Lines Matching full:how

13 perform, specifically, it affects how one goes from internal 32bit accumulator
23 In the present document, our purpose is to show how, reasoning from first
25 naturally at some specific quantization paradigm, and how that can be
28 We also aim to show how that differs from the older, legacy quantization
133     how to eliminate all usage of floating-point arithmetic. That will come
138 Now that we know — equation (3) — how real numbers are to correspond
182 matrix multiplication should do, i.e. how to compute `result_quantized_value`.
202 Equation (5) is the conclusion of this general discussion of how to specify what
227 It is out of scope of the present doc to discuss how to avoid the overhead of
263 So how do we implement the multiplication of a int32 value by a positive real
277 ## How this is implemented in gemmlowp
309 ## How this differs from the older legacy gemmlowp quantization paradigm
320 output stages and how they differ.
333         on how it is used.
335 2.  Note how the order of multiplying by the multipler and adding the
344 Example code showing how to perfom a quantized matrix multiplication in the