2014-02-24 20:31:01 -07:00
|
|
|
// Copyright 2014 The Go Authors. All rights reserved.
|
2014-01-16 21:58:10 -07:00
|
|
|
// Use of this source code is governed by a BSD-style
|
|
|
|
|
// license that can be found in the LICENSE file.
|
|
|
|
|
//
|
|
|
|
|
// System calls and other sys.stuff for AMD64, SunOS
|
|
|
|
|
// /usr/include/sys/syscall.h for syscall numbers.
|
|
|
|
|
//
|
|
|
|
|
|
[dev.cc] runtime: convert assembly files for C to Go transition
The main change is that #include "zasm_GOOS_GOARCH.h"
is now #include "go_asm.h" and/or #include "go_tls.h".
Also, because C StackGuard is now Go _StackGuard,
the assembly name changes from const_StackGuard to
const__StackGuard.
In asm_$GOARCH.s, add new function getg, formerly
implemented in C.
The renamed atomics now have Go wrappers, to get
escape analysis annotations right. Those wrappers
are in CL 174860043.
LGTM=r, aram
R=r, aram
CC=austin, dvyukov, golang-codereviews, iant, khr
https://golang.org/cl/168510043
2014-11-11 15:06:22 -07:00
|
|
|
#include "go_asm.h"
|
|
|
|
|
#include "go_tls.h"
|
2014-09-04 21:05:18 -06:00
|
|
|
#include "textflag.h"
|
2014-01-16 21:58:10 -07:00
|
|
|
|
|
|
|
|
// This is needed by asm_amd64.s
|
|
|
|
|
TEXT runtime·settls(SB),NOSPLIT,$8
|
|
|
|
|
RET
|
|
|
|
|
|
[dev.cc] runtime: convert Solaris port to Go
Memory management was consolitated with the BSD ports, since
it was almost identical.
Assembly thunks are gone, being replaced by the new //go:linkname
feature.
This change supersedes CL 138390043 (runtime: convert solaris
netpoll to Go), which was previously reviewed and tested.
This change is only the first step, the port now builds,
but doesn't run. Binaries fail to exec:
ld.so.1: 6.out: fatal: 6.out: TLS requirement failure : TLS support is unavailable
Killed
This seems to happen because binaries don't link with libc.so
anymore. We will have to solve that in a different CL.
Also this change is just a rough translation of the original
C code, cleanup will come in a different CL.
[This CL is part of the removal of C code from package runtime.
See golang.org/s/dev.cc for an overview.]
LGTM=rsc
R=rsc, dave
CC=golang-codereviews, iant, khr, minux, r, rlh
https://golang.org/cl/174960043
2014-11-13 08:07:10 -07:00
|
|
|
// void libc_miniterrno(void *(*___errno)(void));
|
2014-01-16 21:58:10 -07:00
|
|
|
//
|
|
|
|
|
// Set the TLS errno pointer in M.
|
|
|
|
|
//
|
|
|
|
|
// Called using runtime·asmcgocall from os_solaris.c:/minit.
|
2014-08-27 12:52:25 -06:00
|
|
|
// NOT USING GO CALLING CONVENTION.
|
2014-01-16 21:58:10 -07:00
|
|
|
TEXT runtime·miniterrno(SB),NOSPLIT,$0
|
|
|
|
|
// asmcgocall will put first argument into DI.
|
|
|
|
|
CALL DI // SysV ABI so returns in AX
|
|
|
|
|
get_tls(CX)
|
all: remove 'extern register M *m' from runtime
The runtime has historically held two dedicated values g (current goroutine)
and m (current thread) in 'extern register' slots (TLS on x86, real registers
backed by TLS on ARM).
This CL removes the extern register m; code now uses g->m.
On ARM, this frees up the register that formerly held m (R9).
This is important for NaCl, because NaCl ARM code cannot use R9 at all.
The Go 1 macrobenchmarks (those with per-op times >= 10 µs) are unaffected:
BenchmarkBinaryTree17 5491374955 5471024381 -0.37%
BenchmarkFannkuch11 4357101311 4275174828 -1.88%
BenchmarkGobDecode 11029957 11364184 +3.03%
BenchmarkGobEncode 6852205 6784822 -0.98%
BenchmarkGzip 650795967 650152275 -0.10%
BenchmarkGunzip 140962363 141041670 +0.06%
BenchmarkHTTPClientServer 71581 73081 +2.10%
BenchmarkJSONEncode 31928079 31913356 -0.05%
BenchmarkJSONDecode 117470065 113689916 -3.22%
BenchmarkMandelbrot200 6008923 5998712 -0.17%
BenchmarkGoParse 6310917 6327487 +0.26%
BenchmarkRegexpMatchMedium_1K 114568 114763 +0.17%
BenchmarkRegexpMatchHard_1K 168977 169244 +0.16%
BenchmarkRevcomp 935294971 914060918 -2.27%
BenchmarkTemplate 145917123 148186096 +1.55%
Minux previous reported larger variations, but these were caused by
run-to-run noise, not repeatable slowdowns.
Actual code changes by Minux.
I only did the docs and the benchmarking.
