CP regfile Fmax, early-Z fix + default on, asic_gate 4x4 top/gfx, CI shards + ccache

.github/workflows/ci.yml

@@ -23,8 +23,11 @@ on:
         description: "comma list: smoke,full,nightly (blank = all)"
         default: ""
 
-env:
-  CCACHE_DISABLE: 1
+# ccache stays on: a hosted runner starts every job with an empty cache, so it
+# cannot serve an object built against other headers (the stale-object class
+# seen on persistent machines), and within a job it turns a cell's repeated sim
+# builds and its config-invariant objects (Verilator runtime, host driver) into
+# cache hits.
 
 jobs:
   # ---------------------------------------------------------------------------
@@ -196,16 +199,22 @@ jobs:
 
   # ---------------------------------------------------------------------------
   # tests — one cell per (category, driver, xlen); pytest selects via markers.
+  # A cell whose recorded runtime exceeds ci/testcase.py's SHARD_BUDGET_S is
+  # planned as several shards (`shard` of `shards`), balanced on the per-case
+  # times in ci/runtimes.json, so every cell fits the 3 h timeout.
   # ---------------------------------------------------------------------------
   tests:
     needs: [plan, build]
     if: needs.plan.outputs.run == 'true'
+    name: tests (${{ matrix.category }}, ${{ matrix.driver }}, ${{ matrix.xlen }}${{ matrix.shards && format(', {0}/{1}', matrix.shard, matrix.shards) || '' }})
     strategy:
       fail-fast: false
       matrix:
         include: ${{ fromJSON(needs.plan.outputs.cells) }}
     runs-on: ubuntu-22.04
-    timeout-minutes: 300
+    timeout-minutes: 180
+    env:
+      CELL: ${{ matrix.category }}-${{ matrix.driver }}-${{ matrix.xlen }}${{ matrix.shards && format('-{0}of{1}', matrix.shard, matrix.shards) || '' }}
     steps:
       - uses: actions/checkout@v4
       - name: Setup Vortex
@@ -227,23 +236,18 @@ jobs:
         run: |
           cd build${{ matrix.xlen }}
           chmod -R +x .
-          EXPR="${{ matrix.category }}"
-          [ "${{ matrix.driver }}" != "host" ] && EXPR="$EXPR and ${{ matrix.driver }}"
-          # A cross-cutting check (model_parity, perf_gate) has its own dedicated
-          # cell whose `-m <check>` sweeps every such case catalog-wide; exclude
-          # those cases from every OTHER category's cell so each runs once, not
-          # twice. The check list comes from the planner, never a copy of it.
-          for c in $(python3 ci/testcase.py checks); do
-            [ "${{ matrix.category }}" = "$c" ] && continue
-            EXPR="$EXPR and not $c"
-          done
+          # The planner owns what a cell selects, so the cell runs exactly the
+          # cases it was planned (and, if sharded, split) from.
+          EXPR=$(python3 ci/testcase.py expr --cell="${{ matrix.category }}" --driver="${{ matrix.driver }}")
+          SHARD=()
+          [ -n "${{ matrix.shards }}" ] && SHARD=(--shard="${{ matrix.shard }}/${{ matrix.shards }}")
           VX_XLEN=${{ matrix.xlen }} python3 -m pytest ci -m "$EXPR" --strict-markers \
-            -v \
-            --junitxml="$GITHUB_WORKSPACE/junit-${{ matrix.category }}-${{ matrix.driver }}-${{ matrix.xlen }}.xml"
+            "${SHARD[@]}" -v \
+            --junitxml="$GITHUB_WORKSPACE/junit-$CELL.xml"
       - uses: actions/upload-artifact@v4
         if: always()
         with:
-          name: junit-${{ matrix.category }}-${{ matrix.driver }}-${{ matrix.xlen }}
+          name: junit-${{ env.CELL }}
           path: junit-*.xml
 
   complete:

