Add benchmarking

int8/benchmarks/x5h_mwmx_npu_apm80_12core.yaml

@@ -29,7 +29,7 @@ benchmark:
 
 performance:
   fps: null
-  latency: 38.103941        # APM80 pipeline run (metawaremx_runtime CI); this artifact is ONE segment
+  latency: 195.34        # synced to mwmx2.2 model_list.html (2026-09-22); was 38.103941
   # of a 4-way split model ("custom_seg_split_4_split_2") — not full end-to-end latency.
   # 1-core run for this artifact failed to compile — no figure available for that slice.
 

int8/benchmarks/x5h_mwmx_npu_apm80_1core.yaml

@@ -0,0 +1,76 @@
+hardware:
+  vendor: renesas
+  chip: rcar-x5h
+  cpu: arm-cortex-a720
+  npu: npx6-48k
+  npu_count: 2
+  npu_cores: 12
+  npu_default_freq_mhz: 1066
+  accelerator:
+    - npu
+
+runtime:
+  engine: mwmx
+  toolchain_version: "MWMX SDK v4.35.0"
+  format: onnx
+  execution_provider: npu
+  execution_precision: int8
+
+configuration:
+  npu_instances: 1
+  npu_cores_per_instance: 1
+  npu_freq_mhz: 850
+
+benchmark:
+  type: hil
+  parameters:
+    batch_size: 1
+    input_resolution: [1, 3, 512, 1024]   # explicit in the source checkpoint name (512x1024, Cityscapes)
+
+performance:
+  fps: null
+  latency: 497.55        # synced to mwmx2.2 model_list.html (2026-09-22); no prior local measurement for this core count
+  # of a 4-way split model ("custom_seg_split_4_split_2") — not full end-to-end latency.
+
+metrics:
+  accuracy: null
+  top5_accuracy: null
+
+memory:
+  peak_mb: null
+
+power:
+  avg_w: null
+
+# Exact commands verified against the NNAC "Getting Started" chapter. Rendered
+# by the AI-Dashboard in place of the generic placeholder flow — see
+# downloadRunHTML() / parse_reproduce() in AI-Dashboard/app.js.
+# CAVEAT: this artifact is segment "split_2" of a 4-way split network (see
+# README) — these steps reproduce only this segment's latency, not an
+# end-to-end DeepLabV3+ result. The network config below is required for this
+# model — there is no working default.
+reproduce:
+  steps:
+    - title: Activate the Python environment
+      command: >-
+        Activate the Python virtual environment that has the `hf` CLI
+        (huggingface_hub) and the NNAC toolchain installed, e.g.
+        `source nnac_venv/bin/activate` -- path depends on your toolchain install.
+      kind: note
+    - title: Download the ONNX model and compile config
+      command: hf download Renesas/DeepLabV3Plus-R50-ONNX --repo-type model --include "fp32/*" "compile_config/*" --local-dir ./DeepLabV3Plus-R50-ONNX-fp32
+    - title: Compile with the NNAC toolchain (INT8 auto-cast from the FP32 graph)
+      command: |
+        python3 nnac_frontend/legalize.py -d binary/nnx ./DeepLabV3Plus-R50-ONNX-fp32/fp32/deeplabv3plus_r50_oss_sim_inf.onnx --num-core 1 --network-config ./DeepLabV3Plus-R50-ONNX-fp32/compile_config/network_config.yaml
+    - title: Set up the R-Car X5H board
+      command: Configure the board per the AI Compiler (NNAC) "Getting Started" guide, section 3.4 (host TFTP/NFS setup, bootloader flashing, U-Boot, Linux boot, login) -- exact steps depend on your board/network setup.
+      kind: note
+    - title: Copy the compiled artifact to the board
+      command: Copy ${WORKDIR}/binary/nnx (the working directory from the download/compile steps above) to the board -- method may vary (NFS mount, scp, USB, etc.).
+      kind: note
+    - title: Run on R-Car X5H (single NPU cluster)
+      command: |
+        cd binary
+        ./host_app ./arc_prog_npus ./nnx/deeplabv3plus_r50_oss_sim_inf
+      expected: hash[n] = 0x...(OK) means the run's output matches the reference hash in hash.txt; latency is the NPX execution time reported in cycles and ms.
+  notes: This is one segment (split_2 of 4) of the full segmentation pipeline — not whole-model latency.

