circl/internal/sha3/xor.go +15 -0
- package/govendor/github.com/cloudflare/circl/internal/sha3/xor_generic.go +33 -0
- package/govendor/github.com/cloudflare/circl/internal/sha3/xor_unaligned.go +61 -0
- package/govendor/github.com/cloudflare/circl/math/fp448/fp.go +164 -0
- package/govendor/github.com/cloudflare/circl/math/fp448/fp_amd64.go +43 -0
- package/govendor/github.com/cloudflare/circl/math/fp448/fp_amd64.h +591 -0
- package/govendor/github.com/cloudflare/circl/math/fp448/fp_amd64.s +75 -0
- package/govendor/github.com/cloudflare/circl/math/fp448/fp_generic.go +339 -0
- package/govendor/github.com/cloudflare/circl/math/fp448/fp_noasm.go +12 -0
- package/govendor/github.com/cloudflare/circl/math/fp448/fuzzer.go +75 -0
- package/govendor/github.com/cloudflare/circl/math/integer.go +16 -0
- package/govendor/github.com/cloudflare/circl/math/mlsbset/mlsbset.go +145 -0
- package/govendor/github.com/cloudflare/circl/math/mlsbset/power.go +64 -0
- package/govendor/github.com/cloudflare/circl/math/primes.go +34 -0
- package/govendor/github.com/cloudflare/circl/math/wnaf.go +84 -0
- package/govendor/github.com/cloudflare/circl/sign/ed448/ed448.go +417 -0
- package/govendor/github.com/cloudflare/circl/sign/ed448/signapi.go +89 -0
- package/govendor/github.com/cloudflare/circl/sign/sign.go +119 -0
- package/govendor/github.com/decred/dcrd/dcrec/secp256k1/v4/LICENSE +17 -0
- package/govendor/github.com/decred/dcrd/dcrec/secp256k1/v4/README.md +72 -0
- package/govendor/github.com/decred/dcrd/dcrec/secp256k1/v4/compressedbytepoints.go +18 -0
- package/govendor/github.com/decred/dcrd/dcrec/secp256k1/v4/curve.go +1310 -0
- package/govendor/github.com/decred/dcrd/dcrec/secp256k1/v4/curve_embedded.go +14 -0
- package/govendor/github.com/decred/dcrd/dcrec/secp256k1/v4/curve_precompute.go +14 -0
- package/govendor/github.com/decred/dcrd/dcrec/secp256k1/v4/doc.go +59 -0
- package/govendor/github.com/decred/dcrd/dcrec/secp256k1/v4/ecdh.go +21 -0
- package/govendor/github.com/decred/dcrd/dcrec/secp256k1/v4/ecdsa/README.md +52 -0
- package/govendor/github.com/decred/dcrd/dcrec/secp256k1/v4/ecdsa/doc.go +42 -0
- package/govendor/github.com/decred/dcrd/dcrec/secp256k1/v4/ecdsa/error.go +134 -0
- package/govendor/github.com/decred/dcrd/dcrec/secp256k1/v4/ecdsa/signature.go +1001 -0
- package/govendor/github.com/decred/dcrd/dcrec/secp256k1/v4/ellipticadaptor.go +296 -0
- package/govendor/github.com/decred/dcrd/dcrec/secp256k1/v4/error.go +67 -0
- package/govendor/github.com/decred/dcrd/dcrec/secp256k1/v4/field.go +1696 -0
- package/govendor/github.com/decred/dcrd/dcrec/secp256k1/v4/go.mod +3 -0
- package/govendor/github.com/decred/dcrd/dcrec/secp256k1/v4/loadprecomputed.go +91 -0
- package/govendor/github.com/decred/dcrd/dcrec/secp256k1/v4/modnscalar.go +1105 -0
- package/govendor/github.com/decred/dcrd/dcrec/secp256k1/v4/nonce.go +263 -0
- package/govendor/github.com/decred/dcrd/dcrec/secp256k1/v4/privkey.go +111 -0
- package/govendor/github.com/decred/dcrd/dcrec/secp256k1/v4/pubkey.go +236 -0
- package/govendor/github.com/ebitengine/purego/LICENSE +201 -0
- package/govendor/github.com/ebitengine/purego/README.md +124 -0
- package/govendor/github.com/ebitengine/purego/abi_amd64.h +99 -0
- package/govendor/github.com/ebitengine/purego/abi_arm64.h +39 -0
- package/govendor/github.com/ebitengine/purego/abi_loong64.h +60 -0
- package/govendor/github.com/ebitengine/purego/cdecl.go +11 -0
- package/govendor/github.com/ebitengine/purego/cgo.go +19 -0
- package/govendor/github.com/ebitengine/purego/dlerror.go +17 -0
- package/govendor/github.com/ebitengine/purego/dlfcn.go +99 -0
- package/govendor/github.com/ebitengine/purego/dlfcn_android.go +34 -0
- package/govendor/github.com/ebitengine/purego/dlfcn_darwin.go +20 -0
- package/govendor/github.com/ebitengine/purego/dlfcn_freebsd.go +14 -0
- package/govendor/github.com/ebitengine/purego/dlfcn_linux.go +16 -0
- package/govendor/github.com/ebitengine/purego/dlfcn_netbsd.go +15 -0
- package/govendor/github.com/ebitengine/purego/dlfcn_nocgo_freebsd.go +11 -0
- package/govendor/github.com/ebitengine/purego/dlfcn_nocgo_linux.go +19 -0
- package/govendor/github.com/ebitengine/purego/dlfcn_nocgo_netbsd.go +11 -0
- package/govendor/github.com/ebitengine/purego/dlfcn_playground.go +24 -0
- package/govendor/github.com/ebitengine/purego/dlfcn_stubs.s +22 -0
- package/govendor/github.com/ebitengine/purego/func.go +780 -0
- package/govendor/github.com/ebitengine/purego/gen.go +6 -0
- package/govendor/github.com/ebitengine/purego/go.mod +3 -0
- package/govendor/github.com/ebitengine/purego/go_runtime.go +19 -0
- package/govendor/github.com/ebitengine/purego/internal/cgo/dlfcn_cgo_unix.go +56 -0
- package/govendor/github.com/ebitengine/purego/internal/cgo/empty.go +6 -0
- package/govendor/github.com/ebitengine/purego/internal/cgo/syscall_cgo_unix.go +49 -0
- package/govendor/github.com/ebitengine/purego/internal/fakecgo/abi_amd64.h +99 -0
- package/govendor/github.com/ebitengine/purego/internal/fakecgo/abi_arm64.h +39 -0
- package/govendor/github.com/ebitengine/purego/internal/fakecgo/abi_loong64.h +60 -0
- package/govendor/github.com/ebitengine/purego/internal/fakecgo/abi_ppc64x.h +195 -0
- package/govendor/github.com/ebitengine/purego/internal/fakecgo/abi_riscv64.h +72 -0
- package/govendor/github.com/ebitengine/purego/internal/fakecgo/asm_386.s +29 -0
- package/govendor/github.com/ebitengine/purego/internal/fakecgo/asm_amd64.s +39 -0
- package/govendor/github.com/ebitengine/purego/internal/fakecgo/asm_arm.s +58 -0
- package/govendor/github.com/ebitengine/purego/internal/fakecgo/asm_arm64.s +35 -0
- package/govendor/github.com/ebitengine/purego/internal/fakecgo/asm_loong64.s +38 -0
- package/govendor/github.com/ebitengine/purego/internal/fakecgo/asm_ppc64le.s +79 -0
- package/govendor/github.com/ebitengine/purego/internal/fakecgo/asm_riscv64.s +36 -0
- package/govendor/github.com/ebitengine/purego/internal/fakecgo/asm_s390x.s +52 -0
- package/govendor/github.com/ebitengine/purego/internal/fakecgo/callbacks.go +93 -0
- package/govendor/github.com/ebitengine/purego/internal/fakecgo/doc.go +32 -0
- package/govendor/github.com/ebitengine/purego/internal/fakecgo/fakecgo.go +14 -0
- package/govendor/github.com/ebitengine/purego/internal/fakecgo/freebsd.go +27 -0
- package/govendor/github.com/ebitengine/purego/internal/fakecgo/go_darwin.go +86 -0
- package/govendor/github.com/ebitengine/purego/internal/fakecgo/go_freebsd.go +77 -0
- package/govendor/github.com/ebitengine/purego/internal/fakecgo/go_libinit.go +72 -0
- package/govendor/github.com/ebitengine/purego/internal/fakecgo/go_linux.go +77 -0
- package/govendor/github.com/ebitengine/purego/internal/fakecgo/go_netbsd.go +83 -0
- package/govendor/github.com/ebitengine/purego/internal/fakecgo/go_setenv.go +18 -0
- package/govendor/github.com/ebitengine/purego/internal/fakecgo/go_util.go +38 -0
- package/govendor/github.com/ebitengine/purego/internal/fakecgo/iscgo.go +19 -0
- package/govendor/github.com/ebitengine/purego/internal/fakecgo/libcgo.go +37 -0
- package/govendor/github.com/ebitengine/purego/internal/fakecgo/libcgo_darwin.go +37 -0
- package/govendor/github.com/ebitengine/purego/internal/fakecgo/libcgo_freebsd.go +27 -0
- package/govendor/github.com/ebitengine/purego/internal/fakecgo/libcgo_linux.go +27 -0
- package/govendor/github.com/ebitengine/purego/internal/fakecgo/libcgo_netbsd.go +36 -0
- package/govendor/github.com/ebitengine/purego/internal/fakecgo/linux.go +189 -0
- package/govendor/github.com/ebitengine/purego/internal/fakecgo/netbsd.go +23 -0
- package/govendor/github.com/ebitengine/purego/internal/fakecgo/setenv.go +19 -0
- package/govendor/github.com/ebitengine/purego/internal/fakecgo/threadentry_amd64.go +15 -0
- package/govendor/github.com/ebitengine/purego/internal/fakecgo/threadentry_noasm.go +24 -0
- package/govendor/github.com/ebitengine/purego/internal/fakecgo/trampolines_386.s +130 -0
- package/govendor/github.com/ebitengine/purego/internal/fakecgo/trampolines_amd64.s +124 -0
- package/govendor/github.com/ebitengine/purego/internal/fakecgo/trampolines_arm.s +131 -0
- package/govendor/github.com/ebitengine/purego/internal/fakecgo/trampolines_arm64.s +81 -0
- package/govendor/github.com/ebitengine/purego/internal/fakecgo/trampolines_linux_386.s +78 -0
- package/govendor/github.com/ebitengine/purego/internal/fakecgo/trampolines_linux_amd64.s +69 -0
- package/govendor/github.com/ebitengine/purego/internal/fakecgo/trampolines_linux_arm.s +69 -0
- package/govendor/github.com/ebitengine/purego/internal/fakecgo/trampolines_linux_arm64.s +60 -0
- package/govendor/github.com/ebitengine/purego/internal/fakecgo/trampolines_linux_loong64.s +60 -0
- package/govendor/github.com/ebitengine/purego/internal/fakecgo/trampolines_linux_ppc64le.s +69 -0
- package/govendor/github.com/ebitengine/purego/internal/fakecgo/trampolines_linux_riscv64.s +60 -0
- package/govendor/github.com/ebitengine/purego/internal/fakecgo/trampolines_linux_s390x.s +51 -0
- package/govendor/github.com/ebitengine/purego/internal/fakecgo/trampolines_loong64.s +78 -0
- package/govendor/github.com/ebitengine/purego/internal/fakecgo/trampolines_ppc64le.s +132 -0
- package/govendor/github.com/ebitengine/purego/internal/fakecgo/trampolines_riscv64.s +76 -0
- package/govendor/github.com/ebitengine/purego/internal/fakecgo/trampolines_s390x.s +158 -0
- package/govendor/github.com/ebitengine/purego/internal/fakecgo/zsymbols.go +165 -0
- package/govendor/github.com/ebitengine/purego/internal/fakecgo/zsymbols_darwin.go +59 -0
- package/govendor/github.com/ebitengine/purego/internal/fakecgo/zsymbols_freebsd.go +48 -0
- package/govendor/github.com/ebitengine/purego/internal/fakecgo/zsymbols_linux.go +158 -0
- package/govendor/github.com/ebitengine/purego/internal/fakecgo/zsymbols_netbsd.go +59 -0
- package/govendor/github.com/ebitengine/purego/internal/fakecgo/ztrampolines_darwin.s +19 -0
- package/govendor/github.com/ebitengine/purego/internal/fakecgo/ztrampolines_freebsd.s +16 -0
- package/govendor/github.com/ebitengine/purego/internal/fakecgo/ztrampolines_linux.s +46 -0
- package/govendor/github.com/ebitengine/purego/internal/fakecgo/ztrampolines_netbsd.s +19 -0
- package/govendor/github.com/ebitengine/purego/internal/fakecgo/ztrampolines_stubs.s +55 -0
- package/govendor/github.com/ebitengine/purego/internal/strings/strings.go +40 -0
- package/govendor/github.com/ebitengine/purego/is_ios.go +13 -0
- package/govendor/github.com/ebitengine/purego/nocgo.go +25 -0
- package/govendor/github.com/ebitengine/purego/struct_amd64.go +518 -0
- package/govendor/github.com/ebitengine/purego/struct_arm.go +100 -0
- package/govendor/github.com/ebitengine/purego/struct_arm64.go +757 -0
- package/govendor/github.com/ebitengine/purego/struct_loong64.go +244 -0
- package/govendor/github.com/ebitengine/purego/struct_other.go +54 -0
- package/govendor/github.com/ebitengine/purego/struct_ppc64le.go +180 -0
- package/govendor/github.com/ebitengine/purego/struct_riscv64.go +158 -0
- package/govendor/github.com/ebitengine/purego/struct_s390x.go +158 -0
- package/govendor/github.com/ebitengine/purego/sys_386.s +153 -0
- package/govendor/github.com/ebitengine/purego/sys_amd64.s +206 -0
- package/govendor/github.com/ebitengine/purego/sys_arm.s +153 -0
- package/govendor/github.com/ebitengine/purego/sys_arm64.s +133 -0
- package/govendor/github.com/ebitengine/purego/sys_loong64.s +131 -0
- package/govendor/github.com/ebitengine/purego/sys_ppc64le.s +126 -0
- package/govendor/github.com/ebitengine/purego/sys_riscv64.s +135 -0
- package/govendor/github.com/ebitengine/purego/sys_s390x.s +148 -0
- package/govendor/github.com/ebitengine/purego/sys_unix_386.s +234 -0
- package/govendor/github.com/ebitengine/purego/sys_unix_arm.s +99 -0
- package/govendor/github.com/ebitengine/purego/sys_unix_arm64.s +75 -0
- package/govendor/github.com/ebitengine/purego/sys_unix_loong64.s +79 -0
- package/govendor/github.com/ebitengine/purego/sys_unix_ppc64le.s +114 -0
- package/govendor/github.com/ebitengine/purego/sys_unix_riscv64.s +79 -0
- package/govendor/github.com/ebitengine/purego/sys_unix_s390x.s +111 -0
- package/govendor/github.com/ebitengine/purego/syscall.go +95 -0
- package/govendor/github.com/ebitengine/purego/syscall_32bit.go +97 -0
- package/govendor/github.com/ebitengine/purego/syscall_cgo_linux.go +20 -0
- package/govendor/github.com/ebitengine/purego/syscall_notstackargs.go +55 -0
- package/govendor/github.com/ebitengine/purego/syscall_ppc64le.go +73 -0
- package/govendor/github.com/ebitengine/purego/syscall_stackargs_ppc64le.go +38 -0
- package/govendor/github.com/ebitengine/purego/syscall_stackargs_s390x.go +38 -0
- package/govendor/github.com/ebitengine/purego/syscall_unix.go +392 -0
- package/govendor/github.com/ebitengine/purego/syscall_windows.go +54 -0
- package/govendor/github.com/ebitengine/purego/zcallback_386.s +4014 -0
- package/govendor/github.com/ebitengine/purego/zcallback_amd64.s +2014 -0
- package/govendor/github.com/ebitengine/purego/zcallback_arm.s +4014 -0
- package/govendor/github.com/ebitengine/purego/zcallback_arm64.s +4014 -0
- package/govendor/github.com/ebitengine/purego/zcallback_loong64.s +4014 -0
- package/govendor/github.com/ebitengine/purego/zcallback_ppc64le.s +4014 -0
- package/govendor/github.com/ebitengine/purego/zcallback_riscv64.s +4051 -0
- package/govendor/github.com/ebitengine/purego/zcallback_s390x.s +4015 -0
- package/govendor/github.com/klauspost/compress/.gitattributes +3 -0
- package/govendor/github.com/klauspost/compress/.goreleaser.yml +132 -0
- package/govendor/github.com/klauspost/compress/AGENTS.md +106 -0
- package/govendor/github.com/klauspost/compress/LICENSE +304 -0
- package/govendor/github.com/klauspost/compress/README.md +766 -0
- package/govendor/github.com/klauspost/compress/SECURITY.md +25 -0
- package/govendor/github.com/klauspost/compress/compressible.go +85 -0
- package/govendor/github.com/klauspost/compress/fse/README.md +79 -0
- package/govendor/github.com/klauspost/compress/fse/bitreader.go +122 -0
- package/govendor/github.com/klauspost/compress/fse/bitwriter.go +167 -0
- package/govendor/github.com/klauspost/compress/fse/bytereader.go +47 -0
- package/govendor/github.com/klauspost/compress/fse/compress.go +685 -0
- package/govendor/github.com/klauspost/compress/fse/decompress.go +383 -0
- package/govendor/github.com/klauspost/compress/fse/fse.go +150 -0
- package/govendor/github.com/klauspost/compress/gen.sh +4 -0
- package/govendor/github.com/klauspost/compress/go.mod +3 -0
- package/govendor/github.com/klauspost/compress/huff0/README.md +89 -0
- package/govendor/github.com/klauspost/compress/huff0/bitreader.go +323 -0
- package/govendor/github.com/klauspost/compress/huff0/bitwriter.go +97 -0
- package/govendor/github.com/klauspost/compress/huff0/build_table.go +168 -0
- package/govendor/github.com/klauspost/compress/huff0/compress.go +747 -0
- package/govendor/github.com/klauspost/compress/huff0/decompress.go +1188 -0
- package/govendor/github.com/klauspost/compress/huff0/decompress_amd64.go +128 -0
- package/govendor/github.com/klauspost/compress/huff0/decompress_amd64.s +3476 -0
- package/govendor/github.com/klauspost/compress/huff0/decompress_arm64.go +72 -0
- package/govendor/github.com/klauspost/compress/huff0/decompress_arm64.s +1648 -0
- package/govendor/github.com/klauspost/compress/huff0/decompress_asm.go +252 -0
- package/govendor/github.com/klauspost/compress/huff0/decompress_generic.go +278 -0
- package/govendor/github.com/klauspost/compress/huff0/huff0.go +337 -0
- package/govendor/github.com/klauspost/compress/internal/cpuinfo/cpuinfo.go +34 -0
- package/govendor/github.com/klauspost/compress/internal/cpuinfo/cpuinfo_amd64.go +10 -0
- package/govendor/github.com/klauspost/compress/internal/cpuinfo/cpuinfo_amd64.s +34 -0
- package/govendor/github.com/klauspost/compress/internal/le/le.go +5 -0
- package/govendor/github.com/klauspost/compress/internal/le/unsafe_disabled.go +42 -0
- package/govendor/github.com/klauspost/compress/internal/le/unsafe_enabled.go +52 -0
- package/govendor/github.com/klauspost/compress/internal/regmask/regmask.go +50 -0
- package/govendor/github.com/klauspost/compress/internal/regmask/regmask_386.go +39 -0
- package/govendor/github.com/klauspost/compress/internal/snapref/LICENSE +27 -0
- package/govendor/github.com/klauspost/compress/internal/snapref/decode.go +264 -0
- package/govendor/github.com/klauspost/compress/internal/snapref/decode_other.go +113 -0
- package/govendor/github.com/klauspost/compress/internal/snapref/encode.go +291 -0
- package/govendor/github.com/klauspost/compress/internal/snapref/encode_other.go +250 -0
- package/govendor/github.com/klauspost/compress/internal/snapref/snappy.go +98 -0
- package/govendor/github.com/klauspost/compress/s2sx.mod +3 -0
- package/govendor/github.com/klauspost/compress/s2sx.sum +0 -0
- package/govendor/github.com/klauspost/compress/zstd/README.md +487 -0
- package/govendor/github.com/klauspost/compress/zstd/bitreader.go +176 -0
- package/govendor/github.com/klauspost/compress/zstd/bitwriter.go +151 -0
- package/govendor/github.com/klauspost/compress/zstd/blockdec.go +728 -0
- package/govendor/github.com/klauspost/compress/zstd/blockenc.go +923 -0
- package/govendor/github.com/klauspost/compress/zstd/blocktype_string.go +85 -0
- package/govendor/github.com/klauspost/compress/zstd/bytebuf.go +131 -0
- package/govendor/github.com/klauspost/compress/zstd/bytereader.go +82 -0
- package/govendor/github.com/klauspost/compress/zstd/decodeheader.go +261 -0
- package/govendor/github.com/klauspost/compress/zstd/decoder.go +995 -0
- package/govendor/github.com/klauspost/compress/zstd/decoder_options.go +213 -0
- package/govendor/github.com/klauspost/compress/zstd/dict.go +605 -0
- package/govendor/github.com/klauspost/compress/zstd/enc_base.go +200 -0
- package/govendor/github.com/klauspost/compress/zstd/enc_best.go +572 -0
- package/govendor/github.com/klauspost/compress/zstd/enc_better.go +1262 -0
- package/govendor/github.com/klauspost/compress/zstd/enc_dfast.go +1136 -0
- package/govendor/github.com/klauspost/compress/zstd/enc_fast.go +892 -0
- package/govendor/github.com/klauspost/compress/zstd/enc_jobs.go +352 -0
- package/govendor/github.com/klauspost/compress/zstd/encoder.go +881 -0
- package/govendor/github.com/klauspost/compress/zstd/encoder_options.go +422 -0
- package/govendor/github.com/klauspost/compress/zstd/framedec.go +404 -0
- package/govendor/github.com/klauspost/compress/zstd/frameenc.go +137 -0
- package/govendor/github.com/klauspost/compress/zstd/fse_decoder.go +333 -0
- package/govendor/github.com/klauspost/compress/zstd/fse_decoder_amd64.s +132 -0
- package/govendor/github.com/klauspost/compress/zstd/fse_decoder_arm64.s +148 -0
- package/govendor/github.com/klauspost/compress/zstd/fse_decoder_asm.go +70 -0
- package/govendor/github.com/klauspost/compress/zstd/fse_decoder_generic.go +71 -0
- package/govendor/github.com/klauspost/compress/zstd/fse_encoder.go +701 -0
- package/govendor/github.com/klauspost/compress/zstd/fse_predefined.go +168 -0
- package/govendor/github.com/klauspost/compress/zstd/hash.go +57 -0
- package/govendor/github.com/klauspost/compress/zstd/history.go +119 -0
- package/govendor/github.com/klauspost/compress/zstd/internal/xxhash/LICENSE.txt +22 -0
- package/govendor/github.com/klauspost/compress/zstd/internal/xxhash/README.md +71 -0
- package/govendor/github.com/klauspost/compress/zstd/internal/xxhash/xxhash.go +242 -0
- package/govendor/github.com/klauspost/compress/zstd/internal/xxhash/xxhash_amd64.s +210 -0
- package/govendor/github.com/klauspost/compress/zstd/internal/xxhash/xxhash_arm64.s +184 -0
- package/govendor/github.com/klauspost/compress/zstd/internal/xxhash/xxhash_asm.go +11 -0
- package/govendor/github.com/klauspost/compress/zstd/internal/xxhash/xxhash_other.go +75 -0
- package/govendor/github.com/klauspost/compress/zstd/internal/xxhash/xxhash_safe.go +11 -0
- package/govendor/github.com/klauspost/compress/zstd/matchlen_amd64.go +15 -0
- package/govendor/github.com/klauspost/compress/zstd/matchlen_amd64.s +66 -0
- package/govendor/github.com/klauspost/compress/zstd/matchlen_generic.go +37 -0
- package/govendor/github.com/klauspost/compress/zstd/seqdec.go +620 -0
- package/govendor/github.com/klauspost/compress/zstd/seqdec_amd64.go +115 -0
- package/govendor/github.com/klauspost/compress/zstd/seqdec_amd64.s +4762 -0
- package/govendor/github.com/klauspost/compress/zstd/seqdec_arm64.go +81 -0
- package/govendor/github.com/klauspost/compress/zstd/seqdec_arm64.s +2919 -0
- package/govendor/github.com/klauspost/compress/zstd/seqdec_asm.go +327 -0
- package/govendor/github.com/klauspost/compress/zstd/seqdec_generic.go +258 -0
- package/govendor/github.com/klauspost/compress/zstd/seqenc.go +159 -0
- package/govendor/github.com/klauspost/compress/zstd/simple.go +54 -0
- package/govendor/github.com/klauspost/compress/zstd/snappy.go +447 -0
- package/govendor/github.com/klauspost/compress/zstd/zip.go +141 -0
- package/govendor/github.com/klauspost/compress/zstd/zstd.go +126 -0
- package/govendor/github.com/tetratelabs/wazero/.editorconfig +7 -0
- package/govendor/github.com/tetratelabs/wazero/.gitattributes +2 -0
- package/govendor/github.com/tetratelabs/wazero/.gitmodules +3 -0
- package/govendor/github.com/tetratelabs/wazero/CONTRIBUTING.md +75 -0
- package/govendor/github.com/tetratelabs/wazero/LICENSE +201 -0
- package/govendor/github.com/tetratelabs/wazero/Makefile +385 -0
- package/govendor/github.com/tetratelabs/wazero/NOTICE +2 -0
- package/govendor/github.com/tetratelabs/wazero/RATIONALE.md +1577 -0
- package/govendor/github.com/tetratelabs/wazero/README.md +135 -0
- package/govendor/github.com/tetratelabs/wazero/api/features.go +228 -0
- package/govendor/github.com/tetratelabs/wazero/api/wasm.go +766 -0
- package/govendor/github.com/tetratelabs/wazero/builder.go +367 -0
- package/govendor/github.com/tetratelabs/wazero/cache.go +123 -0
- package/govendor/github.com/tetratelabs/wazero/codecov.yml +9 -0
- package/govendor/github.com/tetratelabs/wazero/config.go +899 -0
- package/govendor/github.com/tetratelabs/wazero/experimental/checkpoint.go +35 -0
- package/govendor/github.com/tetratelabs/wazero/experimental/close.go +63 -0
- package/govendor/github.com/tetratelabs/wazero/experimental/compilationworkers.go +19 -0
- package/govendor/github.com/tetratelabs/wazero/experimental/experimental.go +41 -0
- package/govendor/github.com/tetratelabs/wazero/experimental/features.go +39 -0
- package/govendor/github.com/tetratelabs/wazero/experimental/importresolver.go +19 -0
- package/govendor/github.com/tetratelabs/wazero/experimental/listener.go +324 -0
- package/govendor/github.com/tetratelabs/wazero/experimental/memory.go +52 -0
- package/govendor/github.com/tetratelabs/wazero/experimental/sys/dir.go +92 -0
- package/govendor/github.com/tetratelabs/wazero/experimental/sys/errno.go +98 -0
- package/govendor/github.com/tetratelabs/wazero/experimental/sys/error.go +45 -0
- package/govendor/github.com/tetratelabs/wazero/experimental/sys/file.go +356 -0
- package/govendor/github.com/tetratelabs/wazero/experimental/sys/fs.go +292 -0
- package/govendor/github.com/tetratelabs/wazero/experimental/sys/oflag.go +70 -0
- package/govendor/github.com/tetratelabs/wazero/experimental/sys/poll.go +48 -0
- package/govendor/github.com/tetratelabs/wazero/experimental/sys/syscall_errno.go +106 -0
- package/govendor/github.com/tetratelabs/wazero/experimental/sys/syscall_errno_notwindows.go +13 -0
- package/govendor/github.com/tetratelabs/wazero/experimental/sys/syscall_errno_unsupported.go +7 -0
- package/govendor/github.com/tetratelabs/wazero/experimental/sys/syscall_errno_windows.go +37 -0
- package/govendor/github.com/tetratelabs/wazero/experimental/sys/time.go +10 -0
- package/govendor/github.com/tetratelabs/wazero/experimental/sys/unimplemented.go +160 -0
- package/govendor/github.com/tetratelabs/wazero/fsconfig.go +213 -0
- package/govendor/github.com/tetratelabs/wazero/go.mod +3 -0
- package/govendor/github.com/tetratelabs/wazero/imports/wasi_snapshot_preview1/args.go +97 -0
- package/govendor/github.com/tetratelabs/wazero/imports/wasi_snapshot_preview1/clock.go +116 -0
- package/govendor/github.com/tetratelabs/wazero/imports/wasi_snapshot_preview1/environ.go +100 -0
- package/govendor/github.com/tetratelabs/wazero/imports/wasi_snapshot_preview1/fs.go +2015 -0
- package/govendor/github.com/tetratelabs/wazero/imports/wasi_snapshot_preview1/poll.go +247 -0
