| /external/tensorflow/tensorflow/compiler/xla/service/ |
| D | multi_output_fusion.cc | 138 HloInstruction* MultiOutputFusion::Fuse(HloInstruction* instr1, in Fuse() argument
140 HloInstruction* remaining = instr1; in Fuse()
199 MultiOutputFusion::GetNewFusibles(HloInstruction* instr1, in GetNewFusibles() argument
201 HloInstruction* fusion = instr1; in GetNewFusibles()
203 if (is_fused(instr1)) { in GetNewFusibles()
205 fused = instr1; in GetNewFusibles()
246 void MultiOutputFusion::UpdateBeforeFuse(HloInstruction* instr1, in UpdateBeforeFuse() argument
248 HloInstruction* fusion = instr1; in UpdateBeforeFuse()
250 if (is_fused(instr1)) { in UpdateBeforeFuse()
252 fused = instr1; in UpdateBeforeFuse()
[all …]
|
| D | multi_output_fusion.h | 65 virtual bool ShapesCompatibleForFusion(HloInstruction* instr1,
74 virtual int64 GetProfit(HloInstruction* instr1, HloInstruction* instr2) = 0;
80 virtual bool LegalToFuse(HloInstruction* instr1, HloInstruction* instr2);
84 bool LegalToFuseMainConstraints(HloInstruction* instr1,
89 virtual HloInstruction* Fuse(HloInstruction* instr1, HloInstruction* instr2);
102 HloInstruction* instr1, HloInstruction* instr2,
118 HloInstruction* instr1, HloInstruction* instr2);
136 HloInstruction* instr1; member
140 ToBeFused(HloInstruction* instr1, HloInstruction* instr2, int64 score, in ToBeFused()
142 : instr1(instr1), instr2(instr2), score(score), timestamp(timestamp) {} in ToBeFused()
[all …]
|
| D | hlo_instruction_test.cc | 1829 auto instr1 = HloInstruction::CreateCustomCall(ShapeUtil::MakeShape(F32, {}), in TEST_F() local
1832 auto instr2 = instr1->Clone(); in TEST_F()
1833 EXPECT_TRUE(instr1->Identical(*instr2)); in TEST_F()
1836 instr1->set_window(w); in TEST_F()
1837 EXPECT_FALSE(instr1->Identical(*instr2)); in TEST_F()
1841 auto instr1 = HloInstruction::CreateCustomCall(ShapeUtil::MakeShape(F32, {}), in TEST_F() local
1844 auto instr2 = instr1->Clone(); in TEST_F()
1845 EXPECT_TRUE(instr1->Identical(*instr2)); in TEST_F()
1849 instr1->set_convolution_dimension_numbers(dnums); in TEST_F()
1850 EXPECT_FALSE(instr1->Identical(*instr2)); in TEST_F()
[all …]
|
| D | hlo_verifier.cc | 1413 Status CheckSameChannel(const HloInstruction* instr1, in CheckSameChannel() argument
1415 if (instr1->channel_id() != instr2->channel_id()) { in CheckSameChannel()
1419 instr1->ToString(), *instr1->channel_id(), instr2->ToString(), in CheckSameChannel()
1428 Status CheckSameIsHostTransfer(const HloInstruction* instr1, in CheckSameIsHostTransfer() argument
1431 DynCast<const HloSendRecvInstruction>(instr1); in CheckSameIsHostTransfer()
1441 instr1->ToString(), instr2->ToString()); in CheckSameIsHostTransfer()
|
| /external/compiler-rt/test/profile/ |
| D | instrprof-shared.test | 5 2. libt-no-instr1.so is not instrumented
18 RUN: %clang -o %t.d/libt-no-instr1.so -fPIC -shared %S/Inputs/instrprof-shared-lib.c
23 RUN: %clang_profgen -o %t-instr-no-instr1 -L%t.d -rpath %t.d -lt-no-instr1 %S/Inputs/instrprof-sha…
25 RUN: %clang -o %t-no-instr1-instr -L%t.d -rpath %t.d -lt-instr %S/Inputs/instrprof-shared-main.c
26 RUN: %clang -o %t-no-instr1-no-instr1 -L%t.d -rpath %t.d -lt-no-instr1 %S/Inputs/instrprof-shared-…
27 RUN: %clang -o %t-no-instr1-no-instr2 -L%t.d -rpath %t.d -lt-no-instr2 %S/Inputs/instrprof-shared-…
30 RUN: %clang -o %t-no-instr2-no-instr1 -L%t.d -rpath %t.d -lt-no-instr1 %t.d/instrprof-shared-main-…
34 RUN: env LLVM_PROFILE_FILE=%t-instr-no-instr1.profraw %run %t-instr-no-instr1
