restore large xfer workaround

I'm sick of the way the simple extension builder in setup.py doesn't
understand header dependencies.

CMakeLists.txt

@@ -0,0 +1,12 @@
+cmake_minimum_required(VERSION 3.4...3.18)
+project("_piomatter")
+set(CMAKE_INTERPROCEDURAL_OPTIMIZATION FALSE)
+
+find_package(pybind11 CONFIG)
+pybind11_add_module("_piomatter"
+    src/pymain.cpp src/piolib/pio_rp1.c src/piolib/piolib.c)
+target_include_directories("_piomatter" PRIVATE
+    ${CMAKE_CURRENT_LIST_DIR}/src/include
+    ${CMAKE_CURRENT_LIST_DIR}/src/piolib/include)
+set_property(TARGET "_piomatter" PROPERTY
+    CXX_STANDARD 20)

examples/fbmirror.py

@@ -44,7 +44,7 @@ def main(xoffset, yoffset, width, height, serpentine, rotation, pinout, n_planes
         pixelmap = simple_multilane_mapper(width, height, n_addr_lines, n_lanes)
         geometry = piomatter.Geometry(width=width, height=height, n_planes=n_planes, n_addr_lines=n_addr_lines, n_temporal_planes=n_temporal_planes, n_lanes=n_lanes, map=pixelmap)
     else:
-        geometry = piomatter.Geometry(width=width, height=height, n_planes=n_planes, n_addr_lines=n_addr_lines, n_temporal_planes=n_temporal_planes, rotation=rotation, serpentine=serpentine)
+        geometry = piomatter.Geometry(width=width, height=height, n_planes=n_planes, n_addr_lines=n_addr_lines, n_temporal_planes=n_temporal_planes, rotation=rotation)
     framebuffer = np.zeros(shape=(geometry.height, geometry.width), dtype=dtype)
     matrix = piomatter.PioMatter(colorspace=piomatter.Colorspace.RGB565, pinout=pinout, framebuffer=framebuffer, geometry=geometry)
 

examples/fbmirror_scaled.py

@@ -72,7 +72,7 @@ def main(xoffset, yoffset, scale, width, height, serpentine, rotation, pinout, n
         pixelmap = simple_multilane_mapper(width, height, n_addr_lines, n_lanes)
         geometry = piomatter.Geometry(width=width, height=height, n_planes=n_planes, n_addr_lines=n_addr_lines, n_temporal_planes=n_temporal_planes, n_lanes=n_lanes, map=pixelmap)
     else:
-        geometry = piomatter.Geometry(width=width, height=height, n_planes=n_planes, n_temporal_planes=n_temporal_planes, n_addr_lines=n_addr_lines, rotation=rotation, serpentine=serpentine)
+        geometry = piomatter.Geometry(width=width, height=height, n_planes=n_planes, n_temporal_planes=n_temporal_planes, n_addr_lines=n_addr_lines, rotation=rotation)
     matrix_framebuffer = np.zeros(shape=(geometry.height, geometry.width, 3), dtype=np.uint8)
     matrix = piomatter.PioMatter(colorspace=piomatter.Colorspace.RGB888Packed, pinout=pinout, framebuffer=matrix_framebuffer, geometry=geometry)
 

examples/rainbow_spiral_active3.py

@@ -30,7 +30,7 @@ pixelmap = simple_multilane_mapper(width, height, n_addr_lines, n_lanes)
 geometry = piomatter.Geometry(width=width, height=height, n_addr_lines=n_addr_lines, n_planes=10, n_temporal_planes=4, map=pixelmap, n_lanes=n_lanes)
 framebuffer = np.asarray(canvas) + 0  # Make a mutable copy
 matrix = piomatter.PioMatter(colorspace=piomatter.Colorspace.RGB888Packed,
-                             pinout=piomatter.Pinout.Active3BGR,
+                             pinout=piomatter.Pinout.Active3,
                              framebuffer=framebuffer,
                              geometry=geometry)
 
