Skip to content
This repository was archived by the owner on Apr 2, 2025. It is now read-only.

Huge merge for AMD GPUs #497

Merged
mxz297 merged 217 commits into from
Feb 23, 2022
Merged

Huge merge for AMD GPUs #497

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
217 commits
Select commit Hold shift + click to select a range
490062f
Move load map init and fnbounds init ahead of registering sample source.
mxz297 Jun 19, 2020
e6826a2
More re-ordering in hpcrun init code. The new order is:
mxz297 Jun 19, 2020
b6440b0
added papi_active_flag
Jul 1, 2020
6359ced
solved papi issue with init and deadlock on cupti
Jul 2, 2020
9574574
papi cuda component available
Jul 2, 2020
121e3c0
Rocm support added
dejangrubisic Jul 6, 2020
3124dac
created intel gptin instrumentation files for hpcrun
Jul 9, 2020
fcf0068
updated imports in hpcrun makefile and opencl files
Jul 10, 2020
f794745
updated imports in hpcrun makefile and opencl files
Jul 10, 2020
88518a3
Merge branch 'develop' of https://github.com/HPCToolkit/hpctoolkit in…
Jul 10, 2020
c2cc831
added static_link guard inside opencl-intercept.c
Jul 10, 2020
b7760ce
started MetricAPI XMIT-nvlink and able to collect papi metric from su…
Jul 15, 2020
3507328
added support for gpu-monitors register
Jul 20, 2020
ff72647
adding static-link guard ifor gotcha code in opencl-intercept.c
Aug 3, 2020
f5e091d
fixed merge conflicts with develop
Aug 3, 2020
d5f6b2e
module-ignore-map includes rocm, deleted print_counter function
dejangrubisic Aug 5, 2020
689599e
identation
dejangrubisic Aug 5, 2020
7c889f1
unwind problem solved, papi gets gpu_sync with pointer
dejangrubisic Aug 17, 2020
ec6b1b4
support for extracting lineinfo from Intel GPU binaries inside hpcstruct
Aug 28, 2020
b5b7c1a
merged with develop and fixed merge conflicts
Aug 28, 2020
071b705
Merge branch 'develop' of https://github.com/HPCToolkit/hpctoolkit in…
Sep 2, 2020
7eeca68
updated code for creating loadmaps for opencl kernels
Sep 7, 2020
0894406
updated loadmap names as absolute filepaths for opencl
Sep 7, 2020
b39e759
added bifurcation in hpcstruct for opencl code
Sep 7, 2020
ba49c3c
added intercept to get source code for opencl kernels
Sep 11, 2020
25b5a87
Rewrite configure files and remove all the manual dependencies, at le…
Sep 13, 2020
08bd009
added source code dumps for opencl
Sep 13, 2020
e2b007d
Line formatting
Jokeren Sep 14, 2020
b457605
Merge branch 'opencl_instrumentation' of github.com:aarontcopal2/hpct…
Jokeren Sep 14, 2020
c993e27
added code for generating intel kernel cfg represenation
Sep 14, 2020
0c01712
Merge opencl_instrumentation
Jokeren Sep 14, 2020
54e319b
Merge pull request #2 from aarontcopal2/opencl_instrumentation_fix
Jokeren Sep 14, 2020
c3a0fbb
opencl writing trace files enabled
dejangrubisic Sep 16, 2020
5e01478
merged with develop
Sep 16, 2020
697e738
Init CFG creation; at least it compiles
Jokeren Sep 17, 2020
4bf977e
Change tab to space
Jokeren Sep 17, 2020
302a20c
Works but needs correctness check
Jokeren Sep 17, 2020
b90d4b1
include gpu-trace.h in opencl.c
dejangrubisic Sep 17, 2020
e255da4
Fix bugs
Jokeren Sep 17, 2020
1062358
Merge pull request #3 from aarontcopal2/opencl_instrumentation_fix
Jokeren Sep 17, 2020
66ac87e
Update dwarf.h
Jokeren Sep 18, 2020
540e503
Add Intel copyright for all pti samples
Jokeren Sep 20, 2020
f291d22
Merge cfg analysis for intel and cuda
Jokeren Sep 21, 2020
aeacef0
code for adding block and instruction offsets in struct files for int…
Sep 22, 2020
74b4b83
replacing gotcha intercepts with function overrides for opencl functions
