Gráfico de Commits

850 Commits

Autor SHA1 Mensagem Data
Maneesh Gupta f769320b5c Merge pull request #755 from gargrahul/fix_memcpy_symb_nonzerodev
Fixed hipMemcpyToSymbol doesn't work on GPU other than device 0
2018-11-14 13:22:22 +05:30
Rahul Garg 70bb04cc61 Fixed symbol tracking device index 2018-11-13 07:01:17 +05:30
Rahul Garg ac32566d9b Fixed hipMemcpyToSymbol doesn't work on GPU other than device 0 SWDEV-166881 2018-11-13 00:49:20 +05:30
Alex Voicu 1e64813716 Missing handling nullary __global__ functions for mixed arity cases. 2018-11-12 01:51:59 +00:00
Alex Voicu 01ec23538e Merge branch 'master' of https://github.com/ROCm-Developer-Tools/HIP into feature_get_alignment_and_size_from_metadata 2018-11-11 22:37:09 +00:00
Yaxun Sam Liu 80b382a37a Add more checks for fatbin 2018-11-08 16:46:01 -05:00
Alex Voicu 392e2ba1ba Merge branch 'master' of https://github.com/ROCm-Developer-Tools/HIP into feature_get_alignment_and_size_from_metadata 2018-11-08 06:53:06 -08:00
Maneesh Gupta e510be9395 Merge pull request #730 from yxsamliu/dbg-fatbin
Add more checks and debug outputs for fat binary
2018-11-08 11:44:27 +05:30
Maneesh Gupta 0bafc38135 Merge pull request #721 from fronteer/my-fix
Make correct checking of the returned hipDeviceptr_t from read_global…
2018-11-08 11:42:08 +05:30
Maneesh Gupta 8b1544c911 Merge pull request #742 from gargrahul/fix_module_global_tracking
Fixes global symbols tracking in hip_module
2018-11-08 11:41:35 +05:30
Alex Voicu 1c525d7222 Merge branch 'master' of https://github.com/ROCm-Developer-Tools/HIP into feature_get_alignment_and_size_from_metadata 2018-11-01 23:17:27 +00:00
Siu Chi Chan cdd0109e70 Move the global arrays for hip malloc/free
from a header into a source file such that
there's only an unique copy in an executable
and prevent wasting static memory on the host

Change-Id: Id5b62766f77809c8d7b47892cb7149c490dcbdb9
2018-11-01 16:20:35 -04:00
Maneesh Gupta a75d4c0ecc Merge pull request #733 from ex-rzr/fix_array_alloc_size
Fix allocation size of arrays with multiple and/or non-32-bit channels
2018-10-31 14:08:58 +05:30
Maneesh Gupta c7f2198c6e Merge pull request #714 from aaronenyeshi/remove-static-hip-device
Remove libhip_device.a static library
2018-10-31 14:08:29 +05:30
Maneesh Gupta 8f667984af Merge pull request #713 from scchan/master-hcc_early_finalize_support
Support more than one device code blobs in a single .kernel section
2018-10-31 14:04:06 +05:30
Alex Voicu 1621deee71 If we've already seen a __global__ function we do not need to re-parse 2018-10-30 23:34:27 +00:00
Rahul Garg 592efa017f Fixes global symbols tracking in hip_module 2018-10-31 03:22:38 +05:30
Alex Voicu 1fe33b4738 Handle the very confusing dual encoding of the symbol name. 2018-10-30 01:55:09 +00:00
Anton Gorenko 6e6297f3cd Fix allocation size of arrays with multiple and/or non-32-bit channels
hipMallocArray and hipMalloc3DArray must use sum of bits
of all components.
2018-10-29 18:12:00 +06:00
Alex Voicu b68caf7653 Merge branch 'master' of https://github.com/ROCm-Developer-Tools/HIP into feature_get_alignment_and_size_from_metadata 2018-10-28 17:02:10 +00:00
Alex Voicu 299b9d1382 Rely on code object metadat for kernarg arguments alignof and sizeof. 2018-10-28 17:01:00 +00:00
Yaxun Sam Liu 0027a54d8b Add HIP_DUMP_CODE_OBJECT 2018-10-26 14:14:00 -04:00
Yaxun Sam Liu da6c5c152e Adding checks and debug output for fat binary for hip-clang 2018-10-26 14:13:12 -04:00
Aaron Enye Shi ba43d1ce1b Remove libhip_device.a static library
Move remaining function definitions from device_util.cpp to hip_runtime.h header. We can now remove the static library completely as device_util.cpp was the last file part of target hip::hip_device .
2018-10-25 19:52:35 +00:00
Maneesh Gupta b01ac26948 Make HIP functional again with HCC from ROCm 1.9.x
Change-Id: I214acdfd0b79dcf783993e44fe31baee64fd4dc3
2018-10-24 10:41:56 +05:30
Qianfeng Zhang 443698ce00 Make correct checking of the returned hipDeviceptr_t from read_global_description() 2018-10-23 21:13:11 +08:00
Siu Chi Chan f39873c584 Support more than one bundles in a single .kernel section
When compiling with Early Finalization enabled in HCC,
the resulting .kernel section of the host object now may
contain more than one device code bundles.  This is to
teach the HIP runtime to correctly extract all the
bundles from the .kernel section.
2018-10-18 17:00:27 -04:00
Alex Voicu 602280eb8b Re-sync with upstream. 2018-10-18 12:27:03 +01:00
Maneesh Gupta 9143ae6bdb Merge pull request #692 from whchung/hip-reinit-take2
HIP program state re-initialization logic (take 2)
2018-10-18 12:06:41 +05:30
Maneesh Gupta 95be669f4a Merge pull request #703 from mangupta/stream_create_with_priority
Implementation for stream priority
2018-10-17 10:53:43 +05:30
Alex Voicu 9ec697c620 Re-sync with upstream. 2018-10-10 11:43:49 +01:00
Maneesh Gupta 07ee1f07d8 Implementation for stream priority
- Requires ROCm 1.9.x or higher
- Requires HCC with PR#886 merged

