[Draft] Multi-Group Scaling for CachePool

.gitignore

@@ -1,34 +1,27 @@
-work-dpi/
-dw*
-site/*
-*.ucdb
-work/*
+# Simulation
 transcript
 *.ini
 *.wlf
-*.vstf
-wlft*
-*nfs*
-*.sig
-*.dtb
-*.dasm
-obj_dir/*
-/tmp*
-*.dasm
-build/
-trace_*
-mems
-*log
-.bender
-.dvt
-.settings
-*.o
-install
-.projectBender.lock
-Bender.lock
-logs
-bin
-hardware/deps/*
 *.tdb
+# Toolchain
+install
+# Vis4Mesh (pinned clone, see util/vis4mesh.version -- not tracked here)
+util/vis4mesh
+# Utilization logs
 util/lint/sg_projects
 util/lint/tmp
+# Dependency files
+.bender
+hardware/deps/*
+# Bootrom auto-generated files
+hardware/bootrom/bootdata.cc
+hardware/bootrom/bootdata_bootrom.cc
+hardware/bootrom/bootrom.bin
+hardware/bootrom/bootrom.dump
+hardware/bootrom/bootrom.elf
+hardware/bootrom/bootrom.sv
+# Configuration auto-generated files
+config/cachepool.hjson
+*noc_profiling
+# CI venv (created by iis-env.sh)
+/cachepool

.gitlab-ci.yml

@@ -8,41 +8,91 @@ variables:
   GIT_SUBMODULE_STRATEGY: none
   ROOT_DIR: '$CI_PROJECT_DIR'
   APPS: "tests"
-  PATH: '/home/gitlabci/.cargo/bin:/usr/local/bin:/usr/bin:/usr/sbin:/sbin:/usr/local/condor/bin:/usr/sepp/bin:$CI_PROJECT_DIR/install/verilator/bin:/home/gitlabci/.local/bin'
   OBJCACHE: ''
   CC:  '/usr/pack/gcc-11.2.0-af/linux-x64/bin/gcc'
   CXX: '/usr/pack/gcc-11.2.0-af/linux-x64/bin/g++'
   CMAKE: 'cmake-3.28.3'
   python: 'python3'
   python3: 'python3'
+  # Config to build and test
+  CI_CONFIG: 'cachepool_fpu_4g'
+  SW_PREFIX: 'test-cachepool-'
 
 default:
-  tags: [dolent]
+  tags: [shared]
 
 stages:
   - build
+  - test
 
-.base:
-  artifacts:
-    when: always
-    expire_in: 1 day
-
-build-vsim:
-  extends: .base
+# ---------------------------------------------------------------------------
+# Build stage: compile RTL and software for CI_CONFIG.
+# Sources iis-env.sh, which points the toolchain at the prebuilt install dir
+# (replaces `make quick-tool`'s symlink) and sets up the venv for code-gen deps;
+# test-stage jobs run prebuilt artifacts and don't need either, so iis-env.sh is
+# only sourced here.
+# ---------------------------------------------------------------------------
+build:
   stage: build
-  timeout: 5h
   script:
     - echo "Using CC=$CC"
     - echo "Using CXX=$CXX"
     - test -x "$CC"
     - test -x "$CXX"
-    - make quick-tool
+    - source iis-env.sh
     - make init
     - make dram-build
-    - cd util/auto-benchmark
-    - chmod +x ./run_ci.sh
-    - ./run_ci.sh
+    - make clean generate update-floonoc bootrom vsim sw config=$CI_CONFIG DEBUG=0 noc_profiling=0
+  artifacts:
+    when: always
+    expire_in: 2h
+    paths:
+      # QuestaSim compiled work library
+      - sim/work/
+      # vsim wrapper scripts (exclude sim/bin/logs/ — not needed by test jobs)
+      - sim/bin/cachepool_cluster.vsim
+      - sim/bin/cachepool_cluster.vsim.gui
+      # DPI shared library
+      - sim/work-dpi/
+      # Software binaries for all kernels
+      - software/build/CachePoolTests/
+      # DRAMSys shared libraries and config files (referenced by vsim at runtime)
+      - hardware/deps/dram_rtl_sim/dramsys_lib/DRAMSys/build/lib/
+      - hardware/deps/dram_rtl_sim/dramsys_lib/DRAMSys/configs/
 
