This change is required by AMDMIGraphX.
It was for HCC only. HIP-Clang also needs it for __fp16 since AMDMIGraphX uses it.
Change-Id: Id49322b7b89ef799accdf6b47627a6fce51d1ab5
These might contain garbage causing the runtime to incorrectly parse the state of the texture references.
Change-Id: I93c726fa30b580b3e14c50ac939f3c71b0d1c8d9
* Disable device side malloc
Currently device side malloc is not working and takes excessive
device memory.
Disable it for now until a working malloc is implemented.
Change-Id: I1ad908c1c53a83752383b4be96688a848642c699
This is charrypick of 9ead991784
and https://github.com/ROCm-Developer-Tools/HIP/pull/2009
Fix cmake config file
Removed cmake target files under packaging directory.
Merged cmake config .in files for HIP-Clang and HCC as one.
Use cmake generated target files in both install and packaging.
This makes cmake config file consistent for make install and
make package.
Let device side malloc/free return nullptr and trap
Change-Id: I448f3ea2d4934648089bad371debc203f895cba6
Support hipLaunchCooperativeKernelMultiDevice()
- Add validation logic for MGPU launches to pass a cuda test
Change-Id: Iccca7fde43493fc3bc6685512d39202271ae3e92
If the code object is embedded in an already mapped file, and the
lifetime of the mapped file exceeds the lifetime of the executable,
we do not need to make a copy of the binary.
This allows the ROCR to present the code object URI as
file:///path/to/file#offset=X&size=Y.
libc++ defines fma as template function for auto promotion of mixed-type
arguments. libc++ does not handle _Float16 as _Float16 is not a supported
type by C++ standard. As such, it is unlikely we can commit our fix for
_Float16 to libc++ trunk.
Therefore we handle _Float16 with a template specialization of
__numeric_type in HIP headers.
Change-Id: If01960a657ebf1a7a67463cdcf66fab7458dff3c
Even though the runtime and driver texture object API is one to one, the structs used by these APIs are not. See hipResourceDesc vs HIP_RESOURCE_DESC differences.
These differences are not trivial and most likely won't be able to handled by hipify, so we need new API entry points.
Change-Id: Id4bcb1ad0ae15378dbdb5a2ed07e5ea30f320082
- Use symbol value as the qeury key. Compared to the symbol name, the
symbol value is more robust as developers may use unqualified or
qualified identifiers. It also removes the mangling and/or demangling
requirement for the runtime API.
Change-Id: I9d4259f3842612c7cc98551269fc2092d8b5c19e
This is cherry-picked from PR#1947 that was committed to the
github repo. It allows printf to work with hip-clang and HCC
runtime.
Change-Id: I754753250ea1e694cf3441722e2d4c9d25fa75bc
This also adds declarations of all the missing texture APIs.
hipTexRefSet*() functions need to take a textureReference as a ptr for type erasure to work. Runtime has been modified to accomodate this.
This change only applies to VDI.
Change-Id: Icf43cc5bd44dfc2c39084b7fe56d5a793bf7319f
* Fix cooperative launch APIs to set hipGetLastError
Previously, the cooperative launch APIs did not properly log their
errors in the global hipGetLastError variable before returning back
to the user. As such, the APIs would leave hipSuccess in the
last error, which would break some use cases.
This fixes that problem by making a trampoline function that does
the HIP_INIT_API and ihipLogStatus.
* Add missing flag to the log of multi-GPU launch
What Cuda refers to "linear texture memory" is the OpenCL equivalent of CL_MEM_OBJECT_IMAGE1D_BUFFER. For these types of allocations we should create a typed buffer instead of an image.
Currently there is no check in the texture fetch functions as to what kind of SRD is written into the texture object, so any kind of incorrect programming will cause the TA to hang. Fortunately for us, every one writes correct code :)
Change-Id: I80dab85a992f2c0754ebf303d40ac6b5e045c7c1
Currently the texture C++ API is forwarded to the ihip*Impl() calls, which are not even a part of Cuda. These should be forwarded to their respective Cuda C APIs instead.
This change also fixes a bug with hipUnbindTexture() creating a dangling pointer.
Change-Id: Ifafc9d106855a11bec84a18ea214b3d89e39990d
Similar to the previous patch, this change adds type constraints to texture indirect functions. Since we don't have to deduce the return type for these, we simply just have to check if the user provided a valid channel type.
Change-Id: Ia094bd6126e01df2ea90902c9aa59cb6cfe85773
When sampling a pixel the hw always returns a float4. The type in the texture reference controls the bitcast that we perform before returning the sampled pixel. Creating a texture with an unsupported will lead to potential UB.
This change makes it so that it's only possible to use textures with a type that makes sense. Using something like texture<int, hipTextureType1D, hipReadModeNormalizedFloat> will now lead to a compilation error with a message "Invalid channel type!".
Change-Id: I7fde44cb1d4b9737e0c48c28cb59c018c59ccaa2
HIP-Clang cuda_wrapper headers require clang include path before standard C++ include path.
However libc++ include path requires to be before clang include path.
To workaround this, we pass -isystem with the parent directory of clang include
path instead of the clang include path itself.
This change addresses three things.
First the available APIs are brought up to par with Cuda (missing ones are added and incorrect ones removed).
Second the size of hip/hcc_detail/texture_functions.h. Using some template magic we can bring down the code size down from ~11k lines to only ~900 lines in total.
Third this change fixes some bugs in the declaration of the texture fetch funcitons. Currently the return type for textures with read mode set to hipReadModeNormalizedFloat is not float. This causes pixel data to be lost during the bitcast when the texture pixel element size is less than the size of float.
The new headers will only be enabled for VDI to avoid breaking HCC.
Change-Id: I77cb29293fb79e55681be094c37702a48d80b64c
GCC emits a warning about using static functions like
hipCUDAErrorTohipError inside this function, because it has an
inline directive, but it's not static. Adding static to this function
to silence warnings (and prevent potential problems in the future).
NVCC warned if you tried to use hipOccupancyMaxActiveBlocksPerMultiprocessor
because when passing in a device function pointer, "const void* func" was
insufficient to describe it accurately. Adding a C++ templated class type
definition for this function.
root
Yaxun (Sam) Liu
Vlad Sytchenko
Maneesh Gupta
Michael LIAO
Aryan Salmanpour
Jeff Daily
Rahul Garg
German Andryeyev
lmoriche
ansurya
Sameer Sahasrabuddhe
Siu Chi Chan
Sarbojit2019
Benjamin Sherman
Joseph Greathouse
Nick Curtis
Sarbojit Sarkar
Jatin Chaudhary