List of all items

Structs

• AttnPreVl
• AttnPreVl8
• Engine
• ExactScope
• F32x8
• FaSeqVl
• FaVl8
• GdnChunkBufs
• GdnPrepVl
• GdnPrepVl8
• GdnSeqVl
• GdnVl8
• GdnWVl
• GdnWVl8
• PinnedHostBuf
• VerifyWs
• WPtr8
• cache::Cache
• cache::CacheSnapshot
• cache::DflashTapSink
• cache::HcTapSink
• cache::KvLayer
• cache::KvRing
• cache::LatentKvLayer
• cache::LatentPlaneSnapshot
• cache::LatentTailCapture
• cache::RecurLayer
• cache::ResidentTpKvCache
• cache::ResidentTpKvCacheRank
• cache::TpKvAppendPlan
• cache::TpKvGrowPlan
• cache::TpKvRankAllocationShape
• cache::TpKvTransaction
• decode::GenOutput
• decode::GenParams
• decode::GraphDecodeState
• decode::GraphSession
• decode::Hy3Layer0Stages
• decode_batch::DevPenalty
• decode_batch::DevSamp
• decode_batch::PendingBatchStep
• dflash::ConfidenceHead
• dflash::Dflash2Conv
• dflash::Dflash2Head
• dflash::DflashCfg
• dflash::DflashDraft
• dflash::DflashDrafter
• dflash::DflashKv
• dflash::DflashKvTail
• dflash::DflashLayer
• dflash::DsparkSpecSession
• dflash::MarkovHead
• dsv4_gpu::CmpDev
• dsv4_gpu::DecodeState
• dsv4_gpu::DsparkCaptureOut
• dsv4_gpu::DsparkDev
• dsv4_gpu::DsparkProposal
• dsv4_gpu::DsparkState
• dsv4_gpu::Dsv4Gpu
• dsv4_gpu::Dsv4PenaltyCfg
• dsv4_gpu::Dsv4Phase
• dsv4_gpu::Dsv4SampleCfg
• dsv4_gpu::ForwardOut
• dsv4_gpu::Fp8Dense
• dsv4_gpu::GpuCapture
• dsv4_gpu::IdxDev
• dsv4_gpu::LayerCache
• dsv4_gpu::LayerDev
• dsv4_gpu::MtpDev
• dsv4_gpu::SpecRoundGpu
• dsv4_gpu::SpecRunGpu
• dsv4_gpu::Stage
• dsv4_gpu::StepWs
• dsv4_gpu::VerifyState
• dsv4_gpu::VerifyWs
• eagle::Eagle3Draft
• eagle::Eagle3Scratch
• ep_map::EpMap
• f16_ffi::F16Scratch
• forward::PrimeGateVerdict
• fp8_ffi::Fp8Scratch
• gemma_spec::GemmaDraft
• gemma_spec::GemmaDraftLayer
• gemma_spec::GemmaSpecSession
• gemma_spec::TrimAdapt
• glm5_tp::EpRankSlab
• glm5_tp::Glm5EpExps
• glm5_tp::Glm5TpKda
• glm5_tp::Glm5TpLoadPlan
• glm5_tp::Glm5TpMla
• glm5_tp::Glm5TpModelView
• glm5_tp::Glm5TpRt
• glm_spec::Glm5SpecKnobs
• glm_spec::Glm5SpecSession
• glm_spec::Glm5VerifyCkpt
• graph_update::FaMain
• graph_update::KernelNode
• hybrid::DevExpertFp8BlockScales
• hybrid::DevExpertFp8ProjectionScales
• hybrid::DevExps
• hybrid::DflashTrimHead
• hybrid::DraftGeom
• hybrid::FullAttnLayer
• hybrid::Gemma4E4bLayer
• hybrid::Gemma4E4bModel
• hybrid::Gemma4LayerBits
• hybrid::Gemma4MoeBits
• hybrid::GemmaAux
• hybrid::HybridLayer
• hybrid::HybridModel
• hybrid::LinearAttnLayer
• hybrid::MlaAttnLayer
• hybrid::MlaGeom
• hybrid::MlaIndexer
• hybrid::MlaIndexerGeom
• hybrid::MoeWeights
• hybrid::MtpHead
• hybrid::Step35Aux
• hybrid::Step35MtpGeom
• hybrid::StepEpExps
• hybrid::StepEpGroupedDecode
• hybrid::StepTpAttention
• hybrid::StepTpExps
• hybrid::StepTpKvDeviceAdmission
• hybrid::StepTpQkv
• hybrid_forward::G4DcSlots
• hybrid_forward::HyperPrimeWorkspaceShape
• hybrid_forward::IndexerPlanes
• hybrid_forward::PrimeSlabs
• hybrid_forward::Step35TokenGraphState
• hyper::HcMix
• hyper::HyperDecodeWs
• hyper::HyperHead
• hyper::HyperLayer
• hyper::HyperSite
• hyper::HyperTopology
• kda::KdaAttnLayer
• kda::KdaRowsStash
• kda::KdaScanInputs
• mla::MlaDims
• mla::MlaInputs
• mmq_ffi::DeviceExpertCsr
• mmq_ffi::ExpertCsr
• mmq_ffi::Fp8GroupedWorkspace
• model::EmbedHost
• model::ExpertLayout
• model::Fp8BlockScales
• model::Fp8Weight
• model::HostExpertFp8BlockScales
• model::HostExps
• model::Layer
• model::Model
• moe_cache::BlockId
• moe_cache::MoeSlotCache
• moesd::MoesdLayerUnion
• parallel::GroupedParallelPlan
• parallel::GroupedTopologyRequest
• parallel::ModelParallelContract
• parallel::ParallelPlan
• parallel::StageGroupRequest
• parallel::StageRankGroup
• parallel::TopologyError
• parallel::TopologyRequest