LGTM=dvyukov, iant, minux
R=minux, josharian, iant, dave, bradfitz, dvyukov
CC=golang-codereviews
https://golang.org/cl/109050043
2014-06-26 09:54:39 -06:00
|
|
|
MOVQ g(CX), BX
|
|
|
|
|
MOVQ g_m(BX), BX
|
2014-01-16 21:58:10 -07:00
|
|
|
MOVQ AX, m_perrno(BX)
|
|
|
|
|
RET
|
|
|
|
|
|
|
|
|
|
// int64 runtime·nanotime1(void);
|
|
|
|
|
//
|
|
|
|
|
// clock_gettime(3c) wrapper because Timespec is too large for
|
|
|
|
|
// runtime·nanotime stack.
|
|
|
|
|
//
|
|
|
|
|
// Called using runtime·sysvicall6 from os_solaris.c:/nanotime.
|
2014-08-27 12:52:25 -06:00
|
|
|
// NOT USING GO CALLING CONVENTION.
|
2014-01-16 21:58:10 -07:00
|
|
|
TEXT runtime·nanotime1(SB),NOSPLIT,$0
|
|
|
|
|
// need space for the timespec argument.
|
|
|
|
|
SUBQ $64, SP // 16 bytes will do, but who knows in the future?
|
|
|
|
|
MOVQ $3, DI // CLOCK_REALTIME from <sys/time_impl.h>
|
|
|
|
|
MOVQ SP, SI
|
2015-03-30 05:52:07 -06:00
|
|
|
LEAQ libc_clock_gettime(SB), AX
|
2014-01-16 21:58:10 -07:00
|
|
|
CALL AX
|
|
|
|
|
MOVQ (SP), AX // tv_sec from struct timespec
|
|
|
|
|
IMULQ $1000000000, AX // multiply into nanoseconds
|
|
|
|
|
ADDQ 8(SP), AX // tv_nsec, offset should be stable.
|
|
|
|
|
ADDQ $64, SP
|
|
|
|
|
RET
|
|
|
|
|
|
|
|
|
|
// pipe(3c) wrapper that returns fds in AX, DX.
|
2014-08-27 12:52:25 -06:00
|
|
|
// NOT USING GO CALLING CONVENTION.
|
2014-01-16 21:58:10 -07:00
|
|
|
TEXT runtime·pipe1(SB),NOSPLIT,$0
|
|
|
|
|
SUBQ $16, SP // 8 bytes will do, but stack has to be 16-byte alligned
|
|
|
|
|
MOVQ SP, DI
|
2015-03-30 05:52:07 -06:00
|
|
|
LEAQ libc_pipe(SB), AX
|
2014-01-16 21:58:10 -07:00
|
|
|
CALL AX
|
|
|
|
|
MOVL 0(SP), AX
|
|
|
|
|
MOVL 4(SP), DX
|
|
|
|
|
ADDQ $16, SP
|
|
|
|
|
RET
|
|
|
|
|
|
|
|
|
|
// Call a library function with SysV calling conventions.
|
|
|
|
|
// The called function can take a maximum of 6 INTEGER class arguments,
|
|
|
|
|
// see
|
|
|
|
|
// Michael Matz, Jan Hubicka, Andreas Jaeger, and Mark Mitchell
|
|
|
|
|
// System V Application Binary Interface
|
|
|
|
|
// AMD64 Architecture Processor Supplement
|
|
|
|
|
// section 3.2.3.
|
|
|
|
|
//
|
|
|
|
|
// Called by runtime·asmcgocall or runtime·cgocall.
|
2014-08-27 12:52:25 -06:00
|
|
|
// NOT USING GO CALLING CONVENTION.
|
2014-01-16 21:58:10 -07:00
|
|
|
TEXT runtime·asmsysvicall6(SB),NOSPLIT,$0
|
|
|
|
|
// asmcgocall will put first argument into DI.
|
|
|
|
|
PUSHQ DI // save for later
|
|
|
|
|
MOVQ libcall_fn(DI), AX
|
|
|
|
|
MOVQ libcall_args(DI), R11
|
|
|
|
|
MOVQ libcall_n(DI), R10
|
|
|
|
|
|
|
|
|
|
get_tls(CX)
|
all: remove 'extern register M *m' from runtime
The runtime has historically held two dedicated values g (current goroutine)
and m (current thread) in 'extern register' slots (TLS on x86, real registers
backed by TLS on ARM).
This CL removes the extern register m; code now uses g->m.
On ARM, this frees up the register that formerly held m (R9).
This is important for NaCl, because NaCl ARM code cannot use R9 at all.
The Go 1 macrobenchmarks (those with per-op times >= 10 µs) are unaffected:
BenchmarkBinaryTree17 5491374955 5471024381 -0.37%
BenchmarkFannkuch11 4357101311 4275174828 -1.88%
BenchmarkGobDecode 11029957 11364184 +3.03%
BenchmarkGobEncode 6852205 6784822 -0.98%
BenchmarkGzip 650795967 650152275 -0.10%
BenchmarkGunzip 140962363 141041670 +0.06%
BenchmarkHTTPClientServer 71581 73081 +2.10%
BenchmarkJSONEncode 31928079 31913356 -0.05%
BenchmarkJSONDecode 117470065 113689916 -3.22%
BenchmarkMandelbrot200 6008923 5998712 -0.17%
BenchmarkGoParse 6310917 6327487 +0.26%
BenchmarkRegexpMatchMedium_1K 114568 114763 +0.17%
BenchmarkRegexpMatchHard_1K 168977 169244 +0.16%
BenchmarkRevcomp 935294971 914060918 -2.27%
BenchmarkTemplate 145917123 148186096 +1.55%
Minux previous reported larger variations, but these were caused by
run-to-run noise, not repeatable slowdowns.
Actual code changes by Minux.
I only did the docs and the benchmarking.
LGTM=dvyukov, iant, minux
R=minux, josharian, iant, dave, bradfitz, dvyukov
CC=golang-codereviews
https://golang.org/cl/109050043
2014-06-26 09:54:39 -06:00
|
|
|
MOVQ g(CX), BX
|
|
|
|
|
MOVQ g_m(BX), BX
|
2014-01-16 21:58:10 -07:00
|
|
|
MOVQ m_perrno(BX), DX
|
|
|
|
|
CMPQ DX, $0
|
|
|
|
|
JEQ skiperrno1
|
|
|
|
|
MOVL $0, 0(DX)
|
|
|
|
|
|
|
|
|
|
skiperrno1:
|
|
|
|
|
CMPQ R11, $0
|
|
|
|
|
JEQ skipargs
|
|
|
|
|
// Load 6 args into correspondent registers.
|
|
|
|
|
MOVQ 0(R11), DI
|
|
|
|
|
MOVQ 8(R11), SI
|
|
|
|
|
MOVQ 16(R11), DX
|
|
|
|
|
MOVQ 24(R11), CX
|
|
|
|
|
MOVQ 32(R11), R8
|
|
|
|
|
MOVQ 40(R11), R9
|
|
|
|
|
skipargs:
|
|
|
|
|
|
|
|
|
|
// Call SysV function
|
|
|
|
|
CALL AX
|
|
|
|
|
|
|
|
|
|
// Return result
|
|
|
|
|
POPQ DI
|
|
|
|
|
MOVQ AX, libcall_r1(DI)
|
|
|
|
|
MOVQ DX, libcall_r2(DI)
|
|
|
|
|
|
|
|
|
|
get_tls(CX)
|
all: remove 'extern register M *m' from runtime
The runtime has historically held two dedicated values g (current goroutine)
and m (current thread) in 'extern register' slots (TLS on x86, real registers
backed by TLS on ARM).
This CL removes the extern register m; code now uses g->m.
On ARM, this frees up the register that formerly held m (R9).
This is important for NaCl, because NaCl ARM code cannot use R9 at all.
The Go 1 macrobenchmarks (those with per-op times >= 10 µs) are unaffected:
BenchmarkBinaryTree17 5491374955 5471024381 -0.37%
BenchmarkFannkuch11 4357101311 4275174828 -1.88%
BenchmarkGobDecode 11029957 11364184 +3.03%
BenchmarkGobEncode 6852205 6784822 -0.98%
BenchmarkGzip 650795967 650152275 -0.10%
BenchmarkGunzip 140962363 141041670 +0.06%
BenchmarkHTTPClientServer 71581 73081 +2.10%
BenchmarkJSONEncode 31928079 31913356 -0.05%
BenchmarkJSONDecode 117470065 113689916 -3.22%
BenchmarkMandelbrot200 6008923 5998712 -0.17%
BenchmarkGoParse 6310917 6327487 +0.26%
BenchmarkRegexpMatchMedium_1K 114568 114763 +0.17%
BenchmarkRegexpMatchHard_1K 168977 169244 +0.16%
BenchmarkRevcomp 935294971 914060918 -2.27%
BenchmarkTemplate 145917123 148186096 +1.55%
Minux previous reported larger variations, but these were caused by
run-to-run noise, not repeatable slowdowns.
Actual code changes by Minux.
I only did the docs and the benchmarking.
LGTM=dvyukov, iant, minux
R=minux, josharian, iant, dave, bradfitz, dvyukov
CC=golang-codereviews
https://golang.org/cl/109050043
2014-06-26 09:54:39 -06:00
|
|
|
MOVQ g(CX), BX
|
|
|
|
|
MOVQ g_m(BX), BX
|
2014-01-16 21:58:10 -07:00
|
|
|
MOVQ m_perrno(BX), AX
|
|
|
|
|
CMPQ AX, $0
|
|
|
|
|
JEQ skiperrno2
|
|
|
|
|
MOVL 0(AX), AX
|
|
|
|
|
MOVQ AX, libcall_err(DI)
|
|
|
|
|
|
|
|
|
|
skiperrno2:
|
|
|
|
|
RET
|
|
|
|
|
|
|
|
|
|
// uint32 tstart_sysvicall(M *newm);
|
|
|
|
|
TEXT runtime·tstart_sysvicall(SB),NOSPLIT,$0
|
|
|
|
|
// DI contains first arg newm
|
|
|
|
|
MOVQ m_g0(DI), DX // g
|
|
|
|
|
|
|
|
|
|
// Make TLS entries point at g and m.
|
|
|
|
|
get_tls(BX)
|
|
|
|
|
MOVQ DX, g(BX)
|
all: remove 'extern register M *m' from runtime
The runtime has historically held two dedicated values g (current goroutine)
and m (current thread) in 'extern register' slots (TLS on x86, real registers
backed by TLS on ARM).
This CL removes the extern register m; code now uses g->m.
On ARM, this frees up the register that formerly held m (R9).
This is important for NaCl, because NaCl ARM code cannot use R9 at all.
The Go 1 macrobenchmarks (those with per-op times >= 10 µs) are unaffected:
BenchmarkBinaryTree17 5491374955 5471024381 -0.37%
BenchmarkFannkuch11 4357101311 4275174828 -1.88%
BenchmarkGobDecode 11029957 11364184 +3.03%
BenchmarkGobEncode 6852205 6784822 -0.98%
BenchmarkGzip 650795967 650152275 -0.10%
BenchmarkGunzip 140962363 141041670 +0.06%
BenchmarkHTTPClientServer 71581 73081 +2.10%
BenchmarkJSONEncode 31928079 31913356 -0.05%
BenchmarkJSONDecode 117470065 113689916 -3.22%
BenchmarkMandelbrot200 6008923 5998712 -0.17%
BenchmarkGoParse 6310917 6327487 +0.26%
BenchmarkRegexpMatchMedium_1K 114568 114763 +0.17%
BenchmarkRegexpMatchHard_1K 168977 169244 +0.16%
BenchmarkRevcomp 935294971 914060918 -2.27%
BenchmarkTemplate 145917123 148186096 +1.55%
Minux previous reported larger variations, but these were caused by
run-to-run noise, not repeatable slowdowns.
Actual code changes by Minux.
I only did the docs and the benchmarking.
