The hiptt from dmitrylyakh

hiptt's People

Contributors

Stargazers

Watchers

hiptt's Issues

compile and test

I was just curious to build and run your library. I built the library with the following changes:

change 3.9.0 to 3.9.1
replace g++ with hipcc because there were errors using g++

g++ -o bin/cutt_test -L/opt/rocm-3.9.1/lib/ -lamdhip64 build/cutt_test.o build/TensorTester.o build/CudaMem.o build/CudaUtils.o build/cuttTimer.o -Llib -lcutt -fPIC
/usr/bin/ld: build/TensorTester.o: relocation R_X86_64_32 against symbol `_Z42__device_stub__setTensorCheckPatternKernelPjj' can not be used when making a PIE object; recompile with -fPIE
/usr/bin/ld: build/CudaUtils.o: relocation R_X86_64_32 against `.rodata.str1.1' can not be used when making a PIE object; recompile with -fPIE
/usr/bin/ld: lib/libcutt.a(cuttkernel.o): relocation R_X86_64_32S against `.rodata' can not be used when making a PIE object; recompile with -fPIE
/usr/bin/ld: lib/libcutt.a(cuttGpuModelKernel.o): relocation R_X86_64_32 against symbol `_Z32__device_stub__runCountersKernelPKiiiiPiS1_S1_' can not be used when making a PIE object; recompile with -fPIE

Running the test shows the following message

TensorTester::checkTranspose FAIL at 98 ref 130 data 40
Test 1 failed

For the benchmark, please suggest the options for meaningful results. Thanks.

./bin/cutt_bench
Using Vega 20 [Radeon VII] SM version 9.0
L2 0.00MB
CPU using vector type AVX2 of length 8
0.00 GB/s
0.00 GB/s
0.00 GB/s
0.00 GB/s
scalarCopy 0.000000 GB/s
0.00 GB/s
0.00 GB/s
0.00 GB/s
0.00 GB/s
vectorCopy 0.000000 GB/s
0.00 GB/s
0.00 GB/s
0.00 GB/s
0.00 GB/s
memcpyFloat 0.000000 GB/s
bench OK
seed 1606747073

Recommend Projects