Change-Id: Id7c95ea091ee610e80c9ad815f1cb989cba570ca
2018-10-05 16:27:46 +05:30
Rahul Garg a78d68d639 Corrected the width calculation logic to accomodate multi channels 2018-10-03 12:07:38 +05:30
Rahul Garg 167265c279 Fixed image width for linear resource type texture 2018-10-01 15:28:34 +05:30
Wen-Heng (Jack) Chung e257de95f3 Keep the map which tracks GPU kernel symbols to grow monotonically 2018-09-26 19:49:02 +00:00
Wen-Heng (Jack) Chung 060b3c0bf8 Improve performance of re-initialization logic
Keep track of shared libaries already discovered. Do not build HSA executables
for them.
2018-09-26 19:48:56 +00:00
Wen-Heng (Jack) Chung 319f007bf1 HIP program state re-initialization logic
This commit is to support kernels dynamically loaded thru means such as
dlopen() after HIP runtime initializes.
2018-09-26 19:48:47 +00:00
Rahul Garg bd27310127 Return hipSuccess when sizeBytes=0 in hipMemset 2018-09-26 12:47:36 +05:30
Alex Voicu 26c26f63cb Align with HC Next. 2018-09-17 11:50:29 +03:00
Maneesh Gupta b23609ad42 Merge branch 'master' into support-malloc 2018-09-17 10:17:25 +05:30
Yaxun Sam Liu ce13c2061a Add HIP_DB=fatbin for debugging fat binary issues 2018-08-17 11:53:45 -04:00
Maneesh Gupta e10fa70173 Merge pull request #621 from ROCm-Developer-Tools/disable_startup_loader
Disable startup loader by default and guard with env var
2018-08-10 10:18:25 +05:30
sunway513 7c467220a8 resolve a segfault bug when env var not set; remove startup_kernel_loader class 2018-08-09 16:40:26 +00:00
sunway513 7d776ae665 Add more check to ensure the startup loader only be enabled with the env var set to 1 2018-08-04 01:52:27 +00:00
sunway513 9af862d364 Add startup loader under HIP_STARTUP_LOADER env var, disable by default 2018-08-04 01:48:06 +00:00
Wen-Heng (Jack) Chung 1fe7f1be91 Revert "HIP program state re-initialization logic"
This reverts commit 04640992dc.
2018-08-03 17:03:04 -05:00
Wen-Heng (Jack) Chung f6178c2159 Revert "Improve performance of re-initialization logic"
This reverts commit b883ea759d.
2018-08-03 17:02:58 -05:00
Wen-Heng (Jack) Chung 329f3b7a5e Revert "Keep the map which tracks GPU kernel symbols to grow monotonically"
This reverts commit 8f521edff1.
2018-08-03 17:02:50 -05:00
Sarunya Pumma 84aadb9274 Remove device mapping from shareWithAll memory
When shareWithAll memory (e.g., host memory) is allocated, set appId
in hc::AmPointerInfo to -1 to indicate that this memory is not mapped
to any device.  Peer checking in ihipStream_t::canSeeMemory is not
necessary if memory is shared with all devices.  Thus, it is skipped.

Note that earlier host memory is always mapped to device 0 and HIP
always performs peer checking for all kinds of hipMemcpy.  Since the
peer checking process requires context locking, hipMemcpy from/to host
memory always grabs device 0's context lock.  Therefore, if there is
another thread holding the context lock of device 0 (e.g.,
hipDeviceSynchronize on device 0), hipMemcpy will have to wait for the
lock until it can actually perform memcpy.  This can significantly
deteriorate execution performance.

Signed-off-by: Sarunya Pumma <sarunya.pumma@amd.com>
2018-07-28 23:15:16 -07:00
Yaxun Sam Liu d17dc21c9c Support malloc/free for hip-clang 2018-07-27 16:24:51 -04:00