• plan_backend::ExecutionRewrite
• plan_backend::KernelManifest
• plan_backend::RewriteParityPolicy
• plan_backend::RewriteParityReceipt
• plan_backend::RewriteQualifications
• pp::PeerProbeMetrics
• pp::PendingLogits
• pp::PpNRt
• pp::PpWalkLease
• pp::StageRt
• prime_graph::PrimeGraph
• qwen4exp_gpu::BankFingerprint
• qwen4exp_gpu::DynKCfg
• qwen4exp_gpu::ExternalParts
• qwen4exp_gpu::LayerPlacement
• qwen4exp_gpu::LoadOptions
• qwen4exp_gpu::LoadedCheckpoint
• qwen4exp_gpu::MoeUnionRow
• qwen4exp_gpu::MtpDraftState
• qwen4exp_gpu::PrefillCapture
• qwen4exp_gpu::Qwen4ExpGpu
• qwen4exp_gpu::Qwen4ExpState
• qwen4exp_gpu::SelShapeRow
• qwen4exp_gpu::SpecOpts
• qwen4exp_gpu::SpecReport
• qwen4exp_gpu::SpecSamplerCfg
• qwen4exp_gpu::SpecTraceRound
• qwen4exp_gpu::Tp2Placement
• qwen4exp_gpu::Tp2Shard
• qwen4exp_gpu::VerifyStash
• round_stream::StreamBufs
• sigrouter_contract::ServedLogitRecord
• spec::DsparkAnchorRecord
• spec::OptiForkGateStats
• spec::OptiForkStateIdentity
• spec::SpecBoundaryCapture
• spec::SpecSampling
• spec::SpecSession
• spec::SpecTelemetry
• spill::MemBudget
• spill::SpillBlock
• spill::SpillCtx
• spill::Tiered
• tp::Bf16Matrix
• tp::ColumnParallelResult
• tp::E4m3BlockMatrix
• tp::E4m3ExpertBank
• tp::Nvfp4BlockMatrix
• tp::Nvfp4DeviceRoutesWorkspace
• tp::Nvfp4ExpertBank
• tp::PreparedPeerWeightedRouteCombine
• tp::PreparedStepGroupedExpertParallelGate
• tp::PreparedStepGroupedFp8Gate
• tp::ResidentBf16ColumnParallel
• tp::ResidentBf16RowParallel
• tp::ResidentColumnParallel
• tp::ResidentExpertParallel
• tp::ResidentNvfp4ColumnBankRank
• tp::ResidentNvfp4ColumnParallel
• tp::ResidentNvfp4ExpertParallel
• tp::ResidentNvfp4Rank
• tp::ResidentNvfp4RowBankRank
• tp::ResidentNvfp4RowParallel
• tp::ResidentNvfp4TensorParallel
• tp::ResidentReplicatedBf16SwiGlu
• tp::ResidentReplicatedDeviceRows
• tp::ResidentRowParallel
• tp::ResidentSigmoidTopKRouter
• tp::ResidentStepBf16RowParallel
• tp::ResidentTensorParallel
• tp::ResidentTpExpert
• tp::ResidentTpKvCache
• tp::ResidentTpKvCacheRank
• tp::ResidentTpNvfp4Expert
• tp::RowParallelResult
• tp::SigmoidTopKHostOutput
• tp::StepEpLayerSpec
• tp::StepGroupedFp8ProjectionOutput
• tp::StepTpDecodeV2Ws
• tp::TokenGraph
• tp::TokenGraphBuilder
• tp::TpE4m3HostBounce
• tp::TpKvAppendPlan
• tp::TpKvTransaction
• tp_transport::Hop
• tp_transport::PeerPullLink
• vision::EmbedOverlay
• vision::VisionTower
• vision_gemma::GemmaVisionTower
• vision_gemma::GemmaVisionUnit
• vision_glm5::Glm5VisionTower
• vision_glm5::Glm5VisionUnit
• vision_pre::PlannedVideo
• vision_pre::PlannedVideoGroup
• vision_pre::PreppedImage
• vision_pre::PreppedVideo
• vision_pre::VisionUnit
• vision_step::StepImagePlan
• vision_step::StepVisionTower
• vision_step::StepVisionUnit

Enums

• cache::KvRingAppend
• decode::StopReason
• dflash::DsparkHarvest
• dflash::DsparkVtPolicy
• dsv4_gpu::DW
• dsv4_gpu::DecodePath
• dsv4_gpu::DenseBf16
• dsv4_gpu::DenseView
• dsv4_gpu::Dsv4Vt
• dsv4_gpu::ExpertArm
• dsv4_gpu::ExpertKind
• forward::PrimeGateClass
• glm5_tp::GateRed
• glm5_tp::Glm5LayerClass
• hybrid::Ffn
• hybrid::Mixer
• hybrid::StepEpExpertBank
• hybrid::StepTpExpertBank
• model::GpuTensor
• model::HostBuf
• moe_cache::DispatchSlot
• parallel::HardwareTarget
• plan_backend::DecodeBatchProgram
• plan_backend::RewriteSurface
• plan_backend::RewriteValueKind
• pp::DualPpMode
• pp::PeerProbeStartupPolicy
• pp::RuntimePeerProbeStatus
• qwen4exp_gpu::HeadMode
• qwen4exp_gpu::IdxQMode
• qwen4exp_gpu::Tp2GateRed
• spec::DraftSourceKind
• spec::OptiForkGateMode
• tp::KvRingAppend
• tp::StepTpGateShards
• tp_transport::TpTransport
• vision::OverlayPublish

Traits

• cache::KvDev
• spec::SpecConstraint

Functions

