From 48cc7855d52e0024071447aa061c65c6fa204fcd Mon Sep 17 00:00:00 2001
From: Christophe Paquot <christophe.paquot@amd.com>
Date: Tue, 22 Jun 2021 21:12:08 -0700
Subject: [PATCH 1/6] SWDEV-282961: dependency arrows missing

When building the json data flow, from_us_list has (timestamp, stream_id, thread_id).
stream_id used to be interpreted as from_tid and tid as to_tid. But that's not correct.
stream_id is always a destination and tid is the initiator (source).

Change-Id: I2f5bb86a387b4003b17271c90bdf9de4b59a79bf
---
 bin/sqlitedb.py | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/bin/sqlitedb.py b/bin/sqlitedb.py
index dc5358ff..00a7dba2 100644
--- a/bin/sqlitedb.py
+++ b/bin/sqlitedb.py
@@ -159,11 +159,11 @@ def flow_json(self, base_id, from_pid, from_us_list, to_pid, to_us_dict, corr_id
       for ind in range(len(from_us_list)):
         corr_id = corr_id_list[ind] if (len(corr_id_list) != 0) else ind
         if corr_id in to_us_dict:
-          (from_ts, from_tid, to_tid) = from_us_list[ind]
+          (from_ts, stream_id, tid) = from_us_list[ind]
           to_ts = to_us_dict[corr_id]
           if from_ts > to_ts: from_ts = to_ts
-          fd.write(',{"ts":%d,"ph":"s","cat":"DataFlow","id":%d,"pid":%d,"tid":%d,"name":"dep"}\n' % (from_ts, dep_id, from_pid, from_tid))
-          fd.write(',{"ts":%d,"ph":"t","cat":"DataFlow","id":%d,"pid":%d,"tid":%d,"name":"dep"}\n' % (to_ts, dep_id, to_pid, to_tid))
+          fd.write(',{"ts":%d,"ph":"s","cat":"DataFlow","id":%d,"pid":%d,"tid":%d,"name":"dep"}\n' % (from_ts, dep_id, from_pid, tid))
+          fd.write(',{"ts":%d,"ph":"t","cat":"DataFlow","id":%d,"pid":%d,"tid":%d,"name":"dep"}\n' % (to_ts, dep_id, to_pid, stream_id))
           dep_id += 1
 
   def metadata_json(self, jsonfile, sysinfo_file):

From 9f0ca101ec2286827c9b1cd8ccfa2ee3438f6a15 Mon Sep 17 00:00:00 2001
From: AMD <amd@node-bp126-r03c.amd.com>
Date: Wed, 16 Jun 2021 18:33:58 -0500
Subject: [PATCH 2/6] Add support for gfx90a

Merge gfx90a support from the 'amd-npi' branch.

Change-Id: I9b51711ed4a1d2f1ed42ba9b83cb12136be228b8
(cherry picked from commit 4df3e0bd9ae6e5982b43fd2fc3867cf5f0b87a53)
---
 src/core/metrics.h        |  6 +++---
 test/tool/gfx_metrics.xml | 17 +++++++++++++++++
 test/tool/metrics.xml     |  8 ++++++--
 3 files changed, 26 insertions(+), 5 deletions(-)

