changes for 0.6-2 from 0.6-1

README.md

@@ -1,11 +1,11 @@
-AOMP - V 0.6-1
+AOMP - V 0.6-2
 ==============
 
 AOMP:  AMD OpenMP Compiler
 
 This is README.md for https://github.com/ROCM-Developer-Tools/aomp .  This is the base repository for AOMP,  Use this for issues, documentation, packaging, examples, build.
 
-The last release of AOMP is version 0.6-1.  Currently version 0.6-2 is under development.
+The last release of AOMP is version 0.6-2.  Currently version 0.6-3 is under development.
 
 AOMP is an experimental PROTOTYPE that is intended to support multiple programming models including OpenMP 4.5+,
 , HIP, and cuda clang.  It supports offloading to multiple GPU acceleration targets(multi-target).  It also supports different host platforms such as AMD64, PPC64LE, and AARCH64. (multi-platform). 
@@ -74,24 +74,24 @@ Software License Agreement.
 
 AOMP needs certain support for Linux to function properly, such as the KFD driver for AMD GPUs and CUDA for nvptx. Click [LINUXSUPPORT](LINUXSUPPORT.md) for more information.
 
-## AOMP Install V 0.6-1
+## AOMP Install V 0.6-2
 
 <A NAME="Install">
 
 ### Debian/Ubunutu Install
 
 On Ubuntu 18.04 LTS (bionic beaver), run these commands:
 ```
-wget https://github.com/ROCm-Developer-Tools/aomp/releases/download/r/aomp_Ubuntu1804_0.6-1_amd64.deb
-sudo dpkg -i aomp_Ubuntu1804_0.6-1_amd64.deb
+wget https://github.com/ROCm-Developer-Tools/aomp/releases/download/r/aomp_Ubuntu1804_0.6-2_amd64.deb
+sudo dpkg -i aomp_Ubuntu1804_0.6-2_amd64.deb
 ```
 The AOMP bin directory (which includes the standard clang and llvm binaries) is not intended to be in your PATH for typical operation.
 
 <!--### RPM Install
 For rpm-based Linux distributions, use this rpm
 ```
-wget https://github.com/ROCm-Developer-Tools/aomp/releases/download/r/aomp-0.6-1.x86_64.rpm
-sudo rpm -i aomp-0.6-1.x86_64.rpm
+wget https://github.com/ROCm-Developer-Tools/aomp/releases/download/r/aomp-0.6-2.x86_64.rpm
+sudo rpm -i aomp-0.6-2.x86_64.rpm
 ```
 -->
 ### No root Debian Install
@@ -100,8 +100,8 @@ By default, the packages install their content to the release directory /opt/roc
 
 To install the debian package without root access into your home directory, you can run these commands.
 ```
-   wget https://github.com/ROCm-Developer-Tools/aomp/releases/download/r/aomp_Ubuntu1604_0.6-1_amd64.deb
-   dpkg -x aomp_Ubuntu1604_0.6-1_amd64.deb /tmp/temproot
+   wget https://github.com/ROCm-Developer-Tools/aomp/releases/download/r/aomp_Ubuntu1604_0.6-2_amd64.deb
+   dpkg -x aomp_Ubuntu1604_0.6-2_amd64.deb /tmp/temproot
    mv /tmp/temproot/opt/rocm $HOME
    export PATH=$PATH:$HOME/rocm/aomp/bin
    export AOMP=$HOME/rocm/aomp
@@ -115,8 +115,8 @@ By default, the packages install their content to the release directory /opt/roc
 To install the rpm package without root access into your home directory, you can run these commands.
 ```
    mkdir /tmp/temproot ; cd /tmp/temproot 
-   wget https://github.com/ROCm-Developer-Tools/aomp/releases/download/r/aomp-0.6-1.x86_64.rpm
-   rpm2cpio aomp-0.6-1.x86_64.rpm | cpio -idmv
+   wget https://github.com/ROCm-Developer-Tools/aomp/releases/download/r/aomp-0.6-2.x86_64.rpm
+   rpm2cpio aomp-0.6-2.x86_64.rpm | cpio -idmv
    mv /tmp/temproot/opt/rocm $HOME
    export PATH=$PATH:$HOME/rocm/aomp/bin
    export AOMP=$HOME/rocm/aomp