@@ -66,7 +66,7 @@ def darken_color(hex_color, darkness_factor):
 
     return darkened_hex_color
 
-step_count = 8
+step_count = 4
 darkness_factor = 0.5
 
 clearing = False

examples/triple_matrix_active3_simpletest.py

@@ -1,46 +0,0 @@
-#!/usr/bin/python3
-# SPDX-FileCopyrightText: 2025 Tim Cocks for Adafruit Industries
-#
-# SPDX-License-Identifier: MIT
-"""
-Display a simple test pattern of 3 shapes on three 64x64 matrix panels
-using Active3 compatible connections.
-
-Run like this:
-
-$ python triple_matrix_active3_simpletest.py
-
-"""
-
-import numpy as np
-from PIL import Image, ImageDraw
-
-import adafruit_blinka_raspberry_pi5_piomatter as piomatter
-from adafruit_blinka_raspberry_pi5_piomatter.pixelmappers import simple_multilane_mapper
-
-width = 64
-n_lanes = 6
-n_addr_lines = 5
-height = n_lanes << n_addr_lines
-pen_radius = 1
-
-canvas = Image.new('RGB', (width, height), (0, 0, 0))
-draw = ImageDraw.Draw(canvas)
-
-pixelmap = simple_multilane_mapper(width, height, n_addr_lines, n_lanes)
-geometry = piomatter.Geometry(width=width, height=height, n_addr_lines=n_addr_lines, n_planes=10, n_temporal_planes=4, map=pixelmap, n_lanes=n_lanes)
-framebuffer = np.asarray(canvas) + 0  # Make a mutable copy
-matrix = piomatter.PioMatter(colorspace=piomatter.Colorspace.RGB888Packed,
-                             pinout=piomatter.Pinout.Active3,
-                             framebuffer=framebuffer,
-                             geometry=geometry)
-
-
-draw.rectangle((8, 8, width-8, width-8), fill=0x008800)
-draw.circle((32, 64+32), 22, fill=0x880000)
-draw.polygon([(32, 136), (54, 180), (10, 180)], fill=0x000088)
-
-framebuffer[:] = np.asarray(canvas)
-matrix.show()
-
-input("Press enter to exit")

examples/virtualdisplay.py

@@ -57,7 +57,7 @@ def main(scale, backend, use_xauth, extra_args, rfbport, brightness, width, heig
                                       n_temporal_planes=n_temporal_planes, n_lanes=n_lanes, map=pixelmap)
     else:
         geometry = piomatter.Geometry(width=width, height=height, n_planes=n_planes, n_addr_lines=n_addr_lines,
-                                      n_temporal_planes=n_temporal_planes, rotation=rotation, serpentine=serpentine)
+                                      n_temporal_planes=n_temporal_planes, rotation=rotation)
     framebuffer = np.zeros(shape=(geometry.height, geometry.width, 3), dtype=np.uint8)
     matrix = piomatter.PioMatter(colorspace=piomatter.Colorspace.RGB888Packed, pinout=pinout, framebuffer=framebuffer, geometry=geometry)
 

examples/xdisplay_mirror.py

@@ -52,7 +52,7 @@ def main(width, height, serpentine, rotation, pinout, n_planes,
                                       n_temporal_planes=n_temporal_planes, n_lanes=n_lanes, map=pixelmap)
     else:
         geometry = piomatter.Geometry(width=width, height=height, n_planes=n_planes, n_addr_lines=n_addr_lines,
-                                      n_temporal_planes=n_temporal_planes, rotation=rotation, serpentine=serpentine)
+                                      n_temporal_planes=n_temporal_planes, rotation=rotation)
 
     framebuffer = np.zeros(shape=(geometry.height, geometry.width, 3), dtype=np.uint8)
     matrix = piomatter.PioMatter(colorspace=piomatter.Colorspace.RGB888Packed, pinout=pinout, framebuffer=framebuffer,

pyproject.toml

@@ -3,6 +3,7 @@ requires = [
     "setuptools>=42",
     "pybind11>=2.10.0",
     "setuptools_scm[toml]>=6.2",
+    "cmake>=3.12"
 ]
 build-backend = "setuptools.build_meta"
 

setup.py

@@ -1,36 +1,80 @@
 # Available at setup time due to pyproject.toml
-from pybind11.setup_helpers import Pybind11Extension, build_ext
-from setuptools import setup
+import multiprocessing
+import os
+import subprocess
+import sys
+from pathlib import Path
+
+from setuptools import Extension, setup
+from setuptools.command.build_ext import build_ext
 from setuptools_scm import get_version
 