LGTM=dvyukov, iant, minux
R=minux, josharian, iant, dave, bradfitz, dvyukov
CC=golang-codereviews
https://golang.org/cl/109050043
2014-06-26 09:54:39 -06:00
|
|
|
MOVQ DI, g_m(DX)
|
2014-01-16 21:58:10 -07:00
|
|
|
|
|
|
|
|
// Layout new m scheduler stack on os stack.
|
|
|
|
|
MOVQ SP, AX
|
2014-09-09 12:02:37 -06:00
|
|
|
MOVQ AX, (g_stack+stack_hi)(DX)
|
2014-01-16 21:58:10 -07:00
|
|
|
SUBQ $(0x100000), AX // stack size
|
2014-09-09 12:02:37 -06:00
|
|
|
MOVQ AX, (g_stack+stack_lo)(DX)
|
[dev.cc] runtime: convert Solaris port to Go
Memory management was consolitated with the BSD ports, since
it was almost identical.
Assembly thunks are gone, being replaced by the new //go:linkname
feature.
This change supersedes CL 138390043 (runtime: convert solaris
netpoll to Go), which was previously reviewed and tested.
This change is only the first step, the port now builds,
but doesn't run. Binaries fail to exec:
ld.so.1: 6.out: fatal: 6.out: TLS requirement failure : TLS support is unavailable
Killed
This seems to happen because binaries don't link with libc.so
anymore. We will have to solve that in a different CL.
Also this change is just a rough translation of the original
C code, cleanup will come in a different CL.
[This CL is part of the removal of C code from package runtime.
See golang.org/s/dev.cc for an overview.]
LGTM=rsc
R=rsc, dave
CC=golang-codereviews, iant, khr, minux, r, rlh
https://golang.org/cl/174960043
2014-11-13 08:07:10 -07:00
|
|
|
ADDQ $const__StackGuard, AX
|
2015-01-05 09:29:21 -07:00
|
|
|
MOVQ AX, g_stackguard0(DX)
|
|
|
|
|
MOVQ AX, g_stackguard1(DX)
|
2014-01-16 21:58:10 -07:00
|
|
|
|
|
|
|
|
// Someday the convention will be D is always cleared.
|
|
|
|
|
CLD
|
|
|
|
|
|
|
|
|
|
CALL runtime·stackcheck(SB) // clobbers AX,CX
|
|
|
|
|
CALL runtime·mstart(SB)
|
|
|
|
|
|
|
|
|
|
XORL AX, AX // return 0 == success
|
cmd/cc, runtime: convert C compilers to use Go calling convention
To date, the C compilers and Go compilers differed only in how
values were returned from functions. This made it difficult to call
Go from C or C from Go if return values were involved. It also made
assembly called from Go and assembly called from C different.
This CL changes the C compiler to use the Go conventions, passing
results on the stack, after the arguments.
[Exception: this does not apply to C ... functions, because you can't
know where on the stack the arguments end.]
By doing this, the CL makes it possible to rewrite C functions into Go
one at a time, without worrying about which languages call that
function or which languages it calls.
This CL also updates all the assembly files in package runtime to use
the new conventions. Argument references of the form 40(SP) have
been rewritten to the form name+10(FP) instead, and there are now
Go func prototypes for every assembly function called from C or Go.
This means that 'go vet runtime' checks effectively every assembly
function, and go vet's output was used to automate the bulk of the
conversion.
Some functions, like seek and nsec on Plan 9, needed to be rewritten.
Many assembly routines called from C were reading arguments
incorrectly, using MOVL instead of MOVQ or vice versa, especially on
the less used systems like openbsd.
These were found by go vet and have been corrected too.
If we're lucky, this may reduce flakiness on those systems.
Tested on:
darwin/386
darwin/amd64
linux/arm
linux/386
linux/amd64
If this breaks another system, the bug is almost certainly in the
sys_$GOOS_$GOARCH.s file, since the rest of the CL is tested
by the combination of the above systems.
LGTM=dvyukov, iant
R=golang-codereviews, 0intro, dave, alex.brainman, dvyukov, iant
CC=golang-codereviews, josharian, r
https://golang.org/cl/135830043
2014-08-27 09:32:17 -06:00
|
|
|
MOVL AX, ret+8(FP)
|
2014-01-16 21:58:10 -07:00
|
|
|
RET
|
|
|
|
|
|
|
|
|
|
// Careful, this is called by __sighndlr, a libc function. We must preserve
|
|
|
|
|
// registers as per AMD 64 ABI.
|
|
|
|
|
TEXT runtime·sigtramp(SB),NOSPLIT,$0
|
|
|
|
|
// Note that we are executing on altsigstack here, so we have
|
|
|
|
|
// more stack available than NOSPLIT would have us believe.
|
|
|
|
|
// To defeat the linker, we make our own stack frame with
|
|
|
|
|
// more space:
|
|
|
|
|
SUBQ $184, SP
|
|
|
|
|
|
|
|
|
|
// save registers
|
|
|
|
|
MOVQ BX, 32(SP)
|
|
|
|
|
MOVQ BP, 40(SP)
|
|
|
|
|
MOVQ R12, 48(SP)
|
|
|
|
|
MOVQ R13, 56(SP)
|
|
|
|
|
MOVQ R14, 64(SP)
|
|
|
|
|
MOVQ R15, 72(SP)
|
|
|
|
|
|
|
|
|
|
get_tls(BX)
|
all: remove 'extern register M *m' from runtime
The runtime has historically held two dedicated values g (current goroutine)
and m (current thread) in 'extern register' slots (TLS on x86, real registers
backed by TLS on ARM).
This CL removes the extern register m; code now uses g->m.
On ARM, this frees up the register that formerly held m (R9).
This is important for NaCl, because NaCl ARM code cannot use R9 at all.
