Skip to contentSkip to conversation

speed up lfetch.fault [NULL] (take 2)

PATCH In review David Mosberger 2 participants 2 messages May 4, 2005

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 speed up lfetch.fault [NULL] (take 2)

This is a resend of the patch from March 26, with slight tweaking to omit the instruction needed to initialize p6. Please apply to test-2.6.

Thanks,

	--david

[IA64] Speed up lfetch.fault [NULL]

This patch greatly speeds up the handling of lfetch.fault instructions which result in NaT consumption. Due to the NaT-page mapped at address 0, this is guaranteed to happen when lfetch.fault'ing a NULL pointer. With this patch in place, we can even define prefetch()/prefetchw() as lfetch.fault without significant performance degradation. More importantly, it allows compilers to be more aggressive with using lfetch.fault on pointers that might be NULL.

Index: arch/ia64/kernel/ivt.S
=================================--- f52dd11fe8741b6589019c9fa6cff53491e9e853/arch/ia64/kernel/ivt.S (mode:100644 sha1:d9c05d53435bd1d541538494336cffcc65df83c4) +++ uncommitted/arch/ia64/kernel/ivt.S (mode:100644)
@@ -1199,6 +1199,24 @@
// 0x5600 Entry 26 (size 16 bundles) Nat Consumption (11,23,37,50)
ENTRY(nat_consumption)
DBG_FAULT(26)
+
+ mov r16=cr.ipsr
+ mov r17=cr.isr
+ mov r31=pr // save PR
+ ;;
+ and r18=0xf,r17 // r18 = cr.ipsr.code{3:0}
+ tbit.z p6,p0=r17,IA64_ISR_NA_BIT
+ ;;
+ cmp.ne.or p6,p0=IA64_ISR_CODE_LFETCH,r18
+ dep r16=-1,r16,IA64_PSR_ED_BIT,1
+(p6) br.cond.spnt 1f // branch if (cr.ispr.na = 0 || cr.ipsr.code{3:0} != LFETCH)
+ ;;
+ mov cr.ipsr=r16 // set cr.ipsr.na
+ ;;
+ rfi
+
+1: mov pr=r31,-1
+ ;;
FAULT(26)
END(nat_consumption)

#1 lore raw

David Mosberger wrote on Wednesday, May 04, 2005 7:50 AM

Quoted 3 lines

This is a resend of the patch from March 26, with slight tweaking to omit the instruction needed to initialize p6. Please apply to test-2.6.

David, the pr register need to be restored in the fast path ;-) otherwise, p6 is clobbered.

- Ken