 __version__ = get_version()
 
+# A CMakeExtension needs a sourcedir instead of a file list.
+# The name must be the _single_ output extension from the CMake build.
+# If you need multiple extensions, see scikit-build.
+class CMakeExtension(Extension):
+    def __init__(self, name: str, sourcedir: str = "") -> None:
+        super().__init__(name, sources=[])
+        self.sourcedir = os.fspath(Path(sourcedir).resolve())
+
+
+class CMakeBuild(build_ext):
+    def build_extension(self, ext: CMakeExtension) -> None:
+        # Must be in this form due to bug in .resolve() only fixed in Python 3.10+
+        ext_fullpath = Path.cwd() / self.get_ext_fullpath(ext.name)
+        extdir = ext_fullpath.parent.resolve()
+
+        # Using this requires trailing slash for auto-detection & inclusion of
+        # auxiliary "native" libs
+
+        debug = int(os.environ.get("DEBUG", 0)) if self.debug is None else self.debug
+        cfg = "Debug" if debug else "Release"
+
+        # Set Python_EXECUTABLE instead if you use PYBIND11_FINDPYTHON
+        # EXAMPLE_VERSION_INFO shows you how to pass a value into the C++ code
+        # from Python.
+        cmake_args = [
+            "-GUnix Makefiles",
+            f"-DCMAKE_LIBRARY_OUTPUT_DIRECTORY={extdir}{os.sep}",
+            f"-DPYTHON_EXECUTABLE={sys.executable}",
+            f"-DCMAKE_BUILD_TYPE={cfg}",  # not used on MSVC, but no harm
+        ]
+        build_args = []
+        # Adding CMake arguments set as environment variable
+        # (needed e.g. to build for ARM OSx on conda-forge)
+        if "CMAKE_ARGS" in os.environ:
+            cmake_args += [item for item in os.environ["CMAKE_ARGS"].split(" ") if item]
+
+        build_args += [f"-j{multiprocessing.cpu_count()}"]
+
+        build_temp = Path(self.build_temp) / ext.name
+        if not build_temp.exists():
+            build_temp.mkdir(parents=True)
+
+        subprocess.run(
+            ["cmake", ext.sourcedir, *cmake_args], cwd=build_temp, check=True
+        )
+        subprocess.run(
+            ["make", *build_args], cwd=build_temp, check=True
+        )
+
+
+
 # The main interface is through Pybind11Extension.
 # * You can add cxx_std=11/14/17, and then build_ext can be removed.
 # * You can set include_pybind11=false to add the include directory yourself,
 #   say from a submodule.
 
-ext_modules = [
-    Pybind11Extension("adafruit_blinka_raspberry_pi5_piomatter._piomatter",
-        ["src/pymain.cpp", "src/piolib/piolib.c", "src/piolib/pio_rp1.c"],
-        define_macros = [('VERSION_INFO', __version__)],
-        include_dirs = ['./src/include', './src/piolib/include'],
-        cxx_std=20,
-        # use this setting when debugging
-        extra_compile_args = ["-g3", "-Og"],
-        ),
-]
-
 setup(
     name="Adafruit-Blinka-Raspberry-Pi5-Piomatter",
     version=__version__,
     url="https://github.com/adafruit/Adafruit_Blinka_Raspberry_Pi5_Piomatter",
     description="HUB75 matrix driver for Raspberry Pi 5 using PIO",
     long_description="A pio-based driver",
-    ext_modules=ext_modules,
-    # Currently, build_ext only provides an optional "highest supported C++
-    # level" feature, but in the future it may provide more features.
-    cmdclass={"build_ext": build_ext},
+    ext_modules=[CMakeExtension("adafruit_blinka_raspberry_pi5_piomatter._piomatter")],
+    cmdclass={"build_ext": CMakeBuild},
     zip_safe=False,
     python_requires=">=3.11",
     packages=['adafruit_blinka_raspberry_pi5_piomatter'],

src/include/piomatter/pins.h

@@ -41,25 +41,6 @@ struct adafruit_matrix_bonnet_pinout_bgr {
 };
 
 struct active3_pinout {
-    static constexpr pin_t PIN_RGB[] = {11, 27, 7,  8,  9, 10, 12, 5,  6,
-                                        19, 13, 20, 14, 2, 3,  26, 16, 21};
-    static constexpr pin_t PIN_ADDR[] = {22, 23, 24, 25, 15};
-    static constexpr pin_t PIN_OE = 18;  // /OE: output enable when LOW
-    static constexpr pin_t PIN_CLK = 17; // SRCLK: clocks on RISING edge
-    static constexpr pin_t PIN_LAT = 4;  // RCLK: latches on RISING edge
-
-    static constexpr uint32_t clk_bit = 1u << PIN_CLK;
-    static constexpr uint32_t lat_bit = 1u << PIN_LAT;
-    static constexpr uint32_t oe_bit = 1u << PIN_OE;
-    static constexpr uint32_t oe_active = 0;
-    static constexpr uint32_t oe_inactive = oe_bit;
-
-    static constexpr uint32_t post_oe_delay = 0;
-    static constexpr uint32_t post_latch_delay = 0;
-    static constexpr uint32_t post_addr_delay = 5;
-};
-
-struct active3_pinout_bgr {
     static constexpr pin_t PIN_RGB[] = {7,  27, 11, 10, 9, 8,  6,  5,  12,
                                         20, 13, 19, 3,  2, 14, 21, 16, 26};
     static constexpr pin_t PIN_ADDR[] = {22, 23, 24, 25, 15};
@@ -78,4 +59,23 @@ struct active3_pinout_bgr {
     static constexpr uint32_t post_addr_delay = 5;
 };
 
