And I found the other reason (Re: Found one reason libibnetdisc is slower than subnet_discover)

2010-01-22 Thread Ira Weiny
It looks like subnet_discover was actually allowing n+1 smps on the wire. With this patch: diff --git a/tests/subnet_discover.c b/tests/subnet_discover.c index acc8c23..22b092a 100644 --- a/tests/subnet_discover.c +++ b/tests/subnet_discover.c @@ -189,7 +189,7 @@ static void run_request_queue(in

ib_write_bw hanging when using max max_inline value

2010-01-22 Thread HÃ¥kon Bugge
Hi, I modified ib_write_bw to support inline sizes beyond the default value of 400 (not sure why the program does not allow larger values though). When using the largest max_inline value supported by ConnectX (i.e. the creation of QP goes OK), the test program hangs when exchanging 920 bytes.

Re: Problems using krping

2010-01-22 Thread Steve Wise
lihaidong wrote: Mr.Wise: Sorry to Bother you Another problem. fastreg(Must with local_dma_key to avoid using ib_reg_phys_mr in my case) with read_inv succeed, yet with server_inv failed. What does 'krping: cq completion failed with wr_id 0 status 6 opcode -1 vender_err 78' in client dme