diff --git a/src/core/metrics.h b/src/core/metrics.h
index a221168a..6eeebee3 100644
--- a/src/core/metrics.h
+++ b/src/core/metrics.h
@@ -196,9 +196,9 @@ class MetricsDict {
       xml_->AddConst("top.const.metric", "SE_NUM", agent_info->se_num);
       ImportMetrics(agent_info, "const");
       agent_name_ = agent_info->name;
-      if (std::string("gfx906") == agent_info->name) {
-        ImportMetrics(agent_info, agent_info->name);
-      } else if (std::string("gfx908") == agent_info->name) {
+      if (std::string("gfx906") == agent_info->name ||
+          std::string("gfx908") == agent_info->name ||
+          std::string("gfx90a") == agent_info->name) {
         ImportMetrics(agent_info, agent_info->name);
       } else {
         agent_name_ = agent_info->gfxip;
diff --git a/test/tool/gfx_metrics.xml b/test/tool/gfx_metrics.xml
index 698826c6..c2a79af2 100644
--- a/test/tool/gfx_metrics.xml
+++ b/test/tool/gfx_metrics.xml
@@ -101,3 +101,20 @@
   <metric name="TCC_EA_RDREQ" block=TCC event=38 descr="Number of TCC/EA read requests (either 32-byte or 64-byte)"></metric>
   <metric name="TCC_EA_RDREQ_32B" block=TCC event=39 descr="Number of 32-byte TCC/EA read requests"></metric>
 </gfx908>
+
+<gfx90a base="gfx908">
+  <metric name="SQ_INSTS_VMEM_WR" block=SQ event=53 descr="Number of VMEM write instructions issued (including FLAT). (per-simd, emulated)"></metric>
+  <metric name="SQ_INSTS_VMEM_RD" block=SQ event=54 descr="Number of VMEM read instructions issued (including FLAT). (per-simd, emulated)"></metric>
+  <metric name="SQ_INSTS_SALU" block=SQ event=56 descr="Number of SALU instructions issued. (per-simd, emulated)"></metric>
+  <metric name="SQ_INSTS_SMEM" block=SQ event=57 descr="Number of SMEM instructions issued. (per-simd, emulated)"></metric>
+  <metric name="SQ_INSTS_FLAT" block=SQ event=58 descr="Number of FLAT instructions issued. (per-simd, emulated)"></metric>
+  <metric name="SQ_INSTS_FLAT_LDS_ONLY" block=SQ event=59 descr="Number of FLAT instructions issued that read/wrote only from/to LDS (only works if EARLY_TA_DONE is enabled). (per-simd, emulated)"></metric>
+  <metric name="SQ_INSTS_LDS" block=SQ event=60 descr="Number of LDS instructions issued (including FLAT). (per-simd, emulated)"></metric>
+  <metric name="SQ_INSTS_GDS" block=SQ event=61 descr="Number of GDS instructions issued. (per-simd, emulated)"></metric>
+
+  <metric name="SQ_WAIT_INST_LDS" block=SQ event=91 descr="Number of wave-cycles spent waiting for LDS instruction issue. In units of 4 cycles. (per-simd, nondeterministic)"></metric>
+  <metric name="SQ_ACTIVE_INST_VALU" block=SQ event=99 descr="regspec 71? Number of cycles the SQ instruction arbiter is working on a VALU instruction. (per-simd, nondeterministic)"></metric>
+  <metric name="SQ_INST_CYCLES_SALU" block=SQ event=112 descr="Number of cycles needed to execute non-memory read scalar operations. (per-simd, emulated)"></metric>
+  <metric name="SQ_THREAD_CYCLES_VALU" block=SQ event=113 descr="Number of thread-cycles used to execute VALU operations (similar to INST_CYCLES_VALU but multiplied by # of active threads). (per-simd)"></metric>
+  <metric name="SQ_LDS_BANK_CONFLICT" block=SQ event=121 descr="Number of cycles LDS is stalled by bank conflicts. (emulated)"></metric>
+</gfx90a>
diff --git a/test/tool/metrics.xml b/test/tool/metrics.xml
index c340a439..a920ff04 100644
--- a/test/tool/metrics.xml
+++ b/test/tool/metrics.xml
@@ -65,12 +65,16 @@
   <metric name="TCC_WRREQ_STALL_max" expr=max(TCC_EA_WRREQ_STALL,32) descr="Number of cycles a write request was stalled. Max over TCC instances."></metric>
 </gfx908_expr>
 
+<gfx90a_expr base="gfx908_expr"></gfx90a_expr>
+
 <gfx8 base="gfx8_expr"></gfx8>
 <gfx9 base="gfx9_expr"></gfx9>
-# VG20
+# Vega20
 <gfx906 base="gfx906_expr"></gfx906>
-# MI100
+# Arcturus
 <gfx908 base="gfx908_expr"></gfx908>
+# Aldebaran
+<gfx90a base="gfx90a_expr"></gfx90a>
 
 <global>
   # GPUBusy         The percentage of time GPU was busy.