The Go 1 macrobenchmarks (those with per-op times >= 10 µs) are unaffected:
BenchmarkBinaryTree17 5491374955 5471024381 -0.37%
BenchmarkFannkuch11 4357101311 4275174828 -1.88%
BenchmarkGobDecode 11029957 11364184 +3.03%
BenchmarkGobEncode 6852205 6784822 -0.98%
BenchmarkGzip 650795967 650152275 -0.10%
BenchmarkGunzip 140962363 141041670 +0.06%
BenchmarkHTTPClientServer 71581 73081 +2.10%
BenchmarkJSONEncode 31928079 31913356 -0.05%
BenchmarkJSONDecode 117470065 113689916 -3.22%
BenchmarkMandelbrot200 6008923 5998712 -0.17%
BenchmarkGoParse 6310917 6327487 +0.26%
BenchmarkRegexpMatchMedium_1K 114568 114763 +0.17%
BenchmarkRegexpMatchHard_1K 168977 169244 +0.16%
BenchmarkRevcomp 935294971 914060918 -2.27%
BenchmarkTemplate 145917123 148186096 +1.55%
Minux previous reported larger variations, but these were caused by
run-to-run noise, not repeatable slowdowns.
Actual code changes by Minux.
I only did the docs and the benchmarking.
LGTM=dvyukov, iant, minux
R=minux, josharian, iant, dave, bradfitz, dvyukov
CC=golang-codereviews
https://golang.org/cl/109050043
2014-06-26 09:54:39 -06:00
|
|
|
// check that g exists
|
|
|
|
|
MOVQ g(BX), R10
|
|
|
|
|
CMPQ R10, $0
|
2014-01-16 21:58:10 -07:00
|
|
|
JNE allgood
|
|
|
|
|
MOVQ DI, 0(SP)
|
|
|
|
|
MOVQ $runtime·badsignal(SB), AX
|
|
|
|
|
CALL AX
|
2014-07-01 17:34:06 -06:00
|
|
|
JMP exit
|
2014-01-16 21:58:10 -07:00
|
|
|
|
|
|
|
|
allgood:
|
|
|
|
|
// save g
|
|
|
|
|
MOVQ R10, 80(SP)
|
|
|
|
|
|
|
|
|
|
// Save m->libcall and m->scratch. We need to do this because we
|
|
|
|
|
// might get interrupted by a signal in runtime·asmcgocall.
|
|
|
|
|
|
|
|
|
|
// save m->libcall
|
all: remove 'extern register M *m' from runtime
The runtime has historically held two dedicated values g (current goroutine)
and m (current thread) in 'extern register' slots (TLS on x86, real registers
backed by TLS on ARM).
This CL removes the extern register m; code now uses g->m.
On ARM, this frees up the register that formerly held m (R9).
This is important for NaCl, because NaCl ARM code cannot use R9 at all.
The Go 1 macrobenchmarks (those with per-op times >= 10 µs) are unaffected:
BenchmarkBinaryTree17 5491374955 5471024381 -0.37%
BenchmarkFannkuch11 4357101311 4275174828 -1.88%
BenchmarkGobDecode 11029957 11364184 +3.03%
BenchmarkGobEncode 6852205 6784822 -0.98%
BenchmarkGzip 650795967 650152275 -0.10%
BenchmarkGunzip 140962363 141041670 +0.06%
BenchmarkHTTPClientServer 71581 73081 +2.10%
BenchmarkJSONEncode 31928079 31913356 -0.05%
BenchmarkJSONDecode 117470065 113689916 -3.22%
BenchmarkMandelbrot200 6008923 5998712 -0.17%
BenchmarkGoParse 6310917 6327487 +0.26%
BenchmarkRegexpMatchMedium_1K 114568 114763 +0.17%
BenchmarkRegexpMatchHard_1K 168977 169244 +0.16%
BenchmarkRevcomp 935294971 914060918 -2.27%
BenchmarkTemplate 145917123 148186096 +1.55%
Minux previous reported larger variations, but these were caused by
run-to-run noise, not repeatable slowdowns.
Actual code changes by Minux.
I only did the docs and the benchmarking.
LGTM=dvyukov, iant, minux
R=minux, josharian, iant, dave, bradfitz, dvyukov
CC=golang-codereviews
https://golang.org/cl/109050043
2014-06-26 09:54:39 -06:00
|
|
|
MOVQ g_m(R10), BP
|
2014-01-16 21:58:10 -07:00
|
|
|
LEAQ m_libcall(BP), R11
|
|
|
|
|
MOVQ libcall_fn(R11), R10
|
|
|
|
|
MOVQ R10, 88(SP)
|
|
|
|
|
MOVQ libcall_args(R11), R10
|
|
|
|
|
MOVQ R10, 96(SP)
|
|
|
|
|
MOVQ libcall_n(R11), R10
|
|
|
|
|
MOVQ R10, 104(SP)
|
|
|
|
|
MOVQ libcall_r1(R11), R10
|
|
|
|
|
MOVQ R10, 168(SP)
|
|
|
|
|
MOVQ libcall_r2(R11), R10
|
|
|
|
|
MOVQ R10, 176(SP)
|
|
|
|
|
|
|
|
|
|
// save m->scratch
|
|
|
|
|
LEAQ m_scratch(BP), R11
|
|
|
|
|
MOVQ 0(R11), R10
|
|
|
|
|
MOVQ R10, 112(SP)
|
|
|
|
|
MOVQ 8(R11), R10
|
|
|
|
|
MOVQ R10, 120(SP)
|
|
|
|
|
MOVQ 16(R11), R10
|
|
|
|
|
MOVQ R10, 128(SP)
|
|
|
|
|
MOVQ 24(R11), R10
|
|
|
|
|
MOVQ R10, 136(SP)
|
|
|
|
|
MOVQ 32(R11), R10
|
|
|
|
|
MOVQ R10, 144(SP)
|
|
|
|
|
MOVQ 40(R11), R10
|
|
|
|
|
MOVQ R10, 152(SP)
|
|
|
|
|
|
|
|
|
|
// save errno, it might be EINTR; stuff we do here might reset it.