- package/govendor/github.com/tetratelabs/wazero/imports/wasi_snapshot_preview1/proc.go +44 -0
- package/govendor/github.com/tetratelabs/wazero/imports/wasi_snapshot_preview1/random.go +55 -0
- package/govendor/github.com/tetratelabs/wazero/imports/wasi_snapshot_preview1/sched.go +22 -0
- package/govendor/github.com/tetratelabs/wazero/imports/wasi_snapshot_preview1/sock.go +188 -0
- package/govendor/github.com/tetratelabs/wazero/imports/wasi_snapshot_preview1/wasi.go +314 -0
- package/govendor/github.com/tetratelabs/wazero/internal/descriptor/table.go +149 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/interpreter/compiler.go +3884 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/interpreter/format.go +22 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/interpreter/interpreter.go +4963 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/interpreter/operations.go +2978 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/interpreter/signature.go +825 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/backend/abi.go +170 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/backend/backend.go +3 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/backend/compiler.go +402 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/backend/compiler_lower.go +226 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/backend/go_call.go +33 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/backend/isa/amd64/abi.go +186 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/backend/isa/amd64/abi_entry_amd64.go +9 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/backend/isa/amd64/abi_entry_amd64.s +29 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/backend/isa/amd64/abi_entry_preamble.go +248 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/backend/isa/amd64/abi_go_call.go +440 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/backend/isa/amd64/cond.go +168 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/backend/isa/amd64/ext.go +35 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/backend/isa/amd64/instr.go +2495 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/backend/isa/amd64/instr_encoding.go +1714 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/backend/isa/amd64/lower_constant.go +71 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/backend/isa/amd64/lower_mem.go +187 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/backend/isa/amd64/machine.go +3782 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/backend/isa/amd64/machine_pro_epi_logue.go +334 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/backend/isa/amd64/machine_regalloc.go +352 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/backend/isa/amd64/machine_vec.go +992 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/backend/isa/amd64/operands.go +336 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/backend/isa/amd64/reg.go +181 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/backend/isa/amd64/stack.go +130 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/backend/isa/arm64/abi.go +376 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/backend/isa/arm64/abi_entry_arm64.go +9 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/backend/isa/arm64/abi_entry_arm64.s +29 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/backend/isa/arm64/abi_entry_preamble.go +233 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/backend/isa/arm64/abi_go_call.go +430 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/backend/isa/arm64/cond.go +215 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/backend/isa/arm64/instr.go +2562 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/backend/isa/arm64/instr_encoding.go +2357 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/backend/isa/arm64/lower_constant.go +301 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/backend/isa/arm64/lower_instr.go +2227 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/backend/isa/arm64/lower_instr_operands.go +340 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/backend/isa/arm64/lower_mem.go +451 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/backend/isa/arm64/machine.go +631 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/backend/isa/arm64/machine_pro_epi_logue.go +486 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/backend/isa/arm64/machine_regalloc.go +351 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/backend/isa/arm64/machine_relocation.go +128 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/backend/isa/arm64/reg.go +397 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/backend/isa/arm64/unwind_stack.go +84 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/backend/machine.go +119 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/backend/regalloc/api.go +124 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/backend/regalloc/reg.go +123 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/backend/regalloc/regalloc.go +1189 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/backend/regalloc/regset.go +96 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/backend/vdef.go +19 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/call_engine.go +884 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/engine.go +1008 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/engine_cache.go +347 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/entrypoint_amd64.go +13 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/entrypoint_arm64.go +13 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/entrypoint_other.go +15 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/frontend/frontend.go +663 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/frontend/lower.go +5045 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/frontend/misc.go +10 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/frontend/sort_id.go +13 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/hostmodule.go +88 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/isa_amd64.go +30 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/isa_arm64.go +32 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/isa_other.go +29 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/memmove.go +11 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/module_engine.go +403 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/ssa/basic_block.go +381 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/ssa/basic_block_sort.go +32 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/ssa/builder.go +790 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/ssa/cmp.go +107 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/ssa/funcref.go +12 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/ssa/instructions.go +3024 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/ssa/pass.go +393 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/ssa/pass_blk_layouts.go +334 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/ssa/pass_cfg.go +313 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/ssa/signature.go +49 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/ssa/ssa.go +14 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/ssa/type.go +115 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/ssa/vs.go +114 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/wazevoapi/debug_options.go +189 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/wazevoapi/exitcode.go +150 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/wazevoapi/offsetdata.go +248 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/wazevoapi/perfmap.go +96 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/wazevoapi/perfmap_disabled.go +5 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/wazevoapi/perfmap_enabled.go +5 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/wazevoapi/pool.go +215 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/wazevoapi/ptr.go +15 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/wazevoapi/queue.go +26 -0
- package/govendor/github.com/tetratelabs/wazero/internal/engine/wazevo/wazevoapi/resetmap.go +11 -0
- package/govendor/github.com/tetratelabs/wazero/internal/expctxkeys/checkpoint.go +10 -0
- package/govendor/github.com/tetratelabs/wazero/internal/expctxkeys/close.go +5 -0
- package/govendor/github.com/tetratelabs/wazero/internal/expctxkeys/compilationworkers.go +6 -0
- package/govendor/github.com/tetratelabs/wazero/internal/expctxkeys/expctxkeys.go +2 -0
- package/govendor/github.com/tetratelabs/wazero/internal/expctxkeys/importresolver.go +6 -0
- package/govendor/github.com/tetratelabs/wazero/internal/expctxkeys/listener.go +7 -0
- package/govendor/github.com/tetratelabs/wazero/internal/expctxkeys/memory.go +4 -0
- package/govendor/github.com/tetratelabs/wazero/internal/filecache/compilationcache.go +42 -0
- package/govendor/github.com/tetratelabs/wazero/internal/filecache/file_cache.go +76 -0
- package/govendor/github.com/tetratelabs/wazero/internal/ieee754/ieee754.go +29 -0
- package/govendor/github.com/tetratelabs/wazero/internal/internalapi/internal.go +9 -0
- package/govendor/github.com/tetratelabs/wazero/internal/leb128/leb128.go +285 -0
- package/govendor/github.com/tetratelabs/wazero/internal/moremath/moremath.go +271 -0
- package/govendor/github.com/tetratelabs/wazero/internal/platform/cpuid.go +26 -0
- package/govendor/github.com/tetratelabs/wazero/internal/platform/cpuid_amd64.go +22 -0
- package/govendor/github.com/tetratelabs/wazero/internal/platform/cpuid_arm64.go +13 -0
- package/govendor/github.com/tetratelabs/wazero/internal/platform/cpuid_unsupported.go +5 -0
- package/govendor/github.com/tetratelabs/wazero/internal/platform/crypto.go +17 -0
- package/govendor/github.com/tetratelabs/wazero/internal/platform/mmap_linux.go +72 -0
- package/govendor/github.com/tetratelabs/wazero/internal/platform/mmap_other.go +19 -0
- package/govendor/github.com/tetratelabs/wazero/internal/platform/mmap_unix.go +14 -0
- package/govendor/github.com/tetratelabs/wazero/internal/platform/mmap_unsupported.go +22 -0
- package/govendor/github.com/tetratelabs/wazero/internal/platform/mmap_windows.go +31 -0
- package/govendor/github.com/tetratelabs/wazero/internal/platform/path.go +6 -0
- package/govendor/github.com/tetratelabs/wazero/internal/platform/path_windows.go +17 -0
- package/govendor/github.com/tetratelabs/wazero/internal/platform/platform.go +71 -0
- package/govendor/github.com/tetratelabs/wazero/internal/platform/time.go +76 -0
- package/govendor/github.com/tetratelabs/wazero/internal/platform/time_notwindows.go +11 -0
- package/govendor/github.com/tetratelabs/wazero/internal/platform/time_windows.go +40 -0
- package/govendor/github.com/tetratelabs/wazero/internal/sock/sock.go +89 -0
- package/govendor/github.com/tetratelabs/wazero/internal/sock/sock_supported.go +11 -0
- package/govendor/github.com/tetratelabs/wazero/internal/sock/sock_unsupported.go +10 -0
- package/govendor/github.com/tetratelabs/wazero/internal/sys/fs.go +458 -0
- package/govendor/github.com/tetratelabs/wazero/internal/sys/lazy.go +135 -0
- package/govendor/github.com/tetratelabs/wazero/internal/sys/stdio.go +125 -0
- package/govendor/github.com/tetratelabs/wazero/internal/sys/sys.go +228 -0
- package/govendor/github.com/tetratelabs/wazero/internal/sysfs/adapter.go +105 -0
- package/govendor/github.com/tetratelabs/wazero/internal/sysfs/datasync_linux.go +12 -0
- package/govendor/github.com/tetratelabs/wazero/internal/sysfs/datasync_unsupported.go +14 -0
- package/govendor/github.com/tetratelabs/wazero/internal/sysfs/dir.go +24 -0
- package/govendor/github.com/tetratelabs/wazero/internal/sysfs/dirfs.go +137 -0
- package/govendor/github.com/tetratelabs/wazero/internal/sysfs/file.go +576 -0
- package/govendor/github.com/tetratelabs/wazero/internal/sysfs/file_unix.go +39 -0
- package/govendor/github.com/tetratelabs/wazero/internal/sysfs/file_unsupported.go +28 -0
- package/govendor/github.com/tetratelabs/wazero/internal/sysfs/file_windows.go +133 -0
- package/govendor/github.com/tetratelabs/wazero/internal/sysfs/futimens.go +29 -0
- package/govendor/github.com/tetratelabs/wazero/internal/sysfs/futimens_darwin.go +42 -0
- package/govendor/github.com/tetratelabs/wazero/internal/sysfs/futimens_darwin.s +8 -0
- package/govendor/github.com/tetratelabs/wazero/internal/sysfs/futimens_linux.go +38 -0
- package/govendor/github.com/tetratelabs/wazero/internal/sysfs/futimens_unsupported.go +18 -0
- package/govendor/github.com/tetratelabs/wazero/internal/sysfs/futimens_windows.go +42 -0
- package/govendor/github.com/tetratelabs/wazero/internal/sysfs/ino_unix.go +22 -0
- package/govendor/github.com/tetratelabs/wazero/internal/sysfs/ino_unsupported.go +17 -0
- package/govendor/github.com/tetratelabs/wazero/internal/sysfs/ino_windows.go +28 -0
- package/govendor/github.com/tetratelabs/wazero/internal/sysfs/nonblock_unix.go +17 -0
- package/govendor/github.com/tetratelabs/wazero/internal/sysfs/nonblock_unsupported.go +13 -0
- package/govendor/github.com/tetratelabs/wazero/internal/sysfs/nonblock_windows.go +23 -0
- package/govendor/github.com/tetratelabs/wazero/internal/sysfs/oflag.go +38 -0
- package/govendor/github.com/tetratelabs/wazero/internal/sysfs/open_file_bsd.go +26 -0
- package/govendor/github.com/tetratelabs/wazero/internal/sysfs/open_file_darwin.go +26 -0
- package/govendor/github.com/tetratelabs/wazero/internal/sysfs/open_file_supported.go +30 -0
- package/govendor/github.com/tetratelabs/wazero/internal/sysfs/open_file_unix.go +20 -0
- package/govendor/github.com/tetratelabs/wazero/internal/sysfs/open_file_unsupported.go +16 -0
- package/govendor/github.com/tetratelabs/wazero/internal/sysfs/open_file_windows.go +163 -0
- package/govendor/github.com/tetratelabs/wazero/internal/sysfs/osfile.go +318 -0
- package/govendor/github.com/tetratelabs/wazero/internal/sysfs/poll.go +17 -0
- package/govendor/github.com/tetratelabs/wazero/internal/sysfs/poll_darwin.go +55 -0
- package/govendor/github.com/tetratelabs/wazero/internal/sysfs/poll_darwin.s +8 -0
- package/govendor/github.com/tetratelabs/wazero/internal/sysfs/poll_linux.go +57 -0
- package/govendor/github.com/tetratelabs/wazero/internal/sysfs/poll_unsupported.go +12 -0
- package/govendor/github.com/tetratelabs/wazero/internal/sysfs/poll_windows.go +224 -0
- package/govendor/github.com/tetratelabs/wazero/internal/sysfs/readfs.go +117 -0
- package/govendor/github.com/tetratelabs/wazero/internal/sysfs/rename_unix.go +16 -0
- package/govendor/github.com/tetratelabs/wazero/internal/sysfs/rename_unsupported.go +16 -0
- package/govendor/github.com/tetratelabs/wazero/internal/sysfs/rename_windows.go +55 -0
- package/govendor/github.com/tetratelabs/wazero/internal/sysfs/sock.go +186 -0
- package/govendor/github.com/tetratelabs/wazero/internal/sysfs/sock_supported.go +76 -0
- package/govendor/github.com/tetratelabs/wazero/internal/sysfs/sock_unix.go +48 -0
- package/govendor/github.com/tetratelabs/wazero/internal/sysfs/sock_unsupported.go +79 -0
- package/govendor/github.com/tetratelabs/wazero/internal/sysfs/sock_windows.go +80 -0
- package/govendor/github.com/tetratelabs/wazero/internal/sysfs/stat.go +16 -0
- package/govendor/github.com/tetratelabs/wazero/internal/sysfs/stat_supported.go +40 -0
- package/govendor/github.com/tetratelabs/wazero/internal/sysfs/stat_unsupported.go +40 -0
- package/govendor/github.com/tetratelabs/wazero/internal/sysfs/stat_windows.go +118 -0
- package/govendor/github.com/tetratelabs/wazero/internal/sysfs/sync.go +13 -0
- package/govendor/github.com/tetratelabs/wazero/internal/sysfs/sync_windows.go +20 -0
- package/govendor/github.com/tetratelabs/wazero/internal/sysfs/syscall6_darwin.go +13 -0
- package/govendor/github.com/tetratelabs/wazero/internal/sysfs/sysfs.go +6 -0
- package/govendor/github.com/tetratelabs/wazero/internal/sysfs/unlink_unix.go +17 -0
- package/govendor/github.com/tetratelabs/wazero/internal/sysfs/unlink_unsupported.go +14 -0
- package/govendor/github.com/tetratelabs/wazero/internal/sysfs/unlink_windows.go +25 -0
- package/govendor/github.com/tetratelabs/wazero/internal/u32/u32.go +11 -0
- package/govendor/github.com/tetratelabs/wazero/internal/u64/u64.go +15 -0
- package/govendor/github.com/tetratelabs/wazero/internal/version/version.go +52 -0
- package/govendor/github.com/tetratelabs/wazero/internal/wasip1/args.go +6 -0
- package/govendor/github.com/tetratelabs/wazero/internal/wasip1/clock.go +16 -0
- package/govendor/github.com/tetratelabs/wazero/internal/wasip1/environ.go +6 -0
- package/govendor/github.com/tetratelabs/wazero/internal/wasip1/errno.go +314 -0
- package/govendor/github.com/tetratelabs/wazero/internal/wasip1/fs.go +164 -0
- package/govendor/github.com/tetratelabs/wazero/internal/wasip1/poll.go +15 -0
- package/govendor/github.com/tetratelabs/wazero/internal/wasip1/proc.go +6 -0
- package/govendor/github.com/tetratelabs/wazero/internal/wasip1/random.go +3 -0
- package/govendor/github.com/tetratelabs/wazero/internal/wasip1/rights.go +148 -0
- package/govendor/github.com/tetratelabs/wazero/internal/wasip1/sched.go +3 -0
- package/govendor/github.com/tetratelabs/wazero/internal/wasip1/sock.go +71 -0
- package/govendor/github.com/tetratelabs/wazero/internal/wasip1/wasi.go +26 -0
- package/govendor/github.com/tetratelabs/wazero/internal/wasm/binary/code.go +124 -0
- package/govendor/github.com/tetratelabs/wazero/internal/wasm/binary/const_expr.go +116 -0
- package/govendor/github.com/tetratelabs/wazero/internal/wasm/binary/custom.go +22 -0
- package/govendor/github.com/tetratelabs/wazero/internal/wasm/binary/data.go +79 -0
- package/govendor/github.com/tetratelabs/wazero/internal/wasm/binary/decoder.go +235 -0
- package/govendor/github.com/tetratelabs/wazero/internal/wasm/binary/element.go +250 -0
- package/govendor/github.com/tetratelabs/wazero/internal/wasm/binary/errors.go +11 -0
- package/govendor/github.com/tetratelabs/wazero/internal/wasm/binary/export.go +32 -0
- package/govendor/github.com/tetratelabs/wazero/internal/wasm/binary/function.go +56 -0
- package/govendor/github.com/tetratelabs/wazero/internal/wasm/binary/global.go +50 -0
- package/govendor/github.com/tetratelabs/wazero/internal/wasm/binary/header.go +9 -0
- package/govendor/github.com/tetratelabs/wazero/internal/wasm/binary/import.go +68 -0
- package/govendor/github.com/tetratelabs/wazero/internal/wasm/binary/limits.go +47 -0
- package/govendor/github.com/tetratelabs/wazero/internal/wasm/binary/memory.go +42 -0
- package/govendor/github.com/tetratelabs/wazero/internal/wasm/binary/names.go +151 -0
- package/govendor/github.com/tetratelabs/wazero/internal/wasm/binary/section.go +309 -0
- package/govendor/github.com/tetratelabs/wazero/internal/wasm/binary/table.go +79 -0
- package/govendor/github.com/tetratelabs/wazero/internal/wasm/binary/value.go +97 -0
- package/govendor/github.com/tetratelabs/wazero/internal/wasm/const_expr.go +259 -0
- package/govendor/github.com/tetratelabs/wazero/internal/wasm/counts.go +53 -0
- package/govendor/github.com/tetratelabs/wazero/internal/wasm/engine.go +76 -0
- package/govendor/github.com/tetratelabs/wazero/internal/wasm/exception.go +9 -0
- package/govendor/github.com/tetratelabs/wazero/internal/wasm/func_validation.go +2804 -0
- package/govendor/github.com/tetratelabs/wazero/internal/wasm/function_definition.go +210 -0
- package/govendor/github.com/tetratelabs/wazero/internal/wasm/global.go +55 -0
- package/govendor/github.com/tetratelabs/wazero/internal/wasm/gofunc.go +279 -0
- package/govendor/github.com/tetratelabs/wazero/internal/wasm/host.go +179 -0
- package/govendor/github.com/tetratelabs/wazero/internal/wasm/instruction.go +1962 -0
- package/govendor/github.com/tetratelabs/wazero/internal/wasm/memory.go +475 -0
- package/govendor/github.com/tetratelabs/wazero/internal/wasm/memory_definition.go +128 -0
- package/govendor/github.com/tetratelabs/wazero/internal/wasm/module.go +1359 -0
- package/govendor/github.com/tetratelabs/wazero/internal/wasm/module_instance.go +251 -0
- package/govendor/github.com/tetratelabs/wazero/internal/wasm/module_instance_lookup.go +73 -0
- package/govendor/github.com/tetratelabs/wazero/internal/wasm/store.go +726 -0
- package/govendor/github.com/tetratelabs/wazero/internal/wasm/store_module_list.go +95 -0
- package/govendor/github.com/tetratelabs/wazero/internal/wasm/table.go +321 -0
- package/govendor/github.com/tetratelabs/wazero/internal/wasmdebug/debug.go +170 -0
- package/govendor/github.com/tetratelabs/wazero/internal/wasmdebug/dwarf.go +225 -0
- package/govendor/github.com/tetratelabs/wazero/internal/wasmruntime/errors.go +54 -0
- package/govendor/github.com/tetratelabs/wazero/netlify.toml +15 -0
- package/govendor/github.com/tetratelabs/wazero/runtime.go +399 -0
- package/govendor/github.com/tetratelabs/wazero/sys/clock.go +26 -0
- package/govendor/github.com/tetratelabs/wazero/sys/error.go +83 -0
- package/govendor/github.com/tetratelabs/wazero/sys/stat.go +105 -0
- package/govendor/github.com/tetratelabs/wazero/sys/stat_bsd.go +29 -0
- package/govendor/github.com/tetratelabs/wazero/sys/stat_posix.go +29 -0
- package/govendor/github.com/tetratelabs/wazero/sys/stat_unsupported.go +14 -0
- package/govendor/github.com/tetratelabs/wazero/sys/stat_windows.go +24 -0
- package/govendor/golang.org/x/crypto/LICENSE +27 -0
- package/govendor/golang.org/x/crypto/PATENTS +22 -0
- package/govendor/golang.org/x/crypto/cryptobyte/asn1/asn1.go +46 -0
- package/govendor/golang.org/x/crypto/cryptobyte/asn1.go +825 -0
- package/govendor/golang.org/x/crypto/cryptobyte/builder.go +350 -0
- package/govendor/golang.org/x/crypto/cryptobyte/string.go +183 -0
- package/govendor/golang.org/x/crypto/go.mod +3 -0
- package/govendor/golang.org/x/crypto/pbkdf2/pbkdf2.go +30 -0
- package/govendor/golang.org/x/net/LICENSE +27 -0
- package/govendor/golang.org/x/net/PATENTS +22 -0
- package/govendor/golang.org/x/net/dns/dnsmessage/message.go +2733 -0
- package/govendor/golang.org/x/net/dns/dnsmessage/svcb.go +333 -0
- package/govendor/golang.org/x/net/go.mod +3 -0
- package/govendor/golang.org/x/net/http/httpguts/guts.go +50 -0
- package/govendor/golang.org/x/net/http/httpguts/httplex.go +347 -0
- package/govendor/golang.org/x/net/http2/README.md +19 -0
- package/govendor/golang.org/x/net/http2/ascii.go +65 -0
- package/govendor/golang.org/x/net/http2/ciphers.go +641 -0
- package/govendor/golang.org/x/net/http2/client_conn_pool.go +301 -0
- package/govendor/golang.org/x/net/http2/client_priority_go126.go +20 -0
- package/govendor/golang.org/x/net/http2/client_priority_go127.go +13 -0
- package/govendor/golang.org/x/net/http2/clientconn.go +57 -0
- package/govendor/golang.org/x/net/http2/config.go +171 -0
- package/govendor/golang.org/x/net/http2/config_go125.go +15 -0
- package/govendor/golang.org/x/net/http2/config_go126.go +15 -0
- package/govendor/golang.org/x/net/http2/databuffer.go +146 -0
- package/govendor/golang.org/x/net/http2/errors.go +145 -0
- package/govendor/golang.org/x/net/http2/flow.go +175 -0
- package/govendor/golang.org/x/net/http2/frame.go +1887 -0
- package/govendor/golang.org/x/net/http2/gotrack.go +181 -0
- package/govendor/golang.org/x/net/http2/hpack/encode.go +244 -0
- package/govendor/golang.org/x/net/http2/hpack/hpack.go +522 -0
- package/govendor/golang.org/x/net/http2/hpack/huffman.go +226 -0
- package/govendor/golang.org/x/net/http2/hpack/static_table.go +188 -0
- package/govendor/golang.org/x/net/http2/hpack/tables.go +426 -0
- package/govendor/golang.org/x/net/http2/http2.go +415 -0
- package/govendor/golang.org/x/net/http2/pipe.go +184 -0
- package/govendor/golang.org/x/net/http2/server.go +3239 -0
- package/govendor/golang.org/x/net/http2/server_common.go +268 -0
- package/govendor/golang.org/x/net/http2/server_wrap.go +217 -0
- package/govendor/golang.org/x/net/http2/transport.go +3084 -0
- package/govendor/golang.org/x/net/http2/transport_common.go +516 -0
- package/govendor/golang.org/x/net/http2/transport_wrap.go +583 -0
- package/govendor/golang.org/x/net/http2/unencrypted.go +32 -0
- package/govendor/golang.org/x/net/http2/write.go +381 -0
- package/govendor/golang.org/x/net/http2/writesched.go +253 -0
- package/govendor/golang.org/x/net/http2/writesched_common.go +90 -0
- package/govendor/golang.org/x/net/http2/writesched_priority_rfc7540.go +420 -0
- package/govendor/golang.org/x/net/http2/writesched_priority_rfc9218.go +226 -0
- package/govendor/golang.org/x/net/http2/writesched_random.go +81 -0
