aboutsummaryrefslogtreecommitdiff
diff options
context:
space:
mode:
authorGiorgio Arena <giorgio.arena@arm.com>2021-02-22 14:58:37 +0000
committerMichele Di Giorgio <michele.digiorgio@arm.com>2021-02-22 17:30:33 +0000
commitcd7d1781915451793b2b6a007948b81476b9e3e3 (patch)
tree84c605f3b76a6b9f7b53a1cc668513e7ef048dc4
parent4ed1adcc6cc40237da5e333f6e8400e6ebc9017a (diff)
downloadComputeLibrary-cd7d1781915451793b2b6a007948b81476b9e3e3.tar.gz
Update release notes with performance regression on G77 and updated toolchain
Signed-off-by: Giorgio Arena <giorgio.arena@arm.com> Change-Id: I546bddeab320a1d19a1fe53b2a64ba258fb7b36d Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5145 Tested-by: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Michele Di Giorgio <michele.digiorgio@arm.com>
-rw-r--r--docs/00_introduction.dox13
1 files changed, 7 insertions, 6 deletions
diff --git a/docs/00_introduction.dox b/docs/00_introduction.dox
index dc36c0b4ed..5c6b202139 100644
--- a/docs/00_introduction.dox
+++ b/docs/00_introduction.dox
@@ -49,10 +49,10 @@ In order to facilitate the work of the support team please provide the build inf
For each release we provide some pre-built binaries of the library [here](https://github.com/ARM-software/ComputeLibrary/releases)
These binaries have been built using the following toolchains:
- - Linux armv7a: gcc-linaro-6.3.1-2017.05-x86_64_arm-linux-gnueabihf
- - Linux arm64-v8a: gcc-linaro-6.3.1-2017.05-x86_64_aarch64-linux-gnu
+ - Linux armv7a: gcc-linaro-7.2.1-2017.11-x86_64_arm-linux-gnueabihf
+ - Linux arm64-v8a: gcc-linaro-7.2.1-2017.11-x86_64_aarch64-linux-gnu
- Android armv7a: clang++ / libc++ NDK r18b
- - Android am64-v8a: clang++ / libc++ NDK r18b
+ - Android am64-v8a: clang++ / libc++ NDK r20b
@warning Make sure to use a compatible toolchain to build your application or you will get some std::bad_alloc errors at runtime.
@@ -132,6 +132,7 @@ v21.02 Public major release
- Experimental extension for the OpenCL tuner to tune the batches of work groups distribute to compute units
- Add functionality to load the OpenCL GEMM heuristics at runtime
- The GEMM heuristic file (MLGO) can be used to update the default GEMM heuristics available for OpenCL
+ - Note: there might be performance regressions against v20.08 in Inception v3 using int8 data types on Arm Mali-G77 GPUs. Currently under investigation
v20.11 Public major release
- Various bug fixes.
@@ -1706,10 +1707,10 @@ An example build command with SVE is:
For Android, the library was successfully built and tested using Google's standalone toolchains:
- clang++ from NDK r18b for armv7a
- - clang++ from NDK r18b for arm64-v8a
- - clang++ from NDK r18b for arm64-v8.2-a with FP16 support
+ - clang++ from NDK r20b for arm64-v8a
+ - clang++ from NDK r20b for arm64-v8.2-a with FP16 support
-Here is a guide to <a href="https://developer.android.com/ndk/guides/standalone_toolchain.html">create your Android standalone toolchains from the NDK</a>
+Here is a guide to <a href="https://developer.android.com/ndk/guides/standalone_toolchain.html">create your Android standalone toolchains from the NDK</a>. Minimum NDK version required: r18b
- Download the NDK r18b from here: https://developer.android.com/ndk/downloads/index.html to directory $NDK
- Make sure you have Python 2.7 installed on your machine.