|
|
|
|
|
MOVQ m_perrno(BP), R10
|
|
|
|
|
MOVL 0(R10), R10
|
|
|
|
|
MOVQ R10, 160(SP)
|
|
|
|
|
|
|
|
|
|
MOVQ g(BX), R10
|
|
|
|
|
// g = m->gsignal
|
|
|
|
|
MOVQ m_gsignal(BP), BP
|
|
|
|
|
MOVQ BP, g(BX)
|
|
|
|
|
|
|
|
|
|
// prepare call
|
|
|
|
|
MOVQ DI, 0(SP)
|
|
|
|
|
MOVQ SI, 8(SP)
|
|
|
|
|
MOVQ DX, 16(SP)
|
|
|
|
|
MOVQ R10, 24(SP)
|
|
|
|
|
CALL runtime·sighandler(SB)
|
|
|
|
|
|
|
|
|
|
get_tls(BX)
|
all: remove 'extern register M *m' from runtime
The runtime has historically held two dedicated values g (current goroutine)
and m (current thread) in 'extern register' slots (TLS on x86, real registers
backed by TLS on ARM).
This CL removes the extern register m; code now uses g->m.
On ARM, this frees up the register that formerly held m (R9).
This is important for NaCl, because NaCl ARM code cannot use R9 at all.
The Go 1 macrobenchmarks (those with per-op times >= 10 µs) are unaffected:
BenchmarkBinaryTree17 5491374955 5471024381 -0.37%
BenchmarkFannkuch11 4357101311 4275174828 -1.88%
BenchmarkGobDecode 11029957 11364184 +3.03%
BenchmarkGobEncode 6852205 6784822 -0.98%
BenchmarkGzip 650795967 650152275 -0.10%
BenchmarkGunzip 140962363 141041670 +0.06%
BenchmarkHTTPClientServer 71581 73081 +2.10%
BenchmarkJSONEncode 31928079 31913356 -0.05%
BenchmarkJSONDecode 117470065 113689916 -3.22%
BenchmarkMandelbrot200 6008923 5998712 -0.17%
BenchmarkGoParse 6310917 6327487 +0.26%
BenchmarkRegexpMatchMedium_1K 114568 114763 +0.17%
BenchmarkRegexpMatchHard_1K 168977 169244 +0.16%
BenchmarkRevcomp 935294971 914060918 -2.27%
BenchmarkTemplate 145917123 148186096 +1.55%
Minux previous reported larger variations, but these were caused by
run-to-run noise, not repeatable slowdowns.
Actual code changes by Minux.
I only did the docs and the benchmarking.
LGTM=dvyukov, iant, minux
R=minux, josharian, iant, dave, bradfitz, dvyukov
CC=golang-codereviews
https://golang.org/cl/109050043
2014-06-26 09:54:39 -06:00
|
|
|
MOVQ g(BX), BP
|
|
|
|
|
MOVQ g_m(BP), BP
|
2014-01-16 21:58:10 -07:00
|
|
|
// restore libcall
|
|
|
|
|
LEAQ m_libcall(BP), R11
|
|
|
|
|
MOVQ 88(SP), R10
|
|
|
|
|
MOVQ R10, libcall_fn(R11)
|
|
|
|
|
MOVQ 96(SP), R10
|
|
|
|
|
MOVQ R10, libcall_args(R11)
|
|
|
|
|
MOVQ 104(SP), R10
|
|
|
|
|
MOVQ R10, libcall_n(R11)
|
|
|
|
|
MOVQ 168(SP), R10
|
|
|
|
|
MOVQ R10, libcall_r1(R11)
|
|
|
|
|
MOVQ 176(SP), R10
|
|
|
|
|
MOVQ R10, libcall_r2(R11)
|
|
|
|
|
|
|
|
|
|
// restore scratch
|
|
|
|
|
LEAQ m_scratch(BP), R11
|
|
|
|
|
MOVQ 112(SP), R10
|
|
|
|
|
MOVQ R10, 0(R11)
|
|
|
|
|
MOVQ 120(SP), R10
|
|
|
|
|
MOVQ R10, 8(R11)
|
|
|
|
|
MOVQ 128(SP), R10
|
|
|
|
|
MOVQ R10, 16(R11)
|
|
|
|
|
MOVQ 136(SP), R10
|
|
|
|
|
MOVQ R10, 24(R11)
|
|
|
|
|
MOVQ 144(SP), R10
|
|
|
|
|
MOVQ R10, 32(R11)
|
|
|
|
|
MOVQ 152(SP), R10
|
|
|
|
|
MOVQ R10, 40(R11)
|
|
|
|
|
|
|
|
|
|
// restore errno
|
|
|
|
|
MOVQ m_perrno(BP), R11
|
|
|
|
|
MOVQ 160(SP), R10
|
|
|
|
|
MOVL R10, 0(R11)
|
|
|
|
|
|
|
|
|
|
// restore g
|
|
|
|
|
MOVQ 80(SP), R10
|
|
|
|
|
MOVQ R10, g(BX)
|
|
|
|
|
|
2014-07-01 17:34:06 -06:00
|
|
|
exit:
|
2014-01-16 21:58:10 -07:00
|
|
|
// restore registers
|
|
|
|
|
MOVQ 32(SP), BX
|
|
|
|
|
MOVQ 40(SP), BP
|
|
|
|
|
MOVQ 48(SP), R12
|
|
|
|
|
MOVQ 56(SP), R13
|
|
|
|
|
MOVQ 64(SP), R14
|
|
|
|
|
MOVQ 72(SP), R15
|
|
|
|
|
|
|
|
|
|
ADDQ $184, SP
|
|
|
|
|
RET
|
2014-07-02 19:36:05 -06:00
|
|
|
|
|
|
|
|
// Called from runtime·usleep (Go). Can be called on Go stack, on OS stack,
|
|
|
|
|
// can also be called in cgo callback path without a g->m.
|
|
|
|
|
TEXT runtime·usleep1(SB),NOSPLIT,$0
|
cmd/cc, runtime: convert C compilers to use Go calling convention
To date, the C compilers and Go compilers differed only in how
values were returned from functions. This made it difficult to call
Go from C or C from Go if return values were involved. It also made
assembly called from Go and assembly called from C different.