- package/govendor/golang.org/x/net/http2/writesched_roundrobin.go +121 -0
- package/govendor/golang.org/x/net/idna/idna.go +884 -0
- package/govendor/golang.org/x/net/idna/punycode.go +220 -0
- package/govendor/golang.org/x/net/idna/tables15.0.0.go +5144 -0
- package/govendor/golang.org/x/net/idna/tables17.0.0.go +5302 -0
- package/govendor/golang.org/x/net/idna/trie.go +51 -0
- package/govendor/golang.org/x/net/idna/trieval.go +119 -0
- package/govendor/golang.org/x/net/internal/httpcommon/README.md +7 -0
- package/govendor/golang.org/x/net/internal/httpcommon/ascii.go +53 -0
- package/govendor/golang.org/x/net/internal/httpcommon/gzip.go +132 -0
- package/govendor/golang.org/x/net/internal/httpcommon/headermap.go +115 -0
- package/govendor/golang.org/x/net/internal/httpcommon/request.go +503 -0
- package/govendor/golang.org/x/net/internal/httpsfv/httpsfv.go +665 -0
- package/govendor/golang.org/x/sys/cpu/asm_aix_ppc64.s +17 -0
- package/govendor/golang.org/x/sys/cpu/asm_darwin_arm64_gc.s +12 -0
- package/govendor/golang.org/x/sys/cpu/asm_darwin_x86_gc.s +17 -0
- package/govendor/golang.org/x/sys/cpu/byteorder.go +66 -0
- package/govendor/golang.org/x/sys/cpu/cpu.go +357 -0
- package/govendor/golang.org/x/sys/cpu/cpu_aix.go +33 -0
- package/govendor/golang.org/x/sys/cpu/cpu_arm.go +73 -0
- package/govendor/golang.org/x/sys/cpu/cpu_arm64.go +191 -0
- package/govendor/golang.org/x/sys/cpu/cpu_arm64.s +35 -0
- package/govendor/golang.org/x/sys/cpu/cpu_darwin_arm64.go +67 -0
- package/govendor/golang.org/x/sys/cpu/cpu_darwin_arm64_other.go +31 -0
- package/govendor/golang.org/x/sys/cpu/cpu_darwin_x86.go +61 -0
- package/govendor/golang.org/x/sys/cpu/cpu_gc_arm64.go +12 -0
- package/govendor/golang.org/x/sys/cpu/cpu_gc_riscv64.go +11 -0
- package/govendor/golang.org/x/sys/cpu/cpu_gc_s390x.go +21 -0
- package/govendor/golang.org/x/sys/cpu/cpu_gc_x86.go +15 -0
- package/govendor/golang.org/x/sys/cpu/cpu_gc_x86.s +26 -0
- package/govendor/golang.org/x/sys/cpu/cpu_gccgo_arm64.go +12 -0
- package/govendor/golang.org/x/sys/cpu/cpu_gccgo_s390x.go +22 -0
- package/govendor/golang.org/x/sys/cpu/cpu_gccgo_x86.c +37 -0
- package/govendor/golang.org/x/sys/cpu/cpu_gccgo_x86.go +25 -0
- package/govendor/golang.org/x/sys/cpu/cpu_linux.go +15 -0
- package/govendor/golang.org/x/sys/cpu/cpu_linux_arm.go +39 -0
- package/govendor/golang.org/x/sys/cpu/cpu_linux_arm64.go +120 -0
- package/govendor/golang.org/x/sys/cpu/cpu_linux_loong64.go +22 -0
- package/govendor/golang.org/x/sys/cpu/cpu_linux_mips64x.go +22 -0
- package/govendor/golang.org/x/sys/cpu/cpu_linux_noinit.go +9 -0
- package/govendor/golang.org/x/sys/cpu/cpu_linux_ppc64x.go +34 -0
- package/govendor/golang.org/x/sys/cpu/cpu_linux_riscv64.go +173 -0
- package/govendor/golang.org/x/sys/cpu/cpu_linux_s390x.go +40 -0
- package/govendor/golang.org/x/sys/cpu/cpu_loong64.go +62 -0
- package/govendor/golang.org/x/sys/cpu/cpu_loong64.s +13 -0
- package/govendor/golang.org/x/sys/cpu/cpu_mips64x.go +15 -0
- package/govendor/golang.org/x/sys/cpu/cpu_mipsx.go +11 -0
- package/govendor/golang.org/x/sys/cpu/cpu_netbsd_amd64.go +47 -0
- package/govendor/golang.org/x/sys/cpu/cpu_netbsd_arm64.go +173 -0
- package/govendor/golang.org/x/sys/cpu/cpu_openbsd_arm64.go +65 -0
- package/govendor/golang.org/x/sys/cpu/cpu_openbsd_arm64.s +11 -0
- package/govendor/golang.org/x/sys/cpu/cpu_other_arm.go +9 -0
- package/govendor/golang.org/x/sys/cpu/cpu_other_arm64.go +11 -0
- package/govendor/golang.org/x/sys/cpu/cpu_other_mips64x.go +11 -0
- package/govendor/golang.org/x/sys/cpu/cpu_other_ppc64x.go +12 -0
- package/govendor/golang.org/x/sys/cpu/cpu_other_riscv64.go +11 -0
- package/govendor/golang.org/x/sys/cpu/cpu_other_x86.go +11 -0
- package/govendor/golang.org/x/sys/cpu/cpu_ppc64x.go +16 -0
- package/govendor/golang.org/x/sys/cpu/cpu_riscv64.go +33 -0
- package/govendor/golang.org/x/sys/cpu/cpu_riscv64.s +17 -0
- package/govendor/golang.org/x/sys/cpu/cpu_s390x.go +172 -0
- package/govendor/golang.org/x/sys/cpu/cpu_s390x.s +57 -0
- package/govendor/golang.org/x/sys/cpu/cpu_sparc64.go +12 -0
- package/govendor/golang.org/x/sys/cpu/cpu_wasm.go +17 -0
- package/govendor/golang.org/x/sys/cpu/cpu_windows.go +26 -0
- package/govendor/golang.org/x/sys/cpu/cpu_windows_arm64.go +38 -0
- package/govendor/golang.org/x/sys/cpu/cpu_x86.go +259 -0
- package/govendor/golang.org/x/sys/cpu/cpu_zos.go +10 -0
- package/govendor/golang.org/x/sys/cpu/cpu_zos_s390x.go +25 -0
- package/govendor/golang.org/x/sys/cpu/endian_big.go +10 -0
- package/govendor/golang.org/x/sys/cpu/endian_little.go +10 -0
- package/govendor/golang.org/x/sys/cpu/hwcap_linux.go +71 -0
- package/govendor/golang.org/x/sys/cpu/parse.go +55 -0
- package/govendor/golang.org/x/sys/cpu/proc_cpuinfo_linux.go +53 -0
- package/govendor/golang.org/x/sys/cpu/runtime_auxv.go +16 -0
- package/govendor/golang.org/x/sys/cpu/runtime_auxv_go121.go +18 -0
- package/govendor/golang.org/x/sys/cpu/syscall_aix_gccgo.go +26 -0
- package/govendor/golang.org/x/sys/cpu/syscall_aix_ppc64_gc.go +35 -0
- package/govendor/golang.org/x/sys/cpu/syscall_darwin_arm64_gc.go +54 -0
- package/govendor/golang.org/x/sys/cpu/syscall_darwin_x86_gc.go +98 -0
- package/govendor/golang.org/x/sys/cpu/zcpu_windows.go +48 -0
- package/govendor/golang.org/x/sys/unix/asm_linux_ppc64x.s +20 -0
- package/govendor/golang.org/x/sys/unix/asm_linux_sparc64.s +48 -0
- package/govendor/golang.org/x/sys/unix/asm_zos_s390x.s +4 -186
- package/govendor/golang.org/x/sys/unix/bpxsvc_zos.s +9 -9
- package/govendor/golang.org/x/sys/unix/fcntl_darwin.go +2 -2
- package/govendor/golang.org/x/sys/unix/mkerrors.sh +1 -1
- package/govendor/golang.org/x/sys/unix/ptrace_zos.go +537 -0
- package/govendor/golang.org/x/sys/unix/ptrace_zos_singlestep.go +425 -0
- package/govendor/golang.org/x/sys/unix/syscall_darwin.go +1 -0
- package/govendor/golang.org/x/sys/unix/syscall_freebsd.go +9 -0
- package/govendor/golang.org/x/sys/unix/syscall_linux.go +20 -1
- package/govendor/golang.org/x/sys/unix/syscall_linux_sparc64.go +3 -3
- package/govendor/golang.org/x/sys/unix/syscall_netbsd.go +9 -0
- package/govendor/golang.org/x/sys/unix/syscall_unix_gc.go +1 -1
- package/govendor/golang.org/x/sys/unix/syscall_zos_s390x.go +0 -4
- package/govendor/golang.org/x/sys/unix/xattr_bsd.go +12 -12
- package/govendor/golang.org/x/sys/unix/zerrors_darwin_amd64.go +6 -0
- package/govendor/golang.org/x/sys/unix/zerrors_darwin_arm64.go +6 -0
- package/govendor/golang.org/x/sys/unix/zerrors_linux.go +32 -0
- package/govendor/golang.org/x/sys/unix/zerrors_linux_386.go +2 -0
- package/govendor/golang.org/x/sys/unix/zerrors_linux_amd64.go +2 -0
- package/govendor/golang.org/x/sys/unix/zerrors_linux_arm.go +2 -0
- package/govendor/golang.org/x/sys/unix/zerrors_linux_arm64.go +2 -0
- package/govendor/golang.org/x/sys/unix/zerrors_linux_loong64.go +2 -0
- package/govendor/golang.org/x/sys/unix/zerrors_linux_mips.go +2 -0
- package/govendor/golang.org/x/sys/unix/zerrors_linux_mips64.go +2 -0
- package/govendor/golang.org/x/sys/unix/zerrors_linux_mips64le.go +2 -0
- package/govendor/golang.org/x/sys/unix/zerrors_linux_mipsle.go +2 -0
- package/govendor/golang.org/x/sys/unix/zerrors_linux_ppc.go +2 -0
- package/govendor/golang.org/x/sys/unix/zerrors_linux_ppc64.go +2 -0
- package/govendor/golang.org/x/sys/unix/zerrors_linux_ppc64le.go +2 -0
- package/govendor/golang.org/x/sys/unix/zerrors_linux_riscv64.go +2 -0
- package/govendor/golang.org/x/sys/unix/zerrors_linux_s390x.go +2 -0
- package/govendor/golang.org/x/sys/unix/zerrors_linux_sparc64.go +2 -0
- package/govendor/golang.org/x/sys/unix/zsyscall_darwin_amd64.go +11 -0
- package/govendor/golang.org/x/sys/unix/zsyscall_darwin_arm64.go +11 -0
- package/govendor/golang.org/x/sys/unix/zsyscall_freebsd_386.go +159 -0
- package/govendor/golang.org/x/sys/unix/zsyscall_freebsd_amd64.go +159 -0
- package/govendor/golang.org/x/sys/unix/zsyscall_freebsd_arm.go +159 -0
- package/govendor/golang.org/x/sys/unix/zsyscall_freebsd_arm64.go +159 -0
- package/govendor/golang.org/x/sys/unix/zsyscall_freebsd_riscv64.go +159 -0
- package/govendor/golang.org/x/sys/unix/zsyscall_linux.go +28 -1
- package/govendor/golang.org/x/sys/unix/zsyscall_linux_sparc64.go +3 -3
- package/govendor/golang.org/x/sys/unix/zsyscall_netbsd_386.go +159 -0
- package/govendor/golang.org/x/sys/unix/zsyscall_netbsd_amd64.go +159 -0
- package/govendor/golang.org/x/sys/unix/zsyscall_netbsd_arm.go +159 -0
- package/govendor/golang.org/x/sys/unix/zsyscall_netbsd_arm64.go +159 -0
- package/govendor/golang.org/x/sys/unix/zsysnum_zos_s390x.go +13 -0
- package/govendor/golang.org/x/sys/unix/ztypes_linux.go +65 -36
- package/govendor/golang.org/x/sys/unix/ztypes_linux_386.go +6 -0
- package/govendor/golang.org/x/sys/unix/ztypes_linux_amd64.go +7 -0
- package/govendor/golang.org/x/sys/unix/ztypes_linux_arm.go +6 -0
- package/govendor/golang.org/x/sys/unix/ztypes_linux_arm64.go +7 -0
- package/govendor/golang.org/x/sys/unix/ztypes_linux_loong64.go +7 -0
- package/govendor/golang.org/x/sys/unix/ztypes_linux_mips.go +6 -0
- package/govendor/golang.org/x/sys/unix/ztypes_linux_mips64.go +7 -0
- package/govendor/golang.org/x/sys/unix/ztypes_linux_mips64le.go +7 -0
- package/govendor/golang.org/x/sys/unix/ztypes_linux_mipsle.go +6 -0
- package/govendor/golang.org/x/sys/unix/ztypes_linux_ppc.go +6 -0
- package/govendor/golang.org/x/sys/unix/ztypes_linux_ppc64.go +7 -0
- package/govendor/golang.org/x/sys/unix/ztypes_linux_ppc64le.go +7 -0
- package/govendor/golang.org/x/sys/unix/ztypes_linux_riscv64.go +7 -0
- package/govendor/golang.org/x/sys/unix/ztypes_linux_s390x.go +7 -0
- package/govendor/golang.org/x/sys/unix/ztypes_linux_sparc64.go +10 -5
- package/govendor/golang.org/x/sys/unix/ztypes_zos_s390x.go +78 -0
- package/govendor/golang.org/x/text/.gitattributes +10 -0
- package/govendor/golang.org/x/text/CONTRIBUTING.md +26 -0
- package/govendor/golang.org/x/text/LICENSE +27 -0
- package/govendor/golang.org/x/text/PATENTS +22 -0
- package/govendor/golang.org/x/text/README.md +81 -0
- package/govendor/golang.org/x/text/cases/cases.go +162 -0
- package/govendor/golang.org/x/text/cases/context.go +376 -0
- package/govendor/golang.org/x/text/cases/fold.go +34 -0
- package/govendor/golang.org/x/text/cases/icu.go +61 -0
- package/govendor/golang.org/x/text/cases/info.go +82 -0
- package/govendor/golang.org/x/text/cases/map.go +816 -0
- package/govendor/golang.org/x/text/cases/tables15.0.0.go +2527 -0
- package/govendor/golang.org/x/text/cases/tables17.0.0.go +2642 -0
- package/govendor/golang.org/x/text/cases/trieval.go +217 -0
- package/govendor/golang.org/x/text/codereview.cfg +1 -0
- package/govendor/golang.org/x/text/collate/collate.go +402 -0
- package/govendor/golang.org/x/text/collate/index.go +32 -0
- package/govendor/golang.org/x/text/collate/option.go +239 -0
- package/govendor/golang.org/x/text/collate/sort.go +81 -0
- package/govendor/golang.org/x/text/collate/tables.go +73789 -0
- package/govendor/golang.org/x/text/currency/common.go +67 -0
- package/govendor/golang.org/x/text/currency/currency.go +185 -0
- package/govendor/golang.org/x/text/currency/format.go +220 -0
- package/govendor/golang.org/x/text/currency/query.go +152 -0
- package/govendor/golang.org/x/text/currency/tables.go +2629 -0
- package/govendor/golang.org/x/text/doc.go +16 -0
- package/govendor/golang.org/x/text/feature/plural/common.go +70 -0
- package/govendor/golang.org/x/text/feature/plural/message.go +244 -0
- package/govendor/golang.org/x/text/feature/plural/plural.go +262 -0
- package/govendor/golang.org/x/text/feature/plural/tables.go +552 -0
- package/govendor/golang.org/x/text/go.mod +3 -0
- package/govendor/golang.org/x/text/internal/catmsg/catmsg.go +417 -0
- package/govendor/golang.org/x/text/internal/catmsg/codec.go +407 -0
- package/govendor/golang.org/x/text/internal/catmsg/varint.go +62 -0
- package/govendor/golang.org/x/text/internal/colltab/collelem.go +374 -0
- package/govendor/golang.org/x/text/internal/colltab/colltab.go +105 -0
- package/govendor/golang.org/x/text/internal/colltab/contract.go +145 -0
- package/govendor/golang.org/x/text/internal/colltab/iter.go +178 -0
- package/govendor/golang.org/x/text/internal/colltab/numeric.go +236 -0
- package/govendor/golang.org/x/text/internal/colltab/table.go +275 -0
- package/govendor/golang.org/x/text/internal/colltab/trie.go +159 -0
- package/govendor/golang.org/x/text/internal/colltab/weighter.go +31 -0
- package/govendor/golang.org/x/text/internal/format/format.go +41 -0
- package/govendor/golang.org/x/text/internal/format/parser.go +358 -0
- package/govendor/golang.org/x/text/internal/internal.go +49 -0
- package/govendor/golang.org/x/text/internal/language/common.go +16 -0
- package/govendor/golang.org/x/text/internal/language/compact/compact.go +61 -0
- package/govendor/golang.org/x/text/internal/language/compact/language.go +260 -0
- package/govendor/golang.org/x/text/internal/language/compact/parents.go +120 -0
- package/govendor/golang.org/x/text/internal/language/compact/tables.go +1015 -0
- package/govendor/golang.org/x/text/internal/language/compact/tags.go +91 -0
- package/govendor/golang.org/x/text/internal/language/compact.go +29 -0
- package/govendor/golang.org/x/text/internal/language/compose.go +167 -0
- package/govendor/golang.org/x/text/internal/language/coverage.go +28 -0
- package/govendor/golang.org/x/text/internal/language/language.go +627 -0
- package/govendor/golang.org/x/text/internal/language/lookup.go +412 -0
- package/govendor/golang.org/x/text/internal/language/match.go +226 -0
- package/govendor/golang.org/x/text/internal/language/parse.go +608 -0
- package/govendor/golang.org/x/text/internal/language/tables.go +3494 -0
- package/govendor/golang.org/x/text/internal/language/tags.go +48 -0
- package/govendor/golang.org/x/text/internal/match.go +67 -0
- package/govendor/golang.org/x/text/internal/number/common.go +55 -0
- package/govendor/golang.org/x/text/internal/number/decimal.go +500 -0
- package/govendor/golang.org/x/text/internal/number/format.go +533 -0
- package/govendor/golang.org/x/text/internal/number/number.go +152 -0
- package/govendor/golang.org/x/text/internal/number/pattern.go +485 -0
- package/govendor/golang.org/x/text/internal/number/roundingmode_string.go +30 -0
- package/govendor/golang.org/x/text/internal/number/tables.go +1219 -0
- package/govendor/golang.org/x/text/internal/stringset/set.go +86 -0
- package/govendor/golang.org/x/text/internal/tag/tag.go +100 -0
- package/govendor/golang.org/x/text/language/coverage.go +187 -0
- package/govendor/golang.org/x/text/language/display/dict.go +92 -0
- package/govendor/golang.org/x/text/language/display/display.go +420 -0
- package/govendor/golang.org/x/text/language/display/lookup.go +253 -0
- package/govendor/golang.org/x/text/language/display/tables.go +53114 -0
- package/govendor/golang.org/x/text/language/doc.go +98 -0
- package/govendor/golang.org/x/text/language/language.go +605 -0
- package/govendor/golang.org/x/text/language/match.go +735 -0
- package/govendor/golang.org/x/text/language/parse.go +258 -0
- package/govendor/golang.org/x/text/language/tables.go +298 -0
- package/govendor/golang.org/x/text/language/tags.go +145 -0
- package/govendor/golang.org/x/text/message/catalog/catalog.go +365 -0
- package/govendor/golang.org/x/text/message/catalog/dict.go +133 -0
- package/govendor/golang.org/x/text/secure/bidirule/bidirule.go +346 -0
- package/govendor/golang.org/x/text/transform/transform.go +709 -0
- package/govendor/golang.org/x/text/unicode/bidi/bidi.go +359 -0
- package/govendor/golang.org/x/text/unicode/bidi/bracket.go +335 -0
- package/govendor/golang.org/x/text/unicode/bidi/core.go +1050 -0
- package/govendor/golang.org/x/text/unicode/bidi/prop.go +206 -0
- package/govendor/golang.org/x/text/unicode/bidi/tables15.0.0.go +2042 -0
- package/govendor/golang.org/x/text/unicode/bidi/tables17.0.0.go +2135 -0
- package/govendor/golang.org/x/text/unicode/bidi/trieval.go +48 -0
- package/govendor/golang.org/x/text/unicode/norm/composition.go +524 -0
- package/govendor/golang.org/x/text/unicode/norm/forminfo.go +296 -0
- package/govendor/golang.org/x/text/unicode/norm/input.go +109 -0
- package/govendor/golang.org/x/text/unicode/norm/iter.go +454 -0
- package/govendor/golang.org/x/text/unicode/norm/normalize.go +613 -0
- package/govendor/golang.org/x/text/unicode/norm/readwriter.go +125 -0
- package/govendor/golang.org/x/text/unicode/norm/tables15.0.0.go +7907 -0
- package/govendor/golang.org/x/text/unicode/norm/tables17.0.0.go +8104 -0
- package/govendor/golang.org/x/text/unicode/norm/transform.go +89 -0
- package/govendor/golang.org/x/text/unicode/norm/trie.go +54 -0
- package/govendor/manifest.txt +11 -2
- package/package.json +2 -1
- package/performance/README.md +87 -35
- package/performance/arr-for-of.ts +29 -0
- package/performance/arr-push-local.ts +39 -0
- package/performance/arr-slice-concat.ts +24 -0
- package/performance/async-promise-all.ts +38 -0
- package/performance/bigint-arith.ts +36 -0
- package/performance/buf-encode.ts +29 -0
- package/performance/class-accessors.ts +48 -0
- package/performance/date-format.ts +18 -0
- package/performance/fn-rest-spread.ts +43 -0
- package/performance/gen-next.ts +32 -0
- package/performance/iter-helpers.ts +38 -0
- package/performance/json-any.ts +30 -0
- package/performance/map-iterate.ts +26 -0
- package/performance/map-number-keys.ts +37 -0
- package/performance/num-hash-imul.ts +36 -0
- package/performance/obj-keys-entries.ts +44 -0
- package/performance/obj-optional-chain.ts +42 -0
- package/performance/obj-record-dict.ts +24 -0
- package/performance/obj-spread-destructure.ts +40 -0
- package/performance/regex-replace-fn.ts +24 -0
- package/performance/str-methods.ts +24 -0
- package/performance/str-switch-compare.ts +50 -0
- package/performance/str-unicode.ts +25 -0
- package/performance/typed-array.ts +30 -0
- package/performance/union-typeof.ts +40 -0
- package/performance/wasm-call.ts +63 -0
- package/types/typenative-npm.d.ts +836 -69
- package/types/typenative.d.ts +1224 -23
- package/govendor/golang.org/x/sys/unix/syscall_unix_gc_ppc64x.go +0 -22
- package/govendor/golang.org/x/sys/windows/types_windows_arm.go +0 -35
|
@@ -0,0 +1,2227 @@
|
|
|
1
|
+
package arm64
|
|
2
|
+
|
|
3
|
+
// Files prefixed as lower_instr** do the instruction selection, meaning that lowering SSA level instructions
|
|
4
|
+
// into machine specific instructions.
|
|
5
|
+
//
|
|
6
|
+
// Importantly, what the lower** functions does includes tree-matching; find the pattern from the given instruction tree,
|
|
7
|
+
// and merge the multiple instructions if possible. It can be considered as "N:1" instruction selection.
|
|
8
|
+
|
|
9
|
+
import (
|
|
10
|
+
"fmt"
|
|
11
|
+
"math"
|
|
12
|
+
|
|
13
|
+
"github.com/tetratelabs/wazero/internal/engine/wazevo/backend/regalloc"
|
|
14
|
+
"github.com/tetratelabs/wazero/internal/engine/wazevo/ssa"
|
|
15
|
+
"github.com/tetratelabs/wazero/internal/engine/wazevo/wazevoapi"
|
|
16
|
+
)
|
|
17
|
+
|
|
18
|
+
// LowerSingleBranch implements backend.Machine.
|
|
19
|
+
func (m *machine) LowerSingleBranch(br *ssa.Instruction) {
|
|
20
|
+
switch br.Opcode() {
|
|
21
|
+
case ssa.OpcodeJump:
|
|
22
|
+
_, _, targetBlkID := br.BranchData()
|
|
23
|
+
if br.IsFallthroughJump() {
|
|
24
|
+
return
|
|
25
|
+
}
|
|
26
|
+
b := m.allocateInstr()
|
|
27
|
+
targetBlk := m.compiler.SSABuilder().BasicBlock(targetBlkID)
|
|
28
|
+
if targetBlk.ReturnBlock() {
|
|
29
|
+
b.asRet()
|
|
30
|
+
} else {
|
|
31
|
+
b.asBr(ssaBlockLabel(targetBlk))
|
|
32
|
+
}
|
|
33
|
+
m.insert(b)
|
|
34
|
+
case ssa.OpcodeBrTable:
|
|
35
|
+
m.lowerBrTable(br)
|
|
36
|
+
default:
|
|
37
|
+
panic("BUG: unexpected branch opcode" + br.Opcode().String())
|
|
38
|
+
}
|
|
39
|
+
}
|
|
40
|
+
|
|
41
|
+
func (m *machine) lowerBrTable(i *ssa.Instruction) {
|
|
42
|
+
index, targetBlockIDs := i.BrTableData()
|
|
43
|
+
targetBlockCount := len(targetBlockIDs.View())
|
|
44
|
+
indexOperand := m.getOperand_NR(m.compiler.ValueDefinition(index), extModeNone)
|
|
45
|
+
|
|
46
|
+
// Firstly, we have to do the bounds check of the index, and
|
|
47
|
+
// set it to the default target (sitting at the end of the list) if it's out of bounds.
|
|
48
|
+
|
|
49
|
+
// mov maxIndexReg #maximum_index
|
|
50
|
+
// subs wzr, index, maxIndexReg
|
|
51
|
+
// csel adjustedIndex, maxIndexReg, index, hs ;; if index is higher or equal than maxIndexReg.
|
|
52
|
+
maxIndexReg := m.compiler.AllocateVReg(ssa.TypeI32)
|
|
53
|
+
m.lowerConstantI32(maxIndexReg, int32(targetBlockCount-1))
|
|
54
|
+
subs := m.allocateInstr()
|
|
55
|
+
subs.asALU(aluOpSubS, xzrVReg, indexOperand, operandNR(maxIndexReg), false)
|
|
56
|
+
m.insert(subs)
|
|
57
|
+
csel := m.allocateInstr()
|
|
58
|
+
adjustedIndex := m.compiler.AllocateVReg(ssa.TypeI32)
|
|
59
|
+
csel.asCSel(adjustedIndex, operandNR(maxIndexReg), indexOperand, hs, false)
|
|
60
|
+
m.insert(csel)
|
|
61
|
+
|
|
62
|
+
brSequence := m.allocateInstr()
|
|
63
|
+
|
|
64
|
+
tableIndex := m.addJmpTableTarget(targetBlockIDs)
|
|
65
|
+
brSequence.asBrTableSequence(adjustedIndex, tableIndex, targetBlockCount)
|
|
66
|
+
m.insert(brSequence)
|
|
67
|
+
}
|
|
68
|
+
|
|
69
|
+
// LowerConditionalBranch implements backend.Machine.
|
|
70
|
+
func (m *machine) LowerConditionalBranch(b *ssa.Instruction) {
|
|
71
|
+
cval, args, targetBlkID := b.BranchData()
|
|
72
|
+
if len(args) > 0 {
|
|
73
|
+
panic(fmt.Sprintf(
|
|
74
|
+
"conditional branch shouldn't have args; likely a bug in critical edge splitting: from %s to %s",
|
|
75
|
+
m.currentLabelPos.sb,
|
|
76
|
+
targetBlkID,
|
|
77
|
+
))
|
|
78
|
+
}
|
|
79
|
+
|
|
80
|
+
targetBlk := m.compiler.SSABuilder().BasicBlock(targetBlkID)
|
|
81
|
+
target := ssaBlockLabel(targetBlk)
|
|
82
|
+
cvalDef := m.compiler.ValueDefinition(cval)
|
|
83
|
+
|
|
84
|
+
switch {
|
|
85
|
+
case m.compiler.MatchInstr(cvalDef, ssa.OpcodeIcmp): // This case, we can use the ALU flag set by SUBS instruction.
|
|
86
|
+
cvalInstr := cvalDef.Instr
|
|
87
|
+
x, y, c := cvalInstr.IcmpData()
|
|
88
|
+
cc, signed := condFlagFromSSAIntegerCmpCond(c), c.Signed()
|
|
89
|
+
if b.Opcode() == ssa.OpcodeBrz {
|
|
90
|
+
cc = cc.invert()
|
|
91
|
+
}
|
|
92
|
+
|
|
93
|
+
if !m.tryLowerBandToFlag(x, y) {
|
|
94
|
+
m.lowerIcmpToFlag(x, y, signed)
|
|
95
|
+
}
|
|
96
|
+
cbr := m.allocateInstr()
|
|
97
|
+
cbr.asCondBr(cc.asCond(), target, false /* ignored */)
|
|
98
|
+
m.insert(cbr)
|
|
99
|
+
cvalDef.Instr.MarkLowered()
|
|
100
|
+
case m.compiler.MatchInstr(cvalDef, ssa.OpcodeFcmp): // This case we can use the Fpu flag directly.
|
|
101
|
+
cvalInstr := cvalDef.Instr
|
|
102
|
+
x, y, c := cvalInstr.FcmpData()
|
|
103
|
+
cc := condFlagFromSSAFloatCmpCond(c)
|
|
104
|
+
if b.Opcode() == ssa.OpcodeBrz {
|
|
105
|
+
cc = cc.invert()
|
|
106
|
+
}
|
|
107
|
+
m.lowerFcmpToFlag(x, y)
|
|
108
|
+
cbr := m.allocateInstr()
|
|
109
|
+
cbr.asCondBr(cc.asCond(), target, false /* ignored */)
|
|
110
|
+
m.insert(cbr)
|
|
111
|
+
cvalDef.Instr.MarkLowered()
|
|
112
|
+
default:
|
|
113
|
+
rn := m.getOperand_NR(cvalDef, extModeNone)
|
|
114
|
+
var c cond
|
|
115
|
+
if b.Opcode() == ssa.OpcodeBrz {
|
|
116
|
+
c = registerAsRegZeroCond(rn.nr())
|
|
117
|
+
} else {
|
|
118
|
+
c = registerAsRegNotZeroCond(rn.nr())
|
|
119
|
+
}
|
|
120
|
+
cbr := m.allocateInstr()
|
|
121
|
+
cbr.asCondBr(c, target, false)
|
|
122
|
+
m.insert(cbr)
|
|
123
|
+
}
|
|
124
|
+
}
|
|
125
|
+
|
|
126
|
+
func (m *machine) tryLowerBandToFlag(x, y ssa.Value) (ok bool) {
|
|
127
|
+
xx := m.compiler.ValueDefinition(x)
|
|
128
|
+
yy := m.compiler.ValueDefinition(y)
|
|
129
|
+
if xx.IsFromInstr() && xx.Instr.Constant() && xx.Instr.ConstantVal() == 0 {
|
|
130
|
+
if m.compiler.MatchInstr(yy, ssa.OpcodeBand) {
|
|
131
|
+
bandInstr := yy.Instr
|
|
132
|
+
m.lowerBitwiseAluOp(bandInstr, aluOpAnds, true)
|
|
133
|
+
ok = true
|
|
134
|
+
bandInstr.MarkLowered()
|
|
135
|
+
return
|
|
136
|
+
}
|
|
137
|
+
}
|
|
138
|
+
|
|
139
|
+
if yy.IsFromInstr() && yy.Instr.Constant() && yy.Instr.ConstantVal() == 0 {
|
|
140
|
+
if m.compiler.MatchInstr(xx, ssa.OpcodeBand) {
|
|
141
|
+
bandInstr := xx.Instr
|
|
142
|
+
m.lowerBitwiseAluOp(bandInstr, aluOpAnds, true)
|
|
143
|
+
ok = true
|
|
144
|
+
bandInstr.MarkLowered()
|
|
145
|
+
return
|
|
146
|
+
}
|
|
147
|
+
}
|
|
148
|
+
return
|
|
149
|
+
}
|
|
150
|
+
|
|
151
|
+
// LowerInstr implements backend.Machine.
