diff options
| author | Russ Cox <rsc@golang.org> | 2014-06-26 11:54:39 -0400 |
|---|---|---|
| committer | Russ Cox <rsc@golang.org> | 2014-06-26 11:54:39 -0400 |
| commit | 89f185fe8a036b0fabce30b20c480cf1c832bdd7 (patch) | |
| tree | cd8c84fca5164747bebd852da7edfed132ce4e70 /src/pkg/runtime/sys_linux_arm.s | |
| parent | 2565b5c06086488b2b23d48929803c8c3cec4400 (diff) | |
| download | go-89f185fe8a036b0fabce30b20c480cf1c832bdd7.tar.xz | |
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
Diffstat (limited to 'src/pkg/runtime/sys_linux_arm.s')
| -rw-r--r-- | src/pkg/runtime/sys_linux_arm.s | 19 |
1 files changed, 11 insertions, 8 deletions
diff --git a/src/pkg/runtime/sys_linux_arm.s b/src/pkg/runtime/sys_linux_arm.s index c537a87223..8bfc72b571 100644 --- a/src/pkg/runtime/sys_linux_arm.s +++ b/src/pkg/runtime/sys_linux_arm.s @@ -244,11 +244,12 @@ TEXT runtime·clone(SB),NOSPLIT,$0 BEQ 2(PC) BL runtime·abort(SB) - MOVW 0(R13), m MOVW 4(R13), g + MOVW 0(R13), R8 + MOVW R8, g_m(g) // paranoia; check they are not nil - MOVW 0(m), R0 + MOVW 0(R8), R0 MOVW 0(g), R0 BL runtime·emptyfunc(SB) // fault if stack check is wrong @@ -256,7 +257,8 @@ TEXT runtime·clone(SB),NOSPLIT,$0 // Initialize m->procid to Linux tid MOVW $SYS_gettid, R7 SWI $0 - MOVW R0, m_procid(m) + MOVW g_m(g), R8 + MOVW R0, m_procid(R8) // Call fn MOVW 8(R13), R0 @@ -285,14 +287,14 @@ TEXT runtime·sigaltstack(SB),NOSPLIT,$0 TEXT runtime·sigtramp(SB),NOSPLIT,$24 // this might be called in external code context, - // where g and m are not set. - // first save R0, because runtime·load_gm will clobber it + // where g is not set. + // first save R0, because runtime·load_g will clobber it MOVW R0, 4(R13) MOVB runtime·iscgo(SB), R0 CMP $0, R0 - BL.NE runtime·load_gm(SB) + BL.NE runtime·load_g(SB) - CMP $0, m + CMP $0, g BNE 4(PC) // signal number is already prepared in 4(R13) MOVW $runtime·badsignal(SB), R11 @@ -304,7 +306,8 @@ TEXT runtime·sigtramp(SB),NOSPLIT,$24 MOVW g, 20(R13) // g = m->gsignal - MOVW m_gsignal(m), g + MOVW g_m(g), R8 + MOVW m_gsignal(R8), g // copy arguments for call to sighandler // R0 is already saved above |