Sep 23, 2020
f069a23
Use another dyninst instruction interface to construct gpu instructio…
Jokeren Sep 23, 2020
7e6fcca
opencl_multiplexer: compilable version
dejangrubisic Sep 23, 2020
715c4dd
refactoring opencl-api interface
dejangrubisic Sep 23, 2020
184e54b
refactoring opencl-translate
dejangrubisic Sep 24, 2020
251ebc2
using the same enum for gpu_activity and opencl_activity
dejangrubisic Sep 24, 2020
1b9a39e
1. Refactor gpu binary dump and parsing process 2. Fix memory leak pr…
Jokeren Sep 24, 2020
dafacca
Merge opencl_instrumentation
Jokeren Sep 24, 2020
34a18b9
adding hpcrun metrics for changing scope of metrics
Sep 25, 2020
a793ea0
Merge branch 'opencl_instrumentation' of https://github.com/HPCToolki…
Sep 25, 2020
f032286
operation_channel_signal_consumer_when_full
dejangrubisic Sep 25, 2020
c94c525
Dump debug info extracted from GTPin's API
Jokeren Sep 26, 2020
679a5c7
operation_channel is working, item must has return_channel*, activity
dejangrubisic Sep 26, 2020
39f1a8f
Fix hpcstruct
Jokeren Sep 26, 2020
46d3b4d
All consume/produce are properly matched, gpu_trace_fini(NULL);
dejangrubisic Sep 27, 2020
e442df7
Fix opencl apis
Jokeren Sep 27, 2020
618daf7
Further refine hpcstruct
Jokeren Sep 27, 2020
ef9c061
changing the cct_node creation for opencl instructions. ccts will be …
Sep 28, 2020
eb471a3
merged changes from remote opencl-instrumentation branch
Sep 28, 2020
0693153
refactoring opencl instrumentation code. instrumented metrics not vis…
Sep 29, 2020
13986f9
Merge branch 'opencl_instrumentation' of github.com:HPCToolkit/hpctoo…
Jokeren Sep 29, 2020
4b0c48a
Refactoring gtpin code. 1. Using structured blocks and insts to store…
Jokeren Sep 29, 2020
d15ab0c
Tracing is working: gpu_trace_fini is called from Monitoring thread, …
dejangrubisic Sep 30, 2020
1402825
Use opencl runtime stack for gtpin
Jokeren Sep 30, 2020
5d0f747
removing opencl-intercept files and some unwanted functions related t…
Sep 30, 2020
3fcf585
Fix opencl api by adding a clCreateCommandQueueWithProperties wrapper
Jokeren Oct 2, 2020
e5eeb39
1. Fix duplicate correlation id in gtpin; 2. Fix clCreateCommandQueue…
Jokeren Oct 3, 2020
6a1c23a
Enable instruction metrics only when gtpin is enabled
Jokeren Oct 3, 2020
b65bf7f
Modify gtpin log to include correlation
Jokeren Oct 3, 2020
19afe11
Fix initialization order warning
Jokeren Oct 3, 2020
e0a9008
opencl: traceOrdered flag added
dejangrubisic Oct 3, 2020
fb0acea
improved multi-threaded tracing, requires testing
Oct 3, 2020
51b26b6
added atomicity for num_threads
Oct 3, 2020
6e545e6
main: control_knob_init must go before hpcrun_registered_sources_init
Oct 3, 2020
43b81e3
working version
Oct 3, 2020
fe71181
gpu_trace_demultiplexer implemented
Oct 4, 2020
586686e
gpu_trace_set_apply version -> making empty files
Oct 4, 2020
bb4aaae
working version 2
Oct 4, 2020
b5f00aa
new_channel_set->channel_set_ptr = gpu_trace_channel_stack_alloc(stre…
Oct 4, 2020
5c7b3ba
gpu-trace-demultiplexer: Final version
dejangrubisic Oct 4, 2020
f4329b9
traceOrdered bit added in core_profile_data
dejangrubisic Oct 4, 2020
2c021fa
Fix papi compilation error
Oct 5, 2020
0add9f0
Fix opencl configuration
Oct 5, 2020
74e54dd
Revert "Fix papi compilation error"
Oct 5, 2020
4e74316
Setting defaults for STREAMS_PER_THREAD=4, MAX_COMPLETION_CALLBACK_TH…
dejangrubisic Oct 5, 2020
cc205b5
Refactoring of func names
dejangrubisic Oct 5, 2020
19c7f2e
instrumentation mode for opencl can be enabled by passing -e gpu=open…
Oct 5, 2020
0adbcfd
merge changes with remote origin
Oct 5, 2020
c9f4cbc