• arm_step37_serving_defaults
• bf16_tcols_red_fused_dispatches
• bf16_tcols_wide_dispatches
• bf16_tcols_x1_dispatches
• cache::cache_bytes_per_token
• cache::cache_bytes_per_token_for_layers
• cache::cache_bytes_per_token_for_plan
• cache::cache_ring_bytes_per_token
• cache::cache_ring_bytes_per_token_for_layers
• cache::cache_ring_bytes_per_token_for_plan
• cache::cache_ring_row_cap
• cache::cache_ring_row_cap_for_plan
• cache::gkv_on
• cache::index_plane_physical_row
• cache::index_ring_default_rows
• cache::index_ring_rows
• cache::index_ring_rows_for
• cache::index_ring_take
• cache::kv_blk_bytes
• cache::kv_cache_formats
• cache::kv_fp8_on
• cache::latent_kv_bytes_per_token_for_plan
• cache::set_swa_ring_default
• cache::swa_retain_from
• cache::swa_ring_on
• cache::swa_ring_rows
• cache::tp_kv_rank_allocation_shape
• cache::wkv_on
• decode_batch::batch_phase_on
• decode_batch::batch_phase_report
• dflash::dflash2_walk_greedy
• dflash::dflash2_walk_greedy_q
• dflash::dflash2_walk_sampled
• dflash::dspark_accept_prefix
• dflash::dspark_confidence_vt
• dflash::dspark_slot_confidence_vt
• dflash::dspark_spec_prompt_fits
• dflash::dspark_strategy_census
• dflash::load_drafter
• dflash::rejection_accept_len
• dflash::resolve_tap_layers
• dflash::tail_geometry_ok
• dsv4_ffi::ck
• dsv4_ffi::memra_dsv4_act_quant
• dsv4_ffi::memra_dsv4_act_quant_fp8
• dsv4_ffi::memra_dsv4_add_inplace
• dsv4_ffi::memra_dsv4_argmax
• dsv4_ffi::memra_dsv4_build_idx
• dsv4_ffi::memra_dsv4_build_idx_redirect
• dsv4_ffi::memra_dsv4_combine_rows
• dsv4_ffi::memra_dsv4_combine_rows_m
• dsv4_ffi::memra_dsv4_compressor_pool
• dsv4_ffi::memra_dsv4_cvt_bf16
• dsv4_ffi::memra_dsv4_dots_f32
• dsv4_ffi::memra_dsv4_dots_f32_mrow
• dsv4_ffi::memra_dsv4_dots_f32_rowblk
• dsv4_ffi::memra_dsv4_dots_f32acc
• dsv4_ffi::memra_dsv4_dots_f32acc_mrow
• dsv4_ffi::memra_dsv4_embed_rows
• dsv4_ffi::memra_dsv4_fp4_act_quant
• dsv4_ffi::memra_dsv4_fp4_gemm
• dsv4_ffi::memra_dsv4_fp4_gemm_sel
• dsv4_ffi::memra_dsv4_fp4_gemm_sel_g
• dsv4_ffi::memra_dsv4_gather_bf16
• dsv4_ffi::memra_dsv4_gather_row_by_idx
• dsv4_ffi::memra_dsv4_gather_rows_u8
• dsv4_ffi::memra_dsv4_gemm_bf16
• dsv4_ffi::memra_dsv4_gemv_bf16
• dsv4_ffi::memra_dsv4_gemv_bf16_m
• dsv4_ffi::memra_dsv4_gemv_fp8
• dsv4_ffi::memra_dsv4_gemv_fp8_m
• dsv4_ffi::memra_dsv4_hadamard
• dsv4_ffi::memra_dsv4_hc_collapse
• dsv4_ffi::memra_dsv4_hc_expand
• dsv4_ffi::memra_dsv4_hc_head_pre
• dsv4_ffi::memra_dsv4_hc_head_pre_m
• dsv4_ffi::memra_dsv4_hc_mean
• dsv4_ffi::memra_dsv4_hc_post
• dsv4_ffi::memra_dsv4_hc_pre_fused
• dsv4_ffi::memra_dsv4_hc_sinkhorn
• dsv4_ffi::memra_dsv4_hc_sinkhorn_m
• dsv4_ffi::memra_dsv4_headrms
• dsv4_ffi::memra_dsv4_headrms_f32acc
• dsv4_ffi::memra_dsv4_indexer_score
• dsv4_ffi::memra_dsv4_indexer_score_f32acc
• dsv4_ffi::memra_dsv4_mxfp4_deq_bf16
• dsv4_ffi::memra_dsv4_nvfp4_deq_bf16
• dsv4_ffi::memra_dsv4_nvtx_pop
• dsv4_ffi::memra_dsv4_nvtx_push
• dsv4_ffi::memra_dsv4_place_cols
• dsv4_ffi::memra_dsv4_repeat_hc
• dsv4_ffi::memra_dsv4_rmsnorm
• dsv4_ffi::memra_dsv4_rmsnorm_f32acc
• dsv4_ffi::memra_dsv4_rope
• dsv4_ffi::memra_dsv4_rope_at
• dsv4_ffi::memra_dsv4_route
• dsv4_ffi::memra_dsv4_route_m
• dsv4_ffi::memra_dsv4_rowsq_scale
• dsv4_ffi::memra_dsv4_rowsq_scale_f32acc
• dsv4_ffi::memra_dsv4_scatter_add
• dsv4_ffi::memra_dsv4_scatter_rows
• dsv4_ffi::memra_dsv4_sink_attn
• dsv4_ffi::memra_dsv4_sink_attn_dec
• dsv4_ffi::memra_dsv4_sink_attn_dec_f32acc
• dsv4_ffi::memra_dsv4_sink_attn_dec_mq
• dsv4_ffi::memra_dsv4_sink_attn_dec_mq_f32acc
• dsv4_ffi::memra_dsv4_swiglu
• dsv4_ffi::memra_dsv4_take_cols
• dsv4_ffi::memra_dsv4_topk_idx
• dsv4_gpu::dsv4_penalize_row
• dsv4_gpu::dsv4_phase_report
• dsv4_gpu::dsv4_phase_reset
