List of all items

Structs

• AttnPreVl
• AttnPreVl8
• Engine
• F32x8
• FaSeqVl
• FaVl8
• GdnChunkBufs
• GdnPrepVl
• GdnPrepVl8
• GdnSeqVl
• GdnVl8
• GdnWVl
• GdnWVl8
• WPtr8
• cache::Cache
• cache::CacheSnapshot
• cache::DflashTapSink
• cache::KvLayer
• cache::KvRing
• cache::RecurLayer
• decode::GenOutput
• decode::GenParams
• decode::GraphDecodeState
• decode::GraphSession
• decode::Hy3Layer0Stages
• dflash::DflashCfg
• dflash::DflashDraft
• dflash::DflashKv
• dflash::DflashLayer
• dflash::MarkovHead
• eagle::Eagle3Draft
• eagle::Eagle3Scratch
• f16_ffi::F16Scratch
• forward::PrimeGateVerdict
• fp8_ffi::Fp8Scratch
• gemma_spec::GemmaDraft
• gemma_spec::GemmaDraftLayer
• gemma_spec::TrimAdapt
• graph_update::FaMain
• graph_update::KernelNode
• hybrid::DevExps
• 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::MoeWeights
• hybrid::MtpHead
• hybrid::Step35Aux
• hybrid::Step35MtpGeom
• hybrid_forward::G4DcSlots
• hybrid_forward::PrimeSlabs
• mla::MlaDims
• mla::MlaInputs
• model::EmbedHost
• model::ExpertLayout
• model::Fp8BlockScales
• model::Fp8Weight
• model::HostExps
• model::Layer
• model::Model
• moe_cache::BlockId
• moe_cache::MoeSlotCache
• pp::PendingLogits
• pp::PpNRt
• pp::StageRt
• prime_graph::PrimeGraph
• round_stream::StreamBufs
• spec::SpecSampling
• spec::SpecSession
• spec::SpecTelemetry
• spill::MemBudget
• spill::SpillBlock
• spill::SpillCtx
• spill::Tiered

Enums

• cache::KvRingAppend
• decode::StopReason
• forward::PrimeGateClass
• hybrid::Ffn
• hybrid::Mixer
• model::GpuTensor
• model::HostBuf
• moe_cache::DispatchSlot

Traits

• cache::KvDev
• spec::SpecConstraint

Functions

• cache::cache_bytes_per_token
• cache::cache_ring_bytes_per_token
• cache::cache_ring_row_cap
• cache::gkv_on
• cache::kv_blk_bytes
• cache::kv_cache_formats
• cache::kv_fp8_on
• cache::swa_ring_on
• cache::swa_ring_rows
• cache::wkv_on
• decode_batch::batch_phase_on
• decode_batch::batch_phase_report
• 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
• 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
• hybrid::draft_head_tensor
• hybrid_forward::prime_chunk_ranges
• hybrid_forward::prime_chunk_tokens
• 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
• 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_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
• pp::batch_pp_on
• pp::layer_engine
• pp::new_cache
• pp::pp2_overlap
• pp::pp2_split
• pp::pp2_streams_off
• pp::pp_cuts
• 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::prime_pipe_on
• pp::prime_pipe_overlaps
• pp::prime_pp_on
• pp::prime_split_chunks
• pp::refuse_unsplit_if_remote
• pp::restore_cache_checkpoint
• 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
• round_stream::kv_len_ptr_table
• router_batch_on
• router_kernel_on
• router_prefill_exact_on
• spill::disk_tier_enabled
• spill::place_expert

Type Aliases

• model::TensorMap
• pp::Pp2Rt

Statics

• FA_SMEM_TKV_DEFAULT
• FA_SP512