hcc

hcc copied to clipboard

Copy data from host to hc::array_view using hc::copy fails without calling synchronize

1

I would expect that this will work but it causes segmentation fault: ``` c++ int arr_host[100]; hc::array_view arr(100); hc::copy(arr_host, arr); ``` Seems I need to call synchronize before copy: ```...

misos1

commit 9a0c2a5f839e6c9c7025687da8594419fcb7a203 (HEAD -> clang_tot_upgrade, origin/clang_tot_upgrade, origin/HEAD) We are trying to port the ROCm userspace stack to Linux s390x. ROCR-Runtime and ROCT-Thunk-Interface build and install fine. But building hcc with...

gongsu832

HCC compiler internal fault

2

When compiling latest master, an fatal error was triggered in the compiler. This is reproducible across build attempts. ``` fatal error: error in backend: Cannot select: 0x151547b78: f64,ch = load...

madscientist159

False positive dependency

2

Here only second kernel uses "arr" array_view and so there is no dependency between kernels but they are executed serially. Second kernel is waiting until is first finished. But as...

misos1

I need help in understanding why these two statements produce different code. Original: ```cpp unsigned int key[1] = { i < n ? keys[i] : 123456789 }; // INCORRECT ```...

ex-rzr

For example these: ``` c++ extern "C" unsigned int __bitalign_b32(unsigned int src0, unsigned int src1, unsigned int src2) __HC__; extern "C" unsigned int __bytealign_b32(unsigned int src0, unsigned int src1, unsigned...

misos1

Below is reduced test case from more complex code for simplicity. For me it started crashing from N=514 but it is possible that it could do for higher values on...

misos1

Question: Accessing std lib in GPU lambdas

2

Is it possible to access std library functions within hcc lambdas? Use case right now would be ```std::mt19937_64 generator;``` and the like, or is there a "HCC" way to do...

PhilipDeegan

**Issue** Accuracy issue related to nan loss is observed on executing the mxnet training application on gpu where as on cpu we get the approriate output. For example: **Training on...

sriharikarnam

Compiler crash when using std::exp in kernel and compiling with -Ofast flag

4

Yes I know that I should use exp from fast_math or precise_math but compiler should not crash but either print some errors or successfully compile if std::exp is intended to...

misos1