• dsv4_gpu::dsv4_pos_uniform
• dsv4_gpu::dsv4_prof_on
• dsv4_gpu::dsv4_sample_row
• dsv4_gpu::resolve_dense_arm
• dsv4_gpu::resolve_vt
• dsv4_gpu::vt_slot_drafts
• ep_diet_on
• ep_grouped_prime_on
• ep_map::ep_map_env
• ep_map::resolve_ep_map_env
• f16_ffi::bf16_tc_dispatches
• f16_ffi::pp_bf16_enabled
• f16_ffi::pp_f16_capacity_ok
• f16_ffi::pp_f16_enabled
• fa512_hp_on
• fa512_min_tkv
• fa512_wide_warps
• fa_deep_at_pub
• fa_f16pv_on
• fa_seqs_eligible
• fa_split_keys_pub
• fa_v4_at_pub
• fa_vec_min_tkv
• faw_hp_on
• forward::argmax
• forward::prime_gate_verdict
• forward::top2
• fp8_ffi::blk_native_nan_refused
• fp8_ffi::fp8_blk_gpu_enabled
• fp8_ffi::fp8_blk_mmq_native_enabled
• fp8_ffi::fp8_mmq_enabled
• fp8_ffi::fp8_mmq_hits
• fp8_ffi::fp8_mmq_ledger
• fp8_ffi::note_blk_native_nan_refused
• fp8_ffi::pp_fp8_enabled
• fp8_ffi::st_e4m3_blk_enabled
• fp8_ffi::st_e4m3_enabled
• glm5_tp::gate_red
• glm5_tp::gate_same_device
• glm5_tp::glm5_ep_diet_bulk_returns
• glm5_tp::glm5_ep_diet_dispatches
• glm5_tp::glm5_ep_diet_fanout_uploads_avoided
• glm5_tp::glm5_ep_diet_peer_roundtrips_avoided
• glm5_tp::glm5_ep_grouped_prime_dispatches
• glm5_tp::glm5_ep_peer_slot_dispatches
• glm5_tp::glm5_tp_armed
• glm5_tp::glm5_tp_env_raw
• glm5_tp::parse_glm5_tp_layer_specs
• glm5_tp::prepare_glm5_tp_load
• glm5_tp::refuse_glm5_tp_door_composition
• glm_spec::glm5_spec_on
• glm_spec::glm5_spec_prefix_on
• glm_spec::glm5_spec_tp_on
• glm_spec::glm5_verify_batch_on
• graph_update::fa_apply
• graph_update::fa_plan
• graph_update::kernel_nodes
• graph_update::node_census
• graph_update::read_i32_arg
• graph_update::read_ptr_arg
• graph_update::set_exec_params
• graph_update::write_i32_arg
• htod_diet_avoided
• htod_diet_on
• hybrid::draft_head_tensor
• hybrid_forward::align_prime_ranges_to_gdn
• hybrid_forward::hyper_prime_call_rows
• hybrid_forward::hyper_prime_ranges
• hybrid_forward::prime_chunk_ranges
• hybrid_forward::prime_chunk_tokens
• hyper::collapse
• hyper::contract_mean
• hyper::expand
• hyper::post
• hyper::post_t1_ws
• hyper::pre
• hyper::pre_exact
• hyper::pre_t1_ws
• kda::kda_attn
• kda::kda_attn_decode
• kda::kda_attn_prime
• kda::kda_decode_cached
• kda::kda_decode_cached_stash
• kda::kda_prime_cached
• kda::kda_scan_replay
• kda::kda_verify_rollback_rows
• kda::kda_verify_rollback_rows_on
• kda::kda_verify_rows_cached
• kv_blk_bytes
• kv_cache_formats
• mla::mla_attend_absorbed
• mla::mla_attend_naive
• mla::norm_to_neox_perm
• mla::rope_interleaved
• mla::rope_neox
• mla_ffi::memra_mla_absorb_q_f32
• mla_ffi::memra_mla_absorb_q_split_f32
• mla_ffi::memra_mla_append_latent_f32
• mla_ffi::memra_mla_attn_absorbed_f32
• mla_ffi::memra_mla_attn_gathered_f32
• mla_ffi::memra_mla_decompress_v_f32
• mla_ffi::memra_mla_decompress_v_split_f32
• mla_ffi::memra_mla_index_append_ring_f32
• mla_ffi::memra_mla_kpool_pool_keys_f32
• mla_ffi::memra_mla_kpool_score_f32
• mla_ffi::memra_mla_kpool_score_ref_f32
• mla_ffi::memra_mla_kpool_select_f32
• mla_ffi::memra_mla_kpool_select_ref_f32
• mla_ffi::memra_mla_rope_interleaved_f32
• mla_ffi::memra_mla_split_latent_f32
• mla_tc_prefill_dispatches
• mmq_ffi::memra_accprobe_act_bytes
• mmq_ffi::memra_accprobe_gemm_f32
• mmq_ffi::memra_accprobe_gemm_s32
• mmq_ffi::memra_fp8_blk_count_nan
• mmq_ffi::memra_mmq_fp8_blk
• mmq_ffi::memra_mmq_fp8_blk_act_bytes
• mmq_ffi::memra_mmq_fp8_blk_grouped
• mmq_ffi::memra_mmq_fp8_blk_quantize_act
• mmq_ffi::memra_mmq_fp8_blk_scale_cols
• mmq_ffi::memra_mmq_fp8_blk_scale_rows
• mmq_ffi::memra_mmq_iq4xs_dense
• mmq_ffi::memra_mmq_iq_experts
• mmq_ffi::memra_mmq_iq_experts_act_bytes
• mmq_ffi::memra_mmq_iq_fused_act_quant
• mmq_ffi::memra_mmq_iq_quantize_act
• mmq_ffi::memra_mmq_nvfp4
