atsamd2x: improve USBCDC

[sago35]

This PR improves the USBCDC speed of atsamd2x.
The atsamd5x version is #1481 .

I checked it on Windows 10 and found the following

• 13 byte : fmt.Printf("hello world\r\n")
  • old : 648.298 us
  • new : 88.284 us
• 11 byte : println("hello world")
  • old : 592.844 us
  • new : 22.224 us
• 1024 byte + "\r\n"
  • old : 51.735 ms
  • new : 1.906 ms
• 1024 * 10 byte + "\r\n"
  • old : 513.601 ms
  • new : 18.888ms

[sago35]

The differences from atsamd5x version are as follows.

diff --git a/5x.txt b/2x.txt
index 023b2d8..8f36d1a 100644
--- a/5x.txt
+++ b/2x.txt
@@ -1,96 +1,96 @@
 // USBCDC is the USB CDC aka serial over USB interface on the SAMD21.
 type USBCDC struct {
 	Buffer  *RingBuffer
 	TxIdx   volatile.Register8
 	waitTxc bool
 	sent    bool
 }
 
 const (
 	usbcdcTxSizeMask uint8 = 0x3F
 	usbcdcTxBankMask uint8 = ^usbcdcTxSizeMask
 	usbcdcTxBank1st  uint8 = 0x00
 	usbcdcTxBank2nd  uint8 = usbcdcTxSizeMask + 1
 )
 
 // Flush flushes buffered data.
 func (usbcdc *USBCDC) Flush() error {
 	if usbLineInfo.lineState > 0 {
 		idx := usbcdc.TxIdx.Get()
 		sz := idx & usbcdcTxSizeMask
 		bk := idx & usbcdcTxBankMask
 		if 0 < sz {
 
 			if usbcdc.waitTxc {
 				// waiting for the next flush(), because the transmission is not complete
 				return nil
 			}
 			usbcdc.waitTxc = true
 
 			// set the data
 			usbEndpointDescriptors[usb_CDC_ENDPOINT_IN].DeviceDescBank[1].ADDR.Set(uint32(uintptr(unsafe.Pointer(&udd_ep_in_cache_buffer[usb_CDC_ENDPOINT_IN][bk]))))
 			if bk == usbcdcTxBank1st {
 				usbcdc.TxIdx.Set(usbcdcTxBank2nd)
 			} else {
 				usbcdc.TxIdx.Set(usbcdcTxBank1st)
 			}
 
 			// clean multi packet size of bytes already sent
 			usbEndpointDescriptors[usb_CDC_ENDPOINT_IN].DeviceDescBank[1].PCKSIZE.ClearBits(usb_DEVICE_PCKSIZE_MULTI_PACKET_SIZE_Mask << usb_DEVICE_PCKSIZE_MULTI_PACKET_SIZE_Pos)
 
 			// set count of bytes to be sent
 			usbEndpointDescriptors[usb_CDC_ENDPOINT_IN].DeviceDescBank[1].PCKSIZE.ClearBits(usb_DEVICE_PCKSIZE_BYTE_COUNT_Mask << usb_DEVICE_PCKSIZE_BYTE_COUNT_Pos)
 			usbEndpointDescriptors[usb_CDC_ENDPOINT_IN].DeviceDescBank[1].PCKSIZE.SetBits((uint32(sz) & usb_DEVICE_PCKSIZE_BYTE_COUNT_Mask) << usb_DEVICE_PCKSIZE_BYTE_COUNT_Pos)
 
 			// clear transfer complete flag
-			setEPINTFLAG(usb_CDC_ENDPOINT_IN, sam.USB_DEVICE_ENDPOINT_EPINTFLAG_TRCPT1)
+			setEPINTFLAG(usb_CDC_ENDPOINT_IN, sam.USB_DEVICE_EPINTFLAG_TRCPT1)
 
 			// send data by setting bank ready
-			setEPSTATUSSET(usb_CDC_ENDPOINT_IN, sam.USB_DEVICE_ENDPOINT_EPSTATUSSET_BK1RDY)
+			setEPSTATUSSET(usb_CDC_ENDPOINT_IN, sam.USB_DEVICE_EPSTATUSSET_BK1RDY)
 			UART0.sent = true
 		}
 	}
 	return nil
 }
 
 // WriteByte writes a byte of data to the USB CDC interface.
 func (usbcdc *USBCDC) WriteByte(c byte) error {
 	// Supposedly to handle problem with Windows USB serial ports?
 	if usbLineInfo.lineState > 0 {
 		ok := false
 		for {
 			mask := interrupt.Disable()
 
 			idx := UART0.TxIdx.Get()
 			if (idx & usbcdcTxSizeMask) < usbcdcTxSizeMask {
 				udd_ep_in_cache_buffer[usb_CDC_ENDPOINT_IN][idx] = c
 				UART0.TxIdx.Set(idx + 1)
 				ok = true
 			}
 
 			interrupt.Restore(mask)
 
 			if ok {
 				break
 			} else {
 				mask := interrupt.Disable()
 				if UART0.sent {
 					if UART0.waitTxc {
-						if (getEPINTFLAG(usb_CDC_ENDPOINT_IN) & sam.USB_DEVICE_ENDPOINT_EPINTFLAG_TRCPT1) != 0 {
-							setEPSTATUSCLR(usb_CDC_ENDPOINT_IN, sam.USB_DEVICE_ENDPOINT_EPSTATUSCLR_BK1RDY)
-							setEPINTFLAG(usb_CDC_ENDPOINT_IN, sam.USB_DEVICE_ENDPOINT_EPINTFLAG_TRCPT1)
+						if (getEPINTFLAG(usb_CDC_ENDPOINT_IN) & sam.USB_DEVICE_EPINTFLAG_TRCPT1) != 0 {
+							setEPSTATUSCLR(usb_CDC_ENDPOINT_IN, sam.USB_DEVICE_EPSTATUSCLR_BK1RDY)
+							setEPINTFLAG(usb_CDC_ENDPOINT_IN, sam.USB_DEVICE_EPINTFLAG_TRCPT1)
 							UART0.waitTxc = false
 							UART0.Flush()
 						}
 					} else {
 						UART0.Flush()
 					}
 				}
 				interrupt.Restore(mask)
 			}
 		}
 	}
 
 	return nil
 }
 

[deadprogram]

Fantastic work @sago35 you are so fast! Almost as fast as the new USBCDC code 🏎️

Now merging. Thanks!

[deadprogram]

Was just looking at https://github.com/tinygo-org/tinygo/blob/dev/src/machine/usb_nrf52840.go which is structured in a very similar way. Seems like a similar change could be made there as well.

[sago35]

@deadprogram
I think nrf can be sped up in the same way.
I have feather-nrf52840, so I'll check it out.

[deadprogram]

Oh please oh please! I really need it.