From a369af3049bc3a6d2faca537f4e47d5a046f01cf Mon Sep 17 00:00:00 2001
From: rachida <rachida.kebichi@amd.com>
Date: Tue, 18 May 2021 20:11:48 -0400
Subject: [PATCH 3/6] SWDEV-282961 Skip barrier events. Process hipMemSet
 events

Marker events inside hcc_ops_trace.txt are from barriers so they are not meant to be stored in ops_patch_data map.
Added support for hipMemset events which are a kind of memory copy.

Change-Id: I213fe959bcd35ff0371613ba5bffd95bc53e06b5
(cherry picked from commit caa5f323007734fd0b14b3fa49618a5d7cc7acdd)
---
 bin/tblextr.py | 15 +++++++++------
 1 file changed, 9 insertions(+), 6 deletions(-)

diff --git a/bin/tblextr.py b/bin/tblextr.py
index 61644e2a..deafb199 100755
--- a/bin/tblextr.py
+++ b/bin/tblextr.py
@@ -354,7 +354,7 @@ def fill_api_db(table_name, db, indir, api_name, api_pid, dep_pid, dep_list, dep
   copy_index = 0
 
   ptrn_val = re.compile(r'(\d+):(\d+) (\d+):(\d+) ([^\(]+)(\(.*)$')
-  hip_mcopy_ptrn = re.compile(r'hipMemcpy')
+  hip_mcopy_ptrn = re.compile(r'hipMemcpy|hipMemset')
   hip_wait_event_ptrn =  re.compile(r'WaitEvent')
   hip_sync_event_ptrn = re.compile(r'hipStreamSynchronize')
   hip_sync_dev_event_ptrn = re.compile(r'hipDeviceSynchronize')
@@ -430,7 +430,6 @@ def fill_api_db(table_name, db, indir, api_name, api_pid, dep_pid, dep_list, dep
         if corr_id == 0: corr_id = record_id
 
         rec_vals.append(corr_id)
-
         # extracting/converting stream id
         (stream_id, stream_found) = get_field(record_args, 'stream')
         if stream_found == 0:
@@ -489,9 +488,6 @@ def fill_api_db(table_name, db, indir, api_name, api_pid, dep_pid, dep_list, dep
           mcopy_found = 1
           op_found = 1
 
-        if op_found:
-          ops_patch_data[(corr_id, proc_id)] = (thread_id, stream_id, kernel_str)
-
         # HIP WaitEvent API
         if wait_event_ptrn.search(record_name):
           op_found = 1
@@ -505,6 +501,9 @@ def fill_api_db(table_name, db, indir, api_name, api_pid, dep_pid, dep_list, dep
           hsa_patch_data[(copy_index, proc_id)] = thread_id
           copy_index += 1
 
+        if op_found:
+          ops_patch_data[(corr_id, proc_id)] = (thread_id, stream_id, kernel_str)
+
         if op_found:
           op_found = 0
           beg_ns = int(rec_vals[0])
@@ -650,6 +649,7 @@ def fill_ops_db(kernel_table_name, mcopy_table_name, db, indir):
         proc_id = int(m.group(3))
 
         # checking name for memcopy pattern
+        is_barrier = 0
         if ptrn_mcopy.search(name):
           rec_table_name = mcopy_table_name
           table_handle = mcopy_table_handle
@@ -664,6 +664,7 @@ def fill_ops_db(kernel_table_name, mcopy_table_name, db, indir):
 
           if ptrn_barrier.search(name):
             name = '"<barrier packet>"'
+            is_barrier = 1
 
         thread_id = 0
         stream_id = 0
@@ -671,7 +672,8 @@ def fill_ops_db(kernel_table_name, mcopy_table_name, db, indir):
           (thread_id, stream_id, name_patch) = ops_patch_data[(corr_id, proc_id)]
           if name_patch != '': name = name_patch
         else:
-          fatal("hcc ops data not found: '" + record + "', " + str(corr_id) + ", " + str(proc_id))
+          if is_barrier: continue
+          else: fatal("hcc ops data not found: '" + record + "', " + str(corr_id) + ", " + str(proc_id))
 
         # activity record
         rec_vals[4] = name                       # Name
@@ -878,3 +880,4 @@ def fill_ops_db(kernel_table_name, mcopy_table_name, db, indir):
 
 sys.exit(0)
 #############################################################
+

From 2f189791a5cf6a6beb0aea9572d8760f792f0b32 Mon Sep 17 00:00:00 2001
From: Icarus Sparry <icarus.sparry@amd.com>
Date: Wed, 7 Jul 2021 15:49:02 +0000
Subject: [PATCH 4/6] Add dependency on rocm-core