• mmq_ffi::memra_mmq_nvfp4_act_bytes
• mmq_ffi::memra_mmq_nvfp4_ex
• mmq_ffi::memra_mmq_nvfp4_ex2
• mmq_ffi::memra_mmq_nvfp4_f8f4
• mmq_ffi::memra_mmq_nvfp4_f8f4_act_bytes
• mmq_ffi::memra_mmq_nvfp4_w4a8
• mmq_ffi::memra_mmq_nvfp4_w4a8_act_bytes
• mmq_ffi::memra_mmq_q45k_act_bytes
• mmq_ffi::memra_mmq_q4_0
• mmq_ffi::memra_mmq_q4_0_act_bytes
• mmq_ffi::memra_mmq_q4_0_fixup_bytes
• mmq_ffi::memra_mmq_q4_0_gemm
• mmq_ffi::memra_mmq_q4_0_gemm_sk
• mmq_ffi::memra_mmq_q4_0_quant_act
• mmq_ffi::memra_mmq_q4_0_set_clc
• mmq_ffi::memra_mmq_q4_K
• mmq_ffi::memra_mmq_q5_K
• mmq_ffi::memra_mmq_q8_0
• mmq_ffi::memra_mmq_q8_0_act_bytes
• mmq_ffi::memra_moe_f16g_dequant
• mmq_ffi::memra_moe_f16g_gather_act
• mmq_ffi::memra_moe_f16g_gemm
• mmq_ffi::memra_moe_f16g_gemm_sk
• mmq_ffi::memra_moe_f16g_h2f
• mmq_ffi::memra_moe_f16g_h2f_scaled
• mmq_ffi::memra_moe_kq_gemm_sk
• mmq_ffi::mmq_iq4xs_enabled
• mmq_ffi::mmq_q4_enabled
• mmq_ffi::mmq_q8_enabled
• mmq_ffi::mmq_residual_k
• mmq_ffi::mmq_w4a8_enabled
• model::full_prec_enabled
• model::repack_nvfp4_split
• model::residency_census_report
• model::rp_enabled
• model::unpack_nvfp4_split
• moe_f16g_direct_on
• moe_f16g_gemma_on
• moe_f16g_mode
• moe_f16g_sk_params
• moe_f16g_tail_on
• moe_fuse_actq_on
• moe_fused_epilogue_dispatches
• moe_grouped_prefill_dispatches
• moe_vrows_dedup_order_dispatches
• moe_vrows_dev_tables_dispatches
• moe_vrows_dispatches
• moe_vrows_down_tmaj_dispatches
• moe_vrows_pack_dispatches
• moe_vrows_pair_overlap
• moe_vrows_router_syncs_avoided
• moe_vrows_slab_reads_avoided
• moesd::begin_capture
• moesd::finish_capture
• parallel::validate_fp8_expert_checkpoint
• parallel::validate_nvfp4_expert_checkpoint
• parallel::validate_step_fp8_checkpoint
• parallel::validate_step_nvfp4_checkpoint
• parallel::validate_step_pp_request
• plan_backend::bind_rewrite_artifact
• plan_backend::decode_batch_program
• plan_backend::decode_batch_unconverted
• plan_backend::execution_rewrites
• plan_backend::gdn_dspark_compatible
• pp::batch_pp_on
• pp::clear_runtime_peer_probe_integrity_degraded
• pp::dual_pp_eligibility
• pp::dual_pp_mode
• pp::dual_pp_mode_resolve
• pp::dual_pp_on
• pp::dual_pp_overlaps
• pp::dual_pp_route
• pp::dual_pp_slot_snapshot
• pp::dual_pp_timing_dropped
• pp::dual_pp_timing_on
• pp::dual_pp_timing_snapshot
• pp::dual_pp_wave_mid
• pp::init_model_transport
• pp::layer_engine
• pp::new_cache
• pp::new_cache_planned
• pp::peer_probe_metrics
• pp::peer_probe_startup_policy
• pp::pp2_overlap
• pp::pp2_overlap_resolve
• pp::pp2_split
• pp::pp2_streams_off
• pp::pp_cuts
• pp::pp_exit_publish
• pp::pp_host_bounce_active
• pp::pp_host_bounce_on
• pp::pp_multi_stream_same_device
• pp::pp_shard_off
• pp::pp_sharded_cross_device
• pp::pp_wave_diagonal
• pp::pp_wave_eligibility
• pp::pp_wave_numeric_eligibility
• pp::pp_wave_on
• pp::pp_wave_on_value
• pp::pp_wave_ranges
• pp::pp_wave_route_enabled
• pp::pp_wave_snapshot
• pp::prime_pipe_on
• pp::prime_pipe_overlaps
• pp::prime_pp_on
• pp::prime_split_chunks
• pp::record_runtime_peer_probe_deferral
• pp::refuse_unsplit_if_remote
• pp::restore_cache_checkpoint
• pp::service_runtime_peer_probe
• pp::spec_pp_on
• pp::stage_of
• pp::step35_prime_batch_splits
• pp::step35_prime_batches
• pp::sync_stages_after_load
• pp::warn_unwired_once
• qwen4exp_gpu::apply_env_seams
• qwen4exp_gpu::build_tp2_shard
• qwen4exp_gpu::gate_gdn_step_kernels
• qwen4exp_gpu::gate_hc_diet_kernels
• qwen4exp_gpu::gate_hc_micro_kernels
• qwen4exp_gpu::gate_kvq_kernels
• qwen4exp_gpu::gate_nvfp4_sel_group
• qwen4exp_gpu::gate_nvfp4_sel_matvec
• qwen4exp_gpu::gate_ple_ngram_cache
• qwen4exp_gpu::gate_qmatvec_bf16
• qwen4exp_gpu::gate_qsa_index_score
