diff options
| author |  Sanjoy Das <sanjoy@google.com> | 2018-07-27 13:24:46 -0700 |
|---|---|---|
| committer |  TensorFlower Gardener <gardener@tensorflow.org> | 2018-07-27 13:28:38 -0700 |
| commit | 388d0d860110a19a9d133fe4de85f8f6fa060cde (patch) | |
| tree | ce12ca4a32266f4a4bc5101e65b0123f810f1373 /tensorflow/compiler/xla/service/gpu/ir_emitter_unnested.cc | |
| parent | 90fe37ab8d056c56a5127e9b7ae237c04a7907ec (diff) | |
Use constant buffer allocations for XLA:CPU
This is simpler than the corresponding change to XLA:GPU because on XLA:CPU all
instructions are codegened so we can always embed a pointer to the constant
global variable directly in the generated LLVM IR.
PiperOrigin-RevId: 206363887
Diffstat (limited to 'tensorflow/compiler/xla/service/gpu/ir_emitter_unnested.cc')
| -rw-r--r-- | tensorflow/compiler/xla/service/gpu/ir_emitter_unnested.cc | 10 |
1 files changed, 6 insertions, 4 deletions
diff --git a/tensorflow/compiler/xla/service/gpu/ir_emitter_unnested.cc b/tensorflow/compiler/xla/service/gpu/ir_emitter_unnested.cc index fb9540b7ef..3a5394dac6 100644 --- a/tensorflow/compiler/xla/service/gpu/ir_emitter_unnested.cc +++ b/tensorflow/compiler/xla/service/gpu/ir_emitter_unnested.cc @@ -60,6 +60,7 @@ limitations under the License. #include "tensorflow/compiler/xla/service/hlo_computation.h" #include "tensorflow/compiler/xla/service/hlo_instruction.h" #include "tensorflow/compiler/xla/service/hlo_opcode.h" +#include "tensorflow/compiler/xla/service/llvm_ir/buffer_assignment_util.h" #include "tensorflow/compiler/xla/service/llvm_ir/dynamic_update_slice_util.h" #include "tensorflow/compiler/xla/service/llvm_ir/fused_ir_emitter.h" #include "tensorflow/compiler/xla/service/llvm_ir/kernel_support_library.h" @@ -2411,8 +2412,8 @@ std::unique_ptr<KernelThunk> IrEmitterUnnested::BuildKernelThunk( llvm::Value* loc; if (slice.allocation()->is_constant()) { loc = ir_emitter_context_->llvm_module()->getGlobalVariable( - llvm_ir::AsStringRef( - ConstantBufferAllocationToGlobalName(*slice.allocation()))); + llvm_ir::AsStringRef(llvm_ir::ConstantBufferAllocationToGlobalName( + *slice.allocation()))); CHECK_NE(loc, nullptr); } else { loc = b_.CreateInBoundsGEP(kernel_args.at(slice.allocation()), @@ -3428,7 +3429,7 @@ Status IrEmitterUnnested::EmitConstantGlobals() { continue; } - const Literal& literal = LiteralForConstantAllocation(allocation); + const Literal& literal = llvm_ir::LiteralForConstantAllocation(allocation); const bool should_emit_initializer = ShouldEmitLiteralInLlvmIr(literal); llvm::ArrayType* global_type = llvm::ArrayType::get(b_.getInt8Ty(), allocation.size()); @@ -3453,7 +3454,8 @@ Status IrEmitterUnnested::EmitConstantGlobals() { global_type, /*isConstant=*/should_emit_initializer, llvm::GlobalValue::ExternalLinkage, /*Initializer=*/initializer, - llvm_ir::AsStringRef(ConstantBufferAllocationToGlobalName(allocation))); + llvm_ir::AsStringRef( + llvm_ir::ConstantBufferAllocationToGlobalName(allocation))); global_for_const->setAlignment(kConstantBufferAlignBytes); ir_emitter_context_->llvm_module()->getGlobalList().push_back( global_for_const); |