Commit grafiek

811 Commits

Auteur SHA1 Bericht Datum
Maneesh Gupta 4a671abf13 Merge branch 'master' into support-malloc
[ROCm/hip commit: 66f863d1f3]
2018-09-17 10:17:25 +05:30
Yaxun Sam Liu 86b18a66ef Add HIP_DB=fatbin for debugging fat binary issues
[ROCm/hip commit: 1299b65e15]
2018-08-17 11:53:45 -04:00
Maneesh Gupta b5940c7969 Merge pull request #621 from ROCm-Developer-Tools/disable_startup_loader
Disable startup loader by default and guard with env var

[ROCm/hip commit: 4cf851f416]
2018-08-10 10:18:25 +05:30
sunway513 cfe96a2db0 resolve a segfault bug when env var not set; remove startup_kernel_loader class
[ROCm/hip commit: 17f38937e0]
2018-08-09 16:40:26 +00:00
sunway513 09a4ad1dcf Add more check to ensure the startup loader only be enabled with the env var set to 1
[ROCm/hip commit: 30dfa6f129]
2018-08-04 01:52:27 +00:00
sunway513 653a04fe98 Add startup loader under HIP_STARTUP_LOADER env var, disable by default
[ROCm/hip commit: 3a68ab4919]
2018-08-04 01:48:06 +00:00
Wen-Heng (Jack) Chung 38b6402860 Revert "HIP program state re-initialization logic"
This reverts commit e5ca9eb081.


[ROCm/hip commit: 2604f33930]
2018-08-03 17:03:04 -05:00
Wen-Heng (Jack) Chung 48406c881c Revert "Improve performance of re-initialization logic"
This reverts commit c6b975bf13.


[ROCm/hip commit: 3426f15171]
2018-08-03 17:02:58 -05:00
Wen-Heng (Jack) Chung 22e4ddf301 Revert "Keep the map which tracks GPU kernel symbols to grow monotonically"
This reverts commit d214b045c3.


[ROCm/hip commit: 136bcc2981]
2018-08-03 17:02:50 -05:00
Sarunya Pumma a68ea730c2 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>


[ROCm/hip commit: 8111fd3b8b]
2018-07-28 23:15:16 -07:00
Yaxun Sam Liu 289bfe5f95 Support malloc/free for hip-clang
[ROCm/hip commit: 02d0e93601]
2018-07-27 16:24:51 -04:00
Rahul Garg c13c5c9476 Null check before setting offset
[ROCm/hip commit: bc4cdf7e41]
2018-07-24 12:25:40 +05:30
Rahul Garg af407ce8be Set offset in hipGetTextureAlignmentOffset
[ROCm/hip commit: 867a4aa971]
2018-07-24 10:11:26 +05:30
Alex Voicu 5739384c3e It is unclear what I was thinking when authoring the original code...
[ROCm/hip commit: 9938edb636]
2018-07-17 14:04:57 +01:00
Maneesh Gupta c86d32350a Merge branch 'master' into move-memcpy
[ROCm/hip commit: b5cfa773ef]
2018-07-17 10:51:42 +05:30
Maneesh Gupta d655537bc6 Merge pull request #515 from ROCm-Developer-Tools/hipclang-add-amdgcn-funcs
Add hipclang amdgcn functions

[ROCm/hip commit: fbbe2599dd]
2018-07-17 09:25:09 +05:30
Maneesh Gupta f62f441ac0 Merge pull request #562 from ROCm-Developer-Tools/fix-build-failure
Fix build failure in code_object_bundle.cpp

[ROCm/hip commit: 06f4579c3a]
2018-07-12 07:49:41 +05:30
Maneesh Gupta eb15d607c9 Merge pull request #546 from gargrahul/fix_bindtex_offset_null_check
Fixed offset null check in bind texture functions

[ROCm/hip commit: 296dce2e2b]
2018-07-11 12:52:31 +05:30
Maneesh Gupta 2d2e264fd2 Merge pull request #545 from ROCm-Developer-Tools/revert-521-temp_fixmemcpy2dasync_trsmissue
Revert "Use memcpy kernel for all pinned memory cases in hipMemcpy2DAsync"

[ROCm/hip commit: 86e10fed99]
2018-07-11 12:52:09 +05:30
Yaxun (Sam) Liu 244c9875e2 Move __hip_hc_memcpy and __hip_hc_memset from device_utils.cpp to device_functions.h as inline functions
[ROCm/hip commit: 8136a348ab]
2018-07-10 18:12:41 -04:00
Yaxun (Sam) Liu 0b0041e82c Fix build failure in code_object_bundle.cpp
[ROCm/hip commit: 7db7cce9e4]
2018-07-10 16:49:59 -04:00
Aaron Enye Shi 17453a6278 Implement hip_ldg Functions into HIP header
Move all the function definitions for hip_ldg.cpp into hip_ldg.h header and enable for HIP clang path.


[ROCm/hip commit: 76f86ef097]
2018-07-05 20:38:46 +00:00
Aaron Enye Shi 8e1a7fdd09 Implement min/max functions in HIP header
Remove using hc::precise_math min and max. Instead we can use ocml directly for device and std:: for host.


[ROCm/hip commit: 47d78e372e]
2018-07-05 20:15:41 +00:00
Aaron Enye Shi 0ea959ba9d Implement Memory Fence Functions in header
Enabled __llvm_fence_* functions for seq_cst.