|
|
152
|
+
func (m *machine) LowerInstr(instr *ssa.Instruction) {
|
|
153
|
+
if l := instr.SourceOffset(); l.Valid() {
|
|
154
|
+
info := m.allocateInstr().asEmitSourceOffsetInfo(l)
|
|
155
|
+
m.insert(info)
|
|
156
|
+
}
|
|
157
|
+
|
|
158
|
+
switch op := instr.Opcode(); op {
|
|
159
|
+
case ssa.OpcodeBrz, ssa.OpcodeBrnz, ssa.OpcodeJump, ssa.OpcodeBrTable:
|
|
160
|
+
panic("BUG: branching instructions are handled by LowerBranches")
|
|
161
|
+
case ssa.OpcodeReturn:
|
|
162
|
+
panic("BUG: return must be handled by backend.Compiler")
|
|
163
|
+
case ssa.OpcodeIadd, ssa.OpcodeIsub:
|
|
164
|
+
m.lowerSubOrAdd(instr, op == ssa.OpcodeIadd)
|
|
165
|
+
case ssa.OpcodeFadd, ssa.OpcodeFsub, ssa.OpcodeFmul, ssa.OpcodeFdiv, ssa.OpcodeFmax, ssa.OpcodeFmin:
|
|
166
|
+
m.lowerFpuBinOp(instr)
|
|
167
|
+
case ssa.OpcodeIconst, ssa.OpcodeF32const, ssa.OpcodeF64const: // Constant instructions are inlined.
|
|
168
|
+
case ssa.OpcodeExitWithCode:
|
|
169
|
+
execCtx, code := instr.ExitWithCodeData()
|
|
170
|
+
m.lowerExitWithCode(m.compiler.VRegOf(execCtx), code)
|
|
171
|
+
case ssa.OpcodeExitIfTrueWithCode:
|
|
172
|
+
execCtx, c, code := instr.ExitIfTrueWithCodeData()
|
|
173
|
+
m.lowerExitIfTrueWithCode(m.compiler.VRegOf(execCtx), c, code)
|
|
174
|
+
case ssa.OpcodeStore, ssa.OpcodeIstore8, ssa.OpcodeIstore16, ssa.OpcodeIstore32:
|
|
175
|
+
m.lowerStore(instr)
|
|
176
|
+
case ssa.OpcodeLoad:
|
|
177
|
+
dst := instr.Return()
|
|
178
|
+
ptr, offset, typ := instr.LoadData()
|
|
179
|
+
m.lowerLoad(ptr, offset, typ, dst)
|
|
180
|
+
case ssa.OpcodeVZeroExtLoad:
|
|
181
|
+
dst := instr.Return()
|
|
182
|
+
ptr, offset, typ := instr.VZeroExtLoadData()
|
|
183
|
+
m.lowerLoad(ptr, offset, typ, dst)
|
|
184
|
+
case ssa.OpcodeUload8, ssa.OpcodeUload16, ssa.OpcodeUload32, ssa.OpcodeSload8, ssa.OpcodeSload16, ssa.OpcodeSload32:
|
|
185
|
+
ptr, offset, _ := instr.LoadData()
|
|
186
|
+
ret := m.compiler.VRegOf(instr.Return())
|
|
187
|
+
m.lowerExtLoad(op, ptr, offset, ret)
|
|
188
|
+
case ssa.OpcodeCall, ssa.OpcodeCallIndirect:
|
|
189
|
+
m.lowerCall(instr)
|
|
190
|
+
case ssa.OpcodeIcmp:
|
|
191
|
+
m.lowerIcmp(instr)
|
|
192
|
+
case ssa.OpcodeVIcmp:
|
|
193
|
+
m.lowerVIcmp(instr)
|
|
194
|
+
case ssa.OpcodeVFcmp:
|
|
195
|
+
m.lowerVFcmp(instr)
|
|
196
|
+
case ssa.OpcodeVCeil:
|
|
197
|
+
m.lowerVecMisc(vecOpFrintp, instr)
|
|
198
|
+
case ssa.OpcodeVFloor:
|
|
199
|
+
m.lowerVecMisc(vecOpFrintm, instr)
|
|
200
|
+
case ssa.OpcodeVTrunc:
|
|
201
|
+
m.lowerVecMisc(vecOpFrintz, instr)
|
|
202
|
+
case ssa.OpcodeVNearest:
|
|
203
|
+
m.lowerVecMisc(vecOpFrintn, instr)
|
|
204
|
+
case ssa.OpcodeVMaxPseudo:
|
|
205
|
+
m.lowerVMinMaxPseudo(instr, true)
|
|
206
|
+
case ssa.OpcodeVMinPseudo:
|
|
207
|
+
m.lowerVMinMaxPseudo(instr, false)
|
|
208
|
+
case ssa.OpcodeBand:
|
|
209
|
+
m.lowerBitwiseAluOp(instr, aluOpAnd, false)
|
|
210
|
+
case ssa.OpcodeBor:
|
|
211
|
+
m.lowerBitwiseAluOp(instr, aluOpOrr, false)
|
|
212
|
+
case ssa.OpcodeBxor:
|
|
213
|
+
m.lowerBitwiseAluOp(instr, aluOpEor, false)
|
|
214
|
+
case ssa.OpcodeIshl:
|
|
215
|
+
m.lowerShifts(instr, extModeNone, aluOpLsl)
|
|
216
|
+
case ssa.OpcodeSshr:
|
|
217
|
+
if instr.Return().Type().Bits() == 64 {
|
|
218
|
+
m.lowerShifts(instr, extModeSignExtend64, aluOpAsr)
|
|
219
|
+
} else {
|
|
220
|
+
m.lowerShifts(instr, extModeSignExtend32, aluOpAsr)
|
|
221
|
+
}
|
|
222
|
+
case ssa.OpcodeUshr:
|
|
223
|
+
if instr.Return().Type().Bits() == 64 {
|
|
224
|
+
m.lowerShifts(instr, extModeZeroExtend64, aluOpLsr)
|
|
225
|
+
} else {
|
|
226
|
+
m.lowerShifts(instr, extModeZeroExtend32, aluOpLsr)
|
|
227
|
+
}
|
|
228
|
+
case ssa.OpcodeRotl:
|
|
229
|
+
m.lowerRotl(instr)
|
|
230
|
+
case ssa.OpcodeRotr:
|
|
231
|
+
m.lowerRotr(instr)
|
|
232
|
+
case ssa.OpcodeSExtend, ssa.OpcodeUExtend:
|
|
233
|
+
from, to, signed := instr.ExtendData()
|
|
234
|
+
m.lowerExtend(instr.Arg(), instr.Return(), from, to, signed)
|
|
235
|
+
case ssa.OpcodeFcmp:
|
|
236
|
+
x, y, c := instr.FcmpData()
|
|
237
|
+
m.lowerFcmp(x, y, instr.Return(), c)
|
|
238
|
+
case ssa.OpcodeImul:
|
|
239
|
+
x, y := instr.Arg2()
|
|
240
|
+
result := instr.Return()
|
|
241
|
+
m.lowerImul(x, y, result)
|
|
242
|
+
case ssa.OpcodeUndefined:
|
|
243
|
+
undef := m.allocateInstr()
|
|
244
|
+
undef.asUDF()
|
|
245
|
+
m.insert(undef)
|
|
246
|
+
case ssa.OpcodeSelect:
|
|
247
|
+
c, x, y := instr.SelectData()
|
|
248
|
+
if x.Type() == ssa.TypeV128 {
|
|
249
|
+
rc := m.getOperand_NR(m.compiler.ValueDefinition(c), extModeNone)
|
|
250
|
+
rn := m.getOperand_NR(m.compiler.ValueDefinition(x), extModeNone)
|
|
251
|
+
rm := m.getOperand_NR(m.compiler.ValueDefinition(y), extModeNone)
|
|
252
|
+
rd := m.compiler.VRegOf(instr.Return())
|
|
253
|
+
m.lowerSelectVec(rc, rn, rm, rd)
|
|
254
|
+
} else {
|
|
255
|
+
m.lowerSelect(c, x, y, instr.Return())
|
|
256
|
+
}
|
|
257
|
+
case ssa.OpcodeClz:
|
|
258
|
+
x := instr.Arg()
|
|
259
|
+
result := instr.Return()
|
|
260
|
+
m.lowerClz(x, result)
|
|
261
|
+
case ssa.OpcodeCtz:
|
|
262
|
+
x := instr.Arg()
|
|
263
|
+
result := instr.Return()
|
|
264
|
+
m.lowerCtz(x, result)
|
|
265
|
+
case ssa.OpcodePopcnt:
|
|
266
|
+
x := instr.Arg()
|
|
267
|
+
result := instr.Return()
|
|
268
|
+
m.lowerPopcnt(x, result)
|
|
269
|
+
case ssa.OpcodeFcvtToSint, ssa.OpcodeFcvtToSintSat:
|
|
270
|
+
x, ctx := instr.Arg2()
|
|
271
|
+
result := instr.Return()
|
|
272
|
+
rn := m.getOperand_NR(m.compiler.ValueDefinition(x), extModeNone)
|
|
273
|
+
rd := m.compiler.VRegOf(result)
|
|
274
|
+
ctxVReg := m.compiler.VRegOf(ctx)
|
|
275
|
+
m.lowerFpuToInt(rd, rn, ctxVReg, true, x.Type() == ssa.TypeF64,
|
|
276
|
+
result.Type().Bits() == 64, op == ssa.OpcodeFcvtToSintSat)
|
|
277
|
+
case ssa.OpcodeFcvtToUint, ssa.OpcodeFcvtToUintSat:
|
|
278
|
+
x, ctx := instr.Arg2()
|
|
279
|
+
result := instr.Return()
|
|
280
|
+
rn := m.getOperand_NR(m.compiler.ValueDefinition(x), extModeNone)
|
|
281
|
+
rd := m.compiler.VRegOf(result)
|
|
282
|
+
ctxVReg := m.compiler.VRegOf(ctx)
|
|
283
|
+
m.lowerFpuToInt(rd, rn, ctxVReg, false, x.Type() == ssa.TypeF64,
|
|
284
|
+
result.Type().Bits() == 64, op == ssa.OpcodeFcvtToUintSat)
|
|
285
|
+
case ssa.OpcodeFcvtFromSint:
|
|
286
|
+
x := instr.Arg()
|
|
287
|
+
result := instr.Return()
|
|
288
|
+
rn := m.getOperand_NR(m.compiler.ValueDefinition(x), extModeNone)
|
|
289
|
+
rd := m.compiler.VRegOf(result)
|
|
290
|
+
m.lowerIntToFpu(rd, rn, true, x.Type() == ssa.TypeI64, result.Type().Bits() == 64)
|
|
291
|
+
case ssa.OpcodeFcvtFromUint:
|
|
292
|
+
x := instr.Arg()
|
|
293
|
+
result := instr.Return()
|
|
294
|
+
rn := m.getOperand_NR(m.compiler.ValueDefinition(x), extModeNone)
|
|
295
|
+
rd := m.compiler.VRegOf(result)
|
|
296
|
+
m.lowerIntToFpu(rd, rn, false, x.Type() == ssa.TypeI64, result.Type().Bits() == 64)
|
|
297
|
+
case ssa.OpcodeFdemote:
|
|
298
|
+
v := instr.Arg()
|
|
299
|
+
rn := m.getOperand_NR(m.compiler.ValueDefinition(v), extModeNone)
|
|
300
|
+
rd := m.compiler.VRegOf(instr.Return())
|
|
301
|
+
cnt := m.allocateInstr()
|
|
302
|
+
cnt.asFpuRR(fpuUniOpCvt64To32, rd, rn, false)
|
|
303
|
+
m.insert(cnt)
|
|
304
|
+
case ssa.OpcodeFpromote:
|
|
305
|
+
v := instr.Arg()
|
|
306
|
+
rn := m.getOperand_NR(m.compiler.ValueDefinition(v), extModeNone)
|
|
307
|
+
rd := m.compiler.VRegOf(instr.Return())
|
|
308
|
+
cnt := m.allocateInstr()
|
|
309
|
+
cnt.asFpuRR(fpuUniOpCvt32To64, rd, rn, true)
|
|
310
|
+
m.insert(cnt)
|
|
311
|
+
case ssa.OpcodeIreduce:
|
|
312
|
+
rn := m.getOperand_NR(m.compiler.ValueDefinition(instr.Arg()), extModeNone)
|
|
313
|
+
retVal := instr.Return()
|
|
314
|
+
rd := m.compiler.VRegOf(retVal)
|
|
315
|
+
|
|
316
|
+
if retVal.Type() != ssa.TypeI32 {
|
|
317
|
+
panic("TODO?: Ireduce to non-i32")
|
|
318
|
+
}
|
|
319
|
+
mov := m.allocateInstr()
|
|
320
|
+
mov.asMove32(rd, rn.reg())
|
|
321
|
+
m.insert(mov)
|
|
322
|
+
case ssa.OpcodeFneg:
|
|
323
|
+
m.lowerFpuUniOp(fpuUniOpNeg, instr.Arg(), instr.Return())
|
|
324
|
+
case ssa.OpcodeSqrt:
|
|
325
|
+
m.lowerFpuUniOp(fpuUniOpSqrt, instr.Arg(), instr.Return())
|
|
326
|
+
case ssa.OpcodeCeil:
|
|
327
|
+
m.lowerFpuUniOp(fpuUniOpRoundPlus, instr.Arg(), instr.Return())
|
|
328
|
+
case ssa.OpcodeFloor:
|
|
329
|
+
m.lowerFpuUniOp(fpuUniOpRoundMinus, instr.Arg(), instr.Return())
|
|
330
|
+
case ssa.OpcodeTrunc:
|
|
331
|
+
m.lowerFpuUniOp(fpuUniOpRoundZero, instr.Arg(), instr.Return())
|
|
332
|
+
case ssa.OpcodeNearest:
|
|
333
|
+
m.lowerFpuUniOp(fpuUniOpRoundNearest, instr.Arg(), instr.Return())
|
|
334
|
+
case ssa.OpcodeFabs:
|
|
335
|
+
m.lowerFpuUniOp(fpuUniOpAbs, instr.Arg(), instr.Return())
|
|
336
|
+
case ssa.OpcodeBitcast:
|
|
337
|
+
m.lowerBitcast(instr)
|
|
338
|
+
case ssa.OpcodeFcopysign:
|
|
339
|
+
x, y := instr.Arg2()
|
|
340
|
+
m.lowerFcopysign(x, y, instr.Return())
|
|
341
|
+
case ssa.OpcodeSdiv, ssa.OpcodeUdiv:
|
|
342
|
+
x, y, ctx := instr.Arg3()
|
|
343
|
+
ctxVReg := m.compiler.VRegOf(ctx)
|
|
344
|
+
rn := m.getOperand_NR(m.compiler.ValueDefinition(x), extModeNone)
|
|
345
|
+
rm := m.getOperand_NR(m.compiler.ValueDefinition(y), extModeNone)
|
|
346
|
+
rd := m.compiler.VRegOf(instr.Return())
|
|
347
|
+
m.lowerIDiv(ctxVReg, rd, rn, rm, x.Type() == ssa.TypeI64, op == ssa.OpcodeSdiv)
|
|
348
|
+
case ssa.OpcodeSrem, ssa.OpcodeUrem:
|
|
349
|
+
x, y, ctx := instr.Arg3()
|
|
350
|
+
ctxVReg := m.compiler.VRegOf(ctx)
|
|
351
|
+
rn := m.getOperand_NR(m.compiler.ValueDefinition(x), extModeNone)
|
|
352
|
+
rm := m.getOperand_NR(m.compiler.ValueDefinition(y), extModeNone)
|
|
353
|
+
rd := m.compiler.VRegOf(instr.Return())
|
|
354
|
+
m.lowerIRem(ctxVReg, rd, rn.nr(), rm, x.Type() == ssa.TypeI64, op == ssa.OpcodeSrem)
|
|
355
|
+
case ssa.OpcodeVconst:
|
|
356
|
+
result := m.compiler.VRegOf(instr.Return())
|
|
357
|
+
lo, hi := instr.VconstData()
|
|
358
|
+
v := m.allocateInstr()
|
|
359
|
+
v.asLoadFpuConst128(result, lo, hi)
|
|
360
|
+
m.insert(v)
|
|
361
|
+
case ssa.OpcodeVbnot:
|
|
362
|
+
x := instr.Arg()
|
|
363
|
+
ins := m.allocateInstr()
|
|
364
|
+
rn := m.getOperand_NR(m.compiler.ValueDefinition(x), extModeNone)
|
|
365
|
+
rd := m.compiler.VRegOf(instr.Return())
|
|
366
|
+
ins.asVecMisc(vecOpNot, rd, rn, vecArrangement16B)
|
|
367
|
+
m.insert(ins)
|
|
368
|
+
case ssa.OpcodeVbxor:
|
|
369
|
+
x, y := instr.Arg2()
|
|
370
|
+
m.lowerVecRRR(vecOpEOR, x, y, instr.Return(), vecArrangement16B)
|
|
371
|
+
case ssa.OpcodeVbor:
|
|
372
|
+
x, y := instr.Arg2()
|
|
373
|
+
m.lowerVecRRR(vecOpOrr, x, y, instr.Return(), vecArrangement16B)
|
|
374
|
+
case ssa.OpcodeVband:
|
|
375
|
+
x, y := instr.Arg2()
|
|
376
|
+
m.lowerVecRRR(vecOpAnd, x, y, instr.Return(), vecArrangement16B)
|
|
377
|
+
case ssa.OpcodeVbandnot:
|
|
378
|
+
x, y := instr.Arg2()
|
|
379
|
+
m.lowerVecRRR(vecOpBic, x, y, instr.Return(), vecArrangement16B)
|
|
380
|
+
case ssa.OpcodeVbitselect:
|
|
381
|
+
c, x, y := instr.SelectData()
|
|
382
|
+
rn := m.getOperand_NR(m.compiler.ValueDefinition(x), extModeNone)
|
|
383
|
+
rm := m.getOperand_NR(m.compiler.ValueDefinition(y), extModeNone)
|
|
384
|
+
creg := m.getOperand_NR(m.compiler.ValueDefinition(c), extModeNone)
|
|
385
|
+
tmp := m.compiler.AllocateVReg(ssa.TypeV128)
|
|
386
|
+
|
|
387
|
+
// creg is overwritten by BSL, so we need to move it to the result register before the instruction
|
|
388
|
+
// in case when it is used somewhere else.
|
|
389
|
+
mov := m.allocateInstr()
|
|
390
|
+
mov.asFpuMov128(tmp, creg.nr())
|
|
391
|
+
m.insert(mov)
|
|
392
|
+
|
|
393
|
+
ins := m.allocateInstr()
|
|
394
|
+
ins.asVecRRRRewrite(vecOpBsl, tmp, rn, rm, vecArrangement16B)
|
|
395
|
+
m.insert(ins)
|
|
396
|
+
|
|
397
|
+
mov2 := m.allocateInstr()
|
|
398
|
+
rd := m.compiler.VRegOf(instr.Return())
|
|
399
|
+
mov2.asFpuMov128(rd, tmp)
|
|
400
|
+
m.insert(mov2)
|
|
401
|
+
case ssa.OpcodeVanyTrue, ssa.OpcodeVallTrue:
|
|
402
|
+
x, lane := instr.ArgWithLane()
|
|
403
|
+
var arr vecArrangement
|
|
404
|
+
if op == ssa.OpcodeVallTrue {
|
|
405
|
+
arr = ssaLaneToArrangement(lane)
|
|
406
|
+
}
|
|
407
|
+
rm := m.getOperand_NR(m.compiler.ValueDefinition(x), extModeNone)
|
|
408
|
+
rd := m.compiler.VRegOf(instr.Return())
|
|
409
|
+
m.lowerVcheckTrue(op, rm, rd, arr)
|
|
410
|
+
case ssa.OpcodeVhighBits:
|
|
411
|
+
x, lane := instr.ArgWithLane()
|
|
412
|
+
rm := m.getOperand_NR(m.compiler.ValueDefinition(x), extModeNone)
|
|
413
|
+
rd := m.compiler.VRegOf(instr.Return())
|
|
414
|
+
arr := ssaLaneToArrangement(lane)
|
|
415
|
+
m.lowerVhighBits(rm, rd, arr)
|
|
416
|
+
case ssa.OpcodeVIadd:
|
|
417
|
+
x, y, lane := instr.Arg2WithLane()
|
|
418
|
+
arr := ssaLaneToArrangement(lane)
|
|
419
|
+
m.lowerVecRRR(vecOpAdd, x, y, instr.Return(), arr)
|
|
420
|
+
case ssa.OpcodeExtIaddPairwise:
|
|
421
|
+
v, lane, signed := instr.ExtIaddPairwiseData()
|
|
422
|
+
vv := m.getOperand_NR(m.compiler.ValueDefinition(v), extModeNone)
|
|
423
|
+
|
|
424
|
+
tmpLo, tmpHi := operandNR(m.compiler.AllocateVReg(ssa.TypeV128)), operandNR(m.compiler.AllocateVReg(ssa.TypeV128))
|
|
425
|
+
var widen vecOp
|
|
426
|
+
if signed {
|
|
427
|
+
widen = vecOpSshll
|
|
428
|
+
} else {
|
|
429
|
+
widen = vecOpUshll
|
|
430
|
+
}
|
|
431
|
+
|
|
432
|
+
var loArr, hiArr, dstArr vecArrangement
|
|
433
|
+
switch lane {
|
|
434
|
+
case ssa.VecLaneI8x16:
|
|
435
|
+
loArr, hiArr, dstArr = vecArrangement8B, vecArrangement16B, vecArrangement8H
|
|
436
|
+
case ssa.VecLaneI16x8:
|
|
437
|
+
loArr, hiArr, dstArr = vecArrangement4H, vecArrangement8H, vecArrangement4S
|
|
438
|
+
case ssa.VecLaneI32x4:
|
|
439
|
+
loArr, hiArr, dstArr = vecArrangement2S, vecArrangement4S, vecArrangement2D
|
|
440
|
+
default:
|
|
441
|
+
panic("unsupported lane " + lane.String())
|
|
442
|
+
}
|
|
443
|
+
|
|
444
|
+
widenLo := m.allocateInstr().asVecShiftImm(widen, tmpLo.nr(), vv, operandShiftImm(0), loArr)
|
|
445
|
+
widenHi := m.allocateInstr().asVecShiftImm(widen, tmpHi.nr(), vv, operandShiftImm(0), hiArr)
|
|
446
|
+
addp := m.allocateInstr().asVecRRR(vecOpAddp, m.compiler.VRegOf(instr.Return()), tmpLo, tmpHi, dstArr)
|
|
447
|
+
m.insert(widenLo)
|
|
448
|
+
m.insert(widenHi)
|
|
449
|
+
m.insert(addp)
|
|
450
|
+
|
|
451
|
+
case ssa.OpcodeVSaddSat:
|
|
452
|
+
x, y, lane := instr.Arg2WithLane()
|
|
453
|
+
arr := ssaLaneToArrangement(lane)
|
|
454
|
+
m.lowerVecRRR(vecOpSqadd, x, y, instr.Return(), arr)
|
|
455
|
+
case ssa.OpcodeVUaddSat:
|
|
456
|
+
x, y, lane := instr.Arg2WithLane()
|
|
457
|
+
arr := ssaLaneToArrangement(lane)
|
|
458
|
+
m.lowerVecRRR(vecOpUqadd, x, y, instr.Return(), arr)
|
|
459
|
+
case ssa.OpcodeVIsub:
|
|
460
|
+
x, y, lane := instr.Arg2WithLane()
|
|
461
|
+
arr := ssaLaneToArrangement(lane)
|
|
462
|
+
m.lowerVecRRR(vecOpSub, x, y, instr.Return(), arr)
|
|
463
|
+
case ssa.OpcodeVSsubSat:
|
|
464
|
+
x, y, lane := instr.Arg2WithLane()
|
|
465
|
+
arr := ssaLaneToArrangement(lane)
|
|
466
|
+
m.lowerVecRRR(vecOpSqsub, x, y, instr.Return(), arr)
|
|
467
|
+
case ssa.OpcodeVUsubSat:
|
|
468
|
+
x, y, lane := instr.Arg2WithLane()
|
|
469
|
+
arr := ssaLaneToArrangement(lane)
|
|
470
|
+
m.lowerVecRRR(vecOpUqsub, x, y, instr.Return(), arr)
|
|
471
|
+
case ssa.OpcodeVImin:
|
|
472
|
+
x, y, lane := instr.Arg2WithLane()
|
|
473
|
+
arr := ssaLaneToArrangement(lane)
|
|
474
|
+
m.lowerVecRRR(vecOpSmin, x, y, instr.Return(), arr)
|
|
475
|
+
case ssa.OpcodeVUmin:
|
|
476
|
+
x, y, lane := instr.Arg2WithLane()
|
|
477
|
+
arr := ssaLaneToArrangement(lane)
|
|
478
|
+
m.lowerVecRRR(vecOpUmin, x, y, instr.Return(), arr)
|
|
479
|
+
case ssa.OpcodeVImax:
|
|
480
|
+
x, y, lane := instr.Arg2WithLane()
|
|
481
|
+
arr := ssaLaneToArrangement(lane)
|
|
482
|
+
m.lowerVecRRR(vecOpSmax, x, y, instr.Return(), arr)
|
|
483
|
+
case ssa.OpcodeVUmax:
|
|
484
|
+
x, y, lane := instr.Arg2WithLane()
|
|
485
|
+
arr := ssaLaneToArrangement(lane)
|
|
486
|
+
m.lowerVecRRR(vecOpUmax, x, y, instr.Return(), arr)
|
|
487
|
+
case ssa.OpcodeVAvgRound:
|
|
488
|
+
x, y, lane := instr.Arg2WithLane()
|
|
489
|
+
arr := ssaLaneToArrangement(lane)
|
|
490
|
+
m.lowerVecRRR(vecOpUrhadd, x, y, instr.Return(), arr)
|
|
491
|
+
case ssa.OpcodeVImul:
|
|
492
|
+
x, y, lane := instr.Arg2WithLane()
|
|
493
|
+
arr := ssaLaneToArrangement(lane)
|
|
494
|
+
rn := m.getOperand_NR(m.compiler.ValueDefinition(x), extModeNone)
|
|
495
|
+
rm := m.getOperand_NR(m.compiler.ValueDefinition(y), extModeNone)
|
|
496
|
+
rd := m.compiler.VRegOf(instr.Return())
|
|
497
|
+
m.lowerVIMul(rd, rn, rm, arr)
|
|
498
|
+
case ssa.OpcodeVIabs:
|
|
499
|
+
m.lowerVecMisc(vecOpAbs, instr)
|
|
500
|
+
case ssa.OpcodeVIneg:
|
|
501
|
+
m.lowerVecMisc(vecOpNeg, instr)
|
|
502
|
+
case ssa.OpcodeVIpopcnt:
|
|
503
|
+
m.lowerVecMisc(vecOpCnt, instr)
|
|
504
|
+
case ssa.OpcodeVIshl,
|
|
505
|
+
ssa.OpcodeVSshr, ssa.OpcodeVUshr:
|
|
506
|
+
x, y, lane := instr.Arg2WithLane()
|
|
507
|
+
arr := ssaLaneToArrangement(lane)
|
|
508
|
+
rn := m.getOperand_NR(m.compiler.ValueDefinition(x), extModeNone)
|
|
509
|
+
rm := m.getOperand_NR(m.compiler.ValueDefinition(y), extModeNone)
|
|
510
|
+
rd := m.compiler.VRegOf(instr.Return())
|
|
511
|
+
m.lowerVShift(op, rd, rn, rm, arr)
|
|
512
|
+
case ssa.OpcodeVSqrt:
|
|
513
|
+
m.lowerVecMisc(vecOpFsqrt, instr)
|
|
514
|
+
case ssa.OpcodeVFabs:
|
|
515
|
+
m.lowerVecMisc(vecOpFabs, instr)
|
|
516
|
+
case ssa.OpcodeVFneg:
|
|
517
|
+
m.lowerVecMisc(vecOpFneg, instr)
|
|
518
|
+
case ssa.OpcodeVFmin:
|
|
519
|
+
x, y, lane := instr.Arg2WithLane()
|
|
520
|
+
arr := ssaLaneToArrangement(lane)
|
|
521
|
+
m.lowerVecRRR(vecOpFmin, x, y, instr.Return(), arr)
|
|
522
|
+
case ssa.OpcodeVFmax:
|
|
523
|
+
x, y, lane := instr.Arg2WithLane()
|
|
524
|
+
arr := ssaLaneToArrangement(lane)
|
|
525
|
+
m.lowerVecRRR(vecOpFmax, x, y, instr.Return(), arr)
|
|
526
|
+
case ssa.OpcodeVFadd:
|
|
527
|
+
x, y, lane := instr.Arg2WithLane()
|
|
528
|
+
arr := ssaLaneToArrangement(lane)
|
|
529
|
+
m.lowerVecRRR(vecOpFadd, x, y, instr.Return(), arr)
|
|
530
|
+
case ssa.OpcodeVFsub:
|
|
531
|
+
x, y, lane := instr.Arg2WithLane()
|
|
532
|
+
arr := ssaLaneToArrangement(lane)
|
|
533
|
+
m.lowerVecRRR(vecOpFsub, x, y, instr.Return(), arr)
|
|
534
|
+
case ssa.OpcodeVFmul:
|
|
535
|
+
x, y, lane := instr.Arg2WithLane()
|
|
536
|
+
arr := ssaLaneToArrangement(lane)
|
|
537
|
+
m.lowerVecRRR(vecOpFmul, x, y, instr.Return(), arr)
|
|
538
|
+
case ssa.OpcodeSqmulRoundSat:
|
|
539
|
+
x, y, lane := instr.Arg2WithLane()
|
|
540
|
+
arr := ssaLaneToArrangement(lane)
|
|
541
|
+
m.lowerVecRRR(vecOpSqrdmulh, x, y, instr.Return(), arr)
|
|
542
|
+
case ssa.OpcodeVFdiv:
|
|
543
|
+
x, y, lane := instr.Arg2WithLane()
|
|
544
|
+
arr := ssaLaneToArrangement(lane)
|
|
545
|
+
m.lowerVecRRR(vecOpFdiv, x, y, instr.Return(), arr)
|
|
546
|
+
case ssa.OpcodeVFcvtToSintSat, ssa.OpcodeVFcvtToUintSat:
|
|
547
|
+
x, lane := instr.ArgWithLane()
|
|
548
|
+
arr := ssaLaneToArrangement(lane)
|
|
549
|
+
rn := m.getOperand_NR(m.compiler.ValueDefinition(x), extModeNone)
|
|
550
|
+
rd := m.compiler.VRegOf(instr.Return())
|
|
551
|
+
m.lowerVfpuToInt(rd, rn, arr, op == ssa.OpcodeVFcvtToSintSat)
|
|
552
|
+
case ssa.OpcodeVFcvtFromSint, ssa.OpcodeVFcvtFromUint:
|
|
553
|
+
x, lane := instr.ArgWithLane()
|
|
554
|
+
arr := ssaLaneToArrangement(lane)
|
|
555
|
+
rn := m.getOperand_NR(m.compiler.ValueDefinition(x), extModeNone)
|
|
556
|
+
rd := m.compiler.VRegOf(instr.Return())
|
|
557
|
+
m.lowerVfpuFromInt(rd, rn, arr, op == ssa.OpcodeVFcvtFromSint)
|
|
558
|
+
case ssa.OpcodeSwidenLow, ssa.OpcodeUwidenLow:
|
|
559
|
+
x, lane := instr.ArgWithLane()
|
|
560
|
+
rn := m.getOperand_NR(m.compiler.ValueDefinition(x), extModeNone)
|
|
561
|
+
rd := m.compiler.VRegOf(instr.Return())
|
|
562
|
+
|
|
563
|
+
var arr vecArrangement
|
|
564
|
+
switch lane {
|
|
565
|
+
case ssa.VecLaneI8x16:
|
|
566
|
+
arr = vecArrangement8B
|
|
567
|
+
case ssa.VecLaneI16x8:
|
|
568
|
+
arr = vecArrangement4H
|
|
569
|
+
case ssa.VecLaneI32x4:
|
|
570
|
+
arr = vecArrangement2S
|
|
571
|
+
}
|
|
572
|
+
|
|
573
|
+
shll := m.allocateInstr()
|
|
574
|
+
if signed := op == ssa.OpcodeSwidenLow; signed {
|
|
575
|
+
shll.asVecShiftImm(vecOpSshll, rd, rn, operandShiftImm(0), arr)
|
|
576
|
+
} else {
|
|
577
|
+
shll.asVecShiftImm(vecOpUshll, rd, rn, operandShiftImm(0), arr)
|
|
578
|
+
}
|
|
579
|
+
m.insert(shll)
|
|
580
|
+
case ssa.OpcodeSwidenHigh, ssa.OpcodeUwidenHigh:
|
|
581
|
+
x, lane := instr.ArgWithLane()
|
|
582
|
+
rn := m.getOperand_NR(m.compiler.ValueDefinition(x), extModeNone)
|
|
583
|
+
rd := m.compiler.VRegOf(instr.Return())
|
|
584
|
+
|
|
585
|
+
arr := ssaLaneToArrangement(lane)
|
|
586
|
+
|
|
587
|
+
shll := m.allocateInstr()
|
|
588
|
+
if signed := op == ssa.OpcodeSwidenHigh; signed {
|
|
589
|
+
shll.asVecShiftImm(vecOpSshll, rd, rn, operandShiftImm(0), arr)
|
|
590
|
+
} else {
|
|
591
|
+
shll.asVecShiftImm(vecOpUshll, rd, rn, operandShiftImm(0), arr)
|
|
592
|
+
}
|
|
593
|
+
m.insert(shll)
|
|
594
|
+
|
|
595
|
+
case ssa.OpcodeSnarrow, ssa.OpcodeUnarrow:
|
|
596
|
+
x, y, lane := instr.Arg2WithLane()
|
|
597
|
+
var arr, arr2 vecArrangement
|
|
598
|
+
switch lane {
|
|
599
|
+
case ssa.VecLaneI16x8: // I16x8
|
|
600
|
+
arr = vecArrangement8B
|
|
601
|
+
arr2 = vecArrangement16B // Implies sqxtn2.
