-
Notifications
You must be signed in to change notification settings - Fork 136
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Basic build for gfx1010 #926
Comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
I'm trying to build this library so that I can link it to rocBLAS and have a functioning gemm implementation in HIP (similarly to this PR) for gfx1010/gfx1012.
I saw a commit titled
Kernels now working on gfx1010.
and I'm wondering what set of environment variables and compilers is needed to accomplish this.I'm using Linux Mint 19.3 with ROCm 3.3.0 on a RX 5500 XT (gfx1012) and RX 5700 XT (gfx1010). Host processor is Ryzen 9 3900X.
I first tried
python3 ../Tensile/bin/Tensile ../Tensile/Configs/rocblas_sgemm_asm_only.yaml ./
Tensile_default_output.TXT.zip
Most notably, I don't think that HCC supports gfx1010 and up, so I see a lot of compiler errors that look like
Additionally, the generated
run.sh
script tries to set the graphics card clock to an invalid value.I also tried setting
--cxx-compiler
tohipcc
with$HIP_PLATFORM
set toclang
but CMake-generated flags include-hc
which isn't recognized by the compiler. I could get a little further by editing TensileCreateLibrary.py to pass in-D__HIP_VDI__
, manually editing the generatedflags.make
in the build files, and setting the benchmark config to not build a new client, but I inevitably run into one of the above issues.Is there a recommended way to directly build a basic gemm kernel in pure HIP that can be used by rocBLAS without using the benchmarking driver program?
The text was updated successfully, but these errors were encountered: