|
From: <sv...@va...> - 2013-10-23 16:23:42
|
Author: cborntra
Date: Wed Oct 23 16:23:30 2013
New Revision: 13688
Log:
We need to have the host code close to the generated code to improve
branch prediction performance on some s390 systems. This change will
move valgrind text close to the initial aspace start for valgrind
allocations.
Modified:
trunk/configure.ac
Modified: trunk/configure.ac
==============================================================================
--- trunk/configure.ac (original)
+++ trunk/configure.ac Wed Oct 23 16:23:30 2013
@@ -564,9 +564,10 @@
VGCONF_ARCH_SEC=""
VGCONF_PLATFORM_PRI_CAPS="S390X_LINUX"
VGCONF_PLATFORM_SEC_CAPS=""
- # we want to have the generated code close to the dispatcher
- valt_load_address_pri_norml="0x401000000"
- valt_load_address_pri_inner="0x410000000"
+ # To improve branch prediction hit rate we want to have
+ # the generated code close to valgrind (host) code
+ valt_load_address_pri_norml="0x800000000"
+ valt_load_address_pri_inner="0x810000000"
valt_load_address_sec_norml="0xUNSET"
valt_load_address_sec_inner="0xUNSET"
AC_MSG_RESULT([ok (${ARCH_MAX}-${VGCONF_OS})])
|
|
From: Christian B. <bor...@de...> - 2013-10-24 09:06:57
|
Julian,
please merge into the 3.9 branch.
The nightly run of the testsuite on sless390 got around 25 secs quicker and no
new regressions.
memchecking a git gc of one of my trees improved from 5minutes to 3 minutes. (3.8
is also at 3 minutes).
Christian
On 23/10/13 18:23, sv...@va... wrote:
> Author: cborntra
> Date: Wed Oct 23 16:23:30 2013
> New Revision: 13688
>
> Log:
> We need to have the host code close to the generated code to improve
> branch prediction performance on some s390 systems. This change will
> move valgrind text close to the initial aspace start for valgrind
> allocations.
>
>
> Modified:
> trunk/configure.ac
>
> Modified: trunk/configure.ac
> ==============================================================================
> --- trunk/configure.ac (original)
> +++ trunk/configure.ac Wed Oct 23 16:23:30 2013
> @@ -564,9 +564,10 @@
> VGCONF_ARCH_SEC=""
> VGCONF_PLATFORM_PRI_CAPS="S390X_LINUX"
> VGCONF_PLATFORM_SEC_CAPS=""
> - # we want to have the generated code close to the dispatcher
> - valt_load_address_pri_norml="0x401000000"
> - valt_load_address_pri_inner="0x410000000"
> + # To improve branch prediction hit rate we want to have
> + # the generated code close to valgrind (host) code
> + valt_load_address_pri_norml="0x800000000"
> + valt_load_address_pri_inner="0x810000000"
> valt_load_address_sec_norml="0xUNSET"
> valt_load_address_sec_inner="0xUNSET"
> AC_MSG_RESULT([ok (${ARCH_MAX}-${VGCONF_OS})])
>
>
> ------------------------------------------------------------------------------
> October Webinars: Code for Performance
> Free Intel webinars can help you accelerate application performance.
> Explore tips for MPI, OpenMP, advanced profiling, and more. Get the most from
> the latest Intel processors and coprocessors. See abstracts and register >
> http://pubads.g.doubleclick.net/gampad/clk?id=60135991&iu=/4140/ostg.clktrk
> _______________________________________________
> Valgrind-developers mailing list
> Val...@li...
> https://lists.sourceforge.net/lists/listinfo/valgrind-developers
>
|