|
|
602
|
+
case ssa.VecLaneI32x4:
|
|
603
|
+
arr = vecArrangement4H
|
|
604
|
+
arr2 = vecArrangement8H // Implies sqxtn2.
|
|
605
|
+
default:
|
|
606
|
+
panic("unsupported lane " + lane.String())
|
|
607
|
+
}
|
|
608
|
+
rn := m.getOperand_NR(m.compiler.ValueDefinition(x), extModeNone)
|
|
609
|
+
rm := m.getOperand_NR(m.compiler.ValueDefinition(y), extModeNone)
|
|
610
|
+
rd := m.compiler.VRegOf(instr.Return())
|
|
611
|
+
|
|
612
|
+
tmp := m.compiler.AllocateVReg(ssa.TypeV128)
|
|
613
|
+
|
|
614
|
+
loQxtn := m.allocateInstr()
|
|
615
|
+
hiQxtn := m.allocateInstr()
|
|
616
|
+
if signed := op == ssa.OpcodeSnarrow; signed {
|
|
617
|
+
// Narrow lanes on rn and write them into lower-half of rd.
|
|
618
|
+
loQxtn.asVecMisc(vecOpSqxtn, tmp, rn, arr) // low
|
|
619
|
+
// Narrow lanes on rm and write them into higher-half of rd.
|
|
620
|
+
hiQxtn.asVecMisc(vecOpSqxtn, tmp, rm, arr2) // high (sqxtn2)
|
|
621
|
+
} else {
|
|
622
|
+
// Narrow lanes on rn and write them into lower-half of rd.
|
|
623
|
+
loQxtn.asVecMisc(vecOpSqxtun, tmp, rn, arr) // low
|
|
624
|
+
// Narrow lanes on rm and write them into higher-half of rd.
|
|
625
|
+
hiQxtn.asVecMisc(vecOpSqxtun, tmp, rm, arr2) // high (sqxtn2)
|
|
626
|
+
}
|
|
627
|
+
m.insert(loQxtn)
|
|
628
|
+
m.insert(hiQxtn)
|
|
629
|
+
|
|
630
|
+
mov := m.allocateInstr()
|
|
631
|
+
mov.asFpuMov128(rd, tmp)
|
|
632
|
+
m.insert(mov)
|
|
633
|
+
case ssa.OpcodeFvpromoteLow:
|
|
634
|
+
x, lane := instr.ArgWithLane()
|
|
635
|
+
if lane != ssa.VecLaneF32x4 {
|
|
636
|
+
panic("unsupported lane type " + lane.String())
|
|
637
|
+
}
|
|
638
|
+
ins := m.allocateInstr()
|
|
639
|
+
rn := m.getOperand_NR(m.compiler.ValueDefinition(x), extModeNone)
|
|
640
|
+
rd := m.compiler.VRegOf(instr.Return())
|
|
641
|
+
ins.asVecMisc(vecOpFcvtl, rd, rn, vecArrangement2S)
|
|
642
|
+
m.insert(ins)
|
|
643
|
+
case ssa.OpcodeFvdemote:
|
|
644
|
+
x, lane := instr.ArgWithLane()
|
|
645
|
+
if lane != ssa.VecLaneF64x2 {
|
|
646
|
+
panic("unsupported lane type " + lane.String())
|
|
647
|
+
}
|
|
648
|
+
ins := m.allocateInstr()
|
|
649
|
+
rn := m.getOperand_NR(m.compiler.ValueDefinition(x), extModeNone)
|
|
650
|
+
rd := m.compiler.VRegOf(instr.Return())
|
|
651
|
+
ins.asVecMisc(vecOpFcvtn, rd, rn, vecArrangement2S)
|
|
652
|
+
m.insert(ins)
|
|
653
|
+
case ssa.OpcodeExtractlane:
|
|
654
|
+
x, index, signed, lane := instr.ExtractlaneData()
|
|
655
|
+
|
|
656
|
+
rn := m.getOperand_NR(m.compiler.ValueDefinition(x), extModeNone)
|
|
657
|
+
rd := m.compiler.VRegOf(instr.Return())
|
|
658
|
+
|
|
659
|
+
mov := m.allocateInstr()
|
|
660
|
+
switch lane {
|
|
661
|
+
case ssa.VecLaneI8x16:
|
|
662
|
+
mov.asMovFromVec(rd, rn, vecArrangementB, vecIndex(index), signed)
|
|
663
|
+
case ssa.VecLaneI16x8:
|
|
664
|
+
mov.asMovFromVec(rd, rn, vecArrangementH, vecIndex(index), signed)
|
|
665
|
+
case ssa.VecLaneI32x4:
|
|
666
|
+
mov.asMovFromVec(rd, rn, vecArrangementS, vecIndex(index), signed)
|
|
667
|
+
case ssa.VecLaneI64x2:
|
|
668
|
+
mov.asMovFromVec(rd, rn, vecArrangementD, vecIndex(index), signed)
|
|
669
|
+
case ssa.VecLaneF32x4:
|
|
670
|
+
mov.asVecMovElement(rd, rn, vecArrangementS, vecIndex(0), vecIndex(index))
|
|
671
|
+
case ssa.VecLaneF64x2:
|
|
672
|
+
mov.asVecMovElement(rd, rn, vecArrangementD, vecIndex(0), vecIndex(index))
|
|
673
|
+
default:
|
|
674
|
+
panic("unsupported lane: " + lane.String())
|
|
675
|
+
}
|
|
676
|
+
|
|
677
|
+
m.insert(mov)
|
|
678
|
+
|
|
679
|
+
case ssa.OpcodeInsertlane:
|
|
680
|
+
x, y, index, lane := instr.InsertlaneData()
|
|
681
|
+
rn := m.getOperand_NR(m.compiler.ValueDefinition(x), extModeNone)
|
|
682
|
+
rm := m.getOperand_NR(m.compiler.ValueDefinition(y), extModeNone)
|
|
683
|
+
rd := m.compiler.VRegOf(instr.Return())
|
|
684
|
+
tmpReg := m.compiler.AllocateVReg(ssa.TypeV128)
|
|
685
|
+
|
|
686
|
+
// Initially mov rn to tmp.
|
|
687
|
+
mov1 := m.allocateInstr()
|
|
688
|
+
mov1.asFpuMov128(tmpReg, rn.nr())
|
|
689
|
+
m.insert(mov1)
|
|
690
|
+
|
|
691
|
+
// movToVec and vecMovElement do not clear the remaining bits to zero,
|
|
692
|
+
// thus, we can mov rm in-place to tmp.
|
|
693
|
+
mov2 := m.allocateInstr()
|
|
694
|
+
switch lane {
|
|
695
|
+
case ssa.VecLaneI8x16:
|
|
696
|
+
mov2.asMovToVec(tmpReg, rm, vecArrangementB, vecIndex(index))
|
|
697
|
+
case ssa.VecLaneI16x8:
|
|
698
|
+
mov2.asMovToVec(tmpReg, rm, vecArrangementH, vecIndex(index))
|
|
699
|
+
case ssa.VecLaneI32x4:
|
|
700
|
+
mov2.asMovToVec(tmpReg, rm, vecArrangementS, vecIndex(index))
|
|
701
|
+
case ssa.VecLaneI64x2:
|
|
702
|
+
mov2.asMovToVec(tmpReg, rm, vecArrangementD, vecIndex(index))
|
|
703
|
+
case ssa.VecLaneF32x4:
|
|
704
|
+
mov2.asVecMovElement(tmpReg, rm, vecArrangementS, vecIndex(index), vecIndex(0))
|
|
705
|
+
case ssa.VecLaneF64x2:
|
|
706
|
+
mov2.asVecMovElement(tmpReg, rm, vecArrangementD, vecIndex(index), vecIndex(0))
|
|
707
|
+
}
|
|
708
|
+
m.insert(mov2)
|
|
709
|
+
|
|
710
|
+
// Finally mov tmp to rd.
|
|
711
|
+
mov3 := m.allocateInstr()
|
|
712
|
+
mov3.asFpuMov128(rd, tmpReg)
|
|
713
|
+
m.insert(mov3)
|
|
714
|
+
|
|
715
|
+
case ssa.OpcodeSwizzle:
|
|
716
|
+
x, y, lane := instr.Arg2WithLane()
|
|
717
|
+
rn := m.getOperand_NR(m.compiler.ValueDefinition(x), extModeNone)
|
|
718
|
+
rm := m.getOperand_NR(m.compiler.ValueDefinition(y), extModeNone)
|
|
719
|
+
rd := m.compiler.VRegOf(instr.Return())
|
|
720
|
+
|
|
721
|
+
arr := ssaLaneToArrangement(lane)
|
|
722
|
+
|
|
723
|
+
// tbl <rd>.<arr>, { <rn>.<arr> }, <rm>.<arr>
|
|
724
|
+
tbl1 := m.allocateInstr()
|
|
725
|
+
tbl1.asVecTbl(1, rd, rn, rm, arr)
|
|
726
|
+
m.insert(tbl1)
|
|
727
|
+
|
|
728
|
+
case ssa.OpcodeShuffle:
|
|
729
|
+
x, y, lane1, lane2 := instr.ShuffleData()
|
|
730
|
+
rn := m.getOperand_NR(m.compiler.ValueDefinition(x), extModeNone)
|
|
731
|
+
rm := m.getOperand_NR(m.compiler.ValueDefinition(y), extModeNone)
|
|
732
|
+
rd := m.compiler.VRegOf(instr.Return())
|
|
733
|
+
|
|
734
|
+
m.lowerShuffle(rd, rn, rm, lane1, lane2)
|
|
735
|
+
|
|
736
|
+
case ssa.OpcodeSplat:
|
|
737
|
+
x, lane := instr.ArgWithLane()
|
|
738
|
+
rn := m.getOperand_NR(m.compiler.ValueDefinition(x), extModeNone)
|
|
739
|
+
rd := m.compiler.VRegOf(instr.Return())
|
|
740
|
+
|
|
741
|
+
dup := m.allocateInstr()
|
|
742
|
+
switch lane {
|
|
743
|
+
case ssa.VecLaneI8x16:
|
|
744
|
+
dup.asVecDup(rd, rn, vecArrangement16B)
|
|
745
|
+
case ssa.VecLaneI16x8:
|
|
746
|
+
dup.asVecDup(rd, rn, vecArrangement8H)
|
|
747
|
+
case ssa.VecLaneI32x4:
|
|
748
|
+
dup.asVecDup(rd, rn, vecArrangement4S)
|
|
749
|
+
case ssa.VecLaneI64x2:
|
|
750
|
+
dup.asVecDup(rd, rn, vecArrangement2D)
|
|
751
|
+
case ssa.VecLaneF32x4:
|
|
752
|
+
dup.asVecDupElement(rd, rn, vecArrangementS, vecIndex(0))
|
|
753
|
+
case ssa.VecLaneF64x2:
|
|
754
|
+
dup.asVecDupElement(rd, rn, vecArrangementD, vecIndex(0))
|
|
755
|
+
}
|
|
756
|
+
m.insert(dup)
|
|
757
|
+
|
|
758
|
+
case ssa.OpcodeWideningPairwiseDotProductS:
|
|
759
|
+
x, y := instr.Arg2()
|
|
760
|
+
xx, yy := m.getOperand_NR(m.compiler.ValueDefinition(x), extModeNone),
|
|
761
|
+
m.getOperand_NR(m.compiler.ValueDefinition(y), extModeNone)
|
|
762
|
+
tmp, tmp2 := operandNR(m.compiler.AllocateVReg(ssa.TypeV128)), operandNR(m.compiler.AllocateVReg(ssa.TypeV128))
|
|
763
|
+
m.insert(m.allocateInstr().asVecRRR(vecOpSmull, tmp.nr(), xx, yy, vecArrangement8H))
|
|
764
|
+
m.insert(m.allocateInstr().asVecRRR(vecOpSmull2, tmp2.nr(), xx, yy, vecArrangement8H))
|
|
765
|
+
m.insert(m.allocateInstr().asVecRRR(vecOpAddp, tmp.nr(), tmp, tmp2, vecArrangement4S))
|
|
766
|
+
|
|
767
|
+
rd := m.compiler.VRegOf(instr.Return())
|
|
768
|
+
m.insert(m.allocateInstr().asFpuMov128(rd, tmp.nr()))
|
|
769
|
+
|
|
770
|
+
case ssa.OpcodeLoadSplat:
|
|
771
|
+
ptr, offset, lane := instr.LoadSplatData()
|
|
772
|
+
m.lowerLoadSplat(ptr, offset, lane, instr.Return())
|
|
773
|
+
|
|
774
|
+
case ssa.OpcodeAtomicRmw:
|
|
775
|
+
m.lowerAtomicRmw(instr)
|
|
776
|
+
|
|
777
|
+
case ssa.OpcodeAtomicCas:
|
|
778
|
+
m.lowerAtomicCas(instr)
|
|
779
|
+
|
|
780
|
+
case ssa.OpcodeAtomicLoad:
|
|
781
|
+
m.lowerAtomicLoad(instr)
|
|
782
|
+
|
|
783
|
+
case ssa.OpcodeAtomicStore:
|
|
784
|
+
m.lowerAtomicStore(instr)
|
|
785
|
+
|
|
786
|
+
case ssa.OpcodeFence:
|
|
787
|
+
instr := m.allocateInstr()
|
|
788
|
+
instr.asDMB()
|
|
789
|
+
m.insert(instr)
|
|
790
|
+
|
|
791
|
+
case ssa.OpcodeTailCallReturnCall, ssa.OpcodeTailCallReturnCallIndirect:
|
|
792
|
+
m.lowerTailCall(instr)
|
|
793
|
+
|
|
794
|
+
default:
|
|
795
|
+
panic("TODO: lowering " + op.String())
|
|
796
|
+
}
|
|
797
|
+
m.FlushPendingInstructions()
|
|
798
|
+
}
|
|
799
|
+
|
|
800
|
+
func (m *machine) lowerShuffle(rd regalloc.VReg, rn, rm operand, lane1, lane2 uint64) {
|
|
801
|
+
// `tbl2` requires 2 consecutive registers, so we arbitrarily pick v29, v30.
|
|
802
|
+
vReg, wReg := v29VReg, v30VReg
|
|
803
|
+
|
|
804
|
+
// Initialize v29, v30 to rn, rm.
|
|
805
|
+
movv := m.allocateInstr()
|
|
806
|
+
movv.asFpuMov128(vReg, rn.nr())
|
|
807
|
+
m.insert(movv)
|
|
808
|
+
|
|
809
|
+
movw := m.allocateInstr()
|
|
810
|
+
movw.asFpuMov128(wReg, rm.nr())
|
|
811
|
+
m.insert(movw)
|
|
812
|
+
|
|
813
|
+
// `lane1`, `lane2` are already encoded as two u64s with the right layout:
|
|
814
|
+
// lane1 := lane[7]<<56 | ... | lane[1]<<8 | lane[0]
|
|
815
|
+
// lane2 := lane[15]<<56 | ... | lane[9]<<8 | lane[8]
|
|
816
|
+
// Thus, we can use loadFpuConst128.
|
|
817
|
+
tmp := operandNR(m.compiler.AllocateVReg(ssa.TypeV128))
|
|
818
|
+
lfc := m.allocateInstr()
|
|
819
|
+
lfc.asLoadFpuConst128(tmp.nr(), lane1, lane2)
|
|
820
|
+
m.insert(lfc)
|
|
821
|
+
|
|
822
|
+
// tbl <rd>.16b, { <vReg>.16B, <wReg>.16b }, <tmp>.16b
|
|
823
|
+
tbl2 := m.allocateInstr()
|
|
824
|
+
tbl2.asVecTbl(2, rd, operandNR(vReg), tmp, vecArrangement16B)
|
|
825
|
+
m.insert(tbl2)
|
|
826
|
+
}
|
|
827
|
+
|
|
828
|
+
func (m *machine) lowerVShift(op ssa.Opcode, rd regalloc.VReg, rn, rm operand, arr vecArrangement) {
|
|
829
|
+
var modulo byte
|
|
830
|
+
switch arr {
|
|
831
|
+
case vecArrangement16B:
|
|
832
|
+
modulo = 0x7 // Modulo 8.
|
|
833
|
+
case vecArrangement8H:
|
|
834
|
+
modulo = 0xf // Modulo 16.
|
|
835
|
+
case vecArrangement4S:
|
|
836
|
+
modulo = 0x1f // Modulo 32.
|
|
837
|
+
case vecArrangement2D:
|
|
838
|
+
modulo = 0x3f // Modulo 64.
|
|
839
|
+
default:
|
|
840
|
+
panic("unsupported arrangment " + arr.String())
|
|
841
|
+
}
|
|
842
|
+
|
|
843
|
+
rtmp := operandNR(m.compiler.AllocateVReg(ssa.TypeI64))
|
|
844
|
+
vtmp := operandNR(m.compiler.AllocateVReg(ssa.TypeV128))
|
|
845
|
+
|
|
846
|
+
and := m.allocateInstr()
|
|
847
|
+
and.asALUBitmaskImm(aluOpAnd, rtmp.nr(), rm.nr(), uint64(modulo), true)
|
|
848
|
+
m.insert(and)
|
|
849
|
+
|
|
850
|
+
if op != ssa.OpcodeVIshl {
|
|
851
|
+
// Negate the amount to make this as right shift.
|
|
852
|
+
neg := m.allocateInstr()
|
|
853
|
+
neg.asALU(aluOpSub, rtmp.nr(), operandNR(xzrVReg), rtmp, true)
|
|
854
|
+
m.insert(neg)
|
|
855
|
+
}
|
|
856
|
+
|
|
857
|
+
// Copy the shift amount into a vector register as sshl/ushl requires it to be there.
|
|
858
|
+
dup := m.allocateInstr()
|
|
859
|
+
dup.asVecDup(vtmp.nr(), rtmp, arr)
|
|
860
|
+
m.insert(dup)
|
|
861
|
+
|
|
862
|
+
if op == ssa.OpcodeVIshl || op == ssa.OpcodeVSshr {
|
|
863
|
+
sshl := m.allocateInstr()
|
|
864
|
+
sshl.asVecRRR(vecOpSshl, rd, rn, vtmp, arr)
|
|
865
|
+
m.insert(sshl)
|
|
866
|
+
} else {
|
|
867
|
+
ushl := m.allocateInstr()
|
|
868
|
+
ushl.asVecRRR(vecOpUshl, rd, rn, vtmp, arr)
|
|
869
|
+
m.insert(ushl)
|
|
870
|
+
}
|
|
871
|
+
}
|
|
872
|
+
|
|
873
|
+
func (m *machine) lowerVcheckTrue(op ssa.Opcode, rm operand, rd regalloc.VReg, arr vecArrangement) {
|
|
874
|
+
tmp := operandNR(m.compiler.AllocateVReg(ssa.TypeV128))
|
|
875
|
+
|
|
876
|
+
// Special case VallTrue for i64x2.
|
|
877
|
+
if op == ssa.OpcodeVallTrue && arr == vecArrangement2D {
|
|
878
|
+
// cmeq v3?.2d, v2?.2d, #0
|
|
879
|
+
// addp v3?.2d, v3?.2d, v3?.2d
|
|
880
|
+
// fcmp v3?, v3?
|
|
881
|
+
// cset dst, eq
|
|
882
|
+
|
|
883
|
+
ins := m.allocateInstr()
|
|
884
|
+
ins.asVecMisc(vecOpCmeq0, tmp.nr(), rm, vecArrangement2D)
|
|
885
|
+
m.insert(ins)
|
|
886
|
+
|
|
887
|
+
addp := m.allocateInstr()
|
|
888
|
+
addp.asVecRRR(vecOpAddp, tmp.nr(), tmp, tmp, vecArrangement2D)
|
|
889
|
+
m.insert(addp)
|
|
890
|
+
|
|
891
|
+
fcmp := m.allocateInstr()
|
|
892
|
+
fcmp.asFpuCmp(tmp, tmp, true)
|
|
893
|
+
m.insert(fcmp)
|
|
894
|
+
|
|
895
|
+
cset := m.allocateInstr()
|
|
896
|
+
cset.asCSet(rd, false, eq)
|
|
897
|
+
m.insert(cset)
|
|
898
|
+
|
|
899
|
+
return
|
|
900
|
+
}
|
|
901
|
+
|
|
902
|
+
// Create a scalar value with umaxp or uminv, then compare it against zero.
|
|
903
|
+
ins := m.allocateInstr()
|
|
904
|
+
if op == ssa.OpcodeVanyTrue {
|
|
905
|
+
// umaxp v4?.16b, v2?.16b, v2?.16b
|
|
906
|
+
ins.asVecRRR(vecOpUmaxp, tmp.nr(), rm, rm, vecArrangement16B)
|
|
907
|
+
} else {
|
|
908
|
+
// uminv d4?, v2?.4s
|
|
909
|
+
ins.asVecLanes(vecOpUminv, tmp.nr(), rm, arr)
|
|
910
|
+
}
|
|
911
|
+
m.insert(ins)
|
|
912
|
+
|
|
913
|
+
// mov x3?, v4?.d[0]
|
|
914
|
+
// ccmp x3?, #0x0, #0x0, al
|
|
915
|
+
// cset x3?, ne
|
|
916
|
+
// mov x0, x3?
|
|
917
|
+
|
|
918
|
+
movv := m.allocateInstr()
|
|
919
|
+
movv.asMovFromVec(rd, tmp, vecArrangementD, vecIndex(0), false)
|
|
920
|
+
m.insert(movv)
|
|
921
|
+
|
|
922
|
+
fc := m.allocateInstr()
|
|
923
|
+
fc.asCCmpImm(operandNR(rd), uint64(0), al, 0, true)
|
|
924
|
+
m.insert(fc)
|
|
925
|
+
|
|
926
|
+
cset := m.allocateInstr()
|
|
927
|
+
cset.asCSet(rd, false, ne)
|
|
928
|
+
m.insert(cset)
|
|
929
|
+
}
|
|
930
|
+
|
|
931
|
+
func (m *machine) lowerVhighBits(rm operand, rd regalloc.VReg, arr vecArrangement) {
|
|
932
|
+
r0 := operandNR(m.compiler.AllocateVReg(ssa.TypeI64))
|
|
933
|
+
v0 := operandNR(m.compiler.AllocateVReg(ssa.TypeV128))
|
|
934
|
+
v1 := operandNR(m.compiler.AllocateVReg(ssa.TypeV128))
|
|
935
|
+
|
|
936
|
+
switch arr {
|
|
937
|
+
case vecArrangement16B:
|
|
938
|
+
// sshr v6?.16b, v2?.16b, #7
|
|
939
|
+
// movz x4?, #0x201, lsl 0
|
|
940
|
+
// movk x4?, #0x804, lsl 16
|
|
941
|
+
// movk x4?, #0x2010, lsl 32
|
|
942
|
+
// movk x4?, #0x8040, lsl 48
|
|
943
|
+
// dup v5?.2d, x4?
