| Commit message (Collapse) | Author | Age |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
While overriding it from the default of 1 won't be great for performance, it can
help writing tests.
The CL is organized as follows:
* xla/legacy_flags/debug_options_flags and xla/xla.proto now has a
--xla_force_host_platform_device_count flag which defaults to 1.
* xla/service/platform_util.cc respects this --xla_force_host_platform_device_count
flag.
* xla/service/cpu/... has some changes to generalize infeed and outfeed on CPU
to work with multiple devices.
PiperOrigin-RevId: 214078482
|
|
|
|
|
|
|
|
| |
Previously it was possible that a distinct thread could hop in between the
buffer enqueues done by a tuple-outfeeding thread. This changes the sequence to
enqueue all the tuple buffers as an atomic unit.
PiperOrigin-RevId: 163781804
|
|
|
|
|
|
|
|
|
|
|
| |
Previously the CPU runtime wouldn't know how the data that was being outfed was
laid out by the XLA LayoutAssignment pass, which could result in
transposed-value results.
This also allows us to validate the contract between the host program and the
compiled XLA program with (reified) runtime type checks.
PiperOrigin-RevId: 161895093
|
|
Note: does not yet support nested tuples, for symmetry with the current infeed
limitations.
PiperOrigin-RevId: 161502502
|