Signed-off-by: Icarus Sparry <icarus.sparry@amd.com>
Change-Id: Icb935e9230888fd005d9ca3617e28f6173173cc8
---
 CMakeLists.txt | 9 +++++++--
 1 file changed, 7 insertions(+), 2 deletions(-)

diff --git a/CMakeLists.txt b/CMakeLists.txt
index c34f7cc9..e47f06df 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -204,7 +204,7 @@ else()
 endif()
 message ( "Using CPACK_DEBIAN_PACKAGE_RELEASE ${CPACK_DEBIAN_PACKAGE_RELEASE}" )
 set ( CPACK_DEBIAN_FILE_NAME "DEB-DEFAULT" )
-set ( CPACK_DEBIAN_PACKAGE_DEPENDS "hsa-rocr-dev" )
+set ( CPACK_DEBIAN_PACKAGE_DEPENDS "hsa-rocr-dev, rocm-core" )
 ## Process the Debian install/remove scripts to update the CPACK variables
 configure_file ( ${CMAKE_CURRENT_SOURCE_DIR}/DEBIAN/postinst.in DEBIAN/postinst @ONLY )
 configure_file ( ${CMAKE_CURRENT_SOURCE_DIR}/DEBIAN/prerm.in DEBIAN/prerm @ONLY )
@@ -229,12 +229,17 @@ if ( PROC_RESULT EQUAL "0" AND NOT EVAL_RESULT STREQUAL "" )
   string ( APPEND CPACK_RPM_PACKAGE_RELEASE "%{?dist}" )
 endif()
 set ( CPACK_RPM_FILE_NAME "RPM-DEFAULT" )
-set ( CPACK_RPM_PACKAGE_DEPENDS "hsa-rocr-dev" )
+set ( CPACK_RPM_PACKAGE_REQUIRES "hsa-rocr-dev, rocm-core" )
 ## Process the Rpm install/remove scripts to update the CPACK variables
 configure_file ( "${CMAKE_CURRENT_SOURCE_DIR}/RPM/post.in" RPM/post @ONLY )
 configure_file ( "${CMAKE_CURRENT_SOURCE_DIR}/RPM/postun.in" RPM/postun @ONLY )
 
 set ( CPACK_RPM_POST_INSTALL_SCRIPT_FILE "${CMAKE_CURRENT_BINARY_DIR}/RPM/post" )
 set ( CPACK_RPM_POST_UNINSTALL_SCRIPT_FILE "${CMAKE_CURRENT_BINARY_DIR}/RPM/postun" )
+# Remove dependency on rocm-core if -DROCM_DEP_ROCMCORE=ON not given to cmake
+if(NOT ROCM_DEP_ROCMCORE)
+    string(REGEX REPLACE ",? ?rocm-core" "" CPACK_RPM_PACKAGE_REQUIRES ${CPACK_RPM_PACKAGE_REQUIRES})
+    string(REGEX REPLACE ",? ?rocm-core" "" CPACK_DEBIAN_PACKAGE_DEPENDS ${CPACK_DEBIAN_PACKAGE_DEPENDS})
+endif()
 
 include ( CPack )

From b71b5414d1e1d52d2dfe30b7216fc809613a0f44 Mon Sep 17 00:00:00 2001
From: Ammar ELWazir <Ammar.ELWazir@amd.com>
Date: Fri, 21 Jan 2022 12:05:10 -0600
Subject: [PATCH 5/6] SWDEV-318551: Adding License file for profiler

Making the new License file, Adding support in the CMakeLists.txt

Change-Id: I785035a780fbfc59951fc27d45f9c1869ffb4fb3
---
 CMakeLists.txt | 5 +++++
 LICENSE        | 4 ++--
 2 files changed, 7 insertions(+), 2 deletions(-)

diff --git a/CMakeLists.txt b/CMakeLists.txt
index 68f9b86a..dfb069be 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -191,6 +191,9 @@ set ( CPACK_PACKAGE_CONTACT "ROCm Profiler Support <dl.ROCm-Profiler.support@amd
 set ( CPACK_PACKAGE_DESCRIPTION_SUMMARY "ROCPROFILER library for AMD HSA runtime API extension support" )
 set ( CPACK_RESOURCE_FILE_LICENSE "${CMAKE_CURRENT_SOURCE_DIR}/LICENSE" )
 
