Merge branch 'main' into add_checkout

Author: lshaw8317

.github/workflows/build.yml

@@ -23,7 +23,7 @@ jobs:
             numpy-version: "1.26"
 
     steps:
-    - uses: actions/checkout@v5
+    - uses: actions/checkout@v6
 
     - name: Set up Python ${{ matrix.python-version }}
       uses: actions/setup-python@v6

.github/workflows/cibuildwheels.yml

@@ -60,7 +60,7 @@ jobs:
             artifact_name: "macos-universal2"
     steps:
       - name: Checkout repo
-        uses: actions/checkout@v5
+        uses: actions/checkout@v6
 
       - name: Set up Python
         uses: actions/setup-python@v6
@@ -83,7 +83,7 @@ jobs:
           arch: amd64
 
       - name: Build wheels
-        uses: pypa/cibuildwheel@v3.2
+        uses: pypa/cibuildwheel@v3.3
 
       - name: Make sdist
         if: ${{ matrix.os == 'ubuntu-latest' }}

.github/workflows/wasm.yml

@@ -30,7 +30,7 @@ jobs:
 
     steps:
       - name: Checkout repo
-        uses: actions/checkout@v5
+        uses: actions/checkout@v6
 
       - name: Set up Python
         uses: actions/setup-python@v6

bench/ndarray/roofline-analysis.py

@@ -153,8 +153,9 @@ def setup_arrays(mem_mode: bool):
         shape = (15_000, 15_000)
         large_shape = (2,) + shape
     else:
-        shape = (30_000, 30_000)
-        large_shape = (15,) + shape
+        # shape = (30_000, 30_000)
+        shape = (15_000, 15_000)
+        large_shape = (60,) + shape
 
     nelem = math.prod(shape)
     large_nelem = math.prod(large_shape)

bench/ndarray/roofline-mem-speed-plot.py

@@ -108,6 +108,54 @@ def extract_xy(mem_dict):
         va="top",
     )
 
+    # --- single workload label per workload name (avoid duplicates) ---
+    # Build a map: workload name -> list of (intensity, gflops) across mem_4800/mem_6000
+    workload_map: dict[str, dict[str, list[float]]] = {}
+
+    for workload, metrics in mem_4800.items():
+        intensity = metrics["Intensity"]
+        gflops = metrics["GFLOPS"]
+        if workload not in workload_map:
+            workload_map[workload] = {"intensity": [], "gflops": []}
+        workload_map[workload]["intensity"].append(intensity)
+        workload_map[workload]["gflops"].append(gflops)
+
+    for workload, metrics in mem_6000.items():
+        intensity = metrics["Intensity"]
+        gflops = metrics["GFLOPS"]
+        if workload not in workload_map:
+            workload_map[workload] = {"intensity": [], "gflops": []}
+        workload_map[workload]["intensity"].append(intensity)
+        workload_map[workload]["gflops"].append(gflops)
+
+    # Place a single label per workload at the average intensity and slightly below
+    # the minimum GFLOPS across both memory speeds for that workload.
+    for workload, vals in workload_map.items():
+        intensities = vals["intensity"]
+        gflops_list = vals["gflops"]
+        x_label = sum(intensities) / len(intensities)
+        y_min = min(gflops_list)
+        raw_ypos = y_min * 0.6
+
+        ymin_curr, _ = ax.get_ylim()
+        safe_ypos = max(raw_ypos, ymin_curr * 1.5 if ymin_curr > 0 else raw_ypos)
+
+        # Avoid overlap between matmul1 and matmul2 by using different vertical offsets
+        if workload == "matmul1":
+            safe_ypos *= .8   # push matmul1 a bit higher
+        elif workload == "matmul2":
+            safe_ypos *= 1.2   # keep matmul2 lower
+
+        ax.annotate(
+            workload,
+            (x_label, safe_ypos),
+            ha="center",
+            va="top",
+            fontsize=10,
+            alpha=0.9,
+        )
+    # --------------------------------------------------------------
+
     ax.set_xlabel("Arithmetic Intensity (FLOPs/element)")
     ax.set_ylabel("Performance (GFLOPS/sec)")
     ax.set_title("Memory speed impact on NumPy/NumExpr performance\nAMD 7800X3D (in-memory)")