This section lists the support scope of each UCP function module by the combination of operating system and platform, which helps you confirm whether a function is available on the operating system and the platform you are using. For the platform differences on the model quantization and deployment path, such as quantization strategy, output precision, alignment requirements, and deployment capabilities, see the Platform Differences section.

The sections below are grouped by the two operating systems Linux and QNX, and each group lists, by computing platform, the differences in supported features in the current version of UCP. Function points not mentioned in the sections below are assumed to have no difference in support between platforms and operating systems.

#Linux

Modules	Function	J6E/M	J6B	J6H/P	S6P
Sample	Full process of target detection example	Supported	Not supported	Not supported	Only BPU-related model inference capabilities are supported. Modules and examples that rely on other IPs do not apply
Image processing	GDC, PYM and CODEC processing of 10-bit images	Not supported	Supported	Not supported
	GDC and PYM processing of 12-bit images	Not supported	Supported	Not supported
Plugin operator	FRAME_PADDING_REFLECT_101 edge padding mode in VDSP operator development	Supported	Not supported	Supported

#QNX

The following table lists the limitations common to all computing platforms under QNX:

Modules	Function	J6E/M	J6B	J6H/P	S6P
Sample	Full process of target detection example	Not supported	Not supported	Not supported	Only BPU-related model inference capabilities are supported. Modules and examples that rely on other IPs do not apply
Tool	Perfetto Trace tool in the UCP performance tool	Not supported	Not supported	Not supported
	DEB packaging tool	Not supported	Not supported	Not supported
Operating mode	Relay mode	Not supported	Not supported	Not supported
Plugin operator	GPU custom operator development related to OpenCL	Not supported	Not supported	Not supported

In addition to the above limitations common to QNX, J6B also has other limitations listed in the following table:

Modules	Function	J6B
HPL	HPL API	Not supported
	HPL Sample	Not supported
VP	VP operator based on JPU	Not supported
	Surround view image stitching example	Not supported
	Bilinear pyramid	The maximum supported number of layers is 3, and the maximum size supported by PYM1 is 2592x2592.
Plugin operator	Slam example in DSP	Not supported
	FRAME_PADDING_REFLECT_101 edge padding mode in VDSP operator development	Not supported

#Additional Notes on Platform Capabilities

• The number of DSP cores varies by platform. J6E/M/B provides 1 core and J6H/P provides 2 cores. For the complete feature indicators of each hardware IP, see the Hardware Feature Description section of Overview. For the mapping between core0/core1 in the DSP debugging commands and the hardware cores, as well as the commands that apply to J6P only, see the FAQ section.

• On J6B, VDSP Q8 supports functional safety by default, while VDSP V130 supports limited functional safety only. It can be used only after you adapt the functional safety version of the libraries and header files and enable the related compilation options. For the detailed adaptation steps, see the FAQ section.

#Where to Look Next

• To confirm the number, the maximum and minimum input sizes, and the supported formats of each hardware IP, see the Hardware Feature Description section of Overview.

• To confirm the platform differences on the model quantization and deployment path, see the Platform Differences section.

• If an exception occurs on the board or you need to confirm the meaning of an error code, see the FAQ and Error Code sections.