| /external/tensorflow/tensorflow/lite/g3doc/examples/build/ |
| D | index.md | 31 * **Limited compute capabilities** - Compared to fully equipped servers with
32 multiple CPUs, high memory capacity, and specialized processors such as GPUs
36 * **Size of models** - The overall complexity of a model, including data
37 pre-processing logic and the number of layers in the model, increases the
38 in-memory size of a model. A large model may run unacceptably slow or simply
40 * **Size of data** - The size of input data that can be effectively processed
44 off-device storage and access solutions.
45 * **Supported TensorFlow operations** - TensorFlow Lite runtime environments
51 For more information building effective, compatible, high performance models
53 [Performance best practices](../../performance/best_practices).
[all …]
|
| /external/jemalloc_new/ |
| D | TUNING.md | 1 This document summarizes the common approaches for performance fine tuning with
8 by a few percent, or make favorable trade-offs.
11 ## Notable runtime options for performance tuning
42 operating system, and therefore provides a fairly straightforward trade-off
47 Suggested: tune the values based on the desired trade-offs.
52 However high arena count may also increase overall memory fragmentation,
53 since arenas manage memory independently. When high degree of parallelism
71 * High resource consumption application, prioritizing CPU utilization:
77 * High resource consumption application, prioritizing memory usage:
90 * Extremely conservative -- minimize memory usage at all costs, only suitable when
[all …]
|
| /external/mesa3d/docs/drivers/openswr/ |
| D | faq.rst | 5 --------------------------------
10 * Architecture - given our focus on scientific visualization, our
16 * Historical - Intel had developed a high performance software
19 with Mesa to provide a high quality API layer while at the same
20 time benefiting from the excellent performance the software
24 ------------------------
26 SWR is a tile based immediate mode renderer with a sort-free threading
36 Our pipeline uses just-in-time compiled code for the fetch shader that
46 What's the performance?
47 -----------------------
[all …]
|
| /external/tensorflow/tensorflow/core/profiler/protobuf/ |
| D | overview_page.proto | 39 // Remark text in the performance summary section.
47 // Percentage of device computation that is 16-bit.
49 // Percentage of device computation that is 32-bit.
57 // Percentage of TF-op execution time on the host (excluding the idle time)
60 // Percentage of TF-op execution time on the device (excluding the idle time)
63 // Percentage of TF-op execution time on the device (excluding the idle time)
70 // Overview result for a performance tip to users.
77 // Indicates if kernel launch is a performance bottleneck. Possible values:
78 // "no", "moderate", "high".
80 // A statement that recommends if we need to further investigate kernel-launch
[all …]
|
| /external/tensorflow/tensorflow/compiler/mlir/g3doc/ |
| D | _index.yaml | 4 infrastructure for high-performance ML models in TensorFlow.
6 custom_css_path: /site-assets/css/style.css
8 - heading: MLIR unifies the infrastructure for high-performance ML models in TensorFlow.
10 - description: >
12 intermediate representation (IR) that unifies the infrastructure required to execute high
13 performance machine learning models in TensorFlow and similar ML frameworks. This project
19 - code_block: |
23 %x = call @thingToCall(%arg0) : (i32) -> i32
31 - classname: devsite-landing-row-cards
33 - heading: "Multi-Level Intermediate Representation for Compiler Infrastructure"
[all …]
|
| /external/walt/docs/ |
| D | AudioLatency.md | 9 [High Performance Audio website](http://googlesamples.github.io/android-audio-high-performance/).
21 …e](http://googlesamples.github.io/android-audio-high-performance/guides/audio-output-latency.html#…
40 | :--- | :--- | :--- | ---: | ---: |
49 …onization accuracy is about 1 ms hence the relative error for recording latency can be fairly high.
52 Superpowered Inc. maintains an open source app for measuring round trip audio latency -
|
| /external/autotest/client/cros/ |
| D | perf.py | 2 # Use of this source code is governed by a BSD-style license that can be
13 Provides methods for setting the performance mode of a device.
16 it into a consistent, high performance state during initialization.
26 # Do all performance testing.
71 For now we declare performance results as valid if
72 - we did not have an error before.
73 - the monitoring thread never saw temperatures a throttle
112 Warning: this risks abnormal behavior if machine runs in high load.
|
| /external/oboe/docs/ |
| D | FullGuide.md | 2 Oboe is a C++ library which makes it easy to build high-performance audio apps on Android. Apps com…
6 …ams*, represented by the class `AudioStream`. The read/write calls can be blocking or non-blocking.
