[mpich2-dev] [mpi-all-commits] r3669 - mpich2/branches/dev/knem/src/mpid/ch3/channels/nemesis/nemesis/src
William Gropp
wgropp at illinois.edu
Wed Dec 10 08:22:39 CST 2008
This change reminded me that there are standards in MPICH for
parameter handling to provide some uniformity and (in the full case)
embedded self documentation. Its very easy for a large package like
MPICH2 to accumulate a wide variety of parameters with different
handling and naming conventions, and it is easier on everyone in the
long run if these follow a standard form.
In brief, the standard suggests using MPICH_ as the prefix and using a
common set of routines that can ensure that the values are available
and correct (what happens if I set this value to the string "64k").
Bill
On Dec 9, 2008, at 6:36 PM, bgoglin at mcs.anl.gov wrote:
> Author: bgoglin
> Date: 2008-12-09 18:36:48 -0600 (Tue, 09 Dec 2008)
> New Revision: 3669
>
> Modified:
> mpich2/branches/dev/knem/src/mpid/ch3/channels/nemesis/nemesis/src/
> mpid_nem_lmt_dma.c
> Log:
> set NEM_LMT_DMA_THRESHOLD=<n> in the environment to change the DMA
> threshold, 2MB by default
>
> Modified: mpich2/branches/dev/knem/src/mpid/ch3/channels/nemesis/
> nemesis/src/mpid_nem_lmt_dma.c
> ===================================================================
> --- mpich2/branches/dev/knem/src/mpid/ch3/channels/nemesis/nemesis/
> src/mpid_nem_lmt_dma.c 2008-12-09 22:16:00 UTC (rev 3668)
> +++ mpich2/branches/dev/knem/src/mpid/ch3/channels/nemesis/nemesis/
> src/mpid_nem_lmt_dma.c 2008-12-10 00:36:48 UTC (rev 3669)
> @@ -50,7 +50,12 @@
> int err;
> int i;
> struct knem_cmd_info info;
> + char *env;
>
> + env = getenv("NEM_LMT_DMA_THRESHOLD");
> + if (env)
> + dma_threshold = atol(env);
> +
> knem_fd = open(KNEM_DEVICE_FILENAME, O_RDWR);
> MPIU_ERR_CHKANDJUMP2(knem_fd < 0, mpi_errno, MPI_ERR_OTHER,
> "**shm_open",
> "**shm_open %s %d", KNEM_DEVICE_FILENAME,
> errno);
>
William Gropp
Deputy Director for Research
Institute for Advanced Computing Applications and Technologies
Paul and Cynthia Saylor Professor of Computer Science
University of Illinois Urbana-Champaign
More information about the mpich2-dev
mailing list