diff --git a/src/hal.zig b/src/hal.zig
index 41f4744..ae181da 100644
--- a/src/hal.zig
+++ b/src/hal.zig
@@ -2,6 +2,7 @@ const microzig = @import("microzig");
 const regs = microzig.chip.regsisters;
 pub const gpio = @import("hal/gpio.zig");
 pub const clocks = @import("hal/clocks.zig");
+pub const multicore = @import("hal/multicore.zig");
 
 pub const default_clock_config = clocks.GlobalConfiguration.init(.{
     //.ref = .{ .source = .src_xosc },
diff --git a/src/hal/multicore.zig b/src/hal/multicore.zig
new file mode 100644
index 0000000..5c95c13
--- /dev/null
+++ b/src/hal/multicore.zig
@@ -0,0 +1,109 @@
+const std = @import("std");
+const microzig = @import("microzig");
+const regs = microzig.chip.registers;
+const assert = std.debug.assert;
+
+pub const fifo = struct {
+    /// Check if the FIFO has valid data for reading.
+    pub fn is_read_ready() bool {
+        return regs.SIO.FIFO_ST.read().VLD == 1;
+    }
+
+    /// Read from the FIFO
+    /// Will return null if it is empty.
+    pub fn read() ?u32 {
+        if (!is_read_ready())
+            return null;
+
+        return regs.SIO.FIFO_RD.*;
+    }
+
+    /// Read from the FIFO, waiting for data if there is none.
+    pub fn read_blocking() u32 {
+        while (true) {
+            if (read()) |value| return value;
+            microzig.cpu.wfe();
+        }
+    }
+
+    /// Read from the FIFO, and throw everyting away.
+    pub fn drain() void {
+        while (read()) |_| {}
+    }
+
+    /// Check if the FIFO is ready to receive data.
+    pub fn is_write_ready() bool {
+        return regs.SIO.FIFO_ST.read().RDY == 1;
+    }
+
+    /// Write to the FIFO
+    /// You must check if there is space by calling is_write_ready
+    pub fn write(value: u32) void {
+        regs.SIO.FIFO_WR.* = value;
+        microzig.cpu.sev();
+    }
+
+    /// Write to the FIFO, waiting for room if it is full.
+    pub fn write_blocking(value: u32) void {
+        while (!is_write_ready())
+            std.mem.doNotOptimizeAway(value);
+
+        write(value);
+    }
+};
+
+var core1_stack: [128]u32 = undefined;
+
+/// Runs `entrypoint` on the second core.
+pub fn launchCore1(entrypoint: fn () void) void {
+    launchCore1WithStack(entrypoint, &core1_stack);
+}
+
+pub fn launchCore1WithStack(entrypoint: fn () void, stack: []u32) void {
+    // TODO: disable SIO interrupts
+
+    const wrapper = struct {
+        fn wrapper(_: u32, _: u32, _: u32, _: u32, entry: u32, stack_base: [*]u32) callconv(.C) void {
+            // TODO: protect stack using MPU
+            _ = stack_base;
+            @intToPtr(fn () void, entry)();
+        }
+    }.wrapper;
+
+    // reset the second core
+    regs.PSM.FRCE_OFF.modify(.{ .proc1 = 1 });
+    while (regs.PSM.FRCE_OFF.read().proc1 != 1) microzig.cpu.nop();
+    regs.PSM.FRCE_OFF.modify(.{ .proc1 = 0 });
+
+    stack[stack.len - 2] = @ptrToInt(entrypoint);
+    stack[stack.len - 1] = @ptrToInt(stack.ptr);
+
+    // calculate top of the stack
+    const stack_ptr: u32 =
+        @ptrToInt(stack.ptr) +
+        (stack.len - 2) * @sizeOf(u32); // pop the two elements we "pushed" above
+
+    // after reseting core1 is waiting for this specific sequence
+    const cmds: [6]u32 = .{
+        0,
+        0,
+        1,
+        regs.PPB.VTOR.raw,
+        stack_ptr,
+        @ptrToInt(wrapper),
+    };
+
+    var seq: usize = 0;
+    while (seq < cmds.len) {
+        const cmd = cmds[seq];
+        if (cmd == 0) {
+            // always drain the fifo before sending zero
+            fifo.drain();
+            microzig.cpu.sev();
+        }
+
+        fifo.write_blocking(cmd);
+        // the second core should respond with the same value, if it doesnt't lets start over
+        seq = if (cmd == fifo.read_blocking()) seq + 1 else 0;
+    }
+}