19 (a built-in mic or bluetooth headset) with the *Android device* (the phone or watch) that is runnin…
33 
37 
53 | :------------ | :---------- | :---- |
54 | I16 | int16_t | common 16-bit samples, [Q0.15 format](https://source.android.com/devices/audio/da…
55 | Float | float | -1.0 to +1.0 |
59 AudioFormat dataFormat = stream->getDataFormat();
146 general, the API should be chosen by Oboe to allow for best performance and
[all …]
|
| /external/tensorflow/tensorflow/lite/g3doc/performance/ |
| D | best_practices.md | 1 # Performance best practices
6 Lite model performance.
11 and size. If your task requires high accuracy, then you may need a large and
17 ![Graph of model size vs accuracy](../images/performance/model_size_vs_accuracy.png "Model Size vs …
19 
24 [TensorFlow Hub](https://tfhub.dev/s?deployment-format=lite) lists several other
36 has a built-in profiler that shows per operator profiling statistics. This can
37 help in understanding performance bottlenecks and which operators dominate the
66 TensorFlow Lite supports multi-threaded kernels for many operators. You can
74 Multi-threaded execution, however, comes at the cost of increased performance
[all …]
|
| /external/perfetto/docs/ |
| D | tracing-101.md | 2 *This page provides a birds-eye view of performance analysis.
6 ### Performance subsection
7 Performance analysis is concerned with making software run *better*.
15 Much of the difficulty in improving performance comes from
16 identifying the root cause of performance issues. Modern software systems are
17 complicated, having a lot of components and a web of cross-interactions.
21 **Tracing** and **profiling** are two such widely-used techniques for
22 performance analysis. **Perfetto** is an open-source suite of tools, combining
31 They often include low-level kernel events like scheduler context switches,
33 performance bug is not needed as the trace provides all necessary context.
[all …]
|
| /external/icu/icu4j/perf-tests/src/com/ibm/icu/dev/test/perf/ |
| D | CollationPerformanceTest.java | 5 * Copyright (C) 2002-2007, International Business Machines Corporation and *
29 + "-help Display this message.\n"
30 + "-file file_name utf-16 format file of names.\n"
31 + "-locale name ICU locale to use. Default is en_US\n"
32 + "-rules file_name Collation rules file (overrides locale)\n"
33 …//+ "-langid 0x1234 Windows Language ID number. Default to value for -locale option\n"
35 //+ "-win Run test using Windows native services. (ICU is default)\n"
36 //+ "-unix Run test using Unix strxfrm, strcoll services.\n"
37 …//+ "-uselen Use API with string lengths. Default is null-terminated strings\n"
38 + "-usekeys Run tests using sortkeys rather than strcoll\n"
[all …]
|
| /external/cpuinfo/src/arm/ |
| D | uarch.c | 42 * Core information is ambiguous: some sources specify Cortex-A12, others - Cortex-A17. in cpuinfo_arm_decode_vendor_uarch()
43 * Assume it is Cortex-A12. in cpuinfo_arm_decode_vendor_uarch()
91 case 0xD0E: /* Cortex-A76AE */ in cpuinfo_arm_decode_vendor_uarch()
99 case 0xD41: /* Cortex-A78 */ in cpuinfo_arm_decode_vendor_uarch()
102 case 0xD44: /* Cortex-X1 */ in cpuinfo_arm_decode_vendor_uarch()
105 case 0xD46: /* Cortex-A510 */ in cpuinfo_arm_decode_vendor_uarch()
108 case 0xD47: /* Cortex-A710 */ in cpuinfo_arm_decode_vendor_uarch()
111 case 0xD48: /* Cortex-X2 */ in cpuinfo_arm_decode_vendor_uarch()
186 case 0xD40: /* Kirin 980 Big/Medium cores -> Cortex-A76 */ in cpuinfo_arm_decode_vendor_uarch()
243 /* Unlike Scorpion, Cortex-A5 comes with VFPv4 */ in cpuinfo_arm_decode_vendor_uarch()
[all …]
|
| /external/selinux/secilc/docs/ |
| D | cil_introduction.md | 4 …nguage that sits between one or more high level policy languages (such as the current module langu…
6 …high-level languages that can both consume and produce language constructs with more features than…
8 * Eases the creation of high-level languages, encouraging the creation of more domain specific poli…
10 … analysis of the output of multiple high-level languages using a single analysis tool set without …
13 ------------------
17 …- provide rich semantics needed for cross-language interaction but not for convenience. If a featu…
19 …- provide clear, simple syntax that is easy to parse and to generate by high-level compilers, anal…
21 …- the ultimate goal of CIL is the generation of the policy that will be enforced by the kernel. Th…
23 * The only good binary file format is a non-existent one - CIL is meant for a source policy oriente…
25 * Enable backwards compatibility but don't be a slave to it - source, but not binary, compatibility…
[all …]
|
| /external/rust/crates/glam/ |
| D | ARCHITECTURE.md | 3 This document describes the high-level architecture of `glam`. While `glam` is
11 * Good out of the box performance using SIMD when available
15 * High quality [rustdoc] generated document
17 [standard library]: https://doc.rust-lang.org/std/index.html
18 [API guidelines]: https://rust-lang.github.io/api-guidelines
19 [rustdoc]: https://doc.rust-lang.org/rustdoc/index.html
24 `x86_64` architectures gave better performance than using Rust's built in `f32`
29 …ing with SIMD]: https://bitshifter.github.io/2018/06/04/simd-path-tracing/#converting-vec3-to-sse2.