+# Install license file
+install(FILES ${CPACK_RESOURCE_FILE_LICENSE} DESTINATION share/doc/${CPACK_PACKAGE_NAME})
+
 if ( DEFINED ENV{ROCM_LIBPATCH_VERSION} )
   set ( CPACK_PACKAGE_VERSION "${CPACK_PACKAGE_VERSION}.$ENV{ROCM_LIBPATCH_VERSION}" )
   message ( "Using CPACK_PACKAGE_VERSION ${CPACK_PACKAGE_VERSION}" )
@@ -218,6 +221,8 @@ else()
 endif()
 message ( "Using CPACK_RPM_PACKAGE_RELEASE ${CPACK_RPM_PACKAGE_RELEASE}" )
 
+set( CPACK_RPM_PACKAGE_LICENSE "MIT" )
+
 ## 'dist' breaks manual builds on debian systems due to empty Provides
 execute_process( COMMAND rpm --eval %{?dist}
                  RESULT_VARIABLE PROC_RESULT
diff --git a/LICENSE b/LICENSE
index 8384c985..d5c431d7 100644
--- a/LICENSE
+++ b/LICENSE
@@ -1,4 +1,4 @@
-Copyright (c) 2018 Advanced Micro Devices, Inc. All rights reserved.
+Copyright (c) 2018-2022 Advanced Micro Devices, Inc. All rights reserved.
 [MITx11 License]
 
 Permission is hereby granted, free of charge, to any person obtaining a copy
@@ -17,4 +17,4 @@ FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL THE
 AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
 LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
 OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
-THE SOFTWARE.
+THE SOFTWARE.
\ No newline at end of file

From 5bc8e990b7b3f6bfa07ab91ee8625564a5138d8b Mon Sep 17 00:00:00 2001
From: Aleksey <kikimych@gmail.com>
Date: Thu, 31 Mar 2022 09:04:18 +0300
Subject: [PATCH 6/6] gfx908 counts only half of memmory controlles in
 L2CacheHit metric

---
 test/tool/metrics.xml | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/test/tool/metrics.xml b/test/tool/metrics.xml
index a920ff04..42303bfc 100644
--- a/test/tool/metrics.xml
+++ b/test/tool/metrics.xml
@@ -39,6 +39,9 @@
 </gfx9_expr>
 
 <gfx906_expr base="gfx9_expr">
+  <metric name="TCC_HIT_sum" expr=sum(TCC_HIT,16) descr="Number of cache hits. Sum over TCC instances."></metric>
+  <metric name="TCC_MISS_sum" expr=sum(TCC_MISS,16) descr="Number of cache misses. Sum over TCC instances."></metric>
+
   # EA1
   <metric name="TCC_EA1_RDREQ_32B_sum" expr=sum(TCC_EA1_RDREQ_32B,16) descr="Number of 32-byte TCC/EA read requests. Sum over TCC EA1s."></metric>
   <metric name="TCC_EA1_RDREQ_sum" expr=sum(TCC_EA1_RDREQ,16) descr="Number of TCC/EA read requests (either 32-byte or 64-byte). Sum over TCC EA1s."></metric>
@@ -200,7 +203,7 @@
   <metric
     name="L2CacheHit"
     descr="The percentage of fetch, write, atomic, and other instructions that hit the data in L2 cache. Value range: 0% (no hit) to 100% (optimal)."
-    expr=100*sum(TCC_HIT,16)/(sum(TCC_HIT,16)+sum(TCC_MISS,16))
+    expr=100*TCC_HIT_sum/(TCC_HIT_sum+TCC_MISS_sum)
   ></metric>
 
   # MemUnitBusy     The percentage of GPUTime the memory unit is active. The result includes the stall time (MemUnitStalled). This is measured with all extra fetches and writes and any cache or memory effects taken into account. Value range: 0% to 100% (fetch-bound).