diff options
author | marcel <marcel@FreeBSD.org> | 2003-05-31 20:57:26 +0000 |
---|---|---|
committer | marcel <marcel@FreeBSD.org> | 2003-05-31 20:57:26 +0000 |
commit | e4d5efee3904d9c754b521809c7331fb9bf66ec0 (patch) | |
tree | 8658cb133103786cd18c6b3fab619f69cdda59cd /sys/ia64 | |
parent | d2dc9cb8f4a5d68ca7c05062528637bd3ed4958e (diff) | |
download | FreeBSD-src-e4d5efee3904d9c754b521809c7331fb9bf66ec0.zip FreeBSD-src-e4d5efee3904d9c754b521809c7331fb9bf66ec0.tar.gz |
Some ia32 related finetuning for the EPC syscall path:
o The SDM states that flushing the RSE in the cycle prior to the
call to ia32 code yields the best performance. We don't really
care to much about performance here, but we do the same anyway.
I'm being paranoia and conservative here.
o Only initialize the ia32 state registers, not the registers used
as scratch by the ia32 engine. This saves a couple of loads from
the trapframe, but also helps debugging: we don't clobber useful
debugging data (engineering hints :-)
o Make sure all general registers constituting ia32 state have been
initialized. If there's no useful to be loaded from the trapframe,
clear the register. This avoids accidentally leaking NaT bits.
o Make sure we set ar.k6 prior to clobbering ar.bspstore and also
set ar.k7 prior to setting sp. This fixes a race seen for ia64
native code as well (and previously fixed too).
Diffstat (limited to 'sys/ia64')
-rw-r--r-- | sys/ia64/ia64/syscall.S | 57 | ||||
-rw-r--r-- | sys/ia64/ia64/syscall.s | 57 |
2 files changed, 64 insertions, 50 deletions
diff --git a/sys/ia64/ia64/syscall.S b/sys/ia64/ia64/syscall.S index 5aacd79..33e94ce 100644 --- a/sys/ia64/ia64/syscall.S +++ b/sys/ia64/ia64/syscall.S @@ -101,12 +101,19 @@ gw_ret: ;; } gw_ret_ia32: - mov ar.rnat=r0 - mov ar.rsc=0xc - mov ar.pfs=r0 +{ .mfb + flushrs + nop 0 + nop 0 ;; +} +{ .mfb + nop 0 + nop 0 br.ia.sptk b6 ;; +} + ENTRY(break_sigtramp, 0) { .mib @@ -509,33 +516,24 @@ epc_syscall_setup_ia32: { .mmi mov r30=ar.bspstore ;; - mov ar.bspstore=r21 + mov ar.unat=r17 dep r30=0,r30,0,9 ;; } { .mmi mov ar.k6=r30 - mov ar.unat=r17 - mov r11=r26 + mov ar.bspstore=r21 + mov r11=r0 ;; } - - ld8 r16=[r14],16 - ld8 r17=[r15],16 - ;; - ld8 r18=[r14],16 - ld8 r19=[r15],16 - ;; - ld8 r20=[r14],16 - ld8 r21=[r15],16 - ;; - ld8 r22=[r14],16 - ld8 r23=[r15],16 - ;; - ld8 r24=[r14],16 - ld8 r25=[r15],16 +{ .mmi + ld8 r16=[r14],64 + ld8 r17=[r15],80 + mov r13=r0 ;; - ld8 r26=[r14],16 +} + + ld8 r24=[r14],32 ld8 r27=[r15],16 ;; ld8 r28=[r14],16 @@ -544,18 +542,27 @@ epc_syscall_setup_ia32: ld8 r30=[r14],40 ld8 r31=[r15],40 ;; - ld8 r2=[r14],16 - ld8 r3=[r15],8 + +{ .mmi + ld8 r2=[r14] + ld8 r3=[r15] + mov r14=r0 ;; +} +{ .mmi mov ar.csd=r2 mov ar.ssd=r3 + mov r15=r0 ;; +} + mov r2=ar.k5 - mov psr.l=r11 + mov psr.l=r26 ;; srlz.d add r2=gw_ret_ia32-ia64_gateway_page,r2 ;; + mov ar.rsc=0x0 mov b7=r2 br.ret.sptk b7 ;; diff --git a/sys/ia64/ia64/syscall.s b/sys/ia64/ia64/syscall.s index 5aacd79..33e94ce 100644 --- a/sys/ia64/ia64/syscall.s +++ b/sys/ia64/ia64/syscall.s @@ -101,12 +101,19 @@ gw_ret: ;; } gw_ret_ia32: - mov ar.rnat=r0 - mov ar.rsc=0xc - mov ar.pfs=r0 +{ .mfb + flushrs + nop 0 + nop 0 ;; +} +{ .mfb + nop 0 + nop 0 br.ia.sptk b6 ;; +} + ENTRY(break_sigtramp, 0) { .mib @@ -509,33 +516,24 @@ epc_syscall_setup_ia32: { .mmi mov r30=ar.bspstore ;; - mov ar.bspstore=r21 + mov ar.unat=r17 dep r30=0,r30,0,9 ;; } { .mmi mov ar.k6=r30 - mov ar.unat=r17 - mov r11=r26 + mov ar.bspstore=r21 + mov r11=r0 ;; } - - ld8 r16=[r14],16 - ld8 r17=[r15],16 - ;; - ld8 r18=[r14],16 - ld8 r19=[r15],16 - ;; - ld8 r20=[r14],16 - ld8 r21=[r15],16 - ;; - ld8 r22=[r14],16 - ld8 r23=[r15],16 - ;; - ld8 r24=[r14],16 - ld8 r25=[r15],16 +{ .mmi + ld8 r16=[r14],64 + ld8 r17=[r15],80 + mov r13=r0 ;; - ld8 r26=[r14],16 +} + + ld8 r24=[r14],32 ld8 r27=[r15],16 ;; ld8 r28=[r14],16 @@ -544,18 +542,27 @@ epc_syscall_setup_ia32: ld8 r30=[r14],40 ld8 r31=[r15],40 ;; - ld8 r2=[r14],16 - ld8 r3=[r15],8 + +{ .mmi + ld8 r2=[r14] + ld8 r3=[r15] + mov r14=r0 ;; +} +{ .mmi mov ar.csd=r2 mov ar.ssd=r3 + mov r15=r0 ;; +} + mov r2=ar.k5 - mov psr.l=r11 + mov psr.l=r26 ;; srlz.d add r2=gw_ret_ia32-ia64_gateway_page,r2 ;; + mov ar.rsc=0x0 mov b7=r2 br.ret.sptk b7 ;; |