mirror of
https://github.com/golang/go
synced 2024-09-04 23:44:16 +00:00
runtime: use QPC for nanotime and time.now on windows/arm
The previous implementation of nanotime and time.now used a time source that was updated on the system clock tick, which has a maximum resolution of about 1ms. On 386 and amd64, this time source maps to the system performance counter, so has much higher resolution. On ARM, use QueryPerformanceCounter() to get a high resolution timestamp. Updates #26148 Change-Id: I1abc99baf927a95b472ac05020a7788626c71d08 Reviewed-on: https://go-review.googlesource.com/c/154758 Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org> Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org>
This commit is contained in:
parent
9ded8b0e97
commit
e3b4b7baad
|
@ -198,9 +198,12 @@ func loadOptionalSyscalls() {
|
|||
}
|
||||
_NtWaitForSingleObject = windowsFindfunc(n32, []byte("NtWaitForSingleObject\000"))
|
||||
|
||||
if windowsFindfunc(n32, []byte("wine_get_version\000")) != nil {
|
||||
// running on Wine
|
||||
initWine(k32)
|
||||
underWine := windowsFindfunc(n32, []byte("wine_get_version\000")) != nil
|
||||
if underWine || GOARCH == "arm" {
|
||||
initQPC(k32)
|
||||
}
|
||||
if underWine {
|
||||
initWine()
|
||||
}
|
||||
}
|
||||
|
||||
|
@ -357,7 +360,7 @@ func nowQPC() (sec int64, nsec int32, mono int64) {
|
|||
return
|
||||
}
|
||||
|
||||
func initWine(k32 uintptr) {
|
||||
func initQPC(k32 uintptr) {
|
||||
_GetSystemTimeAsFileTime = windowsFindfunc(k32, []byte("GetSystemTimeAsFileTime\000"))
|
||||
if _GetSystemTimeAsFileTime == nil {
|
||||
throw("could not find GetSystemTimeAsFileTime() syscall")
|
||||
|
@ -394,7 +397,9 @@ func initWine(k32 uintptr) {
|
|||
// We have to do it this way (or similar), since multiplying QPC counter by 100 millions overflows
|
||||
// int64 and resulted time will always be invalid.
|
||||
qpcMultiplier = int64(timediv(1000000000, qpcFrequency, nil))
|
||||
}
|
||||
|
||||
func initWine() {
|
||||
useQPCTime = 1
|
||||
}
|
||||
|
||||
|
|
|
@ -487,115 +487,12 @@ TEXT runtime·read_tls_fallback(SB),NOSPLIT|NOFRAME,$0
|
|||
MOVW R0, (R0)
|
||||
RET
|
||||
|
||||
// See http://www.dcl.hpi.uni-potsdam.de/research/WRK/2007/08/getting-os-information-the-kuser_shared_data-structure/
|
||||
// Must read hi1, then lo, then hi2. The snapshot is valid if hi1 == hi2.
|
||||
#define _INTERRUPT_TIME 0x7ffe0008
|
||||
#define _SYSTEM_TIME 0x7ffe0014
|
||||
#define time_lo 0
|
||||
#define time_hi1 4
|
||||
#define time_hi2 8
|
||||
|
||||
TEXT runtime·nanotime(SB),NOSPLIT,$0-8
|
||||
MOVW $0, R0
|
||||
MOVB runtime·useQPCTime(SB), R0
|
||||
CMP $0, R0
|
||||
BNE useQPC
|
||||
MOVW $_INTERRUPT_TIME, R3
|
||||
loop:
|
||||
MOVW time_hi1(R3), R1
|
||||
MOVW time_lo(R3), R0
|
||||
MOVW time_hi2(R3), R2
|
||||
CMP R1, R2
|
||||
BNE loop
|
||||
|
||||