VX_config.toml

@@ -375,7 +375,9 @@ VX_CFG_RASTER_BLOCK_LOG_SIZE   = 2
 VX_CFG_RASTER_MEM_FIFO_DEPTH  = 8
 VX_CFG_RASTER_QUAD_FIFO_DEPTH = 32
 VX_CFG_RASTER_MEM_QUEUE_SIZE  = 4
-VX_CFG_RASTER_EARLYZ_ENABLE          = false
+# Early-Z occlusion cull: on wherever it is legal. It reads committed depth
+# through the OM's ocache, so it needs both units (VX_define.vh enforces it).
+VX_CFG_RASTER_EARLYZ_ENABLE   = "expr: $VX_CFG_EXT_RASTER_ENABLE and $VX_CFG_EXT_OM_ENABLE"
 
 [om]
 VX_CFG_NUM_OM_CORES       = "expr: max(1, up($VX_CFG_NUM_CORES / 8))"

ci/baselines/perf/core.json

@@ -6,7 +6,7 @@
     },
     "app": "raycast",
     "args": "",
-    "config_hash": "ae710a554c1b4afe",
+    "config_hash": "ee3a227211fa9942",
     "configs": "-DVX_CFG_NUM_THREADS=16",
     "driver": "rtlsim"
   },
@@ -17,7 +17,7 @@
     },
     "app": "raycast",
     "args": "",
-    "config_hash": "204dd696e55b22c0",
+    "config_hash": "c8b680eabeb18655",
     "configs": "-DVX_CFG_NUM_THREADS=4",
     "driver": "rtlsim"
   },
@@ -28,7 +28,7 @@
     },
     "app": "sgemm",
     "args": "-n128",
-    "config_hash": "4e0d1121c7c753fb",
+    "config_hash": "58eb3cc06964207a",
     "configs": "-DVX_CFG_NUM_WARPS=8 -DVX_CFG_NUM_CORES=2 -DVX_CFG_NUM_THREADS=16 -DVX_CFG_L2_ENABLE",
     "driver": "rtlsim"
   },
@@ -39,7 +39,7 @@
     },
     "app": "sgemm",
     "args": "-n128",
-    "config_hash": "2e368e29b49794c1",
+    "config_hash": "aa6f26c7306417a8",
     "configs": "-DVX_CFG_NUM_WARPS=8 -DVX_CFG_NUM_CORES=2 -DVX_CFG_NUM_THREADS=4 -DVX_CFG_L2_ENABLE",
     "driver": "rtlsim"
   },
@@ -50,7 +50,7 @@
     },
     "app": "sgemm",
     "args": "-n128",
-    "config_hash": "8493bfcbea192f2b",
+    "config_hash": "b390d957eec8220f",
     "configs": "-DVX_CFG_NUM_THREADS=16",
     "driver": "rtlsim"
   },
@@ -61,7 +61,7 @@
     },
     "app": "sgemm",
     "args": "-n128",
-    "config_hash": "ef1d8528f632523b",
+    "config_hash": "13d03df76d55105e",
     "configs": "-DVX_CFG_NUM_THREADS=4",
     "driver": "rtlsim"
   },
@@ -72,7 +72,7 @@
     },
     "app": "sgemmx",
     "args": "-n128",
-    "config_hash": "c016962fb7669255",