+# ---------------------------------------------------------------------------
+# Test stage: run each kernel in parallel on a separate runner.
+# Each job downloads the build artifacts, runs one simulation, and checks
+# the output log for failures.
+# ---------------------------------------------------------------------------
+test:
+  stage: test
+  needs: [build]
+  parallel:
+    matrix:
+      - KERNEL:
+          - spin-lock
+          - load-store
+          - fdotp-32b_M32768
+          - gemv_M512_N128_K32
+          - fmatmul-32b_M64_N64_K64
+          - fft-32b_M1024_N16
+          - byte-enable
+          - multi_producer_single_consumer_double_linked_list_M1_N1350_K10
+  script:
+    # The vsim script writes a .rtlbinary marker here; monitor logs need the
+    # core/noc/others subfolders. Build artifacts exclude sim/bin/logs/, so
+    # these must be recreated in the test job too.
+    - mkdir -p sim/bin/logs sim/bin/logs/core sim/bin/logs/noc sim/bin/logs/others
+    - chmod +x sim/bin/cachepool_cluster.vsim
+    - BIN="${SW_PREFIX}${KERNEL}"
+    - sim/bin/cachepool_cluster.vsim software/build/CachePoolTests/$BIN 2>&1 | tee test_${KERNEL}.log
+    - python3 util/auto-benchmark/check-ci.py test_${KERNEL}.log
   artifacts:
+    when: always
+    expire_in: 1 week
     paths:
-      - util/auto-benchmark/logs
+      # Full simulation log
+      - test_*.log
+      # Performance-monitor trace files written by the simulator
+      - sim/bin/logs/

Bender.lock

@@ -16,17 +16,34 @@ packages:
     - common_verification
     - tech_cells_generic
   axi_riscv_atomics:
-    revision: 97dcb14ef057cbe5bd70dda2060b5bb9e7e04c6d
-    version: 0.7.0
+    revision: 97a1dd2ac643c276880420a0cf8eea697f228aa9
+    version: 0.8.3
     source:
       Git: https://github.com/pulp-platform/axi_riscv_atomics.git
     dependencies:
     - axi
     - common_cells
     - common_verification
+  axi_stream:
+    revision: 54891ff40455ca94a37641b9da4604647878cc07
+    version: 0.1.1
+    source:
+      Git: https://github.com/pulp-platform/axi_stream.git
+    dependencies:
+    - common_cells
+  cluster_icache:
+    revision: ce0ed94a5b95f5c76b9fa51940303fcce53f56e5
+    version: null
+    source:
+      Git: https://github.com/pulp-platform/cluster_icache.git
+    dependencies:
+    - axi
+    - common_cells
+    - scm
+    - tech_cells_generic
   common_cells:
-    revision: 9afda9abb565971649c2aa0985639c096f351171
-    version: 1.38.0
+    revision: 9ca8a7655f741e7dd5736669a20a301325194c28
+    version: 1.39.0
     source:
       Git: https://github.com/pulp-platform/common_cells.git
     dependencies:
@@ -45,8 +62,27 @@ packages:
       Git: https://github.com/pulp-platform/dram_rtl_sim.git
     dependencies:
     - axi
+  floo_noc:
+    revision: 8413d9b044a51f958a7d4ef261e2b01e73db4544
+    version: null
+    source:
+      Git: https://github.com/pulp-platform/FlooNoC.git
+    dependencies:
+    - axi
+    - axi_riscv_atomics
+    - common_cells
+    - common_verification
+    - floo_noc_pd
+    - fpnew
+    - idma
+  floo_noc_pd:
+    revision: null
+    version: null
+    source:
+      Path: hardware/deps/floo_noc/./pd
+    dependencies: []
   fpnew:
-    revision: a8e0cba6dd50f357ece73c2c955d96efc3c6c315
+    revision: e5aa6a01b5bbe1675c3aa8872e1203413ded83d1
     version: null
     source:
       Git: https://github.com/pulp-platform/cvfpu.git
@@ -61,14 +97,16 @@ packages:
     dependencies:
     - common_cells
   idma:
-    revision: b31e8f019c657eff4126bc789f0336d403da6766
-    version: 0.4.2
+    revision: 28a36e5e07705549e59fc33db96ab681bc1ca88e
+    version: 0.6.5
     source:
       Git: https://github.com/pulp-platform/iDMA.git
     dependencies:
     - axi
+    - axi_stream
     - common_cells
     - common_verification
+    - obi
     - register_interface
   insitu-cache:
     revision: fa761ddebc946f9b46509d84945bf41ee1a9ec49
@@ -79,6 +117,14 @@ packages:
     - axi
     - common_cells
     - register_interface
+  obi:
+    revision: 0155fc34e900c7c884e081c0a1114a247937ff69
+    version: 0.1.7
+    source:
+      Git: https://github.com/pulp-platform/obi.git
+    dependencies:
+    - common_cells
+    - common_verification
   register_interface:
     revision: 146501d80052b61475cdc333d3aab4cd769fd5dc
     version: 0.3.9
@@ -96,8 +142,15 @@ packages:
     dependencies:
     - common_cells
     - tech_cells_generic
+  scm:
+    revision: 1976c7efb4979271eee2abe262fde0f9a20e2557
+    version: 1.2.1
+    source:
+      Git: https://github.com/pulp-platform/scm.git
+    dependencies:
+    - tech_cells_generic
   spatz:
-    revision: ed25c78dd72d839db8141287f9516d78ee399b93
+    revision: 3a6e6bed2a774a449997e426d893d15f84420518
     version: null
     source:
       Git: https://github.com/pulp-platform/spatz.git

