| Commit message (Collapse) | Author | Age |
|
|
|
| |
PiperOrigin-RevId: 209679086
|
|
|
|
| |
PiperOrigin-RevId: 196531355
|
|
|
|
|
|
|
|
|
|
| |
This isn't necessary today, but it will be after an optimization change I'm
about to make.
LLVM has a constant merging pass too, but one of the motivations here is to
avoid the LLVM compile time overhead of having many large arrays in the IR.
PiperOrigin-RevId: 195032900
|
|
|
|
| |
PiperOrigin-RevId: 189224522
|
|
|
|
| |
PiperOrigin-RevId: 189110935
|
|
|
|
|
|
| |
a hash<StringPiece>).
PiperOrigin-RevId: 189060958
|
|
|
|
|
|
|
| |
hash.h and bfloat16.h. This change introduces a generic mechanism for adapting
types that are meant to be used in tensorflow's error objects.
PiperOrigin-RevId: 188920678
|
|
|
|
|
|
|
|
|
|
| |
chain as
we don't have to link protobuf libraries just to use basic hash functions. The case
for deterministic serialization of protocol buffers if very specific and can be
handled by a new header file.
PiperOrigin-RevId: 188366713
|
|
|
|
| |
PiperOrigin-RevId: 182124532
|
|
|
|
|
|
| |
It is a useful function for computing hash values.
PiperOrigin-RevId: 170013135
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
improvements.
Other improvements:
- Don't show tooltips for nodes and clusters. Previously we'd show a
tooltip containing a pointer value expressed as decimal. Not so
useful.
- Show tooltips on edges with the to/from node names.
- Fix bug wherein if we had
- a node at the "edge" of the graph (so its operands aren't included
unless they're referenced by another node),
- with all of its operands included in the graph save one or more
constants, and
- those constants weren't referenced by any nodes not at the edge of
the graph,
we would incorrectly draw the node as "grayed out", indicating that
one of its operands (namely, its constant operand) wasn't present in
the graph.
This is wrong because constants are inlined into their users, so they
should always count as "displayed" for the purposes of determining
whether a node is grayed out.
PiperOrigin-RevId: 163276108
|
|
|
|
|
|
|
| |
Now that we have tensorflow::hash and it's the default hasher for
FlatMap and FlatSet, we can get rid of HashStr and HashStringPiece.
PiperOrigin-RevId: 161560034
|
|
|
|
|
|
|
|
|
|
|
|
| |
Previously the default hasher for FlatMap and FlatSet used std::hash by
default. Most stdlibs' std::hash for pointers is the identity function.
This works for std::unordered_{set,map} because those are prime-sized
hashtables. But FlatSet and FlatMap are power-of-two-sized, and so the
identity function is bad, bad news.
This change also switches us to use Hash64 for strings, rather than
std::hash.
