The __ocml_sqrt_* functions only accept a single arg,
not two args. Also, change ifdef for if because the flag
__HCC_OR_HIP_CLANG__ is always defined.
Change-Id: Id9dc4eaf0f25df1df27a1bf643cb545ac23653f8
Enable this optimization when the barrier is disabled, since
reuse requires a signal wait.
Use the size of pending AQL signals as the size of signal pool.
Change-Id: I2754a0f8b67e19d2601c58945e10fdf0e8be1624
On ReBar systems the invible heap is not present, so in theory we should
fail creating the suballocation chunk, however PAL doesn't report any
errors.
To make sure we never fail, allow creating the allocation in the visible
heap and system memory.
Change-Id: Iea9cc68d98b9cb396a2b7a37398b98b66274083b
This test does not work if block size is greater than wave size
since it relies on lock-step execution of the kernel in the block.
If there are more than waves in the block, the threads in the block
miss synchronization since one wave may finish before another wave.
Due to this bug, the test fails on GFX10 wave32 mode.
This patch fixes that so that it works for block size greater than
wave size.
Change-Id: Ie0097066081df36cb6fe025a71d0ee5a83ec00a2
Now rocm/rocdevice.cpp also includes comgrctx.hpp, and we don't want to statically link against comgr when buidling shared libs.
Change-Id: Ic330bd860559b3e07b776c951afe6126b0f43f7d
also validate the flag argument passed to hipIpcOpenMemHandle API and return
error if it is not equal to "hipIpcMemLazyEnablePeerAccess" (defined to 0)
and it is the only supported flag currenlty for this API.
SWDEV-253462
Change-Id: Ie1c8b79c680a29dc72bf262cf53ae5e011fb1247
APIs: hipModuleLoad, hipModuleLoadData and hipModuleLoadDataEx,
hipModuleGetGlobal(), hipModuleGetFunction
Functional and negative tests
Repeated call to ModuleLoadXX/ModuleUnloadXX
Few tests are disabled now, will be enabled when functional
SWDEV-238517 for enhancing hip unit tests
Change-Id: I65c12027e32db80213468fdee1c5cc1aa3e60bfd
This is helpfull to do when debugging issues on lowend asics. Navi14 can be emulated as Navi10. So can Navi22 be emulated as Navi21.
Change-Id: I693ffd45a5b03657822afdc872781901bc69b65c
The device side assertion calls printf to write out a message. In the
device compiler, printf is expanded into a series of hostcalls that
transmit the printf payload to the host. This expansion increases the
length of the kernel, resulting in sub-optimal compilation. The
solution is to ensure that the assert() implementation is not inlined
into the kernel.
Change-Id: Ia3a075461a755cf007218f262b0863e1926c76aa
With the PAL_ALWAYS_RESIDENT flag memory objects are resident at allocation time, no need to make them resident again before submit.
Also we should never evict anything with this setting, or we'll generate a VM fault.
Change-Id: Ieacc6af88ab4e09c20efd94100e148b2502e1d70
Rahul Garg
Jason Tang
Aaron En Ye Shi
Tao Sang
Chauncey Hui
German Andryeyev
Sourabh Betigeri
Reshabh Sharma
Alex Xie
Vladislav Sytchenko
Julia Jiang
Yaxun (Sam) Liu
Jason Tang
Sarbojit Sarkar
agodavar
Aryan Salmanpour
Lakhan singh Thakur
Rupam Chetia
Freddy Paul
Sameer Sahasrabuddhe