|
|
944
|
+
// and v6?.16b, v6?.16b, v5?.16b
|
|
945
|
+
// ext v5?.16b, v6?.16b, v6?.16b, #8
|
|
946
|
+
// zip1 v5?.16b, v6?.16b, v5?.16b
|
|
947
|
+
// addv s5?, v5?.8h
|
|
948
|
+
// umov s3?, v5?.h[0]
|
|
949
|
+
|
|
950
|
+
// Right arithmetic shift on the original vector and store the result into v1. So we have:
|
|
951
|
+
// v1[i] = 0xff if vi<0, 0 otherwise.
|
|
952
|
+
sshr := m.allocateInstr()
|
|
953
|
+
sshr.asVecShiftImm(vecOpSshr, v1.nr(), rm, operandShiftImm(7), vecArrangement16B)
|
|
954
|
+
m.insert(sshr)
|
|
955
|
+
|
|
956
|
+
// Load the bit mask into r0.
|
|
957
|
+
m.insertMOVZ(r0.nr(), 0x0201, 0, true)
|
|
958
|
+
m.insertMOVK(r0.nr(), 0x0804, 1, true)
|
|
959
|
+
m.insertMOVK(r0.nr(), 0x2010, 2, true)
|
|
960
|
+
m.insertMOVK(r0.nr(), 0x8040, 3, true)
|
|
961
|
+
|
|
962
|
+
// dup r0 to v0.
|
|
963
|
+
dup := m.allocateInstr()
|
|
964
|
+
dup.asVecDup(v0.nr(), r0, vecArrangement2D)
|
|
965
|
+
m.insert(dup)
|
|
966
|
+
|
|
967
|
+
// Lane-wise logical AND with the bit mask, meaning that we have
|
|
968
|
+
// v[i] = (1 << i) if vi<0, 0 otherwise.
|
|
969
|
+
//
|
|
970
|
+
// Below, we use the following notation:
|
|
971
|
+
// wi := (1 << i) if vi<0, 0 otherwise.
|
|
972
|
+
and := m.allocateInstr()
|
|
973
|
+
and.asVecRRR(vecOpAnd, v1.nr(), v1, v0, vecArrangement16B)
|
|
974
|
+
m.insert(and)
|
|
975
|
+
|
|
976
|
+
// Swap the lower and higher 8 byte elements, and write it into v0, meaning that we have
|
|
977
|
+
// v0[i] = w(i+8) if i < 8, w(i-8) otherwise.
|
|
978
|
+
ext := m.allocateInstr()
|
|
979
|
+
ext.asVecExtract(v0.nr(), v1, v1, vecArrangement16B, uint32(8))
|
|
980
|
+
m.insert(ext)
|
|
981
|
+
|
|
982
|
+
// v = [w0, w8, ..., w7, w15]
|
|
983
|
+
zip1 := m.allocateInstr()
|
|
984
|
+
zip1.asVecPermute(vecOpZip1, v0.nr(), v1, v0, vecArrangement16B)
|
|
985
|
+
m.insert(zip1)
|
|
986
|
+
|
|
987
|
+
// v.h[0] = w0 + ... + w15
|
|
988
|
+
addv := m.allocateInstr()
|
|
989
|
+
addv.asVecLanes(vecOpAddv, v0.nr(), v0, vecArrangement8H)
|
|
990
|
+
m.insert(addv)
|
|
991
|
+
|
|
992
|
+
// Extract the v.h[0] as the result.
|
|
993
|
+
movfv := m.allocateInstr()
|
|
994
|
+
movfv.asMovFromVec(rd, v0, vecArrangementH, vecIndex(0), false)
|
|
995
|
+
m.insert(movfv)
|
|
996
|
+
case vecArrangement8H:
|
|
997
|
+
// sshr v6?.8h, v2?.8h, #15
|
|
998
|
+
// movz x4?, #0x1, lsl 0
|
|
999
|
+
// movk x4?, #0x2, lsl 16
|
|
1000
|
+
// movk x4?, #0x4, lsl 32
|
|
1001
|
+
// movk x4?, #0x8, lsl 48
|
|
1002
|
+
// dup v5?.2d, x4?
|
|
1003
|
+
// lsl x4?, x4?, 0x4
|
|
1004
|
+
// ins v5?.d[1], x4?
|
|
1005
|
+
// and v5?.16b, v6?.16b, v5?.16b
|
|
1006
|
+
// addv s5?, v5?.8h
|
|
1007
|
+
// umov s3?, v5?.h[0]
|
|
1008
|
+
|
|
1009
|
+
// Right arithmetic shift on the original vector and store the result into v1. So we have:
|
|
1010
|
+
// v[i] = 0xffff if vi<0, 0 otherwise.
|
|
1011
|
+
sshr := m.allocateInstr()
|
|
1012
|
+
sshr.asVecShiftImm(vecOpSshr, v1.nr(), rm, operandShiftImm(15), vecArrangement8H)
|
|
1013
|
+
m.insert(sshr)
|
|
1014
|
+
|
|
1015
|
+
// Load the bit mask into r0.
|
|
1016
|
+
m.lowerConstantI64(r0.nr(), 0x0008000400020001)
|
|
1017
|
+
|
|
1018
|
+
// dup r0 to vector v0.
|
|
1019
|
+
dup := m.allocateInstr()
|
|
1020
|
+
dup.asVecDup(v0.nr(), r0, vecArrangement2D)
|
|
1021
|
+
m.insert(dup)
|
|
1022
|
+
|
|
1023
|
+
lsl := m.allocateInstr()
|
|
1024
|
+
lsl.asALUShift(aluOpLsl, r0.nr(), r0, operandShiftImm(4), true)
|
|
1025
|
+
m.insert(lsl)
|
|
1026
|
+
|
|
1027
|
+
movv := m.allocateInstr()
|
|
1028
|
+
movv.asMovToVec(v0.nr(), r0, vecArrangementD, vecIndex(1))
|
|
1029
|
+
m.insert(movv)
|
|
1030
|
+
|
|
1031
|
+
// Lane-wise logical AND with the bitmask, meaning that we have
|
|
1032
|
+
// v[i] = (1 << i) if vi<0, 0 otherwise for i=0..3
|
|
1033
|
+
// = (1 << (i+4)) if vi<0, 0 otherwise for i=3..7
|
|
1034
|
+
and := m.allocateInstr()
|
|
1035
|
+
and.asVecRRR(vecOpAnd, v0.nr(), v1, v0, vecArrangement16B)
|
|
1036
|
+
m.insert(and)
|
|
1037
|
+
|
|
1038
|
+
addv := m.allocateInstr()
|
|
1039
|
+
addv.asVecLanes(vecOpAddv, v0.nr(), v0, vecArrangement8H)
|
|
1040
|
+
m.insert(addv)
|
|
1041
|
+
|
|
1042
|
+
movfv := m.allocateInstr()
|
|
1043
|
+
movfv.asMovFromVec(rd, v0, vecArrangementH, vecIndex(0), false)
|
|
1044
|
+
m.insert(movfv)
|
|
1045
|
+
case vecArrangement4S:
|
|
1046
|
+
// sshr v6?.8h, v2?.8h, #15
|
|
1047
|
+
// movz x4?, #0x1, lsl 0
|
|
1048
|
+
// movk x4?, #0x2, lsl 16
|
|
1049
|
+
// movk x4?, #0x4, lsl 32
|
|
1050
|
+
// movk x4?, #0x8, lsl 48
|
|
1051
|
+
// dup v5?.2d, x4?
|
|
1052
|
+
// lsl x4?, x4?, 0x4
|
|
1053
|
+
// ins v5?.d[1], x4?
|
|
1054
|
+
// and v5?.16b, v6?.16b, v5?.16b
|
|
1055
|
+
// addv s5?, v5?.8h
|
|
1056
|
+
// umov s3?, v5?.h[0]
|
|
1057
|
+
|
|
1058
|
+
// Right arithmetic shift on the original vector and store the result into v1. So we have:
|
|
1059
|
+
// v[i] = 0xffffffff if vi<0, 0 otherwise.
|
|
1060
|
+
sshr := m.allocateInstr()
|
|
1061
|
+
sshr.asVecShiftImm(vecOpSshr, v1.nr(), rm, operandShiftImm(31), vecArrangement4S)
|
|
1062
|
+
m.insert(sshr)
|
|
1063
|
+
|
|
1064
|
+
// Load the bit mask into r0.
|
|
1065
|
+
m.lowerConstantI64(r0.nr(), 0x0000000200000001)
|
|
1066
|
+
|
|
1067
|
+
// dup r0 to vector v0.
|
|
1068
|
+
dup := m.allocateInstr()
|
|
1069
|
+
dup.asVecDup(v0.nr(), r0, vecArrangement2D)
|
|
1070
|
+
m.insert(dup)
|
|
1071
|
+
|
|
1072
|
+
lsl := m.allocateInstr()
|
|
1073
|
+
lsl.asALUShift(aluOpLsl, r0.nr(), r0, operandShiftImm(2), true)
|
|
1074
|
+
m.insert(lsl)
|
|
1075
|
+
|
|
1076
|
+
movv := m.allocateInstr()
|
|
1077
|
+
movv.asMovToVec(v0.nr(), r0, vecArrangementD, vecIndex(1))
|
|
1078
|
+
m.insert(movv)
|
|
1079
|
+
|
|
1080
|
+
// Lane-wise logical AND with the bitmask, meaning that we have
|
|
1081
|
+
// v[i] = (1 << i) if vi<0, 0 otherwise for i in [0, 1]
|
|
1082
|
+
// = (1 << (i+4)) if vi<0, 0 otherwise for i in [2, 3]
|
|
1083
|
+
and := m.allocateInstr()
|
|
1084
|
+
and.asVecRRR(vecOpAnd, v0.nr(), v1, v0, vecArrangement16B)
|
|
1085
|
+
m.insert(and)
|
|
1086
|
+
|
|
1087
|
+
addv := m.allocateInstr()
|
|
1088
|
+
addv.asVecLanes(vecOpAddv, v0.nr(), v0, vecArrangement4S)
|
|
1089
|
+
m.insert(addv)
|
|
1090
|
+
|
|
1091
|
+
movfv := m.allocateInstr()
|
|
1092
|
+
movfv.asMovFromVec(rd, v0, vecArrangementS, vecIndex(0), false)
|
|
1093
|
+
m.insert(movfv)
|
|
1094
|
+
case vecArrangement2D:
|
|
1095
|
+
// mov d3?, v2?.d[0]
|
|
1096
|
+
// mov x4?, v2?.d[1]
|
|
1097
|
+
// lsr x4?, x4?, 0x3f
|
|
1098
|
+
// lsr d3?, d3?, 0x3f
|
|
1099
|
+
// add s3?, s3?, w4?, lsl #1
|
|
1100
|
+
|
|
1101
|
+
// Move the lower 64-bit int into result.
|
|
1102
|
+
movv0 := m.allocateInstr()
|
|
1103
|
+
movv0.asMovFromVec(rd, rm, vecArrangementD, vecIndex(0), false)
|
|
1104
|
+
m.insert(movv0)
|
|
1105
|
+
|
|
1106
|
+
// Move the higher 64-bit int into r0.
|
|
1107
|
+
movv1 := m.allocateInstr()
|
|
1108
|
+
movv1.asMovFromVec(r0.nr(), rm, vecArrangementD, vecIndex(1), false)
|
|
1109
|
+
m.insert(movv1)
|
|
1110
|
+
|
|
1111
|
+
// Move the sign bit into the least significant bit.
|
|
1112
|
+
lsr1 := m.allocateInstr()
|
|
1113
|
+
lsr1.asALUShift(aluOpLsr, r0.nr(), r0, operandShiftImm(63), true)
|
|
1114
|
+
m.insert(lsr1)
|
|
1115
|
+
|
|
1116
|
+
lsr2 := m.allocateInstr()
|
|
1117
|
+
lsr2.asALUShift(aluOpLsr, rd, operandNR(rd), operandShiftImm(63), true)
|
|
1118
|
+
m.insert(lsr2)
|
|
1119
|
+
|
|
1120
|
+
// rd = (r0<<1) | rd
|
|
1121
|
+
lsl := m.allocateInstr()
|
|
1122
|
+
lsl.asALU(aluOpAdd, rd, operandNR(rd), operandSR(r0.nr(), 1, shiftOpLSL), false)
|
|
1123
|
+
m.insert(lsl)
|
|
1124
|
+
default:
|
|
1125
|
+
panic("Unsupported " + arr.String())
|
|
1126
|
+
}
|
|
1127
|
+
}
|
|
1128
|
+
|
|
1129
|
+
func (m *machine) lowerVecMisc(op vecOp, instr *ssa.Instruction) {
|
|
1130
|
+
x, lane := instr.ArgWithLane()
|
|
1131
|
+
arr := ssaLaneToArrangement(lane)
|
|
1132
|
+
ins := m.allocateInstr()
|
|
1133
|
+
rn := m.getOperand_NR(m.compiler.ValueDefinition(x), extModeNone)
|
|
1134
|
+
rd := m.compiler.VRegOf(instr.Return())
|
|
1135
|
+
ins.asVecMisc(op, rd, rn, arr)
|
|
1136
|
+
m.insert(ins)
|
|
1137
|
+
}
|
|
1138
|
+
|
|
1139
|
+
func (m *machine) lowerVecRRR(op vecOp, x, y, ret ssa.Value, arr vecArrangement) {
|
|
1140
|
+
ins := m.allocateInstr()
|
|
1141
|
+
rn := m.getOperand_NR(m.compiler.ValueDefinition(x), extModeNone)
|
|
1142
|
+
rm := m.getOperand_NR(m.compiler.ValueDefinition(y), extModeNone)
|
|
1143
|
+
rd := m.compiler.VRegOf(ret)
|
|
1144
|
+
ins.asVecRRR(op, rd, rn, rm, arr)
|
|
1145
|
+
m.insert(ins)
|
|
1146
|
+
}
|
|
1147
|
+
|
|
1148
|
+
func (m *machine) lowerVIMul(rd regalloc.VReg, rn, rm operand, arr vecArrangement) {
|
|
1149
|
+
if arr != vecArrangement2D {
|
|
1150
|
+
mul := m.allocateInstr()
|
|
1151
|
+
mul.asVecRRR(vecOpMul, rd, rn, rm, arr)
|
|
1152
|
+
m.insert(mul)
|
|
1153
|
+
} else {
|
|
1154
|
+
tmp1 := m.compiler.AllocateVReg(ssa.TypeV128)
|
|
1155
|
+
tmp2 := m.compiler.AllocateVReg(ssa.TypeV128)
|
|
1156
|
+
tmp3 := m.compiler.AllocateVReg(ssa.TypeV128)
|
|
1157
|
+
|
|
1158
|
+
tmpRes := m.compiler.AllocateVReg(ssa.TypeV128)
|
|
1159
|
+
|
|
1160
|
+
// Following the algorithm in https://chromium-review.googlesource.com/c/v8/v8/+/1781696
|
|
1161
|
+
rev64 := m.allocateInstr()
|
|
1162
|
+
rev64.asVecMisc(vecOpRev64, tmp2, rm, vecArrangement4S)
|
|
1163
|
+
m.insert(rev64)
|
|
1164
|
+
|
|
1165
|
+
mul := m.allocateInstr()
|
|
1166
|
+
mul.asVecRRR(vecOpMul, tmp2, operandNR(tmp2), rn, vecArrangement4S)
|
|
1167
|
+
m.insert(mul)
|
|
1168
|
+
|
|
1169
|
+
xtn1 := m.allocateInstr()
|
|
1170
|
+
xtn1.asVecMisc(vecOpXtn, tmp1, rn, vecArrangement2S)
|
|
1171
|
+
m.insert(xtn1)
|
|
1172
|
+
|
|
1173
|
+
addp := m.allocateInstr()
|
|
1174
|
+
addp.asVecRRR(vecOpAddp, tmp2, operandNR(tmp2), operandNR(tmp2), vecArrangement4S)
|
|
1175
|
+
m.insert(addp)
|
|
1176
|
+
|
|
1177
|
+
xtn2 := m.allocateInstr()
|
|
1178
|
+
xtn2.asVecMisc(vecOpXtn, tmp3, rm, vecArrangement2S)
|
|
1179
|
+
m.insert(xtn2)
|
|
1180
|
+
|
|
1181
|
+
// Note: do not write the result directly into result yet. This is the same reason as in bsl.
|
|
1182
|
+
// In short, in UMLAL instruction, the result register is also one of the source register, and
|
|
1183
|
+
// the value on the result register is significant.
|
|
1184
|
+
shll := m.allocateInstr()
|
|
1185
|
+
shll.asVecMisc(vecOpShll, tmpRes, operandNR(tmp2), vecArrangement2S)
|
|
1186
|
+
m.insert(shll)
|
|
1187
|
+
|
|
1188
|
+
umlal := m.allocateInstr()
|
|
1189
|
+
umlal.asVecRRRRewrite(vecOpUmlal, tmpRes, operandNR(tmp3), operandNR(tmp1), vecArrangement2S)
|
|
1190
|
+
m.insert(umlal)
|
|
1191
|
+
|
|
1192
|
+
mov := m.allocateInstr()
|
|
1193
|
+
mov.asFpuMov128(rd, tmpRes)
|
|
1194
|
+
m.insert(mov)
|
|
1195
|
+
}
|
|
1196
|
+
}
|
|
1197
|
+
|
|
1198
|
+
func (m *machine) lowerVMinMaxPseudo(instr *ssa.Instruction, max bool) {
|
|
1199
|
+
x, y, lane := instr.Arg2WithLane()
|
|
1200
|
+
arr := ssaLaneToArrangement(lane)
|
|
1201
|
+
|
|
1202
|
+
rn := m.getOperand_NR(m.compiler.ValueDefinition(x), extModeNone)
|
|
1203
|
+
rm := m.getOperand_NR(m.compiler.ValueDefinition(y), extModeNone)
|
|
1204
|
+
|
|
1205
|
+
// Note: this usage of tmp is important.
|
|
1206
|
+
// BSL modifies the destination register, so we need to use a temporary register so that
|
|
1207
|
+
// the actual definition of the destination register happens *after* the BSL instruction.
|
|
1208
|
+
// That way, we can force the spill instruction to be inserted after the BSL instruction.
|
|
1209
|
+
tmp := m.compiler.AllocateVReg(ssa.TypeV128)
|
|
1210
|
+
|
|
1211
|
+
fcmgt := m.allocateInstr()
|
|
1212
|
+
if max {
|
|
1213
|
+
fcmgt.asVecRRR(vecOpFcmgt, tmp, rm, rn, arr)
|
|
1214
|
+
} else {
|
|
1215
|
+
// If min, swap the args.
|
|
1216
|
+
fcmgt.asVecRRR(vecOpFcmgt, tmp, rn, rm, arr)
|
|
1217
|
+
}
|
|
1218
|
+
m.insert(fcmgt)
|
|
1219
|
+
|
|
1220
|
+
bsl := m.allocateInstr()
|
|
1221
|
+
bsl.asVecRRRRewrite(vecOpBsl, tmp, rm, rn, vecArrangement16B)
|
|
1222
|
+
m.insert(bsl)
|
|
1223
|
+
|
|
1224
|
+
res := operandNR(m.compiler.VRegOf(instr.Return()))
|
|
1225
|
+
mov2 := m.allocateInstr()
|
|
1226
|
+
mov2.asFpuMov128(res.nr(), tmp)
|
|
1227
|
+
m.insert(mov2)
|
|
1228
|
+
}
|
|
1229
|
+
|
|
1230
|
+
func (m *machine) lowerIRem(execCtxVReg regalloc.VReg, rd, rn regalloc.VReg, rm operand, _64bit, signed bool) {
|
|
1231
|
+
div := m.allocateInstr()
|
|
1232
|
+
|
|
1233
|
+
if signed {
|
|
1234
|
+
div.asALU(aluOpSDiv, rd, operandNR(rn), rm, _64bit)
|
|
1235
|
+
} else {
|
|
1236
|
+
div.asALU(aluOpUDiv, rd, operandNR(rn), rm, _64bit)
|
|
1237
|
+
}
|
|
1238
|
+
m.insert(div)
|
|
1239
|
+
|
|
1240
|
+
// Check if rm is zero:
|
|
1241
|
+
m.exitIfNot(execCtxVReg, registerAsRegNotZeroCond(rm.nr()), _64bit, wazevoapi.ExitCodeIntegerDivisionByZero)
|
|
1242
|
+
|
|
1243
|
+
// rd = rn-rd*rm by MSUB instruction.
|
|
1244
|
+
msub := m.allocateInstr()
|
|
1245
|
+
msub.asALURRRR(aluOpMSub, rd, operandNR(rd), rm, rn, _64bit)
|
|
1246
|
+
m.insert(msub)
|
|
1247
|
+
}
|
|
1248
|
+
|
|
1249
|
+
func (m *machine) lowerIDiv(execCtxVReg, rd regalloc.VReg, rn, rm operand, _64bit, signed bool) {
|
|
1250
|
+
div := m.allocateInstr()
|
|
1251
|
+
|
|
1252
|
+
if signed {
|
|
1253
|
+
div.asALU(aluOpSDiv, rd, rn, rm, _64bit)
|
|
1254
|
+
} else {
|
|
1255
|
+
div.asALU(aluOpUDiv, rd, rn, rm, _64bit)
|
|
1256
|
+
}
|
|
1257
|
+
m.insert(div)
|
|
1258
|
+
|
|
1259
|
+
// Check if rm is zero:
|
|
1260
|
+
m.exitIfNot(execCtxVReg, registerAsRegNotZeroCond(rm.nr()), _64bit, wazevoapi.ExitCodeIntegerDivisionByZero)
|
|
1261
|
+
|
|
1262
|
+
if signed {
|
|
1263
|
+
// We need to check the signed overflow which happens iff "math.MinInt{32,64} / -1"
|
|
1264
|
+
minusOneCheck := m.allocateInstr()
|
|
1265
|
+
// Sets eq condition if rm == -1.
|
|
1266
|
+
minusOneCheck.asALU(aluOpAddS, xzrVReg, rm, operandImm12(1, 0), _64bit)
|
|
1267
|
+
m.insert(minusOneCheck)
|
|
1268
|
+
|
|
1269
|
+
ccmp := m.allocateInstr()
|
|
1270
|
+
// If eq condition is set, sets the flag by the result based on "rn - 1", otherwise clears the flag.
|
|
1271
|
+
ccmp.asCCmpImm(rn, 1, eq, 0, _64bit)
|
|
1272
|
+
m.insert(ccmp)
|
|
1273
|
+
|
|
1274
|
+
// Check the overflow flag.
|
|
1275
|
+
m.exitIfNot(execCtxVReg, vs.invert().asCond(), false, wazevoapi.ExitCodeIntegerOverflow)
|
|
1276
|
+
}
|
|
1277
|
+
}
|
|
1278
|
+
|
|
1279
|
+
// exitIfNot emits a conditional branch to exit if the condition is not met.
|
|
1280
|
+
// If `c` (cond type) is a register, `cond64bit` must be chosen to indicate whether the register is 32-bit or 64-bit.
|
|
1281
|
+
// Otherwise, `cond64bit` is ignored.
|
|
1282
|
+
func (m *machine) exitIfNot(execCtxVReg regalloc.VReg, c cond, cond64bit bool, code wazevoapi.ExitCode) {
|
|
1283
|
+
execCtxTmp := m.copyToTmp(execCtxVReg)
|
|
1284
|
+
|
|
1285
|
+
cbr := m.allocateInstr()
|
|
1286
|
+
m.insert(cbr)
|
|
1287
|
+
m.lowerExitWithCode(execCtxTmp, code)
|
|
1288
|
+
// Conditional branch target is after exit.
|
|
1289
|
+
l := m.insertBrTargetLabel()
|
|
1290
|
+
cbr.asCondBr(c, l, cond64bit)
|
|
1291
|
+
}
|
|
1292
|
+
|
|
1293
|
+
func (m *machine) lowerFcopysign(x, y, ret ssa.Value) {
|
|
1294
|
+
rn := m.getOperand_NR(m.compiler.ValueDefinition(x), extModeNone)
|
|
1295
|
+
rm := m.getOperand_NR(m.compiler.ValueDefinition(y), extModeNone)
|
|
1296
|
+
var tmpI, tmpF regalloc.VReg
|
|
1297
|
+
_64 := x.Type() == ssa.TypeF64
|
|
1298
|
+
if _64 {
|
|
1299
|
+
tmpF = m.compiler.AllocateVReg(ssa.TypeF64)
|
|
1300
|
+
tmpI = m.compiler.AllocateVReg(ssa.TypeI64)
|
|
1301
|
+
} else {
|
|
1302
|
+
tmpF = m.compiler.AllocateVReg(ssa.TypeF32)
|
|
1303
|
+
tmpI = m.compiler.AllocateVReg(ssa.TypeI32)
|
|
1304
|
+
}
|
|
1305
|
+
rd := m.compiler.VRegOf(ret)
|
|
1306
|
+
m.lowerFcopysignImpl(rd, rn, rm, tmpI, tmpF, _64)
|
|
1307
|
+
}
|
|
1308
|
+
|
|
1309
|
+
func (m *machine) lowerFcopysignImpl(rd regalloc.VReg, rn, rm operand, tmpI, tmpF regalloc.VReg, _64bit bool) {
|
|
1310
|
+
// This is exactly the same code emitted by GCC for "__builtin_copysign":
|
|
1311
|
+
//
|
|
1312
|
+
// mov x0, -9223372036854775808
|
|
1313
|
+
// fmov d2, x0
|
|
1314
|
+
// vbit v0.8b, v1.8b, v2.8b
|
|
1315
|
+
//
|
|
1316
|
+
|
|
1317
|
+
setMSB := m.allocateInstr()
|
|
1318
|
+
if _64bit {
|
|
1319
|
+
m.lowerConstantI64(tmpI, math.MinInt64)
|
|
1320
|
+
setMSB.asMovToVec(tmpF, operandNR(tmpI), vecArrangementD, vecIndex(0))
|
|
1321
|
+
} else {
|
|
1322
|
+
m.lowerConstantI32(tmpI, math.MinInt32)
|
|
1323
|
+
setMSB.asMovToVec(tmpF, operandNR(tmpI), vecArrangementS, vecIndex(0))
|
|
1324
|
+
}
|
|
1325
|
+
m.insert(setMSB)
|
|
1326
|
+
|
|
1327
|
+
tmpReg := m.compiler.AllocateVReg(ssa.TypeF64)
|
|
1328
|
+
|
|
1329
|
+
mov := m.allocateInstr()
|
|
1330
|
+
mov.asFpuMov64(tmpReg, rn.nr())
|
|
1331
|
+
m.insert(mov)
|
|
1332
|
+
|
|
1333
|
+
vbit := m.allocateInstr()
|
|
1334
|
+
vbit.asVecRRRRewrite(vecOpBit, tmpReg, rm, operandNR(tmpF), vecArrangement8B)
|
|
1335
|
+
m.insert(vbit)
|
|
1336
|
+
|
|
1337
|
+
movDst := m.allocateInstr()
|
|
1338
|
+
movDst.asFpuMov64(rd, tmpReg)
|
|
1339
|
+
m.insert(movDst)
|
|
1340
|
+
}
|
|
1341
|
+
|
|
1342
|
+
func (m *machine) lowerBitcast(instr *ssa.Instruction) {
|
|
1343
|
+
v, dstType := instr.BitcastData()
|
|
1344
|
+
srcType := v.Type()
|
|
1345
|
+
rn := m.getOperand_NR(m.compiler.ValueDefinition(v), extModeNone)
|
|
1346
|
+
rd := m.compiler.VRegOf(instr.Return())
|
|
1347
|
+
srcInt := srcType.IsInt()
|
|
1348
|
+
dstInt := dstType.IsInt()
|
|
1349
|
+
switch {
|
|
1350
|
+
case srcInt && !dstInt: // Int to Float:
|
|
1351
|
+
mov := m.allocateInstr()
|
|
1352
|
+
var arr vecArrangement
|
|
1353
|
+
if srcType.Bits() == 64 {
|
|
1354
|
+
arr = vecArrangementD
|
|
1355
|
+
} else {
|
|
1356
|
+
arr = vecArrangementS
|
|
1357
|
+
}
|
|
1358
|
+
mov.asMovToVec(rd, rn, arr, vecIndex(0))
|
|
1359
|
+
m.insert(mov)
|
|
1360
|
+
case !srcInt && dstInt: // Float to Int:
|
|
1361
|
+
mov := m.allocateInstr()
|
|
1362
|
+
var arr vecArrangement
|
|
1363
|
+
if dstType.Bits() == 64 {
|
|
1364
|
+
arr = vecArrangementD
|
|
1365
|
+
} else {
|
|
1366
|
+
arr = vecArrangementS
|
|
1367
|
+
}
|
|
1368
|
+
mov.asMovFromVec(rd, rn, arr, vecIndex(0), false)
|
|
1369
|
+
m.insert(mov)
|
|
1370
|
+
default:
|
|
1371
|
+
panic("TODO?BUG?")