SOURCEINSTALL.md

@@ -1,4 +1,4 @@
-# Source Install V 0.6-2 (DEV)
+# Source Install V 0.6-3 (DEV)
 
 Build and install from sources is possible.  However, the source build for AOMP is complex for several reasons.
 - Many repos are required.  The clone_aomp.sh script ensures you have all repos and the correct branch.
@@ -40,9 +40,9 @@ To build AOMP from source, run these commands.
 ```
 Depending on your system, the last two commands and the CUDA install could take a very long time. For more information, please refer to the AOMP developers README file located [HERE](bin/README.md).
 
-The source build process above builds the development version of AOMP by checking out the master branch of AOMP.   The development version is the next version to be released.  It is possible that the development version is broken due to regressions that often occur during development.  If you want to build from the sources of a previous release such as 0.6-1, run these commands before running clone_aomp.sh.
+The source build process above builds the development version of AOMP by checking out the master branch of AOMP.   The development version is the next version to be released.  It is possible that the development version is broken due to regressions that often occur during development.  If you want to build from the sources of a previous release such as 0.6-2, run these commands before running clone_aomp.sh.
 ```
-   git checkout rel_0.6.1
+   git checkout rel_0.6-2
    git pull
 ```
 You only need to do this in the AOMP repository. The file "bin/aomp_common_vars" lists the branches of each repository for a particular AOMP release. In the master branch of AOMP, aomp_common_vars lists the development branches. It is a good idea to run clone_aomp.sh twice after you checkout a release to be sure you pulled all the checkouts for a particular release.

bin/debian/changelog

@@ -1,4 +1,4 @@
-aomp (0.6-1) UNRELEASED; urgency=medium
+aomp (0.6-2) UNRELEASED; urgency=medium
 
   * Initial release of aomp is 0.3-2
   *   Please see example in /opt/rocm/aomp/examples/vmuldemo 
@@ -143,4 +143,23 @@ aomp (0.6-1) UNRELEASED; urgency=medium
   *  Added two new smoke tests and improved automation of smoke tests
   *  Corrected mymcpu and mygpu for vega20
   *
- -- Greg Rodgers <[email protected]>  Mon, 15 Apr 2019 13:05:57 -0500
+  * 0.6-2
+  *   Fixed issue with constant size teams and threads.
+  *   Moved to the stable clang/llvm 8.0 code base
+  *   Fixed code in deviceRTLs/amdgcn that set Max_Warp_Number to 16, was 64
+  *   Enable Float16 for 0.6-2, disabled by default in release_80 merge
+  *   Disable META data opt, and provide evar AMDGPU_ENABLE_META_OPT_BUG to enable
+  *   Add archive handling for bc linking.
+  *   For performance, rewrite select_outline_wrapper calls, to be direct calls.
+  *    Example:  change the generated from:
+  *       @_HASHW_DeclareSharedMemory_cpp__omp_outlined___wrapper =
+  *          local_unnamed_addr addrspace(4) constant i64 -4874776124079246075
+  *     call void @select_outline_wrapper(i16 0, i32 %6, i64 -4874776124079246075)
+  *     to:
+  *        call void @DeclareSharedMemory_cpp__omp_outlined___wrapper(i16 0, i32 %6)
+  *  In release_80, Loop_tripcount API is now used, so we need to limit num_groups/teams
+  *     to no more than Max_Teams, fixes assertok_error, and snap4
+  *     Also handle num_teams clause inside loop_tripcount logic.
+  *  BALLOT_SYNC macro replaced with ACTIVEMASK in release_80
+
+ -- Greg Rodgers <[email protected]>  Tue, 30 Apr 2019 11:11:14 -0500