#include "Halide.h" #include using namespace Halide; int main(int argc, char *argv[]) { { Buffer input(1024, 1024, 3); for (int c = 0; c < input.channels(); c++) { for (int y = 0; y < input.height(); y++) { for (int x = 0; x < input.width(); x++) { input(x, y, c) = x + y + c; } } } Var x("x"), y("y"), c("c"); Func clamped_input = Halide::BoundaryConditions::repeat_edge(input); // One of the possible conditions for partitioning loop 'f.s0.x' is // ((f.s0.x + g[0]) <= 1023) which depends on 'g'. Since 'g' is // only allocated inside f.s0.x, partition loops should not use this // condition to compute the epilogue/prologue. Func f("f"), g("g"), h("h"); g(x, y, c) = x + y + c; g(x, y, 0) = x; h(x, y) = clamped_input(x + g(x, y, 0), y, 2); f(x, y, c) = select(h(x, y) < x + y, x + y, y + c); f.compute_root(); Func output("output"); output(x, y, c) = cast(f(x, y, c)); Buffer im = output.realize({1024, 1024, 3}); for (int y = 0; y < input.height(); y++) { for (int x = 0; x < input.width(); x++) { for (int c = 0; c < input.channels(); c++) { float correct = (input(std::min(2 * x, input.width() - 1), y, 2) < x + y) ? x + y : y + c; if (im(x, y, c) != correct) { printf("im(%d, %d, %d) = %f instead of %f\n", x, y, c, im(x, y, c), correct); return 1; } } } } } // A loop partitioning bug from https://github.com/halide/Halide/issues/7742 { Var x, y, x_outer, x_inner, y_x_inner_fused; Func f; f(x, y) = x + y; f.split(x, x_outer, x_inner, 2, TailStrategy::PredicateStores).fuse(y, x_inner, y_x_inner_fused); Pipeline p(f); p.realize({100, 100}); } printf("Success!\n"); return 0; }