This CL changes the C compiler to use the Go conventions, passing
results on the stack, after the arguments.
[Exception: this does not apply to C ... functions, because you can't
know where on the stack the arguments end.]
By doing this, the CL makes it possible to rewrite C functions into Go
one at a time, without worrying about which languages call that
function or which languages it calls.
This CL also updates all the assembly files in package runtime to use
the new conventions. Argument references of the form 40(SP) have
been rewritten to the form name+10(FP) instead, and there are now
Go func prototypes for every assembly function called from C or Go.
This means that 'go vet runtime' checks effectively every assembly
function, and go vet's output was used to automate the bulk of the
conversion.
Some functions, like seek and nsec on Plan 9, needed to be rewritten.
Many assembly routines called from C were reading arguments
incorrectly, using MOVL instead of MOVQ or vice versa, especially on
the less used systems like openbsd.
These were found by go vet and have been corrected too.
If we're lucky, this may reduce flakiness on those systems.
Tested on:
darwin/386
darwin/amd64
linux/arm
linux/386
linux/amd64
If this breaks another system, the bug is almost certainly in the
sys_$GOOS_$GOARCH.s file, since the rest of the CL is tested
by the combination of the above systems.
LGTM=dvyukov, iant
R=golang-codereviews, 0intro, dave, alex.brainman, dvyukov, iant
CC=golang-codereviews, josharian, r
https://golang.org/cl/135830043
2014-08-27 09:32:17 -06:00
|
|
|
MOVL usec+0(FP), DI
|
2014-07-02 19:36:05 -06:00
|
|
|
MOVQ $runtime·usleep2(SB), AX // to hide from 6l
|
|
|
|
|
|
|
|
|
|
// Execute call on m->g0.
|
|
|
|
|
get_tls(R15)
|
|
|
|
|
CMPQ R15, $0
|
[dev.power64] cmd/5a, cmd/6a, cmd/8a, cmd/9a: make labels function-scoped
I removed support for jumping between functions years ago,
as part of doing the instruction layout for each function separately.
Given that, it makes sense to treat labels as function-scoped.
This lets each function have its own 'loop' label, for example.
Makes the assembly much cleaner and removes the last
reason anyone would reach for the 123(PC) form instead.
Note that this is on the dev.power64 branch, but it changes all
the assemblers. The change will ship in Go 1.5 (perhaps after
being ported into the new assembler).
Came up as part of CL 167730043.
LGTM=r
R=r
CC=austin, dave, golang-codereviews, minux
https://golang.org/cl/159670043
2014-10-28 19:50:16 -06:00
|
|
|
JE noswitch
|
2014-07-02 19:36:05 -06:00
|
|
|
|
|
|
|
|
MOVQ g(R15), R13
|
|
|
|
|
CMPQ R13, $0
|
[dev.power64] cmd/5a, cmd/6a, cmd/8a, cmd/9a: make labels function-scoped
I removed support for jumping between functions years ago,
as part of doing the instruction layout for each function separately.
Given that, it makes sense to treat labels as function-scoped.
This lets each function have its own 'loop' label, for example.
Makes the assembly much cleaner and removes the last
reason anyone would reach for the 123(PC) form instead.
Note that this is on the dev.power64 branch, but it changes all
the assemblers. The change will ship in Go 1.5 (perhaps after
being ported into the new assembler).
Came up as part of CL 167730043.
LGTM=r
R=r
CC=austin, dave, golang-codereviews, minux
https://golang.org/cl/159670043
2014-10-28 19:50:16 -06:00
|
|
|
JE noswitch
|
2014-07-02 19:36:05 -06:00
|
|
|
MOVQ g_m(R13), R13
|
|
|
|
|
CMPQ R13, $0
|
[dev.power64] cmd/5a, cmd/6a, cmd/8a, cmd/9a: make labels function-scoped
I removed support for jumping between functions years ago,
as part of doing the instruction layout for each function separately.
Given that, it makes sense to treat labels as function-scoped.
This lets each function have its own 'loop' label, for example.
Makes the assembly much cleaner and removes the last
reason anyone would reach for the 123(PC) form instead.
Note that this is on the dev.power64 branch, but it changes all
the assemblers. The change will ship in Go 1.5 (perhaps after
being ported into the new assembler).
Came up as part of CL 167730043.
LGTM=r
R=r
CC=austin, dave, golang-codereviews, minux
https://golang.org/cl/159670043
2014-10-28 19:50:16 -06:00
|
|
|
JE noswitch
|
2014-07-02 19:36:05 -06:00
|
|
|
// TODO(aram): do something about the cpu profiler here.
|
|
|
|
|
|
|
|
|
|
MOVQ m_g0(R13), R14
|
|
|
|
|
CMPQ g(R15), R14
|
[dev.power64] cmd/5a, cmd/6a, cmd/8a, cmd/9a: make labels function-scoped
I removed support for jumping between functions years ago,
as part of doing the instruction layout for each function separately.