// wintime = R1:R0, multiply by 100
|
||||
MOVW $100, R2
|
||||
MULLU R0, R2, (R4, R3) // R4:R3 = R1:R0 * R2
|
||||
MULA R1, R2, R4, R4
|
||||
|
||||
// wintime*100 = R4:R3
|
||||
MOVW R3, ret_lo+0(FP)
|
||||
MOVW R4, ret_hi+4(FP)
|
||||
RET
|
||||
useQPC:
|
||||
TEXT runtime·nanotime(SB),NOSPLIT|NOFRAME,$0-8
|
||||
B runtime·nanotimeQPC(SB) // tail call
|
||||
RET
|
||||
|
||||
TEXT time·now(SB),NOSPLIT,$0-20
|
||||
MOVW $0, R0
|
||||
MOVB runtime·useQPCTime(SB), R0
|
||||
CMP $0, R0
|
||||
BNE useQPC
|
||||
MOVW $_INTERRUPT_TIME, R3
|
||||
loop:
|
||||
MOVW time_hi1(R3), R1
|
||||
MOVW time_lo(R3), R0
|
||||
MOVW time_hi2(R3), R2
|
||||
CMP R1, R2
|
||||
BNE loop
|
||||
|
||||
// wintime = R1:R0, multiply by 100
|
||||
MOVW $100, R2
|
||||
MULLU R0, R2, (R4, R3) // R4:R3 = R1:R0 * R2
|
||||
MULA R1, R2, R4, R4
|
||||
|
||||
// wintime*100 = R4:R3
|
||||
MOVW R3, mono+12(FP)
|
||||
MOVW R4, mono+16(FP)
|
||||
|
||||
MOVW $_SYSTEM_TIME, R3
|
||||
wall:
|
||||
MOVW time_hi1(R3), R1
|
||||
MOVW time_lo(R3), R0
|
||||
MOVW time_hi2(R3), R2
|
||||
CMP R1, R2
|
||||
BNE wall
|
||||
|
||||
// w = R1:R0 in 100ns untis
|
||||
// convert to Unix epoch (but still 100ns units)
|
||||
#define delta 116444736000000000
|
||||
SUB.S $(delta & 0xFFFFFFFF), R0
|
||||
SBC $(delta >> 32), R1
|
||||
|
||||
// Convert to nSec
|
||||
MOVW $100, R2
|
||||
MULLU R0, R2, (R4, R3) // R4:R3 = R1:R0 * R2
|
||||
MULA R1, R2, R4, R4
|
||||
// w = R2:R1 in nSec
|
||||
MOVW R3, R1 // R4:R3 -> R2:R1
|
||||
MOVW R4, R2
|
||||
|
||||
// multiply nanoseconds by reciprocal of 10**9 (scaled by 2**61)
|
||||
// to get seconds (96 bit scaled result)
|
||||
MOVW $0x89705f41, R3 // 2**61 * 10**-9
|
||||
MULLU R1,R3,(R6,R5) // R7:R6:R5 = R2:R1 * R3
|
||||
MOVW $0,R7
|
||||
MULALU R2,R3,(R7,R6)
|
||||
|
||||
// unscale by discarding low 32 bits, shifting the rest by 29
|
||||
MOVW R6>>29,R6 // R7:R6 = (R7:R6:R5 >> 61)
|
||||
ORR R7<<3,R6
|
||||
MOVW R7>>29,R7
|
||||
|
||||
// subtract (10**9 * sec) from nsec to get nanosecond remainder
|
||||
MOVW $1000000000, R5 // 10**9
|
||||
MULLU R6,R5,(R9,R8) // R9:R8 = R7:R6 * R5
|
||||
MULA R7,R5,R9,R9
|
||||
SUB.S R8,R1 // R2:R1 -= R9:R8
|
||||
SBC R9,R2
|
||||
|
||||
// because reciprocal was a truncated repeating fraction, quotient
|
||||
// may be slightly too small -- adjust to make remainder < 10**9
|
||||
CMP R5,R1 // if remainder > 10**9
|
||||
SUB.HS R5,R1 // remainder -= 10**9
|
||||
ADD.HS $1,R6 // sec += 1
|
||||
|
||||
MOVW R6,sec_lo+0(FP)
|
||||
MOVW R7,sec_hi+4(FP)
|
||||
MOVW R1,nsec+8(FP)
|
||||
RET
|
||||
useQPC:
|
||||
B runtime·nanotimeQPC(SB) // tail call
|
||||
TEXT time·now(SB),NOSPLIT|NOFRAME,$0-20
|
||||
B runtime·nowQPC(SB) // tail call
|
||||
RET
|
||||
|
||||
// save_g saves the g register (R10) into thread local memory
|
||||
|
|
Loading…
Reference in a new issue