+    "config_hash": "68638b6685801699",
     "configs": "-DVX_CFG_NUM_THREADS=16",
     "driver": "rtlsim"
   },
@@ -83,7 +83,7 @@
     },
     "app": "sgemmx",
     "args": "-n128",
-    "config_hash": "65047df50cd355e6",
+    "config_hash": "837240be3049d62e",
     "configs": "-DVX_CFG_NUM_THREADS=4",
     "driver": "rtlsim"
   },
@@ -94,7 +94,7 @@
     },
     "app": "sgemv",
     "args": "-m512 -n512",
-    "config_hash": "70d3eb40917f6df3",
+    "config_hash": "c6a3720118661bf7",
     "configs": "-DVX_CFG_NUM_THREADS=16",
     "driver": "rtlsim"
   },
@@ -105,7 +105,7 @@
     },
     "app": "sgemv",
     "args": "-m512 -n512",
-    "config_hash": "e359f1a79bc49193",
+    "config_hash": "db6bb5af90ecc1a5",
     "configs": "-DVX_CFG_NUM_THREADS=4",
     "driver": "rtlsim"
   },
@@ -116,7 +116,7 @@
     },
     "app": "softmax",
     "args": "-n64",
-    "config_hash": "91488cd182a7671a",
+    "config_hash": "1c0e7742deb14093",
     "configs": "-DVX_CFG_NUM_THREADS=16",
     "driver": "rtlsim"
   },
@@ -127,7 +127,7 @@
     },
     "app": "softmax",
     "args": "-n64",
-    "config_hash": "c8061d3015fc6d7e",
+    "config_hash": "3797eecae3200315",
     "configs": "-DVX_CFG_NUM_THREADS=4",
     "driver": "rtlsim"
   },
@@ -138,7 +138,7 @@
     },
     "app": "stencil3d",
     "args": "-n32",
-    "config_hash": "e659f36c50fcf6bf",
+    "config_hash": "aa768c2a9c65b092",
     "configs": "-DVX_CFG_NUM_THREADS=16",
     "driver": "rtlsim"
   },
@@ -149,7 +149,7 @@
     },
     "app": "stencil3d",
     "args": "-n32",
-    "config_hash": "3bbd7b84bf73c746",
+    "config_hash": "5400d882a6c96ac1",
     "configs": "-DVX_CFG_NUM_THREADS=4",
     "driver": "rtlsim"
   },
@@ -160,7 +160,7 @@
     },
     "app": "vecadd",
     "args": "-n16384",
-    "config_hash": "a5ce692bba42344b",
+    "config_hash": "98b8bdc5e546ea62",
     "configs": "-DVX_CFG_NUM_THREADS=16",
     "driver": "rtlsim"
   },
@@ -171,7 +171,7 @@
     },
     "app": "vecadd",
     "args": "-n16384",
-    "config_hash": "7ad7759234a8fedf",
+    "config_hash": "1510b8438164b252",
     "configs": "-DVX_CFG_NUM_THREADS=4",
     "driver": "rtlsim"
   }

