hcc

hcc copied to clipboard

kernarg buffer allocation strategy improvement

1

- add the capability to allocate kernarg buffer in coarse grained GPU memory to leverage caching (off by default) - implemented a new algorithm to manage kernarg buffers * provides...

scchan

rocm-hcc

SWDEV-221755 - Re-implement deprecated HSAQueue::copy_ext

3

Revert "remove deprecated HSAQueue::copy_ext due to inf recursion" This reverts commit 5cd096ae42277d36d123c827fd8981de288f07a8 and 4ee361270f5938e34b0759c21f59cdcd0a99b021. This change causes a regression on GFX908 for BFloat16 tuning on MIOpen.

aaronenyeshi

Raven Ridge/APUs

16

I understand APUs are officially not supported, but as of _rocm_ 1.9/_kernel_ 4.19 the _runtime_, _thunk_, _opencl_ and _amdkfd_ components are now in place; the biggest missing piece is `hcc`...

lsr0

Enable the new PromoteConstant pass. Required by / depends on https://github.com/RadeonOpenCompute/llvm-project/pull/5 and https://github.com/ROCm-Developer-Tools/HIP/pull/1901.

AlexVlx

Do not use system ld in hc-kernel-assemble

1

It seems hc-kernel-assemble is using the system ld, instead of $BINDIR/lld: https://github.com/RadeonOpenCompute/hcc/blob/44f47d857e1df2b1042a1d8f3a2e3225eee3d908/lib/hc-kernel-assemble.in#L185. This will make it platform dependent (even if it works for the current ubuntu/centos, it doesn't work for...

xw285cornell

I've been compiling the hcc(default branch) with llvm(amd-stg-open branch) on arm64 platform ,ROCT-THUNK-INTERFACE ROCR-RUNTIME, ROCM-DEVICE-LIBS have been compiled and make install .but when got to the hcc ,I follow the...

0hanghang

Hello, I'm facing build errors in manjaro 5.5.7 with gcc & g++ 9.2. Here's a screenshot  Furthermore there is a repo implementing arch PKGBUILD for hcc, I tried with...

BrunoLiegiBastonLiegi

Address a deadlock issue in multi-GPU scenario

5

- Fixes SWDEV-219322

mangupta

Ubuntu 18.04, gcc 8.3 ,g++ 8.3 We build hcc 2.10 from source code. meet such error [100%] Building CXX object lib/CMakeFiles/mcwamp.dir/mcwamp.cpp.o [100%] Building CXX object lib/CMakeFiles/mcwamp.dir/mcwamp.cpp.o In file included from...

wormwang