per stream stats (accel-sim#326)

* Passed cuda_stream_id from accelsim to gpgpusim

* Fixed for (auto k : kernels_info) in main() for multi-stream app

* Passed streamID of last finished kernel to gpgpu_sim::print_stats

* checkout correct gpgpu-sim branch for multi stream stats

* Move Jenkins to bak

* Adding merge_group trigger to worklow

* rename jenkinsfile

* Automated clang-format

* update formatter

* correct format script path

---------

Co-authored-by: Justin Qiao <[email protected]>
Co-authored-by: JRPan <[email protected]>

Author: JRPan

.github/workflows/short-tests.yml

@@ -48,29 +48,22 @@ jobs:
       - name: Run Simulation
         run: echo "skipped Tracer-Simulation. Will perform in merge queue"
   format-code:
-    runs-on: ubuntu-latest
+    if: github.event_name == 'pull_request'
+    runs-on: tgrogers-raid
     needs: [SASS-Simulation, PTX-Simulation, Tracer-Tool]
-
-    permissions:
-      # Give the default GITHUB_TOKEN write permission to commit and push the
-      # added or changed files to the repository.
-      contents: write
-
     steps:
       - uses: actions/checkout@v4
-      # Other steps that change files in the repository go here
-      # …
+        with:
+          ref: ${{github.event.pull_request.head.ref}}
+          repository: ${{github.event.pull_request.head.repo.full_name}}
+          ssh-key: ''
+
       - name: Run clang-format
         run: |
-          sudo apt-get install -y clang-format
-          ./gpu-simulator/format-code.sh
-          ./util/tracer_nvbit/tracer_tool/format-code.sh
-
-      - uses: stefanzweifel/git-auto-commit-action@v5
-        with:
-          # Optional. Commit message for the created commit.
-          # Defaults to "Apply automatic changes"
-          commit_message: Automated clang-format
-          # Optional. Option used by `git-status` to determine if the repository is 
-          # dirty. See https://git-scm.com/docs/git-status#_options
-          status_options: '--untracked-files=no'
+          git config user.name "purdue-jenkins"
+          git config user.email "[email protected]"
+          git remote set-url origin [email protected]:${{github.event.pull_request.head.repo.full_name}}
+          git remote -v
+          /bin/bash ./gpu-simulator/format-code.sh
+          /bin/bash ./util/tracer_nvbit/tracer_tool/format-code.sh
+          if git status --untracked-files=no | grep -q "nothing to commit"; then echo "No changes to commit."; else git commit -a -m "Automated Format"; git push; fi

gpu-simulator/accel-sim.cc

@@ -58,7 +58,9 @@ void accel_sim_framework::simulation_loop() {
       if (!stream_busy && m_gpgpu_sim->can_start_kernel() &&
           !k->was_launched()) {
         std::cout << "launching kernel name: " << k->get_name()
-                  << " uid: " << k->get_uid() << std::endl;
+                  << " uid: " << k->get_uid()
+                  << " cuda_stream_id: " << k->get_cuda_stream_id()
+                  << std::endl;
         m_gpgpu_sim->launch(k);
         k->set_launched();
         busy_streams.push_back(k->get_cuda_stream_id());
@@ -119,12 +121,14 @@ void accel_sim_framework::parse_commandlist() {
 
 void accel_sim_framework::cleanup(unsigned finished_kernel) {
   trace_kernel_info_t *k = NULL;
+  unsigned long long finished_kernel_cuda_stream_id = -1;
   for (unsigned j = 0; j < kernels_info.size(); j++) {
     k = kernels_info.at(j);
     if (k->get_uid() == finished_kernel ||
         m_gpgpu_sim->cycle_insn_cta_max_hit() || !m_gpgpu_sim->active()) {
       for (unsigned int l = 0; l < busy_streams.size(); l++) {
         if (busy_streams.at(l) == k->get_cuda_stream_id()) {
+          finished_kernel_cuda_stream_id = k->get_cuda_stream_id();
           busy_streams.erase(busy_streams.begin() + l);
           break;
         }
@@ -138,7 +142,7 @@ void accel_sim_framework::cleanup(unsigned finished_kernel) {
     }
   }
   assert(k);
-  m_gpgpu_sim->print_stats();
+  m_gpgpu_sim->print_stats(finished_kernel_cuda_stream_id);
 }
 
 unsigned accel_sim_framework::simulate() {

gpu-simulator/accel-sim.h

@@ -67,7 +67,7 @@ class accel_sim_framework {
   unsigned window_size;
   unsigned commandlist_index;
 
-  std::vector<unsigned long> busy_streams;
+  std::vector<unsigned long long> busy_streams;
   std::vector<trace_kernel_info_t *> kernels_info;
   std::vector<trace_command> commandlist;
 

gpu-simulator/trace-driven/trace_driven.cc

@@ -94,7 +94,8 @@ trace_kernel_info_t::trace_kernel_info_t(dim3 gridDim, dim3 blockDim,
                                          trace_parser *parser,
                                          class trace_config *config,
                                          kernel_trace_t *kernel_trace_info)
-    : kernel_info_t(gridDim, blockDim, m_function_info) {
+    : kernel_info_t(gridDim, blockDim, m_function_info,
+                    kernel_trace_info->cuda_stream_id) {
   m_parser = parser;
   m_tconfig = config;
   m_kernel_trace_info = kernel_trace_info;

gpu-simulator/trace-driven/trace_driven.h

@@ -93,7 +93,7 @@ class trace_kernel_info_t : public kernel_info_t {
   void get_next_threadblock_traces(
       std::vector<std::vector<inst_trace_t> *> threadblock_traces);
 
-  unsigned long get_cuda_stream_id() {
+  unsigned long long get_cuda_stream_id() {
     return m_kernel_trace_info->cuda_stream_id;
   }
 

gpu-simulator/trace-parser/trace_parser.cc

@@ -379,7 +379,7 @@ kernel_trace_t *trace_parser::parse_kernel_info(
       } else if (string1 == "nregs") {
         sscanf(line.c_str(), "-nregs = %d", &kernel_info->nregs);
       } else if (string1 == "cuda" && string2 == "stream") {
-        sscanf(line.c_str(), "-cuda stream id = %lu",
+        sscanf(line.c_str(), "-cuda stream id = %llu",
                &kernel_info->cuda_stream_id);
       } else if (string1 == "binary" && string2 == "version") {
         sscanf(line.c_str(), "-binary version = %d",

gpu-simulator/trace-parser/trace_parser.h

@@ -88,7 +88,7 @@ struct kernel_trace_t {
   unsigned tb_dim_z;
   unsigned shmem;
   unsigned nregs;
-  unsigned long cuda_stream_id;
+  unsigned long long cuda_stream_id;
   unsigned binary_verion;
   unsigned enable_lineinfo;
   unsigned trace_verion;