/external/gemmlowp/meta/generators/ |
zip_Nx8_neon.py | 69 def GenerateLoadAggregateStore(emitter, lanes, output_address, alignment): 85 emitter.DereferenceIncrement(output_address, 64)) 89 output_address): 164 emitter.DereferenceIncrement(output_address, 64)) 167 def GenerateAggregatorReduction(emitter, registers, lanes, output_address, 211 emitter.Dereference(output_address, None)) 213 emitter.EmitVStore('1.32', low, emitter.Dereference(output_address, 64)) 216 emitter.DereferenceIncrement(output_address, 64)) 218 emitter.Dereference(output_address, None)) 221 emitter.DereferenceIncrement(output_address, 64) [all...] |
streams_common.py | 88 aggregators, output_address): 92 emitter, registers, aggregators, output_address, 100 aggregators, output_address): 104 emitter, registers, aggregators, output_address, 112 output_address, multiplicative_sum_offset, 137 emitter.Dereference(output_address, 251 output_address = registers.MapOutputParameter('out') 273 output_address) 283 stride, output_address) 287 self.asm_emitter, registers, aggregators, output_address) [all...] |
transform_kernels_common.py | 58 output_address): 74 emitter.EmitVStoreAE(8, elements, load, output_address, None) 75 emitter.EmitPld(output_address) 104 output_address): 143 emitter.EmitVStoreAE(32, elements, load, output_address, None) 144 emitter.EmitPld(output_address) 173 output_address): 212 emitter.EmitVStoreAE(8, elements, load, output_address, None) 213 emitter.EmitPld(output_address) 254 output_address) [all...] |
/external/tensorflow/tensorflow/compiler/xla/service/gpu/ |
ir_emitter.cc | 174 const HloComputation& computation, llvm::Value* output_address, 198 llvm::StoreInst* store = Store(source, output_address); 211 {output_address, source}, 212 {output_address->getType()}, &b_); 217 AtomicRMW(llvm::AtomicRMWInst::Add, output_address, source, 229 AtomicRMW(opcode, output_address, source, 239 AtomicRMW(opcode, output_address, source, 249 // 1. Reads the value from the memory pointed to by output_address and 255 // pointed to by output_address, and compares the value with old_output. If 268 // atomic binary operation. We mask off the last two bits of the output_address [all...] |
ir_emitter.h | 161 // sequentially consistent memory model. `output_address` and `source_address` 163 // atomicAdd(output_address, *source_address). 165 const HloComputation& nested_computation, llvm::Value* output_address, 207 llvm::Value* output_address, 214 llvm::Value* output_address, 229 // sequentially consistent memory model. `output_address` and `source_address` 231 // atomicAdd(output_address, *source_address).
|
ir_emitter_unnested.cc | [all...] |
/external/v8/src/ |
deoptimizer.cc | 85 deoptimizer_->QueueValueForMaterialization(output_address(top_offset_), obj, 98 Address output_address(unsigned output_offset) { function in class:v8::internal::FrameWriter 99 Address output_address = local 101 return output_address; 108 output_address(top_offset_), top_offset_, value, debug_hint); 116 output_address(output_offset), output_offset); [all...] |
deoptimizer.h | 534 void QueueValueForMaterialization(Address output_address, Object* obj, [all...] |
/external/tensorflow/tensorflow/compiler/xla/service/cpu/ |
ir_emitter.cc | [all...] |