1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
|
///
/// Copyright (c) 2019-2023 Arm Limited.
///
/// SPDX-License-Identifier: MIT
///
/// Permission is hereby granted, free of charge, to any person obtaining a copy
/// of this software and associated documentation files (the "Software"), to
/// deal in the Software without restriction, including without limitation the
/// rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
/// sell copies of the Software, and to permit persons to whom the Software is
/// furnished to do so, subject to the following conditions:
///
/// The above copyright notice and this permission notice shall be included in all
/// copies or substantial portions of the Software.
///
/// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
/// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
/// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
/// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
/// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
/// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
/// SOFTWARE.
///
namespace arm_compute
{
/**
@page errata Errata
@tableofcontents
@section S7_1_errata Errata
- (COMPMID-6493) Crash when running Arm Compute Library compiled for SVE2 on a computer that support SVE only.
- Versions: >= v21.02 && <=v23.08
- OSs: Linux, Android.
- Conditions:
- Compile the latest Arm Compute Library for SVE2 (arch=armv8.6-a-sve2).
- multi_isa = 0
- Device with SVE but without SVE2 support.
- Result:
- Crash due to illegal instruction.
- To run SVE only, build with arch="armv8.2-a-sve", arch="armv8.6-a-sve", or with multi_isa=1.
- (COMPMID-6404) Under certain conditions, CLTile may produce incorrect result.
- Versions: >= v19.02 && < v23.08
- OSs: Linux, Android.
- Conditions:
- The size of the lowest dimension of the input tensor is greater than 16 bytes.
- The size of the lowest dimension of the input tensor is not a multiple of 16.
- Result:
- Incorrect result is produced.
- (COMPMID-6271) Under certain conditions, CLArgMinMaxLayer validation tests may fail
- Versions Affected: >= v20.02 && < v23.08
- OSs Affected: Linux
- Conditions:
- Backend: OpenCL
- Axis == 0
- Result:
- Sporadic mismatches only on certain devices
- (COMPMID-5324) Issue identified with direct and depthwise convolutions for certain Arm® Mali™ DDK versions.
- Versions Affected: < v22.08
- Conditions:
- Arm® Mali™ DDK Versions : >= r23p0 && <= r38p0
- Mali™ GPUs: Bifrost GPU family with the exception of G71
- Backend: OpenCL
- Build options Include : "cl-fast-relaxed-math"
- Result: Reduced accuracy issue, while using direct and depthwise convolutions fused with LU_BOUNDED_RELU activation.
- (COMPMID-5134) An issue has been identified when running the graph_deepspeech_v0_4_1 graph example.
- Versions Affected: >= v21.08
- Conditions:
- Data type input: F32
- Backend: OpenCL
- Result: The execution of the graph_deepspeech_v0_4_1 could fail on OpenCL backend for systems with a small RAM. The issue is due to the extra temporary memory required to reshape the network weights
- (COMPMID-4013) Experimented performance regressions for some networks on OpenCL when using Arm® Mali™ DDK r8p0
- Versions Affected: v21.05
- OSs Affected: All
- Conditions:
- Arm® Mali™ DDK r8p0
- (COMPMID-5146) Under certain conditions, CLFullyConnectedLayer quantized tests may fail due to an issue in the test framework.
- Versions Affected: v21.02
- OSs Affected: Linux
- Conditions:
- armv7a architecture
- release mode
- asserts enabled
- (COMPMID-4367) Performance regression in Convolution Layer OpenCL backend on Mali™ G77 when QSYMM8_PER_CHANNEL is used as weights' data type.
- Versions Affected: >= v20.11 && < v21.08
- OSs Affected: All
- Conditions:
- Mali™ G77
- Convolution Layer in use
- OpenCL backend
- Convolution Layer uses QSYMM8_PER_CHANNEL as the data type of its weight
- (COMPMID-4306) A wrong test configuration has been found in CLGEMMMatrixMultiplyReshapedOnlyRHS set of tests.
- Versions Affected: >= v20.11 && < v21.05
- Conditions:
- Data type input: F32/F16
- Fused bounded relu activation with coefficient 'a' being negative
- (COMPMID-5135) Under certain conditions, the validation test case 'CL/DirectConvolutionLayer/Float/FP32/RunSmall9x9\@InputShape=32x37x3x4:StrideX=1:StrideY=1:PadX=0:PadY=0:KernelSize=9:NumKernels=1:DataType=F32:ActivationInfo=LU_BOUNDED_RELU:DataLayout=NHWC' may fail.
- Versions Affected: >= v20.08
- Conditions:
- The validation suite has to run in nightly mode and execute 40k+ test cases before the test mentioned above
- (COMPMID-5136) Under certain conditions, benchmark examples can hang when OpenCL profiling queues are enabled.
- Versions Affected: >= v19.11
- OSs Affected: Linux
- Conditions:
- Arm® Mali™ DDK r1p0 - r8p0, and
- Linux kernel >= 4.4
- (COMPMID-5137) On Android with armv8a/armv8.2-a architecture, Arm® Neon™ validation tests can fail when compiled using Android Ndk
>= r18b in debug mode (https://github.com/android/ndk/issues/1135).
- Versions Affected: >= v19.11
- OSs Affected: Android
- Conditions:
- armv8a/armv8.2-a architecture, and
- Compiled using Android NDK >= r18b in debug mode.
- (COMPMID-4288) An issue has been identified with CLCast.
- Versions Affected: >= v18.11 && < v21.05
- Conditions:
- Data type input: F32
- Data type output: All integer types
- Conversion policy: SATURATE
- Result: OpenCL backend will always wrap around instead of saturating for out-of-range inputs
*/
} // namespace
|
