Data transferred by `MPI_BSEND` may corrupt if all of the following
conditions are met.
- The message size is less than the eager limit.
- The `btl_alloc` function in the BTL interface returns `NULL`
for some reason.
- The MPI program overwrites the send buffer after `MPI_BSEND`
returns.
The problem is in the way of pending a send request in ob1 PML.
The `mca_pml_ob1_send_request_start_copy` function retruns
`OMPI_ERR_OUT_OF_RESOURCE` if `mca_bml_base_alloc` function returns
`des = NULL`. In this case, the send request is added to the
`send_pending` list and `MPI_BSEND` returns immediately. Next time
the `mca_pml_ob1_send_request_start_copy` function tries sending,
the user buffer may have been overwritten by the MPI program.
Call hierarchy of `MPI_BSEND`:
```
MPI_Bsend
mca_pml_ob1_send
if (MCA_PML_BASE_SEND_BUFFERED == sendmode)
mca_pml_ob1_isend
MCA_PML_OB1_SEND_REQUEST_START_W_SEQ
mca_pml_ob1_send_request_start_seq
mca_pml_ob1_send_request_start_btl
if (size <= eager_limit)
if (req_send_mode == MCA_PML_BASE_SEND_BUFFERED)
mca_pml_ob1_send_request_start_copy
mca_bml_base_alloc
btl_alloc
if (OMPI_ERR_OUT_OF_RESOURCE == rc)
add_request_to_send_pending
ompi_request_free
```
To solve this problem, we should save the data to the buffer
attached by `MPI_BUFFER_ATTACH` before leaving `MPI_BSEND`.
This problem was introduced by ob1 optimization (commits 2b57f422
and a06e491c) in v1.8 series.
Signed-off-by: KAWASHIMA Takahiro <t-kawashima@jp.fujitsu.com>
This commit renames the arithmetic atomic operations in opal to
indicate that they return the new value not the old value. This naming
differentiates these routines from new functions that return the old
value.
Signed-off-by: Nathan Hjelm <hjelmn@lanl.gov>
they are supposed to be unsigned, casting them to a signed
value for all atomic operations is as errorprone as handling
them as signed entities.
Signed-off-by: George Bosilca <bosilca@icl.utk.edu>
On start we were not correctly resetting all request fields. This was
leading to a double-completion on persistent receives. This commit
updates the base start code to reset the receive req_bytes_packed and
the send request convertor.
Signed-off-by: Nathan Hjelm <hjelmn@lanl.gov>
The request code was setting the request as pml_complete before
calling MCA_PML_OB1_SEND_REQUEST_MPI_COMPLETE. This was causing
MCA_PML_OB1_SEND_REQUEST_RETURN to be called twice in some cases. The
code now mirrors the recvreq code and only sets the request as pml
complete if the request has not already been freed.
Signed-off-by: Nathan Hjelm <hjelmn@lanl.gov>
* Remodel the request.
Added the wait sync primitive and integrate it into the PML and MTL
infrastructure. The multi-threaded requests are now significantly
less heavy and less noisy (only the threads associated with completed
requests are signaled).
* Fix the condition to release the request.
If during the request completion callback we post another request that
completes right away (such a small send or a match for an unexpected
short message) we will try to complete the second request while holding
the lock for the completion of the first. For performance reasons
(mainly to avoid unlocking and locking the request mutex several times)
we have made the request lock recursive.
This commit contains the following changes:
- pml/ob1: use the bml accessor function when requesting a bml
endpoint. this will ensure that bml endpoints are only created when
needed. for example, a bml endpoint is not requested and not
allocated when receiving an eager message from a peer.
- pml/ob1: change the pml_procs array in the ob1 communicator to a
proc pointer array. at the cost of a single level of extra
redirection this will allow us to allocate pml procs on demand.
- pml/ob1: add an accessor function to access the pml proc structure
for a given peer. this function will allocate the proc if it
doesn't already exist.