• qwen4exp_gpu::gate_qsa_index_topk
• qwen4exp_gpu::gate_route_kernel
• qwen4exp_gpu::gate_sdpa_blocklist
• qwen4exp_gpu::gate_seam_table
• qwen4exp_gpu::idx_sel_audit_stats
• qwen4exp_gpu::idxq_audit_stats
• qwen4exp_gpu::idxq_mode_name
• qwen4exp_gpu::kv_hoist_is_on
• qwen4exp_gpu::kv_quant_is_on
• qwen4exp_gpu::moe_union_cost_probe
• qwen4exp_gpu::peer_kv_max_cap
• qwen4exp_gpu::ple_cache_audit_stats
• qwen4exp_gpu::pool_t_is_on
• qwen4exp_gpu::prof::enable
• qwen4exp_gpu::prof::on
• qwen4exp_gpu::prof::rounds_only
• qwen4exp_gpu::prof::set_rounds_only
• qwen4exp_gpu::prof::split_prefill
• qwen4exp_gpu::prof::take
• qwen4exp_gpu::prof::take_prefill
• qwen4exp_gpu::read_checkpoint
• qwen4exp_gpu::read_checkpoint_with
• qwen4exp_gpu::route_audit_stats
• qwen4exp_gpu::seam_exists
• qwen4exp_gpu::seam_names
• qwen4exp_gpu::seam_state
• qwen4exp_gpu::sel_group_spec
• qwen4exp_gpu::sel_shape_cost_probe
• qwen4exp_gpu::set_decode_graphs
• qwen4exp_gpu::set_gdn_fuse
• qwen4exp_gpu::set_gdn_step
• qwen4exp_gpu::set_hc_diet
• qwen4exp_gpu::set_hc_fused_gate
• qwen4exp_gpu::set_hc_micro
• qwen4exp_gpu::set_idx_cache
• qwen4exp_gpu::set_idx_dev
• qwen4exp_gpu::set_idx_sel
• qwen4exp_gpu::set_idxq
• qwen4exp_gpu::set_kv_hoist
• qwen4exp_gpu::set_kv_quant
• qwen4exp_gpu::set_longatt
• qwen4exp_gpu::set_moe_sel_path
• qwen4exp_gpu::set_ple_cache
• qwen4exp_gpu::set_pool_t
• qwen4exp_gpu::set_prefill_grouped_all
• qwen4exp_gpu::set_proj_stack
• qwen4exp_gpu::set_router_bf16
• qwen4exp_gpu::set_router_dev
• qwen4exp_gpu::set_seam
• qwen4exp_gpu::set_sel_group
• qwen4exp_gpu::set_sel_gufuse
• qwen4exp_gpu::set_sel_v2
• qwen4exp_gpu::set_sel_v3
• qwen4exp_gpu::set_step_ws
• qwen4exp_gpu::set_trunk_bf16
• qwen4exp_gpu::set_verify_fused
• qwen4exp_gpu::set_verify_graphs
• qwen4exp_gpu::set_verify_mt
• qwen4exp_gpu::spec_guard_trunc
• qwen4exp_gpu::tp2_enable_p2p
• qwen4exp_gpu::tp2_expert_split_stats
• qwen4exp_gpu::verify_fused_on
• round_stream::kv_len_ptr_table
• round_stream::kv_len_ptr_table_range
• router_batch_on
• router_kernel_on
• router_prefill_exact_on
• sigrouter_contract::capture_served_logits
• sigrouter_contract::read_served_logits
• sigrouter_contract::served_logit_trace_enabled
• sigrouter_contract::validate_active_count
• sigrouter_contract::verify_host_expf
• spec::dspark_vg_debt_projection
• spec::optipipe_gate_stats
• spec::pen_window_seed
• spec::reset_optipipe_gate_stats
• spec::resolve_draft_source_kind
• spec::sample_boundary_token
• spec::sample_boundary_token_dev
• spec::set_optipipe_controller_threshold
• spec::set_optipipe_gate_mode
• spec::spec_capture_gate_on
• spec::spec_conf_keep
• spec::spec_pen_session_on
• spec::spec_replay_env_enabled
• spec::spec_replay_env_on
• spec::spec_restore_republish_on
• spec::spec_sampled_boundary_on
• spec_phase::spec_trace_level
• spec_phase::spec_trace_on
• spill::disk_tier_enabled
• spill::place_expert
• topk_shards_dispatches
• tp::auto_parallel_devices
• tp::graph_section
• tp::graph_section_absorbing
• tp::graph_section_detached
• tp::graph_section_opts
• tp::moe_residual_host
• tp::nvfp4_matrix_v2_permute
• tp::parallel_ep_device_router_enabled
• tp::parallel_ep_graph_enabled
• tp::parallel_ep_pair_down_enabled
• tp::parallel_ep_q8_act_enabled
• tp::parse_step_ep_layer_specs
• tp::parse_step_tp_layer_specs
• tp::step_ep_device_arithmetic_enabled
• tp::step_ep_layer_specs
• tp::step_expert_activation_host
• tp::step_nvfp4_dev_routes_enabled
• tp::step_tp_bulk_p2p_enabled
• tp::step_tp_dcw_enabled
• tp::step_tp_decode_v2_enabled
• tp::step_tp_dev_router_enabled
• tp::step_tp_f32_mirror_enabled
• tp::step_tp_graph_enabled
• tp::step_tp_layer_specs
• tp::step_tp_native_p2p_enabled
• tp::step_tp_qkv_fused_enabled
• tp::token_graph_build_begin
• tp::token_graph_build_finish
• tp::token_graph_building
• tp_transport::arm_transport
• tp_transport::concat_parts_on_root
• tp_transport::fanout_f32
• tp_transport::fanout_f32_to
• tp_transport::fanout_i32
• tp_transport::gather_parts
• tp_transport::host_row_to
• tp_transport::host_stage_block
• tp_transport::parse_transport
• tp_transport::resolve_transport
• tp_transport::return_block_to_root
• tp_transport::return_row_to_root
• tp_transport::set_transport_names
• tp_transport::tp_host_legs
• tp_transport::tp_host_syncs
• tp_transport::tp_local_copies
• tp_transport::tp_peer_pulls
• tp_transport::tp_pub_events
• tp_transport::tp_xfer_bytes
• tp_transport::transport_census_line
• tp_transport::transport_env
• verify_ws_hits
• vision::overlay_publications
• vision::overlay_publish_mode
• vision::overlay_publish_resolve
• vision_gemma::gemma_decode_data_uri
• vision_gemma::gemma_plan_image
• vision_gemma::gemma_prep_data_uri
• vision_gemma::gemma_prep_image
• vision_gemma::gemma_target_size
• vision_gemma::n_soft_for_grid
• vision_glm5::glm5_content_size
• vision_glm5::glm5_plan_image
• vision_glm5::glm5_prep_data_uri
• vision_glm5::glm5_prep_image
• vision_glm5::glm5_smart_resize
• vision_glm5::glm5_visual_tensors_present
• vision_glm5::n_merged_for_grid
• vision_pre::decode_data_uri
• vision_pre::image_header_dims
• vision_pre::n_tokens_for_grid
• vision_pre::plan_image_bytes
• vision_pre::plan_video_gif
• vision_pre::prep_data_uri
• vision_pre::prep_image_bytes
• vision_pre::prep_video_gif
• vision_pre::smart_resize
• vision_pre::video_max_pixels
• vision_step::step_plan_image
• vision_step::step_prep_image
• vrows_expert_major_order_for_test
• vrows_overlap_counts_for_test

Type Aliases

• glm5_tp::Glm5TpKdaVerifyStash
• glm5_tp::Glm5TpLayerSpec
• model::TensorMap
• pp::Pp2Rt
• tp::StepTpLayerSpec

Statics

• BF16_TCOLS_RED_FUSED_DISPATCHES
• BF16_TCOLS_WIDE_DISPATCHES
• BF16_TCOLS_X1_DISPATCHES
• FA_SMEM_TKV_DEFAULT
• FA_SP512_DEFAULT
• FA_SPW_DEFAULT
• FA_SP_GEMMA
• FA_V4_MAX_DEFAULT
• FA_VEC_MIN_DEFAULT
• FUSED_MR1_DEFAULT
• HTOD_DIET_AVOIDED
• KV_FP8_FORCE
• MLA_TC_PREFILL_DISPATCHES
• MMQ_SK_FORCE
• MOE_FUSED_EPI_DISPATCHES
• MOE_GROUPED_PREFILL_DISPATCHES
• MOE_VROWS_DEDUP_ORDER_DISPATCHES
• MOE_VROWS_DEV_TABLES_DISPATCHES
• MOE_VROWS_DISPATCHES
• MOE_VROWS_DOWN_TMAJ_DISPATCHES
• MOE_VROWS_PACK_DISPATCHES
• MOE_VROWS_PAIR_DISTINCT
• MOE_VROWS_PAIR_VISITS
• MOE_VROWS_ROUTER_SYNCS_AVOIDED
• MOE_VROWS_SLAB_READS_AVOIDED
• PRIME_NANOS
• RMS_BLOCK_DEFAULT
• ROUTER_W8_DEFAULT
• SCRATCH_ALLOC_CALLS
• TOPK_SHARDS_DISPATCHES
• VERIFY_WS_HITS
• cache::KV_FP8_FORCE
• f16_ffi::BF16_TC_DISPATCHES
• glm5_tp::GLM5_EP_DIET_BULK_RETURNS
• glm5_tp::GLM5_EP_DIET_DISPATCHES
• glm5_tp::GLM5_EP_DIET_FANOUT_UPLOADS_AVOIDED
• glm5_tp::GLM5_EP_DIET_PEER_ROUNDTRIPS_AVOIDED
• glm5_tp::GLM5_EP_GROUPED_PRIME_DISPATCHES
• glm5_tp::GLM5_EP_PEER_SLOT_DISPATCHES
• hybrid_forward::HC_DECODE_WS_DISPATCHES
• hyper::HC_FUSED_PRE_DISPATCHES
• kda::KDA_FUSED6_BF16_DISPATCHES
• kda::KDA_FUSED6_DISPATCHES
• mla_ffi::MLA_DECODE_SPLIT_DISPATCHES
• pp::PRIME_PIPE_OVERLAPS
• pp::PRIME_SPLIT_CHUNKS
• pp::STEP35_PRIME_BATCHES
• pp::STEP35_PRIME_BATCH_SPLITS
• tp_transport::TP_HOST_LEGS
• tp_transport::TP_HOST_SYNCS
• tp_transport::TP_LOCAL_COPIES
• tp_transport::TP_PEER_PULLS
• tp_transport::TP_PUB_EVENTS
• tp_transport::TP_XFER_BYTES

Constants

• ARGMAX_NB
• FA_DEEP_MIN_DEFAULT
• FA_VEC_MIN_TKV
• QT_BF16
• QT_F32
• QT_F8_E4M3
• QT_F8_E4M3_BLK
• QT_IQ3_S
• QT_IQ4_XS
• QT_NVFP4
• QT_NVFP4_RP
• QT_NVFP4_V2
• QT_Q2_K
• QT_Q3_K
• QT_Q4_0
• QT_Q4_K
• QT_Q5_K
• QT_Q6_K
• QT_Q8_0
• ROUTER_BATCH_MIN_T
• cache::INDEX_RING_WORKING_ROWS
• cache::PRIME_CHUNK_MAX_TOKENS
• cache::SWA_REWIND_SLACK_ROWS
• decode_batch::BATCH_PHASE_NAMES
• ep_map::EP_MAP_ENV
• ep_map::EP_MAP_ENV_GLM5
• glm5_tp::GLM5_TP_ALLOWED_RANKS
• glm5_tp::GLM5_TP_REFUSED_DOOR_FLAGS
• glm5_tp::GLM5_TP_TRANSPORT_TAG
• hybrid_forward::CUDA_GRID_YZ_MAX
• hybrid_forward::PRIME_CHUNK_LAUNCH_CAP
• hybrid_forward::PRIME_MIN_T
• kda::KDA_HEAD_DIM
• moe_cache::PROJ_DOWN
• moe_cache::PROJ_GATE
• moe_cache::PROJ_UP
• parallel::AUTO_PARALLEL_MAX_CARDS
• parallel::PRODUCT_MAX_CARDS
• parallel::STEP37_TRUNK_LAYERS
• plan_backend::CARRIED_PRIME
• plan_backend::DECODE_BATCH
• plan_backend::DECODE_GRAPH
• plan_backend::GLM5_SPEC
• plan_backend::MTP_SPEC
• plan_backend::NATIVE_EAGER
• plan_backend::PIPELINE
• pp::DUAL_PP_HOST_BOUNCE_REFUSAL
• pp::DUAL_PP_SINGLE_SLOT_REFUSAL
• pp::DUAL_PP_STAGE_NAMES
• pp::PEER_PROBE_REQUIRED_REFUSAL
• pp::PEER_RUNTIME_PROBE_CYCLE_COPIES
• pp::PEER_RUNTIME_PROBE_DEFERRAL_BOUND_INTERVALS
• pp::PEER_RUNTIME_PROBE_INTERVAL_COPIES
• pp::PP_WAVE_MAX_STAGES
• pp::PP_WAVE_W4A16_BF16_REFUSAL
• qwen4exp_gpu::GDN_FUSE_DEFAULT
• qwen4exp_gpu::GDN_STEP_DEFAULT
• qwen4exp_gpu::HC_DIET_DEFAULT
• qwen4exp_gpu::IDX_CACHE_DEFAULT
• qwen4exp_gpu::IDX_SEL_DEFAULT
• qwen4exp_gpu::KV_HOIST_DEFAULT
• qwen4exp_gpu::KV_QUANT_DEFAULT
• qwen4exp_gpu::PLE_CACHE_DEFAULT
• qwen4exp_gpu::POOL_T_DEFAULT
• qwen4exp_gpu::PROJ_STACK_DEFAULT
• qwen4exp_gpu::ROUTER_B16_DEFAULT
• qwen4exp_gpu::ROUTER_DEV_DEFAULT
• qwen4exp_gpu::SEL_GUFUSE_DEFAULT
• qwen4exp_gpu::SEL_V3_DEFAULT
• qwen4exp_gpu::VERIFY_FUSED_DEFAULT
• qwen4exp_gpu::VERIFY_MT_DEFAULT
• spec::PEN_WINDOW_MAX
• spec::SPEC_TELEM_POS
• tp::NVFP4_CANONICAL_ROW_SHARDS
• tp_transport::ROOT
• tp_transport::TRANSPORT_ENV
• tp_transport::TRANSPORT_ENV_GLM5
• vision::V_DEPTH
• vision::V_HEADS
• vision::V_HEAD_DIM
• vision::V_HIDDEN
• vision::V_INTER
• vision::V_MERGE
• vision::V_MERGED_IN
• vision::V_PATCH
• vision::V_PATCH_IN
• vision::V_POS_GRID
• vision::V_TEMPORAL
• vision_gemma::GV_ALIGN
• vision_gemma::GV_DEPTH
• vision_gemma::GV_HEADS
• vision_gemma::GV_HEAD_DIM
• vision_gemma::GV_HIDDEN
• vision_gemma::GV_INTER
• vision_gemma::GV_MAX_TOKENS
• vision_gemma::GV_MERGE
• vision_gemma::GV_MIN_TOKENS
• vision_gemma::GV_OUT
• vision_gemma::GV_PATCH
• vision_gemma::GV_PATCH_IN
• vision_gemma::GV_POS_ROWS
• vision_gemma::GV_TOK_BEGIN
• vision_gemma::GV_TOK_END
• vision_gemma::GV_TOK_SOFT
• vision_glm5::G5V_ALIGN
• vision_glm5::G5V_DEPTH
• vision_glm5::G5V_HEADS
• vision_glm5::G5V_HEAD_DIM
• vision_glm5::G5V_HIDDEN
• vision_glm5::G5V_INTER
• vision_glm5::G5V_LIMIT
• vision_glm5::G5V_MEAN
• vision_glm5::G5V_MERGE
• vision_glm5::G5V_MIN_TOKENS
• vision_glm5::G5V_OUT
• vision_glm5::G5V_PATCH
• vision_glm5::G5V_PATCH_IN
• vision_glm5::G5V_PROBE_TENSOR
• vision_glm5::G5V_PROJ_INTER
• vision_glm5::G5V_SERVE_MAX_TOKENS
• vision_glm5::G5V_STD
• vision_glm5::G5V_TEMPORAL
• vision_glm5::G5V_TOK_IMAGE
• vision_glm5::G5V_TOK_IMAGE_END
• vision_glm5::G5V_TOK_IMAGE_START
• vision_glm5::G5V_TOK_VIDEO
• vision_glm5::G5V_TOK_VIDEO_END
• vision_glm5::G5V_TOK_VIDEO_START
• vision_pre::GIF_MAX_FRAMES
• vision_pre::GIF_MAX_TOTAL_PIXELS
• vision_pre::IMG_MAX_DECODE_PIXELS
• vision_pre::IMG_MAX_RAW_BYTES
• vision_pre::MAX_PIXELS
• vision_pre::MIN_PIXELS
• vision_pre::VID_MAX_FRAMES
• vision_pre::VID_MIN_PIXELS
• vision_step::SV_DEPTH
• vision_step::SV_GRID_MAIN
• vision_step::SV_GRID_TILE
• vision_step::SV_HEADS
• vision_step::SV_HEAD_DIM
• vision_step::SV_HIDDEN
• vision_step::SV_IMAGE_SIZE
• vision_step::SV_INTER
• vision_step::SV_MAIN_ROWS
• vision_step::SV_MAX_IMAGE_SIZE
• vision_step::SV_PATCH
• vision_step::SV_PATCH_IN
• vision_step::SV_POS_GRID
• vision_step::SV_TILE_ROWS
• vision_step::SV_TILE_SIZE