int8/benchmarks/x5h_mwmx_npu_apm80_3core.yaml

@@ -0,0 +1,76 @@
+hardware:
+  vendor: renesas
+  chip: rcar-x5h
+  cpu: arm-cortex-a720
+  npu: npx6-48k
+  npu_count: 2
+  npu_cores: 12
+  npu_default_freq_mhz: 1066
+  accelerator:
+    - npu
+
+runtime:
+  engine: mwmx
+  toolchain_version: "MWMX SDK v4.35.0"
+  format: onnx
+  execution_provider: npu
+  execution_precision: int8
+
+configuration:
+  npu_instances: 1
+  npu_cores_per_instance: 3
+  npu_freq_mhz: 850
+
+benchmark:
+  type: hil
+  parameters:
+    batch_size: 1
+    input_resolution: [1, 3, 512, 1024]   # explicit in the source checkpoint name (512x1024, Cityscapes)
+
+performance:
+  fps: null
+  latency: 278.55        # synced to mwmx2.2 model_list.html (2026-09-22); no prior local measurement for this core count
+  # of a 4-way split model ("custom_seg_split_4_split_2") — not full end-to-end latency.
+
+metrics:
+  accuracy: null
+  top5_accuracy: null
+
+memory:
+  peak_mb: null
+
+power:
+  avg_w: null
+
+# Exact commands verified against the NNAC "Getting Started" chapter. Rendered
+# by the AI-Dashboard in place of the generic placeholder flow — see
+# downloadRunHTML() / parse_reproduce() in AI-Dashboard/app.js.
+# CAVEAT: this artifact is segment "split_2" of a 4-way split network (see
+# README) — these steps reproduce only this segment's latency, not an
+# end-to-end DeepLabV3+ result. The network config below is required for this
+# model — there is no working default.
+reproduce:
+  steps:
+    - title: Activate the Python environment
+      command: >-
+        Activate the Python virtual environment that has the `hf` CLI
+        (huggingface_hub) and the NNAC toolchain installed, e.g.
+        `source nnac_venv/bin/activate` -- path depends on your toolchain install.
+      kind: note
+    - title: Download the ONNX model and compile config
+      command: hf download Renesas/DeepLabV3Plus-R50-ONNX --repo-type model --include "fp32/*" "compile_config/*" --local-dir ./DeepLabV3Plus-R50-ONNX-fp32
+    - title: Compile with the NNAC toolchain (INT8 auto-cast from the FP32 graph)
+      command: |
+        python3 nnac_frontend/legalize.py -d binary/nnx ./DeepLabV3Plus-R50-ONNX-fp32/fp32/deeplabv3plus_r50_oss_sim_inf.onnx --num-core 3 --network-config ./DeepLabV3Plus-R50-ONNX-fp32/compile_config/network_config.yaml
+    - title: Set up the R-Car X5H board
+      command: Configure the board per the AI Compiler (NNAC) "Getting Started" guide, section 3.4 (host TFTP/NFS setup, bootloader flashing, U-Boot, Linux boot, login) -- exact steps depend on your board/network setup.
+      kind: note
+    - title: Copy the compiled artifact to the board
+      command: Copy ${WORKDIR}/binary/nnx (the working directory from the download/compile steps above) to the board -- method may vary (NFS mount, scp, USB, etc.).
+      kind: note
+    - title: Run on R-Car X5H (single NPU cluster, 3 AI cores)
+      command: |
+        cd binary
+        ./host_app ./arc_prog_npus ./nnx/deeplabv3plus_r50_oss_sim_inf
+      expected: hash[n] = 0x...(OK) means the run's output matches the reference hash in hash.txt; latency is the NPX execution time reported in cycles and ms.
+  notes: This is one segment (split_2 of 4) of the full segmentation pipeline — not whole-model latency.

