diff --git a/rp2-pio/examples/hub40screen/hub40screen.go b/rp2-pio/examples/hub40screen/hub40screen.go
new file mode 100644
index 0000000..6dc0332
--- /dev/null
+++ b/rp2-pio/examples/hub40screen/hub40screen.go
@@ -0,0 +1,205 @@
+package main
+
+import (
+	"machine"
+	"runtime"
+	"time"
+	"unsafe"
+
+	pio "github.com/tinygo-org/pio/rp2-pio"
+	"github.com/tinygo-org/pio/rp2-pio/piolib"
+)
+
+func main() {
+	time.Sleep(time.Second)
+	const usePIO = true
+	var p6 Parallel6Bus
+	if usePIO {
+		sm, _ := pio.PIO0.ClaimStateMachine()
+		p6concrete, err := piolib.NewParallel6(sm, 12_000, 0, 6)
+		if err != nil {
+			panic(err)
+		}
+		err = p6concrete.EnableDMA(true)
+		if err != nil {
+			panic(err)
+		}
+		p6 = p6concrete
+	} else {
+		p6 = &bitbangParallel6{
+			rd0: makepinout(0),
+			gd0: makepinout(1),
+			bd0: makepinout(2),
+			rd1: makepinout(3),
+			gd1: makepinout(4),
+			bd1: makepinout(5),
+			clk: makepinout(6),
+		}
+	}
+	var emptyScreen [4][4]uint32
+	s := NewHUB40Screen(p6, makepinout(7), makepinout(8))
+	s.Draw(&emptyScreen)
+	for {
+		// testScreen(s)
+		testScreenBus(s)
+	}
+}
+
+// SetScreen sets a 3-bit RGB value at (ix, iy).
+// ix: 0..15, iy: 0..7
+func SetScreen(screen *[4][4]uint32, ix, iy int, rgb3bits uint8) {
+	if ix < 0 || ix >= 16 || iy < 0 || iy >= 8 {
+		return // or panic
+	}
+	colBlock := ix / 4 // 0..3
+	rowIdx := iy % 4   // 0..3 within the half
+	up := 1 - (iy / 4) // 0 = lower half, 1 = upper half
+
+	groupBase := (ix % 4) * 6 // 6 bits per group
+	bitoff := groupBase + up*3
+
+	mask := uint32(0b111) << bitoff
+	v := &screen[colBlock][rowIdx]
+	*v &^= mask
+	*v |= uint32(rgb3bits&0x7) << bitoff
+}
+
+type Parallel6Bus interface {
+	Tx24(buf []uint32) error
+}
+
+func NewHUB40Screen(bus Parallel6Bus, latch, outputEnable pinout) *HUB40Screen {
+	return &HUB40Screen{
+		lat: latch,
+		oe:  outputEnable,
+		p6:  bus,
+	}
+}
+
+func testScreen(s *HUB40Screen) {
+	var emptyscreen [4][4]uint32
+	var screen [4][4]uint32
+	s.Draw(&emptyscreen)
+	for iy := 0; iy < 8; iy++ {
+		for ix := 0; ix < 16; ix++ {
+			SetScreen(&screen, ix, iy, 0b101)
+			s.Draw(&screen)
+			time.Sleep(30 * time.Millisecond)
+		}
+	}
+	for range 2 {
+		s.Draw(&screen)
+		time.Sleep(time.Second / 3)
+		s.Draw(&emptyscreen)
+		time.Sleep(time.Second / 3)
+	}
+	for i := range screen {
+		for j := range screen[i] {
+			screen[i][j] = 0xffff_ffff
+		}
+	}
+	s.Draw(&screen)
+	time.Sleep(2 * time.Second)
+}
+
+func testScreenBus(s *HUB40Screen) {
+	var screen [4][4]uint32
+	buf := unsafe.Slice(&screen[0][0], 4*4)
+	for j := range buf {
+		for i := 0; i < 24; i++ {
+			mask := uint32(1) << i
+			buf[j] = mask
+			if i >= 24 {
+				println("off screen")
+			}
+			buf[0] |= 1
+			buf[len(buf)-1] |= 1 << 23
+			s.Draw(&screen)
+			time.Sleep(50 * time.Millisecond)
+		}
+	}
+}
+
+// HUB40Screen is a RGB single-bit-per-pixel screen that works on the basis of
+// several shift registers arranges in two rows. Each clock represents 6 bits
+// over the parallel-6 bus, so two LEDs per clock. The device I am using has
+// no markings other than "HUB40" and "OF-20R3A2-0816-V21JC" which is likely not a real model.