|
|
1372
|
+
}
|
|
1373
|
+
}
|
|
1374
|
+
|
|
1375
|
+
func (m *machine) lowerFpuUniOp(op fpuUniOp, in, out ssa.Value) {
|
|
1376
|
+
rn := m.getOperand_NR(m.compiler.ValueDefinition(in), extModeNone)
|
|
1377
|
+
rd := m.compiler.VRegOf(out)
|
|
1378
|
+
|
|
1379
|
+
neg := m.allocateInstr()
|
|
1380
|
+
neg.asFpuRR(op, rd, rn, in.Type().Bits() == 64)
|
|
1381
|
+
m.insert(neg)
|
|
1382
|
+
}
|
|
1383
|
+
|
|
1384
|
+
func (m *machine) lowerFpuToInt(rd regalloc.VReg, rn operand, ctx regalloc.VReg, signed, src64bit, dst64bit, nonTrapping bool) {
|
|
1385
|
+
if !nonTrapping {
|
|
1386
|
+
// First of all, we have to clear the FPU flags.
|
|
1387
|
+
flagClear := m.allocateInstr()
|
|
1388
|
+
flagClear.asMovToFPSR(xzrVReg)
|
|
1389
|
+
m.insert(flagClear)
|
|
1390
|
+
}
|
|
1391
|
+
|
|
1392
|
+
// Then, do the conversion which doesn't trap inherently.
|
|
1393
|
+
cvt := m.allocateInstr()
|
|
1394
|
+
cvt.asFpuToInt(rd, rn, signed, src64bit, dst64bit)
|
|
1395
|
+
m.insert(cvt)
|
|
1396
|
+
|
|
1397
|
+
if !nonTrapping {
|
|
1398
|
+
tmpReg := m.compiler.AllocateVReg(ssa.TypeI64)
|
|
1399
|
+
|
|
1400
|
+
// After the conversion, check the FPU flags.
|
|
1401
|
+
getFlag := m.allocateInstr()
|
|
1402
|
+
getFlag.asMovFromFPSR(tmpReg)
|
|
1403
|
+
m.insert(getFlag)
|
|
1404
|
+
|
|
1405
|
+
execCtx := m.copyToTmp(ctx)
|
|
1406
|
+
_rn := operandNR(m.copyToTmp(rn.nr()))
|
|
1407
|
+
|
|
1408
|
+
// Check if the conversion was undefined by comparing the status with 1.
|
|
1409
|
+
// See https://developer.arm.com/documentation/ddi0595/2020-12/AArch64-Registers/FPSR--Floating-point-Status-Register
|
|
1410
|
+
alu := m.allocateInstr()
|
|
1411
|
+
alu.asALU(aluOpSubS, xzrVReg, operandNR(tmpReg), operandImm12(1, 0), true)
|
|
1412
|
+
m.insert(alu)
|
|
1413
|
+
|
|
1414
|
+
// If it is not undefined, we can return the result.
|
|
1415
|
+
ok := m.allocateInstr()
|
|
1416
|
+
m.insert(ok)
|
|
1417
|
+
|
|
1418
|
+
// Otherwise, we have to choose the status depending on it is overflow or NaN conversion.
|
|
1419
|
+
|
|
1420
|
+
// Comparing itself to check if it is a NaN.
|
|
1421
|
+
fpuCmp := m.allocateInstr()
|
|
1422
|
+
fpuCmp.asFpuCmp(_rn, _rn, src64bit)
|
|
1423
|
+
m.insert(fpuCmp)
|
|
1424
|
+
// If the VC flag is not set (== VS flag is set), it is a NaN.
|
|
1425
|
+
m.exitIfNot(execCtx, vc.asCond(), false, wazevoapi.ExitCodeInvalidConversionToInteger)
|
|
1426
|
+
// Otherwise, it is an overflow.
|
|
1427
|
+
m.lowerExitWithCode(execCtx, wazevoapi.ExitCodeIntegerOverflow)
|
|
1428
|
+
|
|
1429
|
+
// Conditional branch target is after exit.
|
|
1430
|
+
l := m.insertBrTargetLabel()
|
|
1431
|
+
ok.asCondBr(ne.asCond(), l, false /* ignored */)
|
|
1432
|
+
}
|
|
1433
|
+
}
|
|
1434
|
+
|
|
1435
|
+
func (m *machine) lowerIntToFpu(rd regalloc.VReg, rn operand, signed, src64bit, dst64bit bool) {
|
|
1436
|
+
cvt := m.allocateInstr()
|
|
1437
|
+
cvt.asIntToFpu(rd, rn, signed, src64bit, dst64bit)
|
|
1438
|
+
m.insert(cvt)
|
|
1439
|
+
}
|
|
1440
|
+
|
|
1441
|
+
func (m *machine) lowerFpuBinOp(si *ssa.Instruction) {
|
|
1442
|
+
instr := m.allocateInstr()
|
|
1443
|
+
var op fpuBinOp
|
|
1444
|
+
switch si.Opcode() {
|
|
1445
|
+
case ssa.OpcodeFadd:
|
|
1446
|
+
op = fpuBinOpAdd
|
|
1447
|
+
case ssa.OpcodeFsub:
|
|
1448
|
+
op = fpuBinOpSub
|
|
1449
|
+
case ssa.OpcodeFmul:
|
|
1450
|
+
op = fpuBinOpMul
|
|
1451
|
+
case ssa.OpcodeFdiv:
|
|
1452
|
+
op = fpuBinOpDiv
|
|
1453
|
+
case ssa.OpcodeFmax:
|
|
1454
|
+
op = fpuBinOpMax
|
|
1455
|
+
case ssa.OpcodeFmin:
|
|
1456
|
+
op = fpuBinOpMin
|
|
1457
|
+
}
|
|
1458
|
+
x, y := si.Arg2()
|
|
1459
|
+
xDef, yDef := m.compiler.ValueDefinition(x), m.compiler.ValueDefinition(y)
|
|
1460
|
+
rn := m.getOperand_NR(xDef, extModeNone)
|
|
1461
|
+
rm := m.getOperand_NR(yDef, extModeNone)
|
|
1462
|
+
rd := m.compiler.VRegOf(si.Return())
|
|
1463
|
+
instr.asFpuRRR(op, rd, rn, rm, x.Type().Bits() == 64)
|
|
1464
|
+
m.insert(instr)
|
|
1465
|
+
}
|
|
1466
|
+
|
|
1467
|
+
func (m *machine) lowerSubOrAdd(si *ssa.Instruction, add bool) {
|
|
1468
|
+
x, y := si.Arg2()
|
|
1469
|
+
if !x.Type().IsInt() {
|
|
1470
|
+
panic("BUG?")
|
|
1471
|
+
}
|
|
1472
|
+
|
|
1473
|
+
xDef, yDef := m.compiler.ValueDefinition(x), m.compiler.ValueDefinition(y)
|
|
1474
|
+
rn := m.getOperand_NR(xDef, extModeNone)
|
|
1475
|
+
rm, yNegated := m.getOperand_MaybeNegatedImm12_ER_SR_NR(yDef, extModeNone)
|
|
1476
|
+
|
|
1477
|
+
var aop aluOp
|
|
1478
|
+
switch {
|
|
1479
|
+
case add && !yNegated: // rn+rm = x+y
|
|
1480
|
+
aop = aluOpAdd
|
|
1481
|
+
case add && yNegated: // rn-rm = x-(-y) = x+y
|
|
1482
|
+
aop = aluOpSub
|
|
1483
|
+
case !add && !yNegated: // rn-rm = x-y
|
|
1484
|
+
aop = aluOpSub
|
|
1485
|
+
case !add && yNegated: // rn+rm = x-(-y) = x-y
|
|
1486
|
+
aop = aluOpAdd
|
|
1487
|
+
}
|
|
1488
|
+
rd := m.compiler.VRegOf(si.Return())
|
|
1489
|
+
alu := m.allocateInstr()
|
|
1490
|
+
alu.asALU(aop, rd, rn, rm, x.Type().Bits() == 64)
|
|
1491
|
+
m.insert(alu)
|
|
1492
|
+
}
|
|
1493
|
+
|
|
1494
|
+
// InsertMove implements backend.Machine.
|
|
1495
|
+
func (m *machine) InsertMove(dst, src regalloc.VReg, typ ssa.Type) {
|
|
1496
|
+
instr := m.allocateInstr()
|
|
1497
|
+
switch typ {
|
|
1498
|
+
case ssa.TypeI32, ssa.TypeI64:
|
|
1499
|
+
instr.asMove64(dst, src)
|
|
1500
|
+
case ssa.TypeF32, ssa.TypeF64:
|
|
1501
|
+
instr.asFpuMov64(dst, src)
|
|
1502
|
+
case ssa.TypeV128:
|
|
1503
|
+
instr.asFpuMov128(dst, src)
|
|
1504
|
+
default:
|
|
1505
|
+
panic("TODO")
|
|
1506
|
+
}
|
|
1507
|
+
m.insert(instr)
|
|
1508
|
+
}
|
|
1509
|
+
|
|
1510
|
+
func (m *machine) lowerIcmp(si *ssa.Instruction) {
|
|
1511
|
+
x, y, c := si.IcmpData()
|
|
1512
|
+
flag := condFlagFromSSAIntegerCmpCond(c)
|
|
1513
|
+
|
|
1514
|
+
in64bit := x.Type().Bits() == 64
|
|
1515
|
+
var ext extMode
|
|
1516
|
+
if in64bit {
|
|
1517
|
+
if c.Signed() {
|
|
1518
|
+
ext = extModeSignExtend64
|
|
1519
|
+
} else {
|
|
1520
|
+
ext = extModeZeroExtend64
|
|
1521
|
+
}
|
|
1522
|
+
} else {
|
|
1523
|
+
if c.Signed() {
|
|
1524
|
+
ext = extModeSignExtend32
|
|
1525
|
+
} else {
|
|
1526
|
+
ext = extModeZeroExtend32
|
|
1527
|
+
}
|
|
1528
|
+
}
|
|
1529
|
+
|
|
1530
|
+
rn := m.getOperand_NR(m.compiler.ValueDefinition(x), ext)
|
|
1531
|
+
rm := m.getOperand_Imm12_ER_SR_NR(m.compiler.ValueDefinition(y), ext)
|
|
1532
|
+
alu := m.allocateInstr()
|
|
1533
|
+
alu.asALU(aluOpSubS, xzrVReg, rn, rm, in64bit)
|
|
1534
|
+
m.insert(alu)
|
|
1535
|
+
|
|
1536
|
+
cset := m.allocateInstr()
|
|
1537
|
+
cset.asCSet(m.compiler.VRegOf(si.Return()), false, flag)
|
|
1538
|
+
m.insert(cset)
|
|
1539
|
+
}
|
|
1540
|
+
|
|
1541
|
+
func (m *machine) lowerVIcmp(si *ssa.Instruction) {
|
|
1542
|
+
x, y, c, lane := si.VIcmpData()
|
|
1543
|
+
flag := condFlagFromSSAIntegerCmpCond(c)
|
|
1544
|
+
arr := ssaLaneToArrangement(lane)
|
|
1545
|
+
|
|
1546
|
+
rn := m.getOperand_NR(m.compiler.ValueDefinition(x), extModeNone)
|
|
1547
|
+
rm := m.getOperand_NR(m.compiler.ValueDefinition(y), extModeNone)
|
|
1548
|
+
rd := m.compiler.VRegOf(si.Return())
|
|
1549
|
+
|
|
1550
|
+
switch flag {
|
|
1551
|
+
case eq:
|
|
1552
|
+
cmp := m.allocateInstr()
|
|
1553
|
+
cmp.asVecRRR(vecOpCmeq, rd, rn, rm, arr)
|
|
1554
|
+
m.insert(cmp)
|
|
1555
|
+
case ne:
|
|
1556
|
+
cmp := m.allocateInstr()
|
|
1557
|
+
cmp.asVecRRR(vecOpCmeq, rd, rn, rm, arr)
|
|
1558
|
+
m.insert(cmp)
|
|
1559
|
+
not := m.allocateInstr()
|
|
1560
|
+
not.asVecMisc(vecOpNot, rd, operandNR(rd), vecArrangement16B)
|
|
1561
|
+
m.insert(not)
|
|
1562
|
+
case ge:
|
|
1563
|
+
cmp := m.allocateInstr()
|
|
1564
|
+
cmp.asVecRRR(vecOpCmge, rd, rn, rm, arr)
|
|
1565
|
+
m.insert(cmp)
|
|
1566
|
+
case gt:
|
|
1567
|
+
cmp := m.allocateInstr()
|
|
1568
|
+
cmp.asVecRRR(vecOpCmgt, rd, rn, rm, arr)
|
|
1569
|
+
m.insert(cmp)
|
|
1570
|
+
case le:
|
|
1571
|
+
cmp := m.allocateInstr()
|
|
1572
|
+
cmp.asVecRRR(vecOpCmge, rd, rm, rn, arr) // rm, rn are swapped
|
|
1573
|
+
m.insert(cmp)
|
|
1574
|
+
case lt:
|
|
1575
|
+
cmp := m.allocateInstr()
|
|
1576
|
+
cmp.asVecRRR(vecOpCmgt, rd, rm, rn, arr) // rm, rn are swapped
|
|
1577
|
+
m.insert(cmp)
|
|
1578
|
+
case hs:
|
|
1579
|
+
cmp := m.allocateInstr()
|
|
1580
|
+
cmp.asVecRRR(vecOpCmhs, rd, rn, rm, arr)
|
|
1581
|
+
m.insert(cmp)
|
|
1582
|
+
case hi:
|
|
1583
|
+
cmp := m.allocateInstr()
|
|
1584
|
+
cmp.asVecRRR(vecOpCmhi, rd, rn, rm, arr)
|
|
1585
|
+
m.insert(cmp)
|
|
1586
|
+
case ls:
|
|
1587
|
+
cmp := m.allocateInstr()
|
|
1588
|
+
cmp.asVecRRR(vecOpCmhs, rd, rm, rn, arr) // rm, rn are swapped
|
|
1589
|
+
m.insert(cmp)
|
|
1590
|
+
case lo:
|
|
1591
|
+
cmp := m.allocateInstr()
|
|
1592
|
+
cmp.asVecRRR(vecOpCmhi, rd, rm, rn, arr) // rm, rn are swapped
|
|
1593
|
+
m.insert(cmp)
|
|
1594
|
+
}
|
|
1595
|
+
}
|
|
1596
|
+
|
|
1597
|
+
func (m *machine) lowerVFcmp(si *ssa.Instruction) {
|
|
1598
|
+
x, y, c, lane := si.VFcmpData()
|
|
1599
|
+
flag := condFlagFromSSAFloatCmpCond(c)
|
|
1600
|
+
arr := ssaLaneToArrangement(lane)
|
|
1601
|
+
|
|
1602
|
+
rn := m.getOperand_NR(m.compiler.ValueDefinition(x), extModeNone)
|
|
1603
|
+
rm := m.getOperand_NR(m.compiler.ValueDefinition(y), extModeNone)
|
|
1604
|
+
rd := m.compiler.VRegOf(si.Return())
|
|
1605
|
+
|
|
1606
|
+
switch flag {
|
|
1607
|
+
case eq:
|
|
1608
|
+
cmp := m.allocateInstr()
|
|
1609
|
+
cmp.asVecRRR(vecOpFcmeq, rd, rn, rm, arr)
|
|
1610
|
+
m.insert(cmp)
|
|
1611
|
+
case ne:
|
|
1612
|
+
cmp := m.allocateInstr()
|
|
1613
|
+
cmp.asVecRRR(vecOpFcmeq, rd, rn, rm, arr)
|
|
1614
|
+
m.insert(cmp)
|
|
1615
|
+
not := m.allocateInstr()
|
|
1616
|
+
not.asVecMisc(vecOpNot, rd, operandNR(rd), vecArrangement16B)
|
|
1617
|
+
m.insert(not)
|
|
1618
|
+
case ge:
|
|
1619
|
+
cmp := m.allocateInstr()
|
|
1620
|
+
cmp.asVecRRR(vecOpFcmge, rd, rn, rm, arr)
|
|
1621
|
+
m.insert(cmp)
|
|
1622
|
+
case gt:
|
|
1623
|
+
cmp := m.allocateInstr()
|
|
1624
|
+
cmp.asVecRRR(vecOpFcmgt, rd, rn, rm, arr)
|
|
1625
|
+
m.insert(cmp)
|
|
1626
|
+
case mi:
|
|
1627
|
+
cmp := m.allocateInstr()
|
|
1628
|
+
cmp.asVecRRR(vecOpFcmgt, rd, rm, rn, arr) // rm, rn are swapped
|
|
1629
|
+
m.insert(cmp)
|
|
1630
|
+
case ls:
|
|
1631
|
+
cmp := m.allocateInstr()
|
|
1632
|
+
cmp.asVecRRR(vecOpFcmge, rd, rm, rn, arr) // rm, rn are swapped
|
|
1633
|
+
m.insert(cmp)
|
|
1634
|
+
}
|
|
1635
|
+
}
|
|
1636
|
+
|
|
1637
|
+
func (m *machine) lowerVfpuToInt(rd regalloc.VReg, rn operand, arr vecArrangement, signed bool) {
|
|
1638
|
+
cvt := m.allocateInstr()
|
|
1639
|
+
if signed {
|
|
1640
|
+
cvt.asVecMisc(vecOpFcvtzs, rd, rn, arr)
|
|
1641
|
+
} else {
|
|
1642
|
+
cvt.asVecMisc(vecOpFcvtzu, rd, rn, arr)
|
|
1643
|
+
}
|
|
1644
|
+
m.insert(cvt)
|
|
1645
|
+
|
|
1646
|
+
if arr == vecArrangement2D {
|
|
1647
|
+
narrow := m.allocateInstr()
|
|
1648
|
+
if signed {
|
|
1649
|
+
narrow.asVecMisc(vecOpSqxtn, rd, operandNR(rd), vecArrangement2S)
|
|
1650
|
+
} else {
|
|
1651
|
+
narrow.asVecMisc(vecOpUqxtn, rd, operandNR(rd), vecArrangement2S)
|
|
1652
|
+
}
|
|
1653
|
+
m.insert(narrow)
|
|
1654
|
+
}
|
|
1655
|
+
}
|
|
1656
|
+
|
|
1657
|
+
func (m *machine) lowerVfpuFromInt(rd regalloc.VReg, rn operand, arr vecArrangement, signed bool) {
|
|
1658
|
+
cvt := m.allocateInstr()
|
|
1659
|
+
if signed {
|
|
1660
|
+
cvt.asVecMisc(vecOpScvtf, rd, rn, arr)
|
|
1661
|
+
} else {
|
|
1662
|
+
cvt.asVecMisc(vecOpUcvtf, rd, rn, arr)
|
|
1663
|
+
}
|
|
1664
|
+
m.insert(cvt)
|
|
1665
|
+
}
|
|
1666
|
+
|
|
1667
|
+
func (m *machine) lowerShifts(si *ssa.Instruction, ext extMode, aluOp aluOp) {
|
|
1668
|
+
x, amount := si.Arg2()
|
|
1669
|
+
rn := m.getOperand_NR(m.compiler.ValueDefinition(x), ext)
|
|
1670
|
+
rm := m.getOperand_ShiftImm_NR(m.compiler.ValueDefinition(amount), ext, x.Type().Bits())
|
|
1671
|
+
rd := m.compiler.VRegOf(si.Return())
|
|
1672
|
+
|
|
1673
|
+
alu := m.allocateInstr()
|
|
1674
|
+
alu.asALUShift(aluOp, rd, rn, rm, x.Type().Bits() == 64)
|
|
1675
|
+
m.insert(alu)
|
|
1676
|
+
}
|
|
1677
|
+
|
|
1678
|
+
func (m *machine) lowerBitwiseAluOp(si *ssa.Instruction, op aluOp, ignoreResult bool) {
|
|
1679
|
+
x, y := si.Arg2()
|
|
1680
|
+
|
|
1681
|
+
xDef, yDef := m.compiler.ValueDefinition(x), m.compiler.ValueDefinition(y)
|
|
1682
|
+
rn := m.getOperand_NR(xDef, extModeNone)
|
|
1683
|
+
|
|
1684
|
+
var rd regalloc.VReg
|
|
1685
|
+
if ignoreResult {
|
|
1686
|
+
rd = xzrVReg
|
|
1687
|
+
} else {
|
|
1688
|
+
rd = m.compiler.VRegOf(si.Return())
|
|
1689
|
+
}
|
|
1690
|
+
|
|
1691
|
+
_64 := x.Type().Bits() == 64
|
|
1692
|
+
alu := m.allocateInstr()
|
|
1693
|
+
if instr := yDef.Instr; instr != nil && instr.Constant() {
|
|
1694
|
+
c := instr.ConstantVal()
|
|
1695
|
+
if isBitMaskImmediate(c, _64) {
|
|
1696
|
+
// Constant bit wise operations can be lowered to a single instruction.
|
|
1697
|
+
alu.asALUBitmaskImm(op, rd, rn.nr(), c, _64)
|
|
1698
|
+
m.insert(alu)
|
|
1699
|
+
return
|
|
1700
|
+
}
|
|
1701
|
+
}
|
|
1702
|
+
|
|
1703
|
+
rm := m.getOperand_SR_NR(yDef, extModeNone)
|
|
1704
|
+
alu.asALU(op, rd, rn, rm, _64)
|
|
1705
|
+
m.insert(alu)
|
|
1706
|
+
}
|
|
1707
|
+
|
|
1708
|
+
func (m *machine) lowerRotl(si *ssa.Instruction) {
|
|
1709
|
+
x, y := si.Arg2()
|
|
1710
|
+
r := si.Return()
|
|
1711
|
+
_64 := r.Type().Bits() == 64
|
|
1712
|
+
|
|
1713
|
+
rn := m.getOperand_NR(m.compiler.ValueDefinition(x), extModeNone)
|
|
1714
|
+
rm := m.getOperand_NR(m.compiler.ValueDefinition(y), extModeNone)
|
|
1715
|
+
var tmp regalloc.VReg
|
|
1716
|
+
if _64 {
|
|
1717
|
+
tmp = m.compiler.AllocateVReg(ssa.TypeI64)
|
|
1718
|
+
} else {
|
|
1719
|
+
tmp = m.compiler.AllocateVReg(ssa.TypeI32)
|
|
1720
|
+
}
|
|
1721
|
+
rd := m.compiler.VRegOf(r)
|
|
1722
|
+
|
|
1723
|
+
// Encode rotl as neg + rotr: neg is a sub against the zero-reg.
|
|
1724
|
+
m.lowerRotlImpl(rd, rn, rm, tmp, _64)
|
|
1725
|
+
}
|
|
1726
|
+
|
|
1727
|
+
func (m *machine) lowerRotlImpl(rd regalloc.VReg, rn, rm operand, tmp regalloc.VReg, is64bit bool) {
|
|
1728
|
+
// Encode rotl as neg + rotr: neg is a sub against the zero-reg.
|
|
1729
|
+
neg := m.allocateInstr()
|
|
1730
|
+
neg.asALU(aluOpSub, tmp, operandNR(xzrVReg), rm, is64bit)
|
|
1731
|
+
m.insert(neg)
|
|
1732
|
+
alu := m.allocateInstr()
|
|
1733
|
+
alu.asALU(aluOpRotR, rd, rn, operandNR(tmp), is64bit)
|
|
1734
|
+
m.insert(alu)
|
|
1735
|
+
}
|
|
1736
|
+
|
|
1737
|
+
func (m *machine) lowerRotr(si *ssa.Instruction) {
|
|
1738
|
+
x, y := si.Arg2()
|
|
1739
|
+
|
|
1740
|
+
xDef, yDef := m.compiler.ValueDefinition(x), m.compiler.ValueDefinition(y)
|
|
1741
|
+
rn := m.getOperand_NR(xDef, extModeNone)
|
|
1742
|
+
rm := m.getOperand_NR(yDef, extModeNone)
|
|
1743
|
+
rd := m.compiler.VRegOf(si.Return())
|
|
1744
|
+
|
|
1745
|
+
alu := m.allocateInstr()
|
|
1746
|
+
alu.asALU(aluOpRotR, rd, rn, rm, si.Return().Type().Bits() == 64)
|
|
1747
|
+
m.insert(alu)
|
|
1748
|
+
}
|
|
1749
|
+
|
|
1750
|
+
func (m *machine) lowerExtend(arg, ret ssa.Value, from, to byte, signed bool) {
|
|
1751
|
+
rd := m.compiler.VRegOf(ret)
|
|
1752
|
+
def := m.compiler.ValueDefinition(arg)
|
|
1753
|
+
|
|
1754
|
+
if instr := def.Instr; !signed && from == 32 && instr != nil {
|
|
1755
|
+
// We can optimize out the unsigned extend because:
|
|
1756
|
+
// Writes to the W register set bits [63:32] of the X register to zero
|
|
1757
|
+
// https://developer.arm.com/documentation/den0024/a/An-Introduction-to-the-ARMv8-Instruction-Sets/The-ARMv8-instruction-sets/Distinguishing-between-32-bit-and-64-bit-A64-instructions
|
|
1758
|
+
switch instr.Opcode() {
|
|
1759
|
+
case
|
|
1760
|
+
ssa.OpcodeIadd, ssa.OpcodeIsub, ssa.OpcodeLoad,
|
|
1761
|
+
ssa.OpcodeBand, ssa.OpcodeBor, ssa.OpcodeBnot,
|
|
1762
|
+
ssa.OpcodeIshl, ssa.OpcodeUshr, ssa.OpcodeSshr,
|
|
1763
|
+
ssa.OpcodeRotl, ssa.OpcodeRotr,
|
|
1764
|
+
ssa.OpcodeUload8, ssa.OpcodeUload16, ssa.OpcodeUload32:
|
|
1765
|
+
// So, if the argument is the result of a 32-bit operation, we can just copy the register.
|
|
1766
|
+
// It is highly likely that this copy will be optimized out after register allocation.
|
|
1767
|
+
rn := m.compiler.VRegOf(arg)
|
|
1768
|
+
mov := m.allocateInstr()
|
|
1769
|
+
// Note: do not use move32 as it will be lowered to a 32-bit move, which is not copy (that is actually the impl of UExtend).
|
|
1770
|
+
mov.asMove64(rd, rn)
|
|
1771
|
+
m.insert(mov)
|
|
1772
|
+
return
|
|
1773
|
+
default:
|
|
1774
|
+
}
|
|
1775
|
+
}
|
|
1776
|
+
rn := m.getOperand_NR(def, extModeNone)
|
|
1777
|
+
|
|
1778
|
+
ext := m.allocateInstr()
|
|
1779
|
+
ext.asExtend(rd, rn.nr(), from, to, signed)
|
|
1780
|
+
m.insert(ext)
|
|
1781
|
+
}
|
|
1782
|
+
|
|
1783
|
+
func (m *machine) lowerFcmp(x, y, result ssa.Value, c ssa.FloatCmpCond) {
|
|
1784
|
+
rn, rm := m.getOperand_NR(m.compiler.ValueDefinition(x), extModeNone), m.getOperand_NR(m.compiler.ValueDefinition(y), extModeNone)
|
|
1785
|
+
|
|
1786
|
+
fc := m.allocateInstr()
|
|
1787
|
+
fc.asFpuCmp(rn, rm, x.Type().Bits() == 64)
|
|
1788
|
+
m.insert(fc)
|
|
1789
|
+
|
|
1790
|
+
cset := m.allocateInstr()
|
|
1791
|
+
cset.asCSet(m.compiler.VRegOf(result), false, condFlagFromSSAFloatCmpCond(c))
|
|
1792
|
+
m.insert(cset)
|
|
1793
|
+
}
|
|
1794
|
+
|
|
1795
|
+
func (m *machine) lowerImul(x, y, result ssa.Value) {
|
|
1796
|
+
rd := m.compiler.VRegOf(result)
|
|
1797
|
+
rn := m.getOperand_NR(m.compiler.ValueDefinition(x), extModeNone)
|
|
1798
|
+
rm := m.getOperand_NR(m.compiler.ValueDefinition(y), extModeNone)
|
|
1799
|
+
|
|
1800
|
+
// TODO: if this comes before Add/Sub, we could merge it by putting it into the place of xzrVReg.
