OpenGrok
Home
Sort by relevance
Sort by last modified time
Full Search
Definition
Symbol
File Path
History
|
|
Help
Searched
refs:request_index
(Results
1 - 5
of
5
) sorted by null
/external/v8/src/builtins/
builtins-dataview.cc
163
Handle<Object>
request_index
,
166
isolate,
request_index
,
167
Object::ToIndex(isolate,
request_index
,
171
if (!TryNumberToSize(*
request_index
, &get_index)) {
249
Handle<Object>
request_index
,
252
isolate,
request_index
,
253
Object::ToIndex(isolate,
request_index
,
258
if (!TryNumberToSize(*
request_index
, &get_index)) {
/external/tensorflow/tensorflow/contrib/verbs/
rdma.cc
474
uint32_t
request_index
= imm_data;
475
RdmaTensorRequest* request = rc->GetTensorRequest(
request_index
);
673
uint32_t
request_index
= request_serial_++;
677
RdmaTensorRequest request(
request_index
, key, step_id, this, dst_dev,
679
auto it = request_table_.emplace(
request_index
, request);
683
void RdmaChannel::RemoveTensorRequest(uint32_t
request_index
) {
685
request_table_.erase(
request_index
);
688
RdmaTensorRequest* RdmaChannel::GetTensorRequest(uint32_t
request_index
) {
690
RequestTable::iterator iter = request_table_.find(
request_index
);
934
const Tensor& in, uint32_t
request_index
,
[
all
...]
rdma.h
107
// type|name_size|name|step_id|
request_index
|remote_addr/checksum|rkey|...
435
void RemoveTensorRequest(uint32_t
request_index
);
436
RdmaTensorRequest* GetTensorRequest(uint32_t
request_index
);
441
void RemoveTensorResponse(uint32_t
request_index
);
patch_notes_verbs_with_0_copies.md
28
3. When the sender writes the tensor content to the result tensor, no additional data is being written with it. That means we need to reside on ibverbs immediate (uint32_t) to indicate which request we are responding to (in order to trigger the receive callback). The easiest and most elegant way is to key the recv callback with a unique
request_index
(uint32_t), instead of the current key_with_step_id (string).
48
The protocol messages themselves will remain mostly unchanged at the first stage, but will be used differently, as described below. The current messages structures already have most of the required fields for the new implementation. The only change is the "buffer_size" field which is no longer used since we are no longer sending additional information with the tensor, and thus it is now always equal to the "tensor_bytes" field. Instead, we use that field to pass the "
request_index
".
52
| type | name_size | name | step_id |
request_index
| remote_addr | rkey | is_dead | data_type | tensor_shape | tensor_bytes |
60
*
request_index
- Request index.
65
*
request_index
- Request index.
71
*
request_index
- Request index.
README.md
140
| type | name_size | name | step_id |
request_index
| remote_addr/checksum | rkey | is_dead | data_type | tensor_shape | tensor_bytes | error_status |
148
*
request_index
- Request index.
153
*
request_index
- Request index.
160
*
request_index
- Request index.
166
*
request_index
- Request index.
Completed in 62 milliseconds