Signed-off-by: Nathan Hjelm <hjelmn@lanl.gov>
This commit fixes several bugs in the static request objects used by
ob1 for blocking send/receive operations.
- Fix memory leak when using MPI_THREAD_MULTIPLE. Requests were
allocated off the free list but were destructed and NOT returned.
- Fix double-destruct of static objects. There is no reason to
CONSTRUCT/DESTUCT the static object for each send/receive
operation. This adds overhead and no benefit. To keep the code
clean helper functions have been added to finalize ob1 send/receive
requests.
- Remove now unnecessary include of alloca.h.
Signed-off-by: Nathan Hjelm <hjelmn@me.com>
Use of the old ompi_free_list_t and ompi_free_list_item_t is
deprecated. These classes will be removed in a future commit.
This commit updates the entire code base to use opal_free_list_t and
opal_free_list_item_t.
Notes:
OMPI_FREE_LIST_*_MT -> opal_free_list_* (uses opal_using_threads ())
Signed-off-by: Nathan Hjelm <hjelmn@lanl.gov>
WHAT: Open our low-level communication infrastructure by moving all necessary components (btl/rcache/allocator/mpool) down in OPAL
All the components required for inter-process communications are currently deeply integrated in the OMPI layer. Several groups/institutions have express interest in having a more generic communication infrastructure, without all the OMPI layer dependencies. This communication layer should be made available at a different software level, available to all layers in the Open MPI software stack. As an example, our ORTE layer could replace the current OOB and instead use the BTL directly, gaining access to more reactive network interfaces than TCP. Similarly, external software libraries could take advantage of our highly optimized AM (active message) communication layer for their own purpose. UTK with support from Sandia, developped a version of Open MPI where the entire communication infrastucture has been moved down to OPAL (btl/rcache/allocator/mpool). Most of the moved components have been updated to match the new schema, with few exceptions (mainly BTLs where I have no way of compiling/testing them). Thus, the completion of this RFC is tied to being able to completing this move for all BTLs. For this we need help from the rest of the Open MPI community, especially those supporting some of the BTLs. A non-exhaustive list of BTLs that qualify here is: mx, portals4, scif, udapl, ugni, usnic.
This commit was SVN r32317.
Per RFC. There are two optimizations in this commit:
- Allocate requests for blocking sends and receives on the stack. This
bypasses the request free list and saves two atomics on the critical path.
This change improves the small message ping-pong by 50-200ns on both AMD
and Intel CPUs.
- For small messages try to use the btl sendi function before intializing a
send request. If the sendi fails or the btl does not have a sendi function
silently fallback on the standard send path.
cmr=v1.7.5:reviewer=brbarret
This commit was SVN r30343.
http://www.open-mpi.org/community/lists/devel/2013/10/13072.php
Add support for pinning GPU Direct RDMA in openib BTL for better small message latency of GPU buffers.
Note that none of this is compiled in unless CUDA-aware support is requested.
This commit was SVN r29680.
configure-time dynamic allocation of flags. The net result for platforms
which only support BTL-based communication is a reduction of 8*nprocs bytes
per process. Platforms which support both MTLs and BTLs will not see
a space reduction, but will now be able to safely run both the MTL and BTL
side-by-side, which will prove useful.
This commit was SVN r29100.
value to signal that the operation of retrieving the element from the free list
failed. However in this case the returned pointer was set to NULL as well, so the
error code was redundant. Moreover, this was a continuous source of warnings when
the picky mode is on.
The attached parch remove the rc argument from the OMPI_FREE_LIST_GET and
OMPI_FREE_LIST_WAIT macros, and change to check if the item is NULL instead of
using the return code.
This commit was SVN r28722.
This macro is only used on the failure path so the additional if statement
should not have any affect on performance.
cmr:v1.7
This commit was SVN r28292.
Roll in the ORTE state machine. Remove last traces of opal_sos. Remove UTK epoch code.
Please see the various emails about the state machine change for details. I'll send something out later with more info on the new arch.
This commit was SVN r26242.