Bender.yml

@@ -10,13 +10,14 @@ dependencies:
   axi_riscv_atomics:   { git: "https://github.com/pulp-platform/axi_riscv_atomics.git",  version: 0.7.0                }
   common_cells:        { git: "https://github.com/pulp-platform/common_cells.git",       version: 1.28.0               }
   FPnew:               { git: "https://github.com/pulp-platform/cvfpu.git",              rev:     pulp-v0.1.3          }
-  idma:                { git: "https://github.com/pulp-platform/iDMA.git",               version: 0.4.2                }
   register_interface:  { git: "https://github.com/pulp-platform/register_interface.git", version: 0.3.8                }
   riscv-dbg:           { git: "https://github.com/pulp-platform/riscv-dbg.git",          version: 0.7.0                }
   tech_cells_generic:  { git: "https://github.com/pulp-platform/tech_cells_generic.git", version: 0.2.11               }
   Insitu-Cache:        { git: "https://github.com/pulp-platform/Insitu-Cache.git",       rev:     zexin/cachepool_dev  }
-  spatz:               { git: "https://github.com/pulp-platform/spatz.git",              rev:     cachepool-32b        }
   dram_rtl_sim:        { git: "https://github.com/pulp-platform/dram_rtl_sim.git",       rev:     cachepool            }
+  floo_noc:            { git: "https://github.com/pulp-platform/FlooNoC.git",            rev:     cachepool            }
+  cluster_icache:      { git: "https://github.com/pulp-platform/cluster_icache.git",     rev:     main                 }
+  spatz:               { git: "https://github.com/pulp-platform/spatz.git",              rev:     cachepool-32b        }
 
 workspace:
   checkout_dir: "./hardware/deps"
@@ -25,6 +26,8 @@ export_include_dirs:
 
 sources:
   # Level 0
+  - hardware/src/reqrsp_to_reg.sv
+  - hardware/src/reqrsp_w2n_converter.sv
   - hardware/src/reqrsp_xbar.sv
   - hardware/src/tcdm_cache_interco.sv
   - hardware/src/tcdm_id_remapper.sv
@@ -35,16 +38,20 @@ sources:
   - hardware/cachepool_peripheral/cachepool_peripheral.sv
   # Bootrom
   - hardware/bootrom/bootrom.sv
-  # Barrier
-  - hardware/src/cachepool_tile_barrier.sv
-  - hardware/src/cachepool_cluster_barrier.sv
   # Level 1
+  - hardware/generated/floo_cachepool_noc_pkg.sv
   - hardware/src/cachepool_pkg.sv
   - hardware/src/cachepool_cc.sv
+  # Barrier
+  - hardware/src/cachepool_tile_barrier.sv
+  - hardware/src/cachepool_cluster_barrier.sv
+  # ICache
+  - hardware/src/axi_hier_interco.sv
   # Level 2
   - hardware/src/cachepool_tile.sv
   # Level 3
   - hardware/src/cachepool_group.sv
+  - hardware/src/cachepool_group_noc_wrapper.sv
 
   - hardware/src/cachepool_cluster.sv
   # Level 4
@@ -53,4 +60,6 @@ sources:
   - target: cachepool_test
     files:
     - hardware/tb/axi_uart.sv
+    - hardware/tb/cachepool_monitor.sv
+    - hardware/tb/cachepool_noc_profiling.sv
     - hardware/tb/tb_cachepool.sv