+struct active3_pinout_bgr {
+    static constexpr pin_t PIN_RGB[] = {11, 27, 7,  8,  9, 10, 12, 5,  6,
+                                        19, 13, 20, 14, 2, 3,  26, 16, 21};
+    static constexpr pin_t PIN_ADDR[] = {22, 23, 24, 25, 15};
+    static constexpr pin_t PIN_OE = 18;  // /OE: output enable when LOW
+    static constexpr pin_t PIN_CLK = 17; // SRCLK: clocks on RISING edge
+    static constexpr pin_t PIN_LAT = 4;  // RCLK: latches on RISING edge
+
+    static constexpr uint32_t clk_bit = 1u << PIN_CLK;
+    static constexpr uint32_t lat_bit = 1u << PIN_LAT;
+    static constexpr uint32_t oe_bit = 1u << PIN_OE;
+    static constexpr uint32_t oe_active = 0;
+    static constexpr uint32_t oe_inactive = oe_bit;
+
+    static constexpr uint32_t post_oe_delay = 0;
+    static constexpr uint32_t post_latch_delay = 0;
+    static constexpr uint32_t post_addr_delay = 5;
+};
+
 } // namespace piomatter

src/include/piomatter/piomatter.h

@@ -14,32 +14,24 @@ namespace piomatter {
 
 static int pio_sm_xfer_data_large(PIO pio, int sm, int direction, size_t size,
                                   uint32_t *databuf) {
-#if 0
-    // it would be NICE to gracefully fall back to blocked transfer, but sadly
-    // once the large xfer ioctl fails, future small xfers fail too.
-    static enum { UNKNOWN, OK, BAD } large_xfer_status = UNKNOWN;
-    printf("large_xfer_status=%d\n", large_xfer_status);
-    if (large_xfer_status != BAD) {
-        int r = pio_sm_xfer_data(pio, sm, direction, size, databuf);
-        if (large_xfer_status == UNKNOWN && r != 0) {
-            large_xfer_status = BAD;
-            fprintf(stderr,
-                    "Transmission limit workaround engaged. May reduce quality of "
-                    "output.\nSee https://github.com/raspberrypi/utils/issues/123 "
-                    "for details.\n");
-        } else {
-            if (large_xfer_status == UNKNOWN && r == 0) {
-                large_xfer_status = OK;
-            }
-            return r;
-        }
+    int r = pio_sm_xfer_data(pio, sm, direction, size, databuf);
+    if (r != -EPERM) {
+        // Kernels that don't support large transfers return EPERM
+        return r;
+    }
+    static bool once = false;
+    if (!once) {
+        once = true;
+        fprintf(stderr,
+                "Transmission limit workaround engaged. May reduce quality of "
+                "output.\nSee https://github.com/raspberrypi/utils/issues/123 "
+                "for details.");
     }
-#endif
     constexpr size_t MAX_XFER = 65532;
     while (size) {
         size_t xfersize = std::min(size_t{MAX_XFER}, size);
-        int r = pio_sm_xfer_data(pio, sm, direction, xfersize, databuf);
-        if (r != 0) {
+        r = pio_sm_xfer_data(pio, sm, direction, xfersize, databuf);
+        if (r) {
             return r;
         }
         size -= xfersize;
@@ -220,9 +212,9 @@ struct piomatter : piomatter_base {
                 int r = pio_sm_xfer_data_large(pio, sm, PIO_DIR_TO_SM, datasize,
                                                dataptr);
                 if (r != 0) {
-                    pending_error_errno.store(errno);
-                    printf("xfer_data() returned error %d (errno=%s)\n", r,
-                           strerror(errno));
+                    pending_error_errno.store(-r);
+                    printf("xfer_data() returned error %d (%s)\n", r,
+                           strerror(-r));
                 }
                 t1 = monotonicns64();
                 if (t0 != t1) {