36 RUN: env LLVM_PROFILE_FILE=%t-no-instr1-instr.profraw %run %t-no-instr1-instr
38 RUN: env LLVM_PROFILE_FILE=%t-no-instr1-no-instr1.profraw %run %t-no-instr1-no-instr1
[all …]
|
| /external/llvm-project/compiler-rt/test/profile/Posix/ |
| D | instrprof-shared.test | 5 2. libt-no-instr1.so is not instrumented
18 RUN: %clang -o %t.d/libt-no-instr1.so -fPIC -shared %S/../Inputs/instrprof-shared-lib.c
23 RUN: %clang_profgen -o %t-instr-no-instr1 -L%t.d -rpath %t.d -lt-no-instr1 %S/../Inputs/instrprof-…
25 RUN: %clang -o %t-no-instr1-instr -L%t.d -rpath %t.d -lt-instr %S/../Inputs/instrprof-shared-main.c
26 RUN: %clang -o %t-no-instr1-no-instr1 -L%t.d -rpath %t.d -lt-no-instr1 %S/../Inputs/instrprof-shar…
27 RUN: %clang -o %t-no-instr1-no-instr2 -L%t.d -rpath %t.d -lt-no-instr2 %S/../Inputs/instrprof-shar…
30 RUN: %clang -o %t-no-instr2-no-instr1 -L%t.d -rpath %t.d -lt-no-instr1 %t.d/instrprof-shared-main-…
34 RUN: env LLVM_PROFILE_FILE=%t-instr-no-instr1.profraw %run %t-instr-no-instr1
36 RUN: env LLVM_PROFILE_FILE=%t-no-instr1-instr.profraw %run %t-no-instr1-instr
38 RUN: env LLVM_PROFILE_FILE=%t-no-instr1-no-instr1.profraw %run %t-no-instr1-no-instr1
[all …]
|
| /external/tensorflow/tensorflow/compiler/xla/service/gpu/ |
| D | gpu_fusible.cc | 165 bool ShapesCompatibleForMultiOutputFusion(const HloInstruction& instr1, in ShapesCompatibleForMultiOutputFusion() argument
182 auto* instr_1 = GetRealHeroForMultiOutputFusion(instr1); in ShapesCompatibleForMultiOutputFusion()
370 bool FusionWouldBeTooLarge(const HloInstruction& instr1, in FusionWouldBeTooLarge() argument
373 if (SharedMemoryUsage(instr1) + SharedMemoryUsage(instr2) > in FusionWouldBeTooLarge()
375 VLOG(5) << "Shared memory usage of fusion of " << instr1.ToString() in FusionWouldBeTooLarge()
396 int64 num_output_buffers = ShapeUtil::SubshapeCount(instr1.shape()) + in FusionWouldBeTooLarge()
406 if (instr1.operand_count() + instr2.operand_count() - 1 + in FusionWouldBeTooLarge()
412 << "(" << instr1.ToString() << " ) = " << instr1.operand_count() in FusionWouldBeTooLarge()
421 absl::flat_hash_set<const HloInstruction*> operands(instr1.operands().begin(), in FusionWouldBeTooLarge()
422 instr1.operands().end()); in FusionWouldBeTooLarge()
[all …]
|
| D | multi_output_fusion.cc | 53 bool LegalToFuse(HloInstruction* instr1, HloInstruction* instr2) { in LegalToFuse() argument
60 CHECK(instr1->opcode() == HloOpcode::kFusion); in LegalToFuse()
62 instr1->fusion_kind() != instr2->fusion_kind()) || in LegalToFuse()
64 instr1->IsLoopFusion())) { in LegalToFuse()
71 if (instr1->fused_expression_root()->opcode() == in LegalToFuse()
79 return !FusionWouldBeTooLarge(*instr1, *instr2); in LegalToFuse()
|
| D | gpu_fusible.h | 65 bool FusionWouldBeTooLarge(const HloInstruction& instr1,
86 bool ShapesCompatibleForMultiOutputFusion(const HloInstruction& instr1,
|
| /external/mesa3d/src/compiler/nir/ |
| D | nir_opt_vectorize.c | 99 instrs_equal(const nir_instr *instr1, const nir_instr *instr2) in instrs_equal() argument
101 switch (instr1->type) { in instrs_equal()
103 nir_alu_instr *alu1 = nir_instr_as_alu(instr1); in instrs_equal()
165 instr_try_combine(struct nir_shader *nir, nir_instr *instr1, nir_instr *instr2, in instr_try_combine() argument
168 assert(instr1->type == nir_instr_type_alu); in instr_try_combine()
170 nir_alu_instr *alu1 = nir_instr_as_alu(instr1); in instr_try_combine()
189 nir_builder_init(&b, nir_cf_node_get_function(&instr1->block->cf_node)); in instr_try_combine()
