Fix wrong mixing of current device and stream device in
ihipModuleLaunchKernel() and hipLaunchCooperativeKernel().
Fix missing hipSetDevice() in hipMemcpyWithStream* tests.
Change-Id: I09333bb40d239bb42c832df5ea16d17eeaeff5e7
hipAddressModeClamp corresponds to CL_ADDRESS_CLAMP_TO_EDGE.
This change fixes incorrect sampling near texture edges.
https://github.com/ROCm-Developer-Tools/HIP/pull/2229
Change-Id: Ia2af9f3cb582a74af7b1f65bad1ed0310eab2628
Change-Id: I5daeacd9dd5c6ce7f914d6e6e45dd41fb2a675a5
hipMemRangeGetAttributes was returning hipErrorInvalidValue due to improper
mapping of the arguments to cudaMemRangeGetAttributes.
If the width is 0, hipMemset2D or 3D APIs should return success
doing nothing
Change-Id: I97f8825d090a73ee63145e52d2b6d2cc0eddade3
Change-Id: I9d62fee70929ad0c79b947aa4cbbc8ab898fb64a
HIP supports emitting two types of static libraries. One type
will export host functions and is compatible with host linkers.
The second type exports device functions, but is generated with
ar manually. Also, add a README with steps on how to run these
samples with Makefile or CMake.
Change-Id: I1be15c2884583b370092bc8e4bf04f726f8f5a27
Remove hipSimpleIpcMem.cpp as the IPC APIs are expected to work
with different processes and not the same process. Also, modify
the hipMultiProcIpcMem.cpp to test the offset scenarios
Change-Id: Ia161fc9ab39e27f22d8ef7268e0d7669ece5cdce
Change-Id: Ic3cb79ef4aca40b4a504483eb81fd83bb8a201d9
[dtest] Tests for hipMemcpy2DFromArray
and hipMemcpy2DFromArrayAsync APIs
1. Negative scenarios for both the APIs
2. Functional size check scenarios for both APIs
3. Invoke the API from Peer Device Context
4. Simple scenarios for both APIs
5. Pinned Host Memory from same & multi GPU
Change-Id: I80c84ec3f9722d18683d31be65caa1bd64054da6
This test is incorrect about _Float16 overloading. Need to keep the test
passing for both old and new compiler until the compiler change is in
mainline.
Change-Id: I2f9be8957282fd0579c74f1e9f23e65c4484f3a2
hipIpcOpenMemHandle should return the device pointer which is
similar to the base ptr of the original allocation even if the offset
to the original pointer is passed to hipIpcGetMemHandle
Change-Id: I99c0553e8c67c15b5fed880b6a4c74bce39c3aee
Device enqueue has an option to execute scheduler on the current
queue and it's enabled by default. Make sure scratch is allocated
on the current queue for that case. Add max vgpr tracking per
program to adjust scratch size accordingly.
Change-Id: I2a6d796913a4551a1e7f343a2465d589eec60d8a
Add concurrentManagedAccess detection in hipMallocManaged test.
Skip test when device doesn't support concurrentManagedAccess.
Change-Id: Ie54046feef3baba857a7068972ec1fc1a60c2dfd
Early notification allows CPU command state to be processed earlier.
Otherwise state update will be delayed until the actual possible
wait.
Change-Id: I0e9d6c2ad88e3723eab2f6dc9a20f288f6c1ca0c
MT doesn't use GPU waits, but CPU for sync between engines.
Change the threshold values for CPU waits for direct dispatch.
That will bring behavior closer to MT.
Change-Id: Ia41c3cb812614962aff2746b6cf858f1bf77dda2
Satyanvesh Dittakavi
Tao Sang
Anton Gorenko
cjatin
Chauncey Hui
pghafari
Rahul Garg
German Andryeyev
Aaron En Ye Shi
Julia Jiang
Yaxun (Sam) Liu
DURGESH KROTTAPALLI
Sarbojit Sarkar
Anusha GodavarthySurya
sumanthtg
Sourabh Betigeri