Use a global map to store correlation id; the gtpin callback thread c…
Jokeren Oct 5, 2020
8de8345
Merge branch 'opencl_instrumentation' of github.com:HPCToolkit/hpctoo…
Jokeren Oct 5, 2020
4a494b7
Enable instruction metrics in gtpin
Jokeren Oct 5, 2020
7fcd34d
changing CL_TARGET_OPENCL_VERSION to 2.2. Some hacks for opencl profi…
Oct 5, 2020
cdf3da6
merged with remote branch
Oct 5, 2020
d77e81e
thread_fini_action: commented out opencl_api_finalize(NULL), if calle…
dejangrubisic Oct 6, 2020
136bc34
opencl finalizer-> thread_finalize, process_finalize
dejangrubisic Oct 6, 2020
abb1756
opencl H2D calls that bypass clEnqueueWriteBuffer and directly call c…
Oct 7, 2020
13e9b86
NOT_COMPLETE:Merging gpu-trace-demultiplexer with opencl_instrumentation
dejangrubisic Oct 7, 2020
feafe49
merged changes of gpu-trace-demultiplexer with opencl-instrumentation
Oct 8, 2020
44e9cf8
added NULL checks to some opencl function intercepts
Oct 8, 2020
8b91f71
opencl H2D and D2H operations made via clEnqueueMapBuffer also will b…
Oct 8, 2020
12f31a4
adding context id and stream id for H2D calls from clSetKernelArgs
Oct 9, 2020
a722602
fixing error with stream id in opencl-api.c
Oct 9, 2020
a17e84c
fix opencl_translate
dejangrubisic Oct 9, 2020
01cc72f
fix opencl_translate
dejangrubisic Oct 9, 2020
d8fd4c6
working version: opencl_activity_translate: gpu_start_time = gpu_subm…
dejangrubisic Oct 9, 2020
020f32d
refactored opencl-activity-translate. Duplicated function code has be…
Oct 12, 2020
d6680f6
removing start_time, end_time for the memory transfers that happens a…
Oct 12, 2020
f86ebe3
adding context_id and stream_id for opencl memory transfers and kerne…
Oct 12, 2020
286dc2b
adding support for monitoring kernel executions from clEnqueueTask API
Oct 13, 2020
0532ec0
changed the order of bytes and submit time in gpu_memcpy_t to match w…
Oct 13, 2020
8bd55e2
adding opencl-queue-map files to remote
Oct 13, 2020
9afb68b
Fix trace data collection and flush activity data when using operatio…
Jokeren Oct 13, 2020
13ea0bb
Merge remote
Jokeren Oct 13, 2020
eae3adc
Code cleanup for opencl
Jokeren Oct 14, 2020
8c0447f
Fix warnings and bugs
Jokeren Oct 14, 2020
9adb6d1
Remove log files generated on jlse
Jokeren Oct 14, 2020
74cbfec
Rename activity-multiplexer to operation-multiplexer
Jokeren Oct 15, 2020
6a68e40
Flush all activities to activity channels before a thread exits
Jokeren Oct 15, 2020
553a4f0
pull from develop
Oct 15, 2020
3ab858b
printf refactoring
dejangrubisic Oct 15, 2020
1975715
merge fixes with develop
Oct 15, 2020
79abee8
Use operation channel to attribute gtpin activities
Jokeren Oct 16, 2020
4bc1d42
merged with develop
Oct 21, 2020
d1d9ee6
Merge branch 'opencl_instrumentation' of https://github.com/HPCToolki…
Oct 21, 2020
ac0fbe0
Adjust gtpin callstack to handle both runtime first and gtpin first c…
Jokeren Oct 22, 2020
4b53905
Remove kernel suffix as gtpin writes a elf per kernel
Jokeren Oct 22, 2020
83a0b1b
Merge develop
Jokeren Oct 23, 2020
abc11e0
Fix bugs
Jokeren Oct 23, 2020
5399efd
Not write gpu elf files to disk
Jokeren Oct 23, 2020
1a57f54
Use ls -s to get file size; change size unit from byte to kb
Jokeren Oct 23, 2020
8864f9b
Revert "Use ls -s to get file size; change size unit from byte to kb"
Jokeren Oct 24, 2020
c0eb7ff
Enable hpcstruct to analyze intel gpubins in parallel: 1. change size…
Jokeren Oct 24, 2020
9115507
Let dyninst return instruction size. Without a dummy buffer, dyninst …
Jokeren Oct 24, 2020
d1e7945
Assign correct instruction arch
Jokeren Oct 24, 2020
da12281
Papi process only gpu on monitor gpu