int8/benchmarks/x5h_mwmx_npu_apm80_4core.yaml

@@ -0,0 +1,76 @@
+hardware:
+  vendor: renesas
+  chip: rcar-x5h
+  cpu: arm-cortex-a720
+  npu: npx6-48k
+  npu_count: 2
+  npu_cores: 12
+  npu_default_freq_mhz: 1066
+  accelerator:
+    - npu
+
+runtime:
+  engine: mwmx
+  toolchain_version: "MWMX SDK v4.35.0"
+  format: onnx
+  execution_provider: npu
+  execution_precision: int8
+
+configuration:
+  npu_instances: 1
+  npu_cores_per_instance: 4
+  npu_freq_mhz: 850
+
+benchmark:
+  type: hil
+  parameters:
+    batch_size: 1
+    input_resolution: [1, 3, 512, 1024]   # explicit in the source checkpoint name (512x1024, Cityscapes)
+
+performance:
+  fps: null
+  latency: 230.27        # synced to mwmx2.2 model_list.html (2026-09-22); no prior local measurement for this core count
+  # of a 4-way split model ("custom_seg_split_4_split_2") — not full end-to-end latency.
+
+metrics:
+  accuracy: null
+  top5_accuracy: null
+
+memory:
+  peak_mb: null
+
+power:
+  avg_w: null
+
+# Exact commands verified against the NNAC "Getting Started" chapter. Rendered
+# by the AI-Dashboard in place of the generic placeholder flow — see
+# downloadRunHTML() / parse_reproduce() in AI-Dashboard/app.js.
+# CAVEAT: this artifact is segment "split_2" of a 4-way split network (see
+# README) — these steps reproduce only this segment's latency, not an
+# end-to-end DeepLabV3+ result. The network config below is required for this
+# model — there is no working default.
+reproduce:
+  steps:
+    - title: Activate the Python environment
+      command: >-
+        Activate the Python virtual environment that has the `hf` CLI
+        (huggingface_hub) and the NNAC toolchain installed, e.g.
+        `source nnac_venv/bin/activate` -- path depends on your toolchain install.
+      kind: note
+    - title: Download the ONNX model and compile config
+      command: hf download Renesas/DeepLabV3Plus-R50-ONNX --repo-type model --include "fp32/*" "compile_config/*" --local-dir ./DeepLabV3Plus-R50-ONNX-fp32
+    - title: Compile with the NNAC toolchain (INT8 auto-cast from the FP32 graph)
+      command: |
+        python3 nnac_frontend/legalize.py -d binary/nnx ./DeepLabV3Plus-R50-ONNX-fp32/fp32/deeplabv3plus_r50_oss_sim_inf.onnx --num-core 4 --network-config ./DeepLabV3Plus-R50-ONNX-fp32/compile_config/network_config.yaml
+    - title: Set up the R-Car X5H board
+      command: Configure the board per the AI Compiler (NNAC) "Getting Started" guide, section 3.4 (host TFTP/NFS setup, bootloader flashing, U-Boot, Linux boot, login) -- exact steps depend on your board/network setup.
+      kind: note
+    - title: Copy the compiled artifact to the board
+      command: Copy ${WORKDIR}/binary/nnx (the working directory from the download/compile steps above) to the board -- method may vary (NFS mount, scp, USB, etc.).
+      kind: note
+    - title: Run on R-Car X5H (single NPU cluster, 4 AI cores)
+      command: |
+        cd binary
+        ./host_app ./arc_prog_npus ./nnx/deeplabv3plus_r50_oss_sim_inf
+      expected: hash[n] = 0x...(OK) means the run's output matches the reference hash in hash.txt; latency is the NPX execution time reported in cycles and ms.
+  notes: This is one segment (split_2 of 4) of the full segmentation pipeline — not whole-model latency.