ci/baselines/perf/dxa.json

@@ -6,7 +6,7 @@
     },
     "app": "dxa_copy_mcast",
     "args": "",
-    "config_hash": "d8dc71acb72624de",
+    "config_hash": "cbcd10be884ceab7",
     "configs": "-DVX_CFG_EXT_DXA_ENABLE",
     "driver": "rtlsim"
   },
@@ -17,7 +17,7 @@
     },
     "app": "dxa_copy",
     "args": "-d1",
-    "config_hash": "84239768cfc34b0d",
+    "config_hash": "3a1ec1b81b5b3baf",
     "configs": "-DVX_CFG_EXT_DXA_ENABLE -DVX_CFG_NUM_THREADS=16",
     "driver": "rtlsim"
   },
@@ -28,7 +28,7 @@
     },
     "app": "dxa_copy",
     "args": "-d1",
-    "config_hash": "e315513bdc6942fb",
+    "config_hash": "3a888067fcc449d6",
     "configs": "-DVX_CFG_EXT_DXA_ENABLE -DVX_CFG_NUM_THREADS=4",
     "driver": "rtlsim"
   },
@@ -39,7 +39,7 @@
     },
     "app": "sgemm2_dxa_mcast",
     "args": "-n128",
-    "config_hash": "d6edd97b6570efdd",
+    "config_hash": "14050e4dcef65266",
     "configs": "-DVX_CFG_EXT_DXA_ENABLE -DVX_CFG_NUM_THREADS=16",
     "driver": "rtlsim"
   },
@@ -50,7 +50,7 @@
     },
     "app": "sgemm2_dxa_mcast",
     "args": "-n128",
-    "config_hash": "f638e7687fcfdf72",
+    "config_hash": "a55b900d67aa3657",
     "configs": "-DVX_CFG_EXT_DXA_ENABLE -DVX_CFG_NUM_THREADS=4",
     "driver": "rtlsim"
   },
@@ -61,7 +61,7 @@
     },
     "app": "sgemm2_dxa",
     "args": "-n128 -t4 -m1",
-    "config_hash": "538516a52b3616f9",
+    "config_hash": "36f5886fdc97e040",
     "configs": "-DVX_CFG_EXT_DXA_ENABLE -DVX_CFG_NUM_THREADS=16",
     "driver": "rtlsim"
   },
@@ -72,7 +72,7 @@
     },
     "app": "sgemm2_dxa",
     "args": "-n128 -t4 -m1",
-    "config_hash": "b68ce24c030ee769",
+    "config_hash": "9ccf890d5ac45a62",
     "configs": "-DVX_CFG_EXT_DXA_ENABLE -DVX_CFG_NUM_THREADS=4",
     "driver": "rtlsim"
   },
@@ -83,7 +83,7 @@
     },
     "app": "sgemm_tcu_wg_dxa_mcast",
     "args": "-m 128 -n 128 -k 64",
-    "config_hash": "f2390772053e2d69",
+    "config_hash": "ee7678a7e7e9b79b",
     "configs": "-DVX_CFG_EXT_TCU_ENABLE -DVX_CFG_EXT_DXA_ENABLE -DVX_CFG_NUM_ALU_BLOCKS=1 -DVX_CFG_NUM_LSU_BLOCKS=1 -DVX_CFG_NUM_FPU_BLOCKS=1 -DVX_CFG_NUM_THREADS=16",
     "driver": "rtlsim"
   },
@@ -94,7 +94,7 @@
     },
     "app": "sgemm_tcu_wg_dxa_mcast",
     "args": "-m 128 -n 128 -k 64",
-    "config_hash": "66b8301aa43d2744",
+    "config_hash": "c79015e4caec7b93",
     "configs": "-DVX_CFG_EXT_TCU_ENABLE -DVX_CFG_EXT_DXA_ENABLE -DVX_CFG_NUM_ALU_BLOCKS=1 -DVX_CFG_NUM_LSU_BLOCKS=1 -DVX_CFG_NUM_FPU_BLOCKS=1 -DVX_CFG_NUM_THREADS=4",
     "driver": "rtlsim"
   },
@@ -105,7 +105,7 @@
     },
     "app": "sgemm_tcu_wg_dxa",
     "args": "-m 128 -n 128 -k 64",
-    "config_hash": "ba8978d35dc6e582",
+    "config_hash": "5a35710ef3aec3a6",
     "configs": "-DVX_CFG_ISSUE_WIDTH=4 -DVX_CFG_EXT_TCU_ENABLE -DVX_CFG_TCU_WGMMA_ENABLE -DVX_CFG_EXT_DXA_ENABLE -DWGMMA_NRC=16 -DVX_CFG_NUM_ALU_BLOCKS=1 -DVX_CFG_NUM_LSU_BLOCKS=1 -DVX_CFG_NUM_FPU_BLOCKS=1 -DVX_CFG_NUM_WARPS=8 -DVX_CFG_NUM_THREADS=32",
     "driver": "rtlsim"
   },
@@ -116,7 +116,7 @@
     },
     "app": "sgemm_tcu_wg_sp_dxa",
     "args": "-m 128 -n 128 -k 128",
-    "config_hash": "9480be10a4650239",
+    "config_hash": "d12fad3fd43e61c6",
     "configs": "-DVX_CFG_ISSUE_WIDTH=4 -DVX_CFG_EXT_TCU_ENABLE -DVX_CFG_TCU_WGMMA_ENABLE -DVX_CFG_TCU_SPARSE_ENABLE -DVX_CFG_EXT_DXA_ENABLE -DWGMMA_NRC=32 -DVX_CFG_NUM_ALU_BLOCKS=1 -DVX_CFG_NUM_LSU_BLOCKS=1 -DVX_CFG_NUM_FPU_BLOCKS=1 -DVX_CFG_NUM_WARPS=8 -DVX_CFG_NUM_THREADS=32",
     "driver": "rtlsim"
   }