67 ## High level structure
|
| /external/perfetto/src/trace_processor/stdlib/common/ |
| D | cpus.sql | 1 --
2 -- Copyright 2023 The Android Open Source Project
3 --
4 -- Licensed under the Apache License, Version 2.0 (the "License");
5 -- you may not use this file except in compliance with the License.
6 -- You may obtain a copy of the License at
7 --
8 -- https://www.apache.org/licenses/LICENSE-2.0
9 --
10 -- Unless required by applicable law or agreed to in writing, software
[all …]
|
| /external/tensorflow/tensorflow/lite/delegates/gpu/ |
| D | README.md | 7 GPUs are designed to have high throughput for massively parallelizable
8 workloads. Thus, they are well-suited for deep neural nets which consists of a
12 run fast enough and now become suitable for real-time applications if it was not
15 GPUs do their computation with 16-bit or 32-bit floating point numbers and do
16 not require quantization for optimal performance unlike the CPUs. If
26 TFLite on GPU supports the following ops in 16-bit and 32-bit float precision:
32 * `DEPTHWISE_CONV_2D v1-2`
46 * `RESIZE_BILINEAR v1-3`
56 [the documentation](https://www.tensorflow.org/lite/performance/gpu).
81 WriteToInputTensor(interpreter->typed_input_tensor<float>(0));
[all …]
|
| /external/tensorflow/tensorflow/lite/g3doc/examples/segmentation/ |
| D | overview.md | 9 The model will create a mask over the target objects with high accuracy.
11 <img src="images/segmentation.gif" class="attempt-right" />
22 You can leverage the out-of-box API from
34 <a class="button button-primary" href="https://github.com/tensorflow/examples/tree/master/lite/exam…
37 <a class="button button-primary" href="https://github.com/tensorflow/examples/tree/master/lite/exam…
45 <a class="button button-primary" href="https://tfhub.dev/tensorflow/lite-model/deeplabv3/1/metadata…
50 _DeepLab_ is a state-of-art deep learning model for semantic image segmentation,
66 …ects at multiple scales with filters at multiple sampling rates and effective fields-of-views.</li>
67 …-level feature [5, 6] to capture longer range information. We also include batch normalization [7]…
68 …in this encoder-decoder structure one can arbitrarily control the resolution of extracted encoder …
[all …]
|
| /external/oss-fuzz/projects/opensips/ |
| D | patch.diff | 1 diff --git a/Makefile.conf.template b/Makefile.conf.template
3 --- a/Makefile.conf.template
5 @@ -69,17 +69,19 @@ exclude_modules?= aaa_diameter aaa_radius auth_jwt b2b_logic_xml cachedb_cassand
9 -DEFS+= -DPKG_MALLOC #Use a faster malloc
10 -DEFS+= -DSHM_MMAP #Use mmap instead of SYSV shared memory
11 -DEFS+= -DUSE_MCAST #Compile in support for IP Multicast
12 +#DEFS+= -DPKG_MALLOC #Use a faster malloc
13 +#DEFS+= -DSHM_MMAP #Use mmap instead of SYSV shared memory
14 +#DEFS+= -DUSE_MCAST #Compile in support for IP Multicast
15 DEFS+= -DDISABLE_NAGLE #Disable the TCP NAgle Algorithm ( lower delay )
[all …]
|
| /external/rust/crates/libz-sys/ |
| D | Cargo.toml.orig | 2 name = "libz-sys"
6 license = "MIT OR Apache-2.0"
7 repository = "https://github.com/rust-lang/libz-sys"
8 description = "Low-level bindings to the system libz library (also known as zlib)."