int8/benchmarks/x5h_mwmx_npu_apm80_6core.yaml

@@ -0,0 +1,76 @@
+hardware:
+  vendor: renesas
+  chip: rcar-x5h
+  cpu: arm-cortex-a720
+  npu: npx6-48k
+  npu_count: 2
+  npu_cores: 12
+  npu_default_freq_mhz: 1066
+  accelerator:
+    - npu
+
+runtime:
+  engine: mwmx
+  toolchain_version: "MWMX SDK v4.35.0"
+  format: onnx
+  execution_provider: npu
+  execution_precision: int8
+
+configuration:
+  npu_instances: 1
+  npu_cores_per_instance: 6
+  npu_freq_mhz: 850
+
+benchmark:
+  type: hil
+  parameters:
+    batch_size: 1
+    input_resolution: [1, 3, 512, 1024]   # explicit in the source checkpoint name (512x1024, Cityscapes)
+
+performance:
+  fps: null
+  latency: 218.57        # synced to mwmx2.2 model_list.html (2026-09-22); no prior local measurement for this core count
+  # of a 4-way split model ("custom_seg_split_4_split_2") — not full end-to-end latency.
+
+metrics:
+  accuracy: null
+  top5_accuracy: null
+
+memory:
+  peak_mb: null
+
+power:
+  avg_w: null
+
+# Exact commands verified against the NNAC "Getting Started" chapter. Rendered
+# by the AI-Dashboard in place of the generic placeholder flow — see
+# downloadRunHTML() / parse_reproduce() in AI-Dashboard/app.js.
+# CAVEAT: this artifact is segment "split_2" of a 4-way split network (see
+# README) — these steps reproduce only this segment's latency, not an
+# end-to-end DeepLabV3+ result. The network config below is required for this
+# model — there is no working default.
+reproduce:
+  steps:
+    - title: Activate the Python environment
+      command: >-
+        Activate the Python virtual environment that has the `hf` CLI
+        (huggingface_hub) and the NNAC toolchain installed, e.g.
+        `source nnac_venv/bin/activate` -- path depends on your toolchain install.
+      kind: note
+    - title: Download the ONNX model and compile config
+      command: hf download Renesas/DeepLabV3Plus-R50-ONNX --repo-type model --include "fp32/*" "compile_config/*" --local-dir ./DeepLabV3Plus-R50-ONNX-fp32
+    - title: Compile with the NNAC toolchain (INT8 auto-cast from the FP32 graph)
+      command: |
+        python3 nnac_frontend/legalize.py -d binary/nnx ./DeepLabV3Plus-R50-ONNX-fp32/fp32/deeplabv3plus_r50_oss_sim_inf.onnx --num-core 6 --network-config ./DeepLabV3Plus-R50-ONNX-fp32/compile_config/network_config.yaml
+    - title: Set up the R-Car X5H board
+      command: Configure the board per the AI Compiler (NNAC) "Getting Started" guide, section 3.4 (host TFTP/NFS setup, bootloader flashing, U-Boot, Linux boot, login) -- exact steps depend on your board/network setup.
+      kind: note
+    - title: Copy the compiled artifact to the board
+      command: Copy ${WORKDIR}/binary/nnx (the working directory from the download/compile steps above) to the board -- method may vary (NFS mount, scp, USB, etc.).
+      kind: note
+    - title: Run on R-Car X5H (single NPU cluster, 6 AI cores)
+      command: |
+        cd binary
+        ./host_app ./arc_prog_npus ./nnx/deeplabv3plus_r50_oss_sim_inf
+      expected: hash[n] = 0x...(OK) means the run's output matches the reference hash in hash.txt; latency is the NPX execution time reported in cycles and ms.
+  notes: This is one segment (split_2 of 4) of the full segmentation pipeline — not whole-model latency.