dejangrubisic Nov 4, 2020
5436ff8
fixing merge conflicts with develop
Nov 14, 2020
8a2511e
adding guards/comments in intel iga code. With this, we can run openc…
Nov 16, 2020
9277ef6
Merged with develop
dejangrubisic Nov 21, 2020
4ae1958
Merged with opencl_develop, debugging in progress
dejangrubisic Nov 21, 2020
80f0aff
checkpoint 1
dejangrubisic Nov 23, 2020
e85d88b
checkpoint 2
dejangrubisic Nov 24, 2020
c7d7ef4
checkpoint 3
dejangrubisic Nov 24, 2020
33eb351
Workable version
dejangrubisic Nov 24, 2020
338717c
papi-c-cupti works
dejangrubisic Nov 25, 2020
e139a47
Deleting redundant function
dejangrubisic Dec 13, 2020
5b37714
Debug 0
dejangrubisic Dec 19, 2020
252e599
Merged with rocm/3.10
dejangrubisic Jan 13, 2021
a38e4f1
support for papi-rocm, problems:loading amd libraries.so
dejangrubisic Jan 13, 2021
69af41d
hpcrun main adopted to libmonitor pthread_create start
dejangrubisic Feb 5, 2021
6e8f38d
merged with develop
dejangrubisic Feb 5, 2021
a2309f5
added HPCRUN_SS_PAPI_C_ROCM to MY_CPP_DEFINES: this is needed for pap…
dejangrubisic Feb 5, 2021
5ec69fb
hpcrun/main.c monitor_init_process splited in two phases
dejangrubisic Feb 6, 2021
96d911f
refactoring monitor_init_process_deferred
dejangrubisic Feb 6, 2021
ce51ae9
hpcrun_prepare_measurement_subsystem must be executed exactly once
dejangrubisic Feb 9, 2021
33084d0
hpcrun_prepare_measurement_subsystem handles forks is_child = true
dejangrubisic Feb 14, 2021
2ed48d2
Tell libmonitor is not tracking monitoring and tracing threads, howev…
dejangrubisic Mar 4, 2021
95d267f
mem_pool_initialized is thread local var, module ignore hpcrun_malloc
dejangrubisic Apr 7, 2021
eb63568
Try to add hip_dev_sync in rocm_papi_read
dejangrubisic May 10, 2021
af1a0fa
rocm-component synchronised on read with hip_dev_sync
dejangrubisic May 18, 2021
06a2b41
Merge branch 'papi_fix_libmonitor_merged_dev' of https://github.com/d…
dejangrubisic Sep 10, 2021
6d4149a
counters works on rocm-4.3.1. ufront
dejangrubisic Sep 15, 2021
fbbc0db
simplify registration of ompt callbacks
jmellorcrummey Oct 7, 2021
54d4ec6
add initial handshake from fnbounds2 to application
jmellorcrummey Oct 8, 2021
5b333e2
1st draft of hpctoolkit ompt support for AMD gpus
jmellorcrummey Oct 21, 2021
ff7e7db
Merge branch 'master' into ompt-amd
jmellorcrummey Oct 29, 2021
bd34914
draft of ompt emi interface implementation
jmellorcrummey Oct 31, 2021
1766f98
add new files for ompt activity processing
jmellorcrummey Oct 31, 2021
2f76238
convert gpu activity correlation_id to 64 bits.
jmellorcrummey Oct 31, 2021
8870f38
Merge branch 'master' into ompt-amd
jmellorcrummey Oct 31, 2021
20b578e
Merge branch 'gpu64' into ompt-amd
jmellorcrummey Nov 1, 2021
fe8dda1
refine gpu memory op to alloc or delete
jmellorcrummey Nov 1, 2021
2126d7d
add new gpu=openmp argument to set up openmp target monitoring
jmellorcrummey Nov 30, 2021
3f08f5c
turn off activity monitoring
jmellorcrummey Nov 30, 2021
227314d
adjust white space
jmellorcrummey Nov 30, 2021
134976c
Merge branch 'master' into ompt-amd
jmellorcrummey Nov 30, 2021
163bf99
Merge branch 'master' of https://github.com/hpctoolkit/hpctoolkit
jmellorcrummey Jan 21, 2022
478b03c
Merge branch 'master' into ompt-amd
jmellorcrummey Jan 21, 2022
df5fee7
compilable version rocprofiler
dejangrubisic Apr 28, 2021
46517d1
OnLoadToolProp added to HPCRUN_NAMES
dejangrubisic May 4, 2021
32713b0
Added -G 'OnLoad' -G 'OnUnloadTool'
dejangrubisic May 27, 2021
e587f29
Prototype support for rocprofiler
mxz297 May 29, 2021
65d0e8e
Enable both roctracer and rocprofiler.