Given that, it makes sense to treat labels as function-scoped.
This lets each function have its own 'loop' label, for example.
Makes the assembly much cleaner and removes the last
reason anyone would reach for the 123(PC) form instead.
Note that this is on the dev.power64 branch, but it changes all
the assemblers. The change will ship in Go 1.5 (perhaps after
being ported into the new assembler).
Came up as part of CL 167730043.
LGTM=r
R=r
CC=austin, dave, golang-codereviews, minux
https://golang.org/cl/159670043
2014-10-28 19:50:16 -06:00
|
|
|
JNE switch
|
2014-07-02 19:36:05 -06:00
|
|
|
// executing on m->g0 already
|
|
|
|
|
CALL AX
|
|
|
|
|
RET
|
|
|
|
|
|
[dev.power64] cmd/5a, cmd/6a, cmd/8a, cmd/9a: make labels function-scoped
I removed support for jumping between functions years ago,
as part of doing the instruction layout for each function separately.
Given that, it makes sense to treat labels as function-scoped.
This lets each function have its own 'loop' label, for example.
Makes the assembly much cleaner and removes the last
reason anyone would reach for the 123(PC) form instead.
Note that this is on the dev.power64 branch, but it changes all
the assemblers. The change will ship in Go 1.5 (perhaps after
being ported into the new assembler).
Came up as part of CL 167730043.
LGTM=r
R=r
CC=austin, dave, golang-codereviews, minux
https://golang.org/cl/159670043
2014-10-28 19:50:16 -06:00
|
|
|
switch:
|
2014-07-02 19:36:05 -06:00
|
|
|
// Switch to m->g0 stack and back.
|
|
|
|
|
MOVQ (g_sched+gobuf_sp)(R14), R14
|
|
|
|
|
MOVQ SP, -8(R14)
|
|
|
|
|
LEAQ -8(R14), SP
|
|
|
|
|
CALL AX
|
|
|
|
|
MOVQ 0(SP), SP
|
|
|
|
|
RET
|
|
|
|
|
|
[dev.power64] cmd/5a, cmd/6a, cmd/8a, cmd/9a: make labels function-scoped
I removed support for jumping between functions years ago,
as part of doing the instruction layout for each function separately.
Given that, it makes sense to treat labels as function-scoped.
This lets each function have its own 'loop' label, for example.
Makes the assembly much cleaner and removes the last
reason anyone would reach for the 123(PC) form instead.
Note that this is on the dev.power64 branch, but it changes all
the assemblers. The change will ship in Go 1.5 (perhaps after
being ported into the new assembler).
Came up as part of CL 167730043.
LGTM=r
R=r
CC=austin, dave, golang-codereviews, minux
https://golang.org/cl/159670043
2014-10-28 19:50:16 -06:00
|
|
|
noswitch:
|
2014-07-02 19:36:05 -06:00
|
|
|
// Not a Go-managed thread. Do not switch stack.
|
|
|
|
|
CALL AX
|
|
|
|
|
RET
|
|
|
|
|
|
|
|
|
|
// Runs on OS stack. duration (in µs units) is in DI.
|
|
|
|
|
TEXT runtime·usleep2(SB),NOSPLIT,$0
|
2015-03-30 05:52:07 -06:00
|
|
|
LEAQ libc_usleep(SB), AX
|
2014-07-02 19:36:05 -06:00
|
|
|
CALL AX
|
|
|
|
|
RET
|
|
|
|
|
|
|
|
|
|
// Runs on OS stack, called from runtime·osyield.
|
|
|
|
|
TEXT runtime·osyield1(SB),NOSPLIT,$0
|
2015-03-30 05:52:07 -06:00
|
|
|
LEAQ libc_sched_yield(SB), AX
|
2014-07-02 19:36:05 -06:00
|
|
|
CALL AX
|
|
|
|
|
RET
|
runtime: implement time.now in assembly on plan9, solaris, windows
These all used a C implementation that contained 64-bit divide by 1000000000.
On 32-bit systems that ends up in the 64-bit C divide support, which makes
other calls and ends up using a fair amount of stack. We could convert them
to Go but then they'd still end up in software 64-bit divide code. That would
be okay, because Go code can split the stack, but it's still unnecessary.
Write time·now in assembly, just like on all the other systems, and use the
actual hardware support for 64/32 -> 64/32 division. This cuts the software
routines out entirely.
The actual code to do the division is copied and pasted from the sys_darwin_*.s files.
LGTM=alex.brainman
R=golang-codereviews, alex.brainman
CC=aram, golang-codereviews, iant, khr, r
https://golang.org/cl/136300043
2014-09-07 21:40:59 -06:00
|
|
|
|
|
|
|
|
// func now() (sec int64, nsec int32)
|
|
|
|
|
TEXT time·now(SB),NOSPLIT,$8-12
|
|
|
|
|
CALL runtime·nanotime(SB)
|
|
|
|
|
MOVQ 0(SP), AX
|
|
|
|
|
|
|
|
|
|
// generated code for
|
|
|
|
|
// func f(x uint64) (uint64, uint64) { return x/1000000000, x%100000000 }
|
|
|
|
|
// adapted to reduce duplication
|
|
|
|
|
MOVQ AX, CX
|
|
|
|
|
MOVQ $1360296554856532783, AX
|
|
|
|
|
MULQ CX
|
|
|
|
|
ADDQ CX, DX
|
|
|
|
|
RCRQ $1, DX
|
|
|
|
|
SHRQ $29, DX
|
|
|
|
|
MOVQ DX, sec+0(FP)
|
|
|
|
|
IMULQ $1000000000, DX
|
|
|
|
|
SUBQ DX, CX
|
|
|
|
|
MOVL CX, nsec+8(FP)
|
|
|
|
|
RET
|