|
|
1801
|
+
|
|
1802
|
+
mul := m.allocateInstr()
|
|
1803
|
+
mul.asALURRRR(aluOpMAdd, rd, rn, rm, xzrVReg, x.Type().Bits() == 64)
|
|
1804
|
+
m.insert(mul)
|
|
1805
|
+
}
|
|
1806
|
+
|
|
1807
|
+
func (m *machine) lowerClz(x, result ssa.Value) {
|
|
1808
|
+
rd := m.compiler.VRegOf(result)
|
|
1809
|
+
rn := m.getOperand_NR(m.compiler.ValueDefinition(x), extModeNone)
|
|
1810
|
+
clz := m.allocateInstr()
|
|
1811
|
+
clz.asBitRR(bitOpClz, rd, rn.nr(), x.Type().Bits() == 64)
|
|
1812
|
+
m.insert(clz)
|
|
1813
|
+
}
|
|
1814
|
+
|
|
1815
|
+
func (m *machine) lowerCtz(x, result ssa.Value) {
|
|
1816
|
+
rd := m.compiler.VRegOf(result)
|
|
1817
|
+
rn := m.getOperand_NR(m.compiler.ValueDefinition(x), extModeNone)
|
|
1818
|
+
rbit := m.allocateInstr()
|
|
1819
|
+
_64 := x.Type().Bits() == 64
|
|
1820
|
+
var tmpReg regalloc.VReg
|
|
1821
|
+
if _64 {
|
|
1822
|
+
tmpReg = m.compiler.AllocateVReg(ssa.TypeI64)
|
|
1823
|
+
} else {
|
|
1824
|
+
tmpReg = m.compiler.AllocateVReg(ssa.TypeI32)
|
|
1825
|
+
}
|
|
1826
|
+
rbit.asBitRR(bitOpRbit, tmpReg, rn.nr(), _64)
|
|
1827
|
+
m.insert(rbit)
|
|
1828
|
+
|
|
1829
|
+
clz := m.allocateInstr()
|
|
1830
|
+
clz.asBitRR(bitOpClz, rd, tmpReg, _64)
|
|
1831
|
+
m.insert(clz)
|
|
1832
|
+
}
|
|
1833
|
+
|
|
1834
|
+
func (m *machine) lowerPopcnt(x, result ssa.Value) {
|
|
1835
|
+
// arm64 doesn't have an instruction for population count on scalar register,
|
|
1836
|
+
// so we use the vector instruction `cnt`.
|
|
1837
|
+
// This is exactly what the official Go implements bits.OneCount.
|
|
1838
|
+
// For example, "func () int { return bits.OneCount(10) }" is compiled as
|
|
1839
|
+
//
|
|
1840
|
+
// MOVD $10, R0 ;; Load 10.
|
|
1841
|
+
// FMOVD R0, F0
|
|
1842
|
+
// VCNT V0.B8, V0.B8
|
|
1843
|
+
// UADDLV V0.B8, V0
|
|
1844
|
+
//
|
|
1845
|
+
// In aarch64 asm, FMOVD is encoded as `ins`, VCNT is `cnt`,
|
|
1846
|
+
// and the registers may use different names. In our encoding we use the following
|
|
1847
|
+
// instructions:
|
|
1848
|
+
//
|
|
1849
|
+
// ins v0.d[0], x0 ;; mov from GPR to vec (FMOV above) is encoded as INS
|
|
1850
|
+
// cnt v0.16b, v0.16b ;; we use vec arrangement 16b
|
|
1851
|
+
// uaddlv h0, v0.8b ;; h0 is still v0 with the dest width specifier 'H', implied when src arrangement is 8b
|
|
1852
|
+
// mov x5, v0.d[0] ;; finally we mov the result back to a GPR
|
|
1853
|
+
//
|
|
1854
|
+
|
|
1855
|
+
rd := m.compiler.VRegOf(result)
|
|
1856
|
+
rn := m.getOperand_NR(m.compiler.ValueDefinition(x), extModeNone)
|
|
1857
|
+
|
|
1858
|
+
rf1 := operandNR(m.compiler.AllocateVReg(ssa.TypeF64))
|
|
1859
|
+
ins := m.allocateInstr()
|
|
1860
|
+
ins.asMovToVec(rf1.nr(), rn, vecArrangementD, vecIndex(0))
|
|
1861
|
+
m.insert(ins)
|
|
1862
|
+
|
|
1863
|
+
rf2 := operandNR(m.compiler.AllocateVReg(ssa.TypeF64))
|
|
1864
|
+
cnt := m.allocateInstr()
|
|
1865
|
+
cnt.asVecMisc(vecOpCnt, rf2.nr(), rf1, vecArrangement16B)
|
|
1866
|
+
m.insert(cnt)
|
|
1867
|
+
|
|
1868
|
+
rf3 := operandNR(m.compiler.AllocateVReg(ssa.TypeF64))
|
|
1869
|
+
uaddlv := m.allocateInstr()
|
|
1870
|
+
uaddlv.asVecLanes(vecOpUaddlv, rf3.nr(), rf2, vecArrangement8B)
|
|
1871
|
+
m.insert(uaddlv)
|
|
1872
|
+
|
|
1873
|
+
mov := m.allocateInstr()
|
|
1874
|
+
mov.asMovFromVec(rd, rf3, vecArrangementD, vecIndex(0), false)
|
|
1875
|
+
m.insert(mov)
|
|
1876
|
+
}
|
|
1877
|
+
|
|
1878
|
+
// lowerExitWithCode lowers the lowerExitWithCode takes a context pointer as argument.
|
|
1879
|
+
func (m *machine) lowerExitWithCode(execCtxVReg regalloc.VReg, code wazevoapi.ExitCode) {
|
|
1880
|
+
tmpReg1 := m.compiler.AllocateVReg(ssa.TypeI32)
|
|
1881
|
+
loadExitCodeConst := m.allocateInstr()
|
|
1882
|
+
loadExitCodeConst.asMOVZ(tmpReg1, uint64(code), 0, true)
|
|
1883
|
+
|
|
1884
|
+
setExitCode := m.allocateInstr()
|
|
1885
|
+
mode := m.amodePool.Allocate()
|
|
1886
|
+
*mode = addressMode{
|
|
1887
|
+
kind: addressModeKindRegUnsignedImm12,
|
|
1888
|
+
rn: execCtxVReg, imm: wazevoapi.ExecutionContextOffsetExitCodeOffset.I64(),
|
|
1889
|
+
}
|
|
1890
|
+
setExitCode.asStore(operandNR(tmpReg1), mode, 32)
|
|
1891
|
+
|
|
1892
|
+
// In order to unwind the stack, we also need to push the current stack pointer:
|
|
1893
|
+
tmp2 := m.compiler.AllocateVReg(ssa.TypeI64)
|
|
1894
|
+
movSpToTmp := m.allocateInstr()
|
|
1895
|
+
movSpToTmp.asMove64(tmp2, spVReg)
|
|
1896
|
+
strSpToExecCtx := m.allocateInstr()
|
|
1897
|
+
mode2 := m.amodePool.Allocate()
|
|
1898
|
+
*mode2 = addressMode{
|
|
1899
|
+
kind: addressModeKindRegUnsignedImm12,
|
|
1900
|
+
rn: execCtxVReg, imm: wazevoapi.ExecutionContextOffsetStackPointerBeforeGoCall.I64(),
|
|
1901
|
+
}
|
|
1902
|
+
strSpToExecCtx.asStore(operandNR(tmp2), mode2, 64)
|
|
1903
|
+
// Also the address of this exit.
|
|
1904
|
+
tmp3 := m.compiler.AllocateVReg(ssa.TypeI64)
|
|
1905
|
+
currentAddrToTmp := m.allocateInstr()
|
|
1906
|
+
currentAddrToTmp.asAdr(tmp3, 0)
|
|
1907
|
+
storeCurrentAddrToExecCtx := m.allocateInstr()
|
|
1908
|
+
mode3 := m.amodePool.Allocate()
|
|
1909
|
+
*mode3 = addressMode{
|
|
1910
|
+
kind: addressModeKindRegUnsignedImm12,
|
|
1911
|
+
rn: execCtxVReg, imm: wazevoapi.ExecutionContextOffsetGoCallReturnAddress.I64(),
|
|
1912
|
+
}
|
|
1913
|
+
storeCurrentAddrToExecCtx.asStore(operandNR(tmp3), mode3, 64)
|
|
1914
|
+
|
|
1915
|
+
exitSeq := m.allocateInstr()
|
|
1916
|
+
exitSeq.asExitSequence(execCtxVReg)
|
|
1917
|
+
|
|
1918
|
+
m.insert(loadExitCodeConst)
|
|
1919
|
+
m.insert(setExitCode)
|
|
1920
|
+
m.insert(movSpToTmp)
|
|
1921
|
+
m.insert(strSpToExecCtx)
|
|
1922
|
+
m.insert(currentAddrToTmp)
|
|
1923
|
+
m.insert(storeCurrentAddrToExecCtx)
|
|
1924
|
+
m.insert(exitSeq)
|
|
1925
|
+
}
|
|
1926
|
+
|
|
1927
|
+
func (m *machine) lowerIcmpToFlag(x, y ssa.Value, signed bool) {
|
|
1928
|
+
if x.Type() != y.Type() {
|
|
1929
|
+
panic(
|
|
1930
|
+
fmt.Sprintf("TODO(maybe): support icmp with different types: v%d=%s != v%d=%s",
|
|
1931
|
+
x.ID(), x.Type(), y.ID(), y.Type()))
|
|
1932
|
+
}
|
|
1933
|
+
|
|
1934
|
+
extMod := extModeOf(x.Type(), signed)
|
|
1935
|
+
|
|
1936
|
+
// First operand must be in pure register form.
|
|
1937
|
+
rn := m.getOperand_NR(m.compiler.ValueDefinition(x), extMod)
|
|
1938
|
+
// Second operand can be in any of Imm12, ER, SR, or NR form supported by the SUBS instructions.
|
|
1939
|
+
rm := m.getOperand_Imm12_ER_SR_NR(m.compiler.ValueDefinition(y), extMod)
|
|
1940
|
+
|
|
1941
|
+
alu := m.allocateInstr()
|
|
1942
|
+
// subs zr, rn, rm
|
|
1943
|
+
alu.asALU(
|
|
1944
|
+
aluOpSubS,
|
|
1945
|
+
// We don't need the result, just need to set flags.
|
|
1946
|
+
xzrVReg,
|
|
1947
|
+
rn,
|
|
1948
|
+
rm,
|
|
1949
|
+
x.Type().Bits() == 64,
|
|
1950
|
+
)
|
|
1951
|
+
m.insert(alu)
|
|
1952
|
+
}
|
|
1953
|
+
|
|
1954
|
+
func (m *machine) lowerFcmpToFlag(x, y ssa.Value) {
|
|
1955
|
+
if x.Type() != y.Type() {
|
|
1956
|
+
panic("TODO(maybe): support icmp with different types")
|
|
1957
|
+
}
|
|
1958
|
+
|
|
1959
|
+
rn := m.getOperand_NR(m.compiler.ValueDefinition(x), extModeNone)
|
|
1960
|
+
rm := m.getOperand_NR(m.compiler.ValueDefinition(y), extModeNone)
|
|
1961
|
+
cmp := m.allocateInstr()
|
|
1962
|
+
cmp.asFpuCmp(rn, rm, x.Type().Bits() == 64)
|
|
1963
|
+
m.insert(cmp)
|
|
1964
|
+
}
|
|
1965
|
+
|
|
1966
|
+
func (m *machine) lowerExitIfTrueWithCode(execCtxVReg regalloc.VReg, cond ssa.Value, code wazevoapi.ExitCode) {
|
|
1967
|
+
condDef := m.compiler.ValueDefinition(cond)
|
|
1968
|
+
if !m.compiler.MatchInstr(condDef, ssa.OpcodeIcmp) {
|
|
1969
|
+
panic("TODO: OpcodeExitIfTrueWithCode must come after Icmp at the moment: " + condDef.Instr.Opcode().String())
|
|
1970
|
+
}
|
|
1971
|
+
condDef.Instr.MarkLowered()
|
|
1972
|
+
|
|
1973
|
+
cvalInstr := condDef.Instr
|
|
1974
|
+
x, y, c := cvalInstr.IcmpData()
|
|
1975
|
+
signed := c.Signed()
|
|
1976
|
+
|
|
1977
|
+
if !m.tryLowerBandToFlag(x, y) {
|
|
1978
|
+
m.lowerIcmpToFlag(x, y, signed)
|
|
1979
|
+
}
|
|
1980
|
+
|
|
1981
|
+
// We need to copy the execution context to a temp register, because if it's spilled,
|
|
1982
|
+
// it might end up being reloaded inside the exiting branch.
|
|
1983
|
+
execCtxTmp := m.copyToTmp(execCtxVReg)
|
|
1984
|
+
|
|
1985
|
+
// We have to skip the entire exit sequence if the condition is false.
|
|
1986
|
+
cbr := m.allocateInstr()
|
|
1987
|
+
m.insert(cbr)
|
|
1988
|
+
m.lowerExitWithCode(execCtxTmp, code)
|
|
1989
|
+
// conditional branch target is after exit.
|
|
1990
|
+
l := m.insertBrTargetLabel()
|
|
1991
|
+
cbr.asCondBr(condFlagFromSSAIntegerCmpCond(c).invert().asCond(), l, false /* ignored */)
|
|
1992
|
+
}
|
|
1993
|
+
|
|
1994
|
+
func (m *machine) lowerSelect(c, x, y, result ssa.Value) {
|
|
1995
|
+
cvalDef := m.compiler.ValueDefinition(c)
|
|
1996
|
+
|
|
1997
|
+
var cc condFlag
|
|
1998
|
+
switch {
|
|
1999
|
+
case m.compiler.MatchInstr(cvalDef, ssa.OpcodeIcmp): // This case, we can use the ALU flag set by SUBS instruction.
|
|
2000
|
+
cvalInstr := cvalDef.Instr
|
|
2001
|
+
x, y, c := cvalInstr.IcmpData()
|
|
2002
|
+
cc = condFlagFromSSAIntegerCmpCond(c)
|
|
2003
|
+
m.lowerIcmpToFlag(x, y, c.Signed())
|
|
2004
|
+
cvalDef.Instr.MarkLowered()
|
|
2005
|
+
case m.compiler.MatchInstr(cvalDef, ssa.OpcodeFcmp): // This case we can use the Fpu flag directly.
|
|
2006
|
+
cvalInstr := cvalDef.Instr
|
|
2007
|
+
x, y, c := cvalInstr.FcmpData()
|
|
2008
|
+
cc = condFlagFromSSAFloatCmpCond(c)
|
|
2009
|
+
m.lowerFcmpToFlag(x, y)
|
|
2010
|
+
cvalDef.Instr.MarkLowered()
|
|
2011
|
+
default:
|
|
2012
|
+
rn := m.getOperand_NR(cvalDef, extModeNone)
|
|
2013
|
+
if c.Type() != ssa.TypeI32 && c.Type() != ssa.TypeI64 {
|
|
2014
|
+
panic("TODO?BUG?: support select with non-integer condition")
|
|
2015
|
+
}
|
|
2016
|
+
alu := m.allocateInstr()
|
|
2017
|
+
// subs zr, rn, zr
|
|
2018
|
+
alu.asALU(
|
|
2019
|
+
aluOpSubS,
|
|
2020
|
+
// We don't need the result, just need to set flags.
|
|
2021
|
+
xzrVReg,
|
|
2022
|
+
rn,
|
|
2023
|
+
operandNR(xzrVReg),
|
|
2024
|
+
c.Type().Bits() == 64,
|
|
2025
|
+
)
|
|
2026
|
+
m.insert(alu)
|
|
2027
|
+
cc = ne
|
|
2028
|
+
}
|
|
2029
|
+
|
|
2030
|
+
rn := m.getOperand_NR(m.compiler.ValueDefinition(x), extModeNone)
|
|
2031
|
+
rm := m.getOperand_NR(m.compiler.ValueDefinition(y), extModeNone)
|
|
2032
|
+
|
|
2033
|
+
rd := m.compiler.VRegOf(result)
|
|
2034
|
+
switch x.Type() {
|
|
2035
|
+
case ssa.TypeI32, ssa.TypeI64:
|
|
2036
|
+
// csel rd, rn, rm, cc
|
|
2037
|
+
csel := m.allocateInstr()
|
|
2038
|
+
csel.asCSel(rd, rn, rm, cc, x.Type().Bits() == 64)
|
|
2039
|
+
m.insert(csel)
|
|
2040
|
+
case ssa.TypeF32, ssa.TypeF64:
|
|
2041
|
+
// fcsel rd, rn, rm, cc
|
|
2042
|
+
fcsel := m.allocateInstr()
|
|
2043
|
+
fcsel.asFpuCSel(rd, rn, rm, cc, x.Type().Bits() == 64)
|
|
2044
|
+
m.insert(fcsel)
|
|
2045
|
+
default:
|
|
2046
|
+
panic("BUG")
|
|
2047
|
+
}
|
|
2048
|
+
}
|
|
2049
|
+
|
|
2050
|
+
func (m *machine) lowerSelectVec(rc, rn, rm operand, rd regalloc.VReg) {
|
|
2051
|
+
// First check if `rc` is zero or not.
|
|
2052
|
+
checkZero := m.allocateInstr()
|
|
2053
|
+
checkZero.asALU(aluOpSubS, xzrVReg, rc, operandNR(xzrVReg), false)
|
|
2054
|
+
m.insert(checkZero)
|
|
2055
|
+
|
|
2056
|
+
// Then use CSETM to set all bits to one if `rc` is zero.
|
|
2057
|
+
allOnesOrZero := m.compiler.AllocateVReg(ssa.TypeI64)
|
|
2058
|
+
cset := m.allocateInstr()
|
|
2059
|
+
cset.asCSet(allOnesOrZero, true, ne)
|
|
2060
|
+
m.insert(cset)
|
|
2061
|
+
|
|
2062
|
+
// Then move the bits to the result vector register.
|
|
2063
|
+
tmp2 := m.compiler.AllocateVReg(ssa.TypeV128)
|
|
2064
|
+
dup := m.allocateInstr()
|
|
2065
|
+
dup.asVecDup(tmp2, operandNR(allOnesOrZero), vecArrangement2D)
|
|
2066
|
+
m.insert(dup)
|
|
2067
|
+
|
|
2068
|
+
// Now that `tmp2` has either all bits one or zero depending on `rc`,
|
|
2069
|
+
// we can use bsl to select between `rn` and `rm`.
|
|
2070
|
+
ins := m.allocateInstr()
|
|
2071
|
+
ins.asVecRRRRewrite(vecOpBsl, tmp2, rn, rm, vecArrangement16B)
|
|
2072
|
+
m.insert(ins)
|
|
2073
|
+
|
|
2074
|
+
// Finally, move the result to the destination register.
|
|
2075
|
+
mov2 := m.allocateInstr()
|
|
2076
|
+
mov2.asFpuMov128(rd, tmp2)
|
|
2077
|
+
m.insert(mov2)
|
|
2078
|
+
}
|
|
2079
|
+
|
|
2080
|
+
func (m *machine) lowerAtomicRmw(si *ssa.Instruction) {
|
|
2081
|
+
ssaOp, size := si.AtomicRmwData()
|
|
2082
|
+
|
|
2083
|
+
var op atomicRmwOp
|
|
2084
|
+
var negateArg bool
|
|
2085
|
+
var flipArg bool
|
|
2086
|
+
switch ssaOp {
|
|
2087
|
+
case ssa.AtomicRmwOpAdd:
|
|
2088
|
+
op = atomicRmwOpAdd
|
|
2089
|
+
case ssa.AtomicRmwOpSub:
|
|
2090
|
+
op = atomicRmwOpAdd
|
|
2091
|
+
negateArg = true
|
|
2092
|
+
case ssa.AtomicRmwOpAnd:
|
|
2093
|
+
op = atomicRmwOpClr
|
|
2094
|
+
flipArg = true
|
|
2095
|
+
case ssa.AtomicRmwOpOr:
|
|
2096
|
+
op = atomicRmwOpSet
|
|
2097
|
+
case ssa.AtomicRmwOpXor:
|
|
2098
|
+
op = atomicRmwOpEor
|
|
2099
|
+
case ssa.AtomicRmwOpXchg:
|
|
2100
|
+
op = atomicRmwOpSwp
|
|
2101
|
+
default:
|
|
2102
|
+
panic(fmt.Sprintf("unknown ssa atomic rmw op: %s", ssaOp))
|
|
2103
|
+
}
|
|
2104
|
+
|
|
2105
|
+
addr, val := si.Arg2()
|
|
2106
|
+
addrDef, valDef := m.compiler.ValueDefinition(addr), m.compiler.ValueDefinition(val)
|
|
2107
|
+
rn := m.getOperand_NR(addrDef, extModeNone)
|
|
2108
|
+
rt := m.compiler.VRegOf(si.Return())
|
|
2109
|
+
rs := m.getOperand_NR(valDef, extModeNone)
|
|
2110
|
+
|
|
2111
|
+
_64 := si.Return().Type().Bits() == 64
|
|
2112
|
+
var tmp regalloc.VReg
|
|
2113
|
+
if _64 {
|
|
2114
|
+
tmp = m.compiler.AllocateVReg(ssa.TypeI64)
|
|
2115
|
+
} else {
|
|
2116
|
+
tmp = m.compiler.AllocateVReg(ssa.TypeI32)
|
|
2117
|
+
}
|
|
2118
|
+
m.lowerAtomicRmwImpl(op, rn.nr(), rs.nr(), rt, tmp, size, negateArg, flipArg, _64)
|
|
2119
|
+
}
|
|
2120
|
+
|
|
2121
|
+
func (m *machine) lowerAtomicRmwImpl(op atomicRmwOp, rn, rs, rt, tmp regalloc.VReg, size uint64, negateArg, flipArg, dst64bit bool) {
|
|
2122
|
+
switch {
|
|
2123
|
+
case negateArg:
|
|
2124
|
+
neg := m.allocateInstr()
|
|
2125
|
+
neg.asALU(aluOpSub, tmp, operandNR(xzrVReg), operandNR(rs), dst64bit)
|
|
2126
|
+
m.insert(neg)
|
|
2127
|
+
case flipArg:
|
|
2128
|
+
flip := m.allocateInstr()
|
|
2129
|
+
flip.asALU(aluOpOrn, tmp, operandNR(xzrVReg), operandNR(rs), dst64bit)
|
|
2130
|
+
m.insert(flip)
|
|
2131
|
+
default:
|
|
2132
|
+
tmp = rs
|
|
2133
|
+
}
|
|
2134
|
+
|
|
2135
|
+
rmw := m.allocateInstr()
|
|
2136
|
+
rmw.asAtomicRmw(op, rn, tmp, rt, size)
|
|
2137
|
+
m.insert(rmw)
|
|
2138
|
+
}
|
|
2139
|
+
|
|
2140
|
+
func (m *machine) lowerAtomicCas(si *ssa.Instruction) {
|
|
2141
|
+
addr, exp, repl := si.Arg3()
|
|
2142
|
+
size := si.AtomicTargetSize()
|
|
2143
|
+
|
|
2144
|
+
addrDef, expDef, replDef := m.compiler.ValueDefinition(addr), m.compiler.ValueDefinition(exp), m.compiler.ValueDefinition(repl)
|
|
2145
|
+
rn := m.getOperand_NR(addrDef, extModeNone)
|
|
2146
|
+
rt := m.getOperand_NR(replDef, extModeNone)
|
|
2147
|
+
rs := m.getOperand_NR(expDef, extModeNone)
|
|
2148
|
+
tmp := m.compiler.AllocateVReg(si.Return().Type())
|
|
2149
|
+
|
|
2150
|
+
_64 := si.Return().Type().Bits() == 64
|
|
2151
|
+
// rs is overwritten by CAS, so we need to move it to the result register before the instruction
|
|
2152
|
+
// in case when it is used somewhere else.
|
|
2153
|
+
mov := m.allocateInstr()
|
|
2154
|
+
if _64 {
|
|
2155
|
+
mov.asMove64(tmp, rs.nr())
|
|
2156
|
+
} else {
|
|
2157
|
+
mov.asMove32(tmp, rs.nr())
|
|
2158
|
+
}
|
|
2159
|
+
m.insert(mov)
|
|
2160
|
+
|
|
2161
|
+
m.lowerAtomicCasImpl(rn.nr(), tmp, rt.nr(), size)
|
|
2162
|
+
|
|
2163
|
+
mov2 := m.allocateInstr()
|
|
2164
|
+
rd := m.compiler.VRegOf(si.Return())
|
|
2165
|
+
if _64 {
|
|
2166
|
+
mov2.asMove64(rd, tmp)
|
|
2167
|
+
} else {
|
|
2168
|
+
mov2.asMove32(rd, tmp)
|
|
2169
|
+
}
|
|
2170
|
+
m.insert(mov2)
|
|
2171
|
+
}
|
|
2172
|
+
|
|
2173
|
+
func (m *machine) lowerAtomicCasImpl(rn, rs, rt regalloc.VReg, size uint64) {
|
|
2174
|
+
cas := m.allocateInstr()
|
|
2175
|
+
cas.asAtomicCas(rn, rs, rt, size)
|
|
2176
|
+
m.insert(cas)
|
|
2177
|
+
}
|
|
2178
|
+
|
|
2179
|
+
func (m *machine) lowerAtomicLoad(si *ssa.Instruction) {
|
|
2180
|
+
addr := si.Arg()
|
|
2181
|
+
size := si.AtomicTargetSize()
|
|
2182
|
+
|
|
2183
|
+
addrDef := m.compiler.ValueDefinition(addr)
|
|
2184
|
+
rn := m.getOperand_NR(addrDef, extModeNone)
|
|
2185
|
+
rt := m.compiler.VRegOf(si.Return())
|
|
2186
|
+
|
|
2187
|
+
m.lowerAtomicLoadImpl(rn.nr(), rt, size)
|
|
2188
|
+
}
|
|
2189
|
+
|
|
2190
|
+
func (m *machine) lowerAtomicLoadImpl(rn, rt regalloc.VReg, size uint64) {
|
|
2191
|
+
ld := m.allocateInstr()
|
|
2192
|
+
ld.asAtomicLoad(rn, rt, size)
|
|
2193
|
+
m.insert(ld)
|
|
2194
|
+
}
|
|
2195
|
+
|
|
2196
|
+
func (m *machine) lowerAtomicStore(si *ssa.Instruction) {
|
|
2197
|
+
addr, val := si.Arg2()
|
|
2198
|
+
size := si.AtomicTargetSize()
|
|
2199
|
+
|
|
2200
|
+
addrDef := m.compiler.ValueDefinition(addr)
|
|
2201
|
+
valDef := m.compiler.ValueDefinition(val)
|
|
2202
|
+
rn := m.getOperand_NR(addrDef, extModeNone)
|
|
2203
|
+
rt := m.getOperand_NR(valDef, extModeNone)
|
|
2204
|
+
|
|
2205
|
+
m.lowerAtomicStoreImpl(rn, rt, size)
|
|
2206
|
+
}
|
|
2207
|
+
|
|
2208
|
+
func (m *machine) lowerAtomicStoreImpl(rn, rt operand, size uint64) {
|
|
2209
|
+
ld := m.allocateInstr()
|
|
2210
|
+
ld.asAtomicStore(rn, rt, size)
|
|
2211
|
+
m.insert(ld)
|
|
2212
|
+
}
|
|
2213
|
+
|
|
2214
|
+
// copyToTmp copies the given regalloc.VReg to a temporary register. This is called before cbr to avoid the regalloc issue
|
|
2215
|
+
// e.g. reload happening in the middle of the exit sequence which is not the path the normal path executes
|
|
2216
|
+
func (m *machine) copyToTmp(v regalloc.VReg) regalloc.VReg {
|
|
2217
|
+
typ := m.compiler.TypeOf(v)
|
|
2218
|
+
mov := m.allocateInstr()
|
|
2219
|
+
tmp := m.compiler.AllocateVReg(typ)
|
|
2220
|
+
if typ.IsInt() {
|
|
2221
|
+
mov.asMove64(tmp, v)
|
|
2222
|
+
} else {
|
|
2223
|
+
mov.asFpuMov128(tmp, v)
|
|
2224
|
+
}
|
|
2225
|
+
m.insert(mov)
|
|
2226
|
+
return tmp
|
|
2227
|
+
}
|