[object Object][object Object][object Object]undefined[object Object]

After matrix computation is completed, quantize the result, and then move the processed result to GM. There are 9 quantization modes:

• NoQuant: Disables the quantization function.
• F322F16: Quantizes float to half. The quantization result does not support the INF_NAN mode.
• F322BF16: Quantizes float to bfloat16_t. The quantization result does not support the INF_NAN mode.
• DEQF16: Quantizes int32_t to half. The quantization result does not support the INF_NAN mode.
• VDEQF16: Quantizes int32_t to half. The quantization result does not support the INF_NAN mode.
• QF322B8_PRE: Quantizes float to uint8_t/int8_t. Scalar quantization.
• VQF322B8_PRE: Quantizes float to uint8_t/int8_t. Vector quantization.
• REQ8: Quantizes int32_t to uint8_t/int8_t. Scalar quantization.
• VREQ8: Quantizes int32_t to uint8_t/int8_t. Vector quantization.

[object Object]

• Regular computation

  [object Object]

• Synchronous computation

  [object Object]

[object Object][object Object]undefined[object Object]

None

[object Object]

PIPE_MTE1

[object Object]

• The start addresses of dst and src must be 32-byte aligned.
• For operand address overlap constraints, refer to .

[object Object]