Uses new CUDA IPC support. Also, a few minor changes in PML to take
advantage of it.
This code has no effect unless user asks for it explicitly via
configure arguments. Otherwise, it is either #ifdef'ed out or
not compiled.
This commit was SVN r26039.
No need for any CMRs to 1.5... that was already done in CMR 2728.
This commit was SVN r24545.
The following SVN revision numbers were found above:
r22841 --> open-mpi/ompi@b400b84162
step is the configure and Fortran mojo that Jeff will put in. Until then I
guess the Fortran interface is broken (at least all functions using the hidden
count firld in the MPI_Status).
This commit was SVN r23467.
(OMPI_ERR_* = OPAL_SOS_GET_ERR_CODE(ret)), since the return value could be a
SOS-encoded error. The OPAL_SOS_GET_ERR_CODE() takes in a SOS error and returns
back the native error code.
* Since OPAL_SUCCESS is preserved by SOS, also change all calls of the form
(OPAL_ERROR == ret) to (OPAL_SUCCESS != ret). We thus avoid having to
decode 'ret' to get the native error code.
This commit was SVN r23162.
OMPI
and a language agnostic part in OPAL. The convertor is completely
moved into OPAL. This offers several benefits as described in RFC
http://www.open-mpi.org/community/lists/devel/2009/07/6387.php
namely:
- Fewer basic types (int* and float* types, boolean and wchar
- Fixing naming scheme to ompi-nomenclature.
- Usability outside of the ompi-layer.
- Due to the fixed nature of simple opal types, their information is
completely
known at compile time and therefore constified
- With fewer datatypes (22), the actual sizes of bit-field types may be
reduced
from 64 to 32 bits, allowing reorganizing the opal_datatype
structure, eliminating holes and keeping data required in convertor
(upon send/recv) in one cacheline...
This has implications to the convertor-datastructure and other parts
of the code.
- Several performance tests have been run, the netpipe latency does not
change with
this patch on Linux/x86-64 on the smoky cluster.
- Extensive tests have been done to verify correctness (no new
regressions) using:
1. mpi_test_suite on linux/x86-64 using clean ompi-trunk and
ompi-ddt:
a. running both trunk and ompi-ddt resulted in no differences
(except for MPI_SHORT_INT and MPI_TYPE_MIX_LB_UB do now run
correctly).
b. with --enable-memchecker and running under valgrind (one buglet
when run with static found in test-suite, commited)
2. ibm testsuite on linux/x86-64 using clean ompi-trunk and ompi-ddt:
all passed (except for the dynamic/ tests failed!! as trunk/MTT)
3. compilation and usage of HDF5 tests on Jaguar using PGI and
PathScale compilers.
4. compilation and usage on Scicortex.
- Please note, that for the heterogeneous case, (-m32 compiled
binaries/ompi), neither
ompi-trunk, nor ompi-ddt branch would successfully launch.
This commit was SVN r21641.
got a whole lot smaller, decreasing the memory footprint of the
running application. How much it's a good question. Here is a
breakdown:
- in mca_bml_base_endpoint_t: 3 *size_t + 1 * uint32_t
- in mca_bml_base_btl_t: 1 * int + 1 * double - 1 * float
+ 6 * size_t + 9 * (void*)
The decrease in mca_bml_base_endpoint_t is for each peer and the
decrease in mca_bml_base_btl_t is for each BTL for each peer.
So, if we consider the most convenient case where there is only
one network between all peers, this decrease the memory foot print
per peer by
9*size_t + 9*(void*) + 2 * int32_t + 1 * double - 1 * float.
On a 64 bits machine this will be 156 bytes per peer.
Now we access all these fields directly from the underlying BTL
structure, and as this structure is common to multiple BML endpoint,
we are a lot more cache friendly. Even if this do not improve the
latency, it makes the SM performance graph a lot smoother.
This commit was SVN r19659.
There was an argument that was barely used, and on return at the PML
level it contained nothing usable. It has been removed, so now we're
using less memory ...
This commit was SVN r19657.
