JIRA : SWDEV-234471
With this change compilation of the ibamdhip64 to be decied by arguments to build_hip_on_rocclr.sh file
build_hip_on_rocclr.sh -r (current) shall conitnue to create libamdhip64.so;
build_hip_on_rocclr.sh -r a (new) shall create libamdhip64.a
other options remain the same
Changed the targets according to requirement of stage2 dependencies.
Changed find_lib with find_package
Adpated the review comments
Change-Id: I719918f73ab33cd8031a32db6fe9b8adaa1b4b87
[ROCm/clr commit: a77c7e21be]
Host memory allocations use shared device context, which allows
to have access from any device. That breaks default queue detection.
Change-Id: I27ad1234d49ab0ea8f5edd6f2e3ffa93c47d4b1f
[ROCm/clr commit: ef75a3c1db]
Add hipHostMallocNumaUser flag to hipHostMalloc() in order to support
numa policy set by user.
Change-Id: I6d70ed539a5f97f27187f2242b68849c0e27e4d6
[ROCm/clr commit: 71c64e98a0]
hipChannelFormatDesc contains size of each channel in bits. There's 8 bits per byte, not 4, so we should divide by 8.
Change-Id: I38ae2d28ad54740823c98908ca2afecad0ba0183
[ROCm/clr commit: 81b15ed36e]
Similar to http://gerrit-git.amd.com/c/compute/ec/vdi/+/375385, except no -Wno-strict-aliasing.
With this change only two warnings show up during the build - -Wsign-compare and -Wmisleading-indentation.
Change-Id: Iffa436c65ab6312aeaa6def71ee6af38ed1b4a4e
[ROCm/clr commit: 509a38f50c]
exposing the following hip APIs which have been previously implemented:
hipMemcpy2DFromArray
hipMemcpy2DFromArrayAsync
hipMemcpyAtoH
hipMemcpyHtoA
hipMemcpyParam2DAsync
Change-Id: I3f8635bd4265d2bf5aa2084f4eabb51f74fb18ba
[ROCm/clr commit: 02c62165b8]
this change follows CUDA convention where lower number is greater priority
Change-Id: I72596a36449e818cbd8c175bf8519c51f46b1610
[ROCm/clr commit: 83b8e1fbac]
Workaround hipStream deadlock issue as the same lock was used twice SWDEV-236746
Change-Id: Icc60104ce6edf4cfd2a3a889bab78a6caadd50b7
[ROCm/clr commit: f3ee29cdb2]
The hipOccupancyMaxPotentialBlockSize API is meant to return the
number of threads for the highest-occupancy workgroup, and the number
of those workgroups. It was previously calculating the number of
maximum-sized workgroups that would fit on a single CU. This is
a mixture of the API we wanted (to calculate max potential block size)
and the MaxBlocksPerMultiprocessor function.
This patch fixes it up so that the internal occupancy calculation
function works for two uses: the traditional function that calculates
the maximum blocks per multiprocessor when a user passes in a fixed
block size (used for hipMaxBlocksPerMultiprocessor style functions)
and a function that calculates the size of a block that would lead
to maximum occupancy, and how many blocks of that size would be
needed to fill the whole GPU (for hipOccupancyMaxPotentialBlockSize
style functions).
This also updates the occupancy calculation function to prepare for
gfx10, which does not have SGPR-based occupancy limits.
Change-Id: Ie007b3f9d5ebc4e166b50a3a051498af35650f35
[ROCm/clr commit: 90453b68d3]
SWDEV-237377 - This fixes time calculation where the event may
be recorded on Null stream and work submitted on other streams
Change-Id: Ie36310dea5cee2fed4a514ed01f04db4b47e571c
[ROCm/clr commit: facb05495f]
Evgeny
Ashutosh Mishra
Sarbojit Sarkar
Tao Sang
kjayapra-amd
German Andryeyev
Aryan Salmanpour
Jatin Chaudhary
agodavar
Vlad Sytchenko
Jason Tang
jujiang
Christophe Paquot
Saleel Kudchadker
Payam
Joseph Greathouse