| /// | |
| /// Copyright (c) 2019-2022 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-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 |