lmbench-users — technical discussion list for lmbench

[Lmbench-users] build issue in ubuntu 22.04

[Deva N. B. <dnb...@gm...>]

It builds and runs well on ubuntu 18.04
I traced the problem to rpc/rpc.h in libc6-dev package
It needs the old libc6-dev(2.27) but with the new ubuntu it has been
updated to 2.35.
what do i do? Is lmbench not supported in newer linux installations? any
workarounds?

[Lmbench-users] [PATCH v0] ILP32 support fix

[Vadim L. <Vad...@ca...>]

This patch adress overflow issue being observed
while running the lat_mem_rd test build against 
ILP32 support at arm64 board.

Vadim Lomovtsev (1):
  ilp32: fix variable overflow

 src/lat_mem_rd.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

-- 
2.5.5

[Lmbench-users] [PATCH] ilp32: fix variable overflow

[Vadim L. <Vad...@ca...>]

While running lat_mem_rd test (build against ilp32 support) at ARM64 system
it was found that test goes into endless loop due to variable overflow.
To fix that it is neccesary to use 64-bit width type for loop control variable.

Signed-off-by: Vadim Lomovtsev <Vad...@ca...>
---
 src/lat_mem_rd.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/src/lat_mem_rd.c b/src/lat_mem_rd.c
index e7412d7..365aa11 100644
--- a/src/lat_mem_rd.c
+++ b/src/lat_mem_rd.c
@@ -159,7 +159,7 @@ step(size_t k)
         } else if (k < 4*1024) {
 		k += 1024;
 	} else {
-		size_t s;
+		unsigned long long s;
 
 		for (s = 4 * 1024; s <= k; s *= 2)
 			;
-- 
2.5.5

[Lmbench-users] [PATCH v0] ILP32 support fix

[Vadim L. <Vad...@ca...>]

This patch adress overflow issue being observed
while running the lat_mem_rd test build against 
ILP32 support at arm64 board.

Vadim Lomovtsev (1):
  ilp32: fix variable overflow

 src/lat_mem_rd.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

-- 
2.5.5

[Lmbench-users] [PATCH] ilp32: fix variable overflow

[Vadim L. <Vad...@ca...>]

While running lat_mem_rd test (build against ilp32 support) at ARM64 system
it was found that test goes into endless loop due to variable overflow.
To fix that it is neccesary to use 64-bit width type for loop control variable.

Signed-off-by: Vadim Lomovtsev <Vad...@ca...>
---
 src/lat_mem_rd.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/src/lat_mem_rd.c b/src/lat_mem_rd.c
index e7412d7..365aa11 100644
--- a/src/lat_mem_rd.c
+++ b/src/lat_mem_rd.c
@@ -159,7 +159,7 @@ step(size_t k)
         } else if (k < 4*1024) {
 		k += 1024;
 	} else {
-		size_t s;
+		unsigned long long s;
 
 		for (s = 4 * 1024; s <= k; s *= 2)
 			;
-- 
2.5.5

Re: [Lmbench-users] [lmbench][question] lat_mem_rd built against ilp32 hangs forever with len=2048

[Lomovtsev, V. <Vad...@ca...>]

It turns to be lmbench related issue.
Thanks.

-Vadim

________________________________________
From: linux-arm-kernel <lin...@li...> on behalf of Lomovtsev, Vadim <Vad...@ca...>
Sent: Thursday, March 31, 2016 17:20
To: lmb...@li...
Cc: Kapoor, Prasun; Norov, Yuri; lin...@vg...; Pinski, Andrew; Prakash, Brahmajyosyula; lin...@li...
Subject: [lmbench][question] lat_mem_rd built against ilp32 hangs forever with len=2048

While running lat_mem_rd test build against ILP32 support at ARM64 board it hangs forever with following logs (CPU load is 100%):

./lat_mem_rd -P 1 2048 16 32
0.00049 1.501
0.00098 1.501
0.00195 1.501
[.. skip]
1024.00000 13.076
1536.00000 13.082
2048.00000 13.076

At the same time lat_mem_rd build with lp64 works fine with exactly same args.

Has anyone face such issue before?

My kernel version is 4.2 and ILP32 patches are under development being backported from http://lkml.iu.edu/hypermail/linux/kernel/1601.1/05453.html.
Sources for 4.5 available at https://github.com/norov/linux

Please let me know if any additional information is necessary.

-Vadim
_______________________________________________
linux-arm-kernel mailing list
lin...@li...
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

[Lmbench-users] [lmbench][question] lat_mem_rd built against ilp32 hangs forever with len=2048

[Lomovtsev, V. <Vad...@ca...>]

While running lat_mem_rd test build against ILP32 support at ARM64 board it hangs forever with following logs (CPU load is 100%):

./lat_mem_rd -P 1 2048 16 32
0.00049 1.501
0.00098 1.501
0.00195 1.501
[.. skip] 
1024.00000 13.076
1536.00000 13.082
2048.00000 13.076 

At the same time lat_mem_rd build with lp64 works fine with exactly same args.

Has anyone face such issue before?

My kernel version is 4.2 and ILP32 patches are under development being backported from http://lkml.iu.edu/hypermail/linux/kernel/1601.1/05453.html.
Sources for 4.5 available at https://github.com/norov/linux

Please let me know if any additional information is necessary.

-Vadim

[Lmbench-users] problem while increasing process size in LMBench

[Madhu <c.r...@gm...>]

Hi All,

     Can anybody help me on LMBench 'Context switching timer' (lat_ctx)
test. In manual, it is mentioned "lat_ctx measures context switching time
for any reasonable number of processes of any reasonable size".
      By default , lmbench tests 'context switching time' for the process
size  2k, 4k, .. upto 64 k with various number of process combinations (ie.
number of process as 2,4,6, upto 96 processes)  .
      But If I try increasing the process size (ie,  2mb, 4mb ) I am not
getting any results or error. on other lesser sizes , I am able to see the
results as the no.of processes and time taken to switch between them. I have
tested this on ARMv6 board having 128 mb ram and in a normal Pentium 4 PC
having 1Gb memory.



Thanks,
Madhu

Re: [Lmbench-users] change log

[pavankumar k. <pav...@gm...>]

Hi

can anyone explain how lat_http works??  after getting socket descriptor
from tcp_connect(), when i tried writing to that socket, it was saying bad
file descriptor. the same tcp_connect() works fine with other benchmarks
like lat_tcp, bw_tcp.
lat_http accepts a port number. so which port number should i give for
that???

thanks
pk

[Lmbench-users] change log

[pavankumar k. <pav...@gm...>]

HI

I have been using  LMBench 3.0a4 for some time. I want to move to LMBench
3.0a8 recent release. where can i get change log for this release.

Thanks in advance
PK

[Lmbench-users] Not enough memory error

[Tong Li <to...@cs...>]

I'm running lmbench3 on an SMP x86_64 system with Linux 2.6.14 and get the 
following the error:

Checking to see if you have 512 MB; please wait for a moment...
3MB OK


../scripts/config-run aborted: Not enough memory, only 1MB available.

My system has sufficient memory, so I don't understand why this is 
happening. It looks like touchRange() in memsize.c is taking too much time 
and I may need to increase TOO_LONG, but I don't know if this is the 
correct way to fix the problem. Does anyone know? Also, I'd like to know 
the rationale in the current code of setting TOO_LONG to 10.

Thanks,

   tong