190 b.cursor = nir_after_instr(instr1); in instr_try_combine()
312 nir_instr_remove(instr1); in instr_try_combine()
372 const nir_instr *instr1 = *(nir_instr **)util_dynarray_begin(arr1); in cmp_func() local
[all …]
|
| D | nir_instr_set.c | 549 nir_instrs_equal(const nir_instr *instr1, const nir_instr *instr2) in nir_instrs_equal() argument
551 assert(instr_can_rewrite(instr1) && instr_can_rewrite(instr2)); in nir_instrs_equal()
553 if (instr1->type != instr2->type) in nir_instrs_equal()
556 switch (instr1->type) { in nir_instrs_equal()
558 nir_alu_instr *alu1 = nir_instr_as_alu(instr1); in nir_instrs_equal()
601 nir_deref_instr *deref1 = nir_instr_as_deref(instr1); in nir_instrs_equal()
645 nir_tex_instr *tex1 = nir_instr_as_tex(instr1); in nir_instrs_equal()
678 nir_load_const_instr *load1 = nir_instr_as_load_const(instr1); in nir_instrs_equal()
700 nir_phi_instr *phi1 = nir_instr_as_phi(instr1); in nir_instrs_equal()
720 nir_intrinsic_instr *intrinsic1 = nir_instr_as_intrinsic(instr1); in nir_instrs_equal()
|
| D | nir.h | 3989 bool nir_instrs_equal(const nir_instr *instr1, const nir_instr *instr2);
|
| /external/llvm-project/lld/ELF/ |
| D | AArch64ErrataFix.cpp | 317 static bool is843419ErratumSequence(uint32_t instr1, uint32_t instr2, in is843419ErratumSequence() argument
319 if (!isADRP(instr1)) in is843419ErratumSequence()
322 uint32_t rn = getRt(instr1); in is843419ErratumSequence()
356 uint32_t instr1 = *instBuf++; in scanCortexA53Errata843419() local
359 if (is843419ErratumSequence(instr1, instr2, instr3)) { in scanCortexA53Errata843419()
363 if (is843419ErratumSequence(instr1, instr2, instr4)) in scanCortexA53Errata843419()
|
| D | ARMErrataFix.cpp | 267 uint32_t instr1 = (hw11 << 16) | hw12; in scanCortexA8Errata657417() local
269 if (!is32bitBranch(instr1) && is32bitBranch(instr2)) { in scanCortexA8Errata657417()
|
| /external/mesa3d/src/gallium/drivers/r600/sfn/ |
| D | sfn_nir_vectorize_vs_inputs.c | 97 r600_io_access_same_var(const nir_instr *instr1, const nir_instr *instr2) in r600_io_access_same_var() argument
99 assert(instr1->type == nir_instr_type_intrinsic && in r600_io_access_same_var()
102 nir_intrinsic_instr *intr1 = nir_instr_as_intrinsic(instr1); in r600_io_access_same_var()
220 const nir_instr *instr1 = *(nir_instr **)util_dynarray_begin(arr1); in r600_cmp_func() local
223 return r600_io_access_same_var(instr1, instr2); in r600_cmp_func()
|
| /external/tensorflow/tensorflow/compiler/tf2xla/ |
| D | xla_compiler_test.cc | 776 auto instr1 = c1.instructions(j); in TEST_F() local
778 instr1.clear_name(); in TEST_F()
779 instr1.clear_id(); in TEST_F()
780 instr1.clear_operand_ids(); in TEST_F()
788 LOG(INFO) << "instr1 = " << instr1.DebugString(); in TEST_F()
790 instr1.AppendPartialToString(&str1); in TEST_F()
|
| /external/llvm-project/llvm/docs/ |
| D | MergeFunctions.rst | 405 instr0 i32 %pf0 instr1 i32 %pf1 instr2 i32 123
411 instr0 i32 %pg0 instr1 i32 %pg0 instr2 i32 123
420 Instructions with opcode "*instr1*" from *f* is *greater* than instructions
421 with opcode "*instr1*" from *g*; here we have equal types and opcodes, but
|
| /external/llvm/docs/ |
| D | MergeFunctions.rst | 415 instr0 i32 %pf0 instr1 i32 %pf1 instr2 i32 123
421 instr0 i32 %pg0 instr1 i32 %pg0 instr2 i32 123
430 Instruction with opcode "*instr1*" from *f* is *greater* than instruction with
431 opcode "*instr1*" from *g*; here we have equal types and opcodes, but "*pf1* is
|