STLINK-V3MODS VCP + BRIDGE VCP get half throughput

[jcphillips]

I’m having an issue with throughput on the STLINK-V3MODS VCP when used with the second BRIDGE UART.  If I set both the VCP and 2nd VCP for 115200 baud, I am only able to simultaneously send 5.76 KiB/s through each (half throughput).  If I use only the VCP at 115200, I’m able to send 11.52 KiB/s (full throughput).

I saw no mention of any such limitation in the documentation.  Am I doing something wrong?

STLINK-V3MODS firmware 3.17.11
Windows 11 version 25H2 (OS Build 26200.8875)

 

[S C]

Hello,

The topic of VCP throughput is mentioned in the STLINK-V3SET User manual (UM2448 rev9 chapter 13.2 STLINK-V3SET debugger/programmer for STM8 and STM32 - User manual). The phrasing is perhaps not perfect, but it is said that “the capability of the receiver to process all the data also affects the communication. With a heavily loaded line, data loss might occur at the STLINK-V3SET side above 12 MHz.”. This means than when the UART line is fully loaded, the STLINK-V3SET starts loosing data at 12MHz. Note that with a loaded PC, the limit might also be lower. The whole system, including the PC application, and the USB stack on both PC and STLINK-V3 side must be able to process all data before the ST-Link embedded Rx buffer gets saturated. Here the bottleneck is not at UART level. Knowing that, you understand that 1x11.52 KiB/s is equivalent to 2x5.76 KiB/s and it won’t be possible to get more.

[TDK]

If the entire system can’t process at 115200 something is wrong. These are not fast speeds.

My modem in the 1990s could do 56k baud.

 

I suspect the underlying issue is something else. Possibly in how the system is processing bytes.

[jcphillips]

I agree with this.  For an additional data point, I first noticed the issue when I was attempting 921.6k baud on both VCP ports.  If I use a single port, I get about 92k bytes/sec as expected.  If I attempt the same on both VCPs, I get half throughput on each.  Afterwards, I repeated the same behavior with 115200 baud.  I suspect I would also get the same behavior at 9600 baud - full throughput on a single port, but half if I use two.  I’ll try that tonight.

[jcphillips]

Confirmed.  This applies even for 9600 baud.  Using both VCPs cuts throughput to half of what should be possible with the baud rate.

 

[S C]

Hello,

I agree my first analysis was wrong for your case, I apologize for that. You are not at the limits of the system, all the more at 9600 baud. Then I have another doubt about my understanding: are you speaking about upstream data flow (from target to PC) or downstream ?

[jcphillips]

I’ve been testing downstream (PC to target) to validate target data reception.  I haven’t tested upstream throughput.

[S C]

OK I understand the point. The downstream is not well-structured at ST-Link level to correctly handle data bursts, I would say it is currently rather designed to handle small commands in this direction, and to favorize the upstream channel (printf functionality for instance). Are you developing an application requiring a high throughput downstream on both VCPs or do you discover the limitation during tests more “stressing” than the final use case ? Unfortunately, the fix is in conflict with other bridge features, so requires arbitration and is not straightforward