+// It seems to be a one-off design for a media company. Data lines are:
+//   - RD0..BD1: Six Red/Green/Blue data lines for row 0 and row 1.
+//   - OE: Output enable. Turns entire display on or off. Is negated.
+//   - LAT: Latch. Latches shift registers values. Basically switches display buffer to the one written most recently.
+type HUB40Screen struct {
+	// data pins.
+	lat pinout
+	oe  pinout
+	p6  Parallel6Bus
+}
+
+func (s *HUB40Screen) Latch() {
+	s.lat.High()
+	runtime.Gosched()
+	s.lat.Low()
+}
+
+func (s *HUB40Screen) Enable(b bool) {
+	println("enable", b)
+	s.oe(!b)
+}
+
+// Write writes the screen data without latching.
+func (s *HUB40Screen) Write(screen *[4][4]uint32) (err error) {
+	buf := unsafe.Slice(&screen[0][0], 4*4)
+	return s.p6.Tx24(buf)
+}
+
+// Draw writes and then latches screen data.
+func (s *HUB40Screen) Draw(screen *[4][4]uint32) (err error) {
+	err = s.Write(screen)
+	if err == nil {
+		s.Latch()
+	}
+	return err
+}
+
+type bitbangParallel6 struct {
+	rd0, gd0, bd0, rd1, gd1, bd1 pinout
+	clk                          pinout
+}
+
+func (s *bitbangParallel6) Tx24(buf []uint32) error {
+	for i := range buf {
+		s.clkout24(buf[i])
+	}
+	return nil
+}
+
+func (s *bitbangParallel6) clkout24(v uint32) {
+	const bits6 = 0b11_1111
+	s.rawclkout6(uint8(v & bits6))
+	s.rawclkout6(uint8((v >> 6) & bits6))
+	s.rawclkout6(uint8((v >> 12) & bits6))
+	s.rawclkout6(uint8((v >> 18) & bits6))
+}
+
+func (s *bitbangParallel6) rawclkout6(rgbBits uint8) {
+	s.clk.Low()
+	s.rd0(rgbBits&(1<<0) != 0)
+	s.gd0(rgbBits&(1<<1) != 0)
+	s.bd0(rgbBits&(1<<2) != 0)
+
+	s.rd1(rgbBits&(1<<3) != 0)
+	s.gd1(rgbBits&(1<<4) != 0)
+	s.bd1(rgbBits&(1<<5) != 0)
+	s.clk.High()
+	s.clk.Low()
+}
+
+type pinout func(level bool)
+
+func (p pinout) High() { p(true) }
+func (p pinout) Low()  { p(false) }
+
+func makepinout(pin machine.Pin) pinout {
+	pin.Configure(machine.PinConfig{Mode: machine.PinOutput})
+	return pin.Set
+}
diff --git a/rp2-pio/piolib/all_generate.go b/rp2-pio/piolib/all_generate.go
index 0bcbcaa..dcc1b7f 100644
--- a/rp2-pio/piolib/all_generate.go
+++ b/rp2-pio/piolib/all_generate.go
@@ -17,6 +17,7 @@ var (
 	errDMAUnavail = errors.New("piolib:DMA channel unavailable")
 )
 
+//go:generate pioasm -o go parallel6.pio         parallel6_pio.go
 //go:generate pioasm -o go parallel8.pio         parallel8_pio.go
 //go:generate pioasm -o go pulsar.pio            pulsar_pio.go
 //go:generate pioasm -o go spi.pio               spi_pio.go
diff --git a/rp2-pio/piolib/parallel6.go b/rp2-pio/piolib/parallel6.go
new file mode 100644
index 0000000..c9dd5c6
--- /dev/null
+++ b/rp2-pio/piolib/parallel6.go
@@ -0,0 +1,115 @@
+package piolib
+
+import (
+	"machine"
+
+	pio "github.com/tinygo-org/pio/rp2-pio"
+)
+
+// Parallel6 implements a 6-parallel bus.