mxz297 Dec 7, 2021
d344958
Start to integrate rocprofiler
mxz297 Dec 18, 2021
e6857db
Set environment variables needed for rocprofiler in hpcrun script
mxz297 Dec 18, 2021
e7024b7
Some code cleanup and add environment variable HPCRUN_PRINT_ROCPROFIL…
mxz297 Dec 18, 2021
30031ce
Refactoring code for roctracer and rocprofiler regarding thread-safety:
mxz297 Dec 22, 2021
0072549
Ensure rocprofiler support to serialize GPU kernel launches
mxz297 Dec 22, 2021
d5b08b6
Refactor gpu_acitvity_t creation and processing to dynamically allocate
mxz297 Jan 5, 2022
b88618e
update to rocm 4.5; no longer works with earlier rocm. (#506)
jmellorcrummey Jan 14, 2022
5788aa2
1. no longer need to disable code object tracking as this option is i…
mxz297 Jan 20, 2022
6008ea3
Use roctracer/rocprofiler to get URIs for AMD GPU binaries
mxz297 Jan 21, 2022
0807053
Remove rocm debug api source files and add a missing environment vari…
mxz297 Jan 21, 2022
5b9dd6c
Fix compilation
mxz297 Jan 21, 2022
d999b53
revert whitespace changes to produce simpler merge
jmellorcrummey Jan 23, 2022
7be834a
remove dead code in ompt-activity-translate.c
jmellorcrummey Jan 24, 2022
3d9994a
selectively issue ompt_trace_flush
jmellorcrummey Jan 24, 2022
3f1817b
not that ompt needs flushing for more than just kernels
jmellorcrummey Jan 24, 2022
867820e
enhance debugging for monitoring GPU intervals
jmellorcrummey Jan 24, 2022
ec481c0
adjust tracing of intervals to avoid interval overlap
jmellorcrummey Jan 24, 2022
9f5e56f
Prototype support for using customized rocprofiler
mxz297 Jan 26, 2022
e82455e
Do no initialize rocprofiler until we find rocprof:: style metrics
mxz297 Jan 31, 2022
dc422ae
fix issues with OMPT device support
jmellorcrummey Feb 7, 2022
12d188e
turn off OMPT logging
jmellorcrummey Feb 7, 2022
8136f07
Refactor thread local memory pool
mxz297 Feb 11, 2022
5c816ad
Refactor AMD and hpcrun initialization interaction
mxz297 Feb 14, 2022
11d4861
Merge AOMP support and rocprofiler support
mxz297 Feb 14, 2022
bdaccde
Fix compilation regarding OMPT placeholders
mxz297 Feb 14, 2022
3075a10
Merge Dejan's work of breaking down and reordering initialization fro…
mxz297 Feb 21, 2022
9c5205d
Merge branch 'develop' into rocprofiler_support
mxz297 Feb 21, 2022
5f72277
No longer need to use customized rocprofiler
mxz297 Feb 21, 2022
f13ad8a
Merge branch 'develop' into rocprofiler_support
mxz297 Feb 22, 2022
511afa5
1. Fix error in Makefile generated by hpcstruct. Double \ is no longe…
mxz297 Feb 23, 2022
111989a
Merge branch 'develop' into rocprofiler_support
mxz297 Feb 23, 2022
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 2 additions & 0 deletions Makefile.in
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -342,6 +342,7 @@ OPT_CUPTI_IFLAGS = @OPT_CUPTI_IFLAGS@
OPT_CUPTI_LDFLAGS = @OPT_CUPTI_LDFLAGS@
OPT_GTPIN = @OPT_GTPIN@
OPT_GTPIN_IFLAGS = @OPT_GTPIN_IFLAGS@
OPT_GTPIN_LDFLAGS = @OPT_GTPIN_LDFLAGS@
OPT_GTPIN_LIBDIR = @OPT_GTPIN_LIBDIR@
OPT_IGC = @OPT_IGC@
OPT_IGC_IFLAGS = @OPT_IGC_IFLAGS@
Expand Down Expand Up @@ -379,6 +380,7 @@ PERFMON_LDFLAGS_STAT = @PERFMON_LDFLAGS_STAT@
PERFMON_LIB = @PERFMON_LIB@
PERF_EVENT_PARANOID = @PERF_EVENT_PARANOID@
RANLIB = @RANLIB@
ROCM_PROFILER_LD_DIR = @ROCM_PROFILER_LD_DIR@
SED = @SED@
SET_MAKE = @SET_MAKE@
SHELL = @SHELL@
Expand Down
Loading