PiperOrigin-RevId: 161553663
|
|
|
|
| |
Change: 147897309
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Also added a benchmark. New code is ~6x the throughput for lengths >= 16;
Run on ... (12 X 3501 MHz CPUs); 2017-01-23T13:53:34.121700508-08:00
CPU: Intel Ivybridge with HyperThreading (6 cores) dL1:32KB dL2:256KB dL3:12MB
Benchmark Time(ns) CPU(ns) Iterations
-----------------------------------------------------------
Old:
BM_CRC/1_mean 4.53 4.53 1529955670 210.559MB/s
BM_CRC/8_mean 7.67 7.67 923461420 995.407MB/s
BM_CRC/64_mean 49.3 49.2 142465300 1.211GB/s
BM_CRC/512_mean 380 380 18472980 1.255GB/s
BM_CRC/4k_mean 3070 3067 2306140 1.244GB/s
BM_CRC/32k_mean 24398 24373 283310 1.252GB/s
BM_CRC/256k_mean 197142 196936 36060 1.240GB/s
New:
BM_CRC/1_mean 4.64 4.63 1441767130 205.946MB/s
BM_CRC/8_mean 8.36 8.35 837511140 913.375MB/s
BM_CRC/64_mean 7.01 7.00 972688290 8.512GB/s
BM_CRC/512_mean 51.8 51.7 100000000 9.223GB/s
BM_CRC/4k_mean 416 416 16838270 9.179GB/s
BM_CRC/32k_mean 3275 3272 2093370 9.327GB/s
BM_CRC/256k_mean 26205 26179 267630 9.326GB/s
name old time/op new time/op delta
BM_CRC/1 4.53ns +- 1% 4.63ns +- 3% +2.30% (p=0.000 n=10+10)
BM_CRC/8 7.64ns +- 1% 8.35ns +- 1% +9.32% (p=0.000 n=9+10)
BM_CRC/64 49.2ns +- 0% 7.0ns +- 0% -85.77% (p=0.000 n=6+10)
BM_CRC/512 380ns +- 0% 52ns +- 1% -86.39% (p=0.000 n=9+10)
BM_CRC/4k 3.07us +- 1% 0.42us +- 1% -86.45% (p=0.000 n=10+10)
BM_CRC/32k 24.4us +- 2% 3.3us +- 1% -86.58% (p=0.000 n=10+10)
BM_CRC/256k 197us +- 3% 26us +- 0% -86.71% (p=0.000 n=10+10)
Change: 146051723
|
|
|
|
|
|
|
|
|
|
| |
std::unordered_map and std::unordered_set.
Add default template argument of std::hash<Key> to gtl::FlatMap and gtl::FlatSet
to better match std::unordered_{map,set}
Improves performance on an RPC-intensive benchmark by ~0.4%
Change: 137754417
|
|
|
|
|
|
|
|
|
|
|
|
| |
Mostly drop-in replacements for std::unordered_map and std::unordered_set,
but much faster (does not do an allocation per entry, and represents
entries in groups of 8 in a flat array, which is much more cache efficient).
Benchmarks not included in this cl show about 3X to 5X performance
improvements over the std::unordered_{set,map} for many kinds of
common maps e.g. std::unordered_mapmap<int64, int64> or
std::unordered_map<string, int64>.
Change: 137401863
|
|
|
|
|
| |
the C++ API's Output/Operation object instead of strings.
Change: 130776638
|
|
|
|
| |
Change: 123900938
|
|
|
|
|
| |
tensorflow/core/ files and build targets.
Change: 113078322
|
|
|
|
| |
Change: 112523833
|
|
|
|
| |
Change: 112481326
|
|
|
|
|
| |
directly so we can drop it from port.h.
Change: 111506675
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Changes:
* error message that refers to removed `DefaultSession` method.
* -Wnull-conversion warnings
* the "_start_time" attr for recvs when the flag "--brain_enable_scheduling_for_recvs" is set.
* typo in tutorial data download progress message.
* a typo ("however their installing"=>"however installing").
* typo, rename "TensorFlow Mechanics" to "How To" to be consistent with the website.
* a typo ("subtact"=>"subtract").
* protobuf examples in comments in tensorflow::Example.proto.
* formula formatting in MNIST beginner tutorial
* negative fraction-of-queue-full stats
* protobuf inclusion path so that Android demo will build under Blaze.
* small typo (moderatly > moderately)
* Session.run() to check that tensor arguments come from the session's graph.
* another six import
* seq2seq typo in bazel command
Base CL: 108349164
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
error handling, updates to website.
Changes:
- Removes redundant reshape from image models by @mrry
- Default TensorBoard to localhost by @danmane
- Reformatting of tensorflow/core by @josh11b
- Make tutorials backwards compatible to 0.5.0 by @girving
- Improve print documentation (md files not updated).
- Add proper scrolling to sitemap by @martinwicke
Base CL: 107956254
|
|
TensorFlow is an open source software library for numerical computation
using data flow graphs.
Base CL: 107276108
|