[ROCm/hip commit: 930a16bccd]
2018-07-04 23:35:24 +00:00
Aaron Enye Shi e8fac0449e Merge branch 'master' into hipclang-add-amdgcn-funcs
[ROCm/hip commit: 2975f2a10a]
2018-07-04 17:36:08 +00:00
Maneesh Gupta 19cb14d54c Update hip_hcc_internal.h
Adding missing include for hip_hcc_internal in order to build with HCC

[ROCm/hip commit: 0c2f985553]
2018-07-04 09:33:51 +05:30
Rahul Garg ff253a279c Fixed offset null check in bind texture functions
[ROCm/hip commit: feff0aeea4]
2018-07-03 08:54:17 +05:30
Rahul Garg c957c42c20 Revert "Use memcpy kernel for all pinned memory cases in hipMemcpy2DAsync"
[ROCm/hip commit: 7cd1d5e644]
2018-07-02 14:32:11 +05:30
Aaron Enye Shi 6e2ee0f857 Implement __shfl_* funcs into HIP headers
[ROCm/hip commit: 9ac31e0bb6]
2018-06-26 18:32:11 +00:00
Aaron Enye Shi e2f8af2f3d Implement __ballot, __any, __all into HIP headers
[ROCm/hip commit: 6dc16bbf04]
2018-06-20 17:39:39 +00:00
Aaron Enye Shi 2a2f34b42d Implement hip_hc.ll into HIP headers
Move all __hip_hc_ir_* functions from hip_hc.ll into HIP header as inline asm. Remove hip_hc.ll and build dependencies from HIP.


[ROCm/hip commit: 2142eb4d12]
2018-06-20 17:39:31 +00:00
Aaron Enye Shi 0fca41a7ce Implement device_functions.cpp into HIP headers
Move all Integer Intrinsics, device_functions.cpp definitions and HIP specific device functions into HIP headers. Implement the device functions using llvm_intrinsics and device-libs functions instead of calling hc::__* functions. Remove device_functions.cpp since everything is now defined in header.


[ROCm/hip commit: e02fc7e680]
2018-06-20 17:39:23 +00:00
Aaron Enye Shi b34868005a Add hipclang amdgcn functions
These are moving from hipclang in device library to hip headers. These are required for the functionality of HIPclang project.


[ROCm/hip commit: c453b42bff]
2018-06-20 17:38:37 +00:00
Maneesh Gupta 27e2af1773 Merge pull request #490 from ROCm-Developer-Tools/feature_decouple_atomics_from_hc
Switch the atomic implementation to use Clang  builtins.

[ROCm/hip commit: 946c8da88a]
2018-06-20 14:16:43 +05:30
Maneesh Gupta 4cdc20a6ce Merge pull request #457 from whchung/hip-reinit
HIP program state re-initialization logic

[ROCm/hip commit: 836627279f]
2018-06-20 09:37:27 +05:30
Wen-Heng (Jack) Chung d214b045c3 Keep the map which tracks GPU kernel symbols to grow monotonically
[ROCm/hip commit: 32789a8b7d]
2018-06-18 16:54:18 -05:00
Wen-Heng (Jack) Chung c6b975bf13 Improve performance of re-initialization logic
Keep track of shared libaries already discovered. Do not build HSA executables
for them.


[ROCm/hip commit: ece4539c1d]
2018-06-15 18:07:33 -05:00
Rahul Garg 388679efc8 TEMP- fix memcpy2dAsync for trsm issue
[ROCm/hip commit: cd23905897]
2018-06-15 16:08:29 +05:30
Wen-Heng (Jack) Chung e5ca9eb081 HIP program state re-initialization logic
This commit is to support kernels dynamically loaded thru means such as
dlopen() after HIP runtime initializes.


[ROCm/hip commit: 379b7a2241]
2018-06-14 15:46:49 +00:00
Rahul Garg 312999de41 Fix stream resolution in memcpy2dasync
[ROCm/hip commit: 069e2c34c9]
2018-06-14 11:58:56 +05:30
Rahul Garg 1d6396dfb9 Fix retrieved locked ptr offset
[ROCm/hip commit: 00f8a36bc7]
2018-06-13 23:10:05 +05:30
Maneesh Gupta b54be20b05 Merge pull request #482 from ROCm-Developer-Tools/feature_clean_up_hip_math
Switch to using ROCDL directly, as opposed to via HC. Add missing bits.

[ROCm/hip commit: 203dd6cb70]
2018-06-06 16:07:22 +05:30
Maneesh Gupta ac027e4092 Merge pull request #497 from gargrahul/fix_memcpy3d_fastpath
Fix hipMemcpy3D for fast path

[ROCm/hip commit: 9e9c039ee4]
2018-06-06 14:44:02 +05:30
Maneesh Gupta f4dd7fd056 Merge pull request #492 from gargrahul/fix_depth_3d_alloc
Fix depth value for 3D allocations

[ROCm/hip commit: 216f34eea8]
2018-06-06 14:41:23 +05:30
Maneesh Gupta f44e77944d Merge pull request #491 from scchan/fix_wait
callback handling: don't need to wait for the thread to become ready

[ROCm/hip commit: 7311b60220]
2018-06-06 14:38:25 +05:30
Rahul Garg e7bc68d347 Fix hipMemcpy3D for fast path
[ROCm/hip commit: a46ff2afd5]
2018-06-05 18:54:33 +05:30
Siu Chi Chan 417dde9d73 remove the _ready flag in ihipStreamCallback_t and the mutex that protects it.
[ROCm/hip commit: a1f3b587fb]
2018-06-04 17:29:04 -04:00
Rahul Garg 6592b35c39 Fix depth value for 3D allocations
[ROCm/hip commit: 276c948a16]
2018-06-04 18:00:22 +05:30
Siu Chi Chan 4b25b76898 callback handler: don't need to wait for the thread to become ready
[ROCm/hip commit: d3a9985f10]
2018-06-02 17:55:37 -04:00
Alex Voicu f7fd20ec17 Switch the atomic implementation to use Clang builtins.
[ROCm/hip commit: 089ab3b947]
2018-06-02 12:27:17 +01:00