9 categories = ["compression", "external-ffi-bindings"]
10 keywords = ["zlib", "zlib-ng"]
16 "/Cargo-zng.toml",
17 "/cargo-zng",
28 # eliminating some high-level functions like gz*, compress* and
33 [build-dependencies]
[all …]
|
| /external/skia/src/core/ |
| D | SkSharedMutex.h | 4 * Use of this source code is governed by a BSD-style license that can be
22 // There are two shared lock implementations one debug the other is high performance. They implement
24 // This is a shared lock implementation similar to pthreads rwlocks. The high performance
26 // http://preshing.com/20150316/semaphores-are-surprisingly-versatile/
|
| /external/fec/ |
| D | simd-viterbi.3 | 1 .TH SIMD-VITERBI 3
10 chainback_viterbi615, delete_viterbi615 -\ IA32 SIMD-assisted Viterbi decoders
53 These functions implement high performance Viterbi decoders for four
78 non-IA32, non-PPC machine, a portable C version is executed
115 decoder instance, a pointer to a user-supplied buffer into which the
120 may be unreliable. The decoded data is written in big-endian order,
121 i.e., the first bit in the frame is written into the high order bit of
140 with a total of 2012 symbols - the last 12 encoded symbols
154 are those of the NASA-JPL convention \fIwithout\fR symbol inversion.
155 The NASA-JPL convention normally inverts the first symbol.
[all …]
|
| /external/rust/crates/aho-corasick/src/packed/teddy/ |
| D | runtime.rs | 8 // variants: Slim 128-bit Teddy (8 buckets), Slim 256-bit Teddy (8 buckets)
9 // and Fat 256-bit Teddy (16 buckets). For each variant, there are three
13 // while at <= len(haystack) - CHUNK_SIZE:
23 // In the code below, a "candidate" corresponds to a single vector with 8-bit
24 // lanes. Each lane is itself an 8-bit bitset, where the ith bit is set in the
31 // 4-byte vector would look like this:
63 /// like Aho-Corasick, it does find occurrences for a small set of patterns
64 /// much more quickly than Aho-Corasick.
86 /// Note that users of the aho-corasick crate cannot get this wrong. Only
101 /// All matches are consistent with the match semantics (leftmost-first or
[all …]
|
| /external/cronet/base/allocator/partition_allocator/partition_alloc_base/time/ |
| D | time_win.cc | 2 // Use of this source code is governed by a BSD-style license that can be
13 // QueryPerformanceCounter is the logical choice for a high-precision timer.
16 // It's 3-4x slower than timeGetTime() on desktops, but can be 10x slower
30 // will only increase the system-wide timer if we're not running on battery
56 // From MSDN, FILETIME "Contains a 64-bit value representing the number of
57 // 100-nanosecond intervals since January 1, 1601 (UTC)."
60 // 100-nanoseconds to microseconds. This only works on little-endian in FileTimeToMicroseconds()
70 PA_DCHECK(CanConvertToFileTime(us)) << "Out-of-range: Cannot convert " << us in MicrosecondsToFileTime()
73 // Multiply by 10 to convert microseconds to 100-nanoseconds. Bit_cast will in MicrosecondsToFileTime()
74 // handle alignment problems. This only works on little-endian machines. in MicrosecondsToFileTime()
[all …]
|
| /external/python/cpython2/Doc/library/ |
| D | hotshot.rst | 2 :mod:`hotshot` --- High performance logging profiler
6 :synopsis: High performance logging profiler, mostly written in C.
15 in C, it should result in a much smaller performance impact than the existing
21 the expense of long data post-processing times. For common usage it is
46 .. _hotshot-objects:
49 ---------------
71 Profile an :keyword:`exec`\ -compatible string in the script environment. The
86 Evaluate an :keyword:`exec`\ -compatible string in a specific environment. The
101 ------------------
125 .. _hotshot-example:
[all …]
|
| /external/bcc/man/man8/ |
| D | readahead.8 | 1 .TH readahead 8 "2020-08-20" "USER COMMANDS"
3 readahead \- Show performance of read-ahead cache
5 .B readahead [-d DURATION]
7 The tool shows the performance of read-ahead caching on the system under a given load to investigat…
24 \-h
27 \-d DURATION
28 Trace the read-ahead caching system for DURATION seconds
31 Trace for 30 seconds and show histogram of page age (ms) in read-ahead cache along with unused pag…
33 .B readahead -d 30
35 The kernel functions instrumented by this program could be high-frequency depending on the profile …
[all …]
|