| /external/gemmlowp/meta/generators/ |
| zip_Nx8_neon.py | 21 def __init__(self, input_address, load, aggregator):
22 self.input_address = input_address
27 def GenerateZipLanes(emitter, registers, zip_lanes, input_address, stride):
34 input_address: register that contains the input address for the first lane.
41 last_address_register = input_address
44 lanes.append(ZipLane(input_address, registers.DoubleRegister(),
77 emitter.DereferenceIncrement(lane.input_address, alignment))
102 emitter.Dereference(lane.input_address, None))
107 emitter.Dereference(lane.input_address, None)
[all...] |
| mul_Nx8_Mx8_neon.py | 20 def __init__(self, input_address):
21 self.input_address = input_address
74 emitter.DereferenceIncrement(left_lanes.input_address, 64))
77 emitter.DereferenceIncrement(right_lanes.input_address, 64))
79 emitter.EmitPldOffset(left_lanes.input_address, emitter.ImmediateConstant(64))
80 emitter.EmitPldOffset(right_lanes.input_address,
120 emitter.DereferenceIncrement(left_lanes.input_address, 64))
123 emitter.DereferenceIncrement(right_lanes.input_address, 64))
125 emitter.EmitPldOffset(left_lanes.input_address, emitter.ImmediateConstant(64)
[all...] |
| streams_common.py | 37 def _GenerateInputs(emitter, registers, lanes_count, input_address, stride):
40 last_address_register = input_address
43 inputs.append(input_address)
69 for (row, input_address) in zip(block, inputs):
70 emitter.EmitVLoadE(8, elements_count, row, input_address, None)
206 elements_count, aggregators, input_address,
219 block, input_address, stride)
250 input_address = registers.MapOutputParameter('in')
272 aggregators, input_address, stride,
282 leftovers, aggregators, input_address,
[all...] |
| transform_kernels_common.py | 57 def Transform(self, emitter, registers, input_address, elements,
64 emitter.EmitVLoadAE(8, elements, load, input_address, None)
65 emitter.EmitPldOffset(input_address, emitter.ImmediateConstant(16))
103 def Transform(self, emitter, registers, input_address, elements,
110 emitter.EmitVLoadAE(8, elements, load, input_address, None)
111 emitter.EmitPldOffset(input_address, emitter.ImmediateConstant(32))
172 def Transform(self, emitter, registers, input_address, elements,
179 emitter.EmitVLoadAE(32, elements, load, input_address, None)
180 emitter.EmitPldOffset(input_address, emitter.ImmediateConstant(64))
253 def Transform(self, emitter, registers, input_address, elements
[all...] |
| neon_emitter.py | 717 input_address, stride):
722 input_deref = self.Dereference(input_address, None)
723 input_deref_increment = self.DereferenceIncrement(input_address, None)
729 self.EmitPld(input_address)
735 self.EmitPld(input_address)
746 self.EmitPld(input_address)
758 self.EmitPld(input_address)
771 self.EmitPld(input_address)
785 self.EmitPld(input_address)
793 self.EmitPld(input_address)
[all...] |
| neon_emitter_64.py | [all...] |
| quantized_mul_kernels_common.py | 19 def _ReadParams(emitter, registers, input_address, elements, min_register):
25 emitter.EmitVLoadAE(registers_count * 4, 32, registers, input_address, 64)
|
| /external/tensorflow/tensorflow/compiler/xla/service/gpu/ |
| ir_emitter_unnested.cc | [all...] |
| ir_emitter.cc | 672 llvm::Value* input_address = local
676 *function, {accumulator_addr, input_address}, accumulator_addr));
[all...] |
| /external/tensorflow/tensorflow/compiler/xla/service/cpu/ |
| ir_emitter.cc | [all...] |