+type Parallel6 struct {
+	sm        pio.StateMachine
+	dma       dmaChannel
+	rgboffset uint8
+}
+
+// NewParallel6 instantiates a 6-parallel bus with pins dataBase..dataBase+5 and clock pin.
+func NewParallel6(sm pio.StateMachine, baud uint32, dataBase machine.Pin, clock machine.Pin) (*Parallel6, error) {
+	sm.TryClaim()
+	Pio := sm.PIO()
+	whole, frac, err := pio.ClkDivFromFrequency(baud, machine.CPUFrequency())
+	if err != nil {
+		return nil, err
+	}
+	rgbOffset, err := Pio.AddProgram(parallel6Instructions, parallel6Origin)
+	if err != nil {
+		return nil, err
+	}
+	pinCfg := machine.PinConfig{Mode: Pio.PinMode()}
+	clock.Configure(pinCfg)
+	clkMask := uint32(1) << clock
+	var pinMask uint32 = clkMask
+	for i := 0; i < 6; i++ {
+		pin := dataBase + machine.Pin(i)
+		pinMask |= 1 << pin
+		pin.Configure(pinCfg)
+	}
+
+	// Configuration.
+	cfg := parallel6ProgramDefaultConfig(rgbOffset)
+	// Statemachine Pin configuration.
+	cfg.SetOutPins(dataBase, 6)
+	cfg.SetOutShift(true, true, 24)
+	cfg.SetSidesetPins(clock)
+
+	cfg.SetClkDivIntFrac(whole, frac)
+	cfg.SetFIFOJoin(pio.FifoJoinTx)
+
+	sm.SetPinsMasked(0, pinMask)
+	sm.SetPindirsMasked(pinMask, pinMask)
+	sm.Init(rgbOffset, cfg)
+	sm.SetEnabled(true)
+	p6 := Parallel6{
+		sm:        sm,
+		rgboffset: rgbOffset,
+	}
+	return &p6, nil
+}
+
+// IsEnabled returns true if the state machine on the Parallel6 is enabled and ready to transmit.
+func (p6 *Parallel6) IsEnabled() bool {
+	return p6.sm.IsEnabled()
+}
+
+// SetEnabled enables or disables the state machine.
+func (p6 *Parallel6) SetEnabled(b bool) {
+	p6.sm.SetEnabled(b)
+}
+
+// Tx24 transmits 6-parallel data over pins. Each 32 bit value contains 24 effective bits
+// making a total of 4 clocks.
+func (p6 *Parallel6) Tx24(data []uint32) (err error) {
+	p6.sm.ClearTxStalled()
+	if p6.IsDMAEnabled() {
+		err = p6.tx24DMA(data)
+	} else {
+		err = p6.tx24(data)
+	}
+	if err != nil {
+		return err
+	}
+	for !p6.sm.HasTxStalled() {
+		gosched() // Block until empty.
+	}
+	return nil
+}
+
+func (p6 *Parallel6) tx24(data []uint32) error {
+	i := 0
+	for i < len(data) {
+		if p6.sm.IsTxFIFOFull() {
+			gosched()
+			continue
+		}
+		p6.sm.TxPut(data[i])
+		i++
+	}
+	return nil
+}
+
+func (p6 *Parallel6) IsDMAEnabled() bool {
+	return p6.dma.helperIsEnabled()
+}
+
+func (p6 *Parallel6) EnableDMA(enabled bool) error {
+	return p6.dma.helperEnableDMA(enabled)
+}
+
+func (p6 *Parallel6) tx24DMA(data []uint32) error {
+	dreq := dmaPIO_TxDREQ(p6.sm)
+	err := p6.dma.Push32(&p6.sm.TxReg().Reg, data, dreq)
+	if err != nil {
+		return err
+	}
+	return nil
+}
diff --git a/rp2-pio/piolib/parallel6.pio b/rp2-pio/piolib/parallel6.pio
new file mode 100644
index 0000000..c401556
--- /dev/null
+++ b/rp2-pio/piolib/parallel6.pio
@@ -0,0 +1,18 @@
+.program parallel6
+.side_set 1
+
+.wrap_target 
+    out pins, 6      side 0   ; write OSR databits into pin
+    nop              side 1
+    nop              side 0
+.wrap
+
+% go {
+//go:build rp2040 || rp2350
+
+package piolib
+
+import (
+	pio "github.com/tinygo-org/pio/rp2-pio"
+)
+%}
\ No newline at end of file
diff --git a/rp2-pio/piolib/parallel6_pio.go b/rp2-pio/piolib/parallel6_pio.go
new file mode 100644
index 0000000..df3b531
--- /dev/null
+++ b/rp2-pio/piolib/parallel6_pio.go
@@ -0,0 +1,27 @@
+// This file is autogenerated by pioasm version 2.2.0; do not edit!
+
+//go:build rp2040 || rp2350
+package piolib
+import (
+	pio "github.com/tinygo-org/pio/rp2-pio"
+)
+// parallel6
+
+const parallel6WrapTarget = 0
+const parallel6Wrap = 2
+
+var parallel6Instructions = []uint16{
+		//     .wrap_target
+		0x6006, //  0: out    pins, 6         side 0
+		0xb042, //  1: nop                    side 1
+		0xa042, //  2: nop                    side 0
+		//     .wrap
+}
+const parallel6Origin = -1
+func parallel6ProgramDefaultConfig(offset uint8) pio.StateMachineConfig {
+	cfg := pio.DefaultStateMachineConfig()
+	cfg.SetWrap(offset+parallel6WrapTarget, offset+parallel6Wrap)
+	cfg.SetSidesetParams(1, false, false)
+	return cfg;
+}
+
diff --git a/rp2-pio/statemachine.go b/rp2-pio/statemachine.go
index 565cb0b..2381ffa 100644
--- a/rp2-pio/statemachine.go
+++ b/rp2-pio/statemachine.go
@@ -167,6 +167,17 @@ func (sm StateMachine) TxFIFOLevel() uint32 {
 	return (sm.pio.hw.FLEVEL.Get() >> uint32(bitoffs)) & mask
 }
 
+// IsTxStalled returns true if state machine has stalled.
+// This value is sticky so it must be cleared with [StateMachine.ClearTxStalled] after reading true to be reset.
+func (sm StateMachine) HasTxStalled() bool {
+	return sm.pio.hw.FDEBUG.HasBits(1 << (rp.PIO0_FDEBUG_TXSTALL_Pos + sm.index))
+}
+
+// ClearTxStalled clears the value of tx stall. See [StateMachine.HasTxStalled].
+func (sm StateMachine) ClearTxStalled() {
+	sm.pio.hw.FDEBUG.Set(1 << (rp.PIO0_FDEBUG_TXSTALL_Pos + sm.index))
+}
+
 // IsTxFIFOEmpty returns true if state machine's TX FIFO is empty.
 func (sm StateMachine) IsTxFIFOEmpty() bool {
 	return (sm.pio.hw.FSTAT.Get() & (1 << (rp.PIO0_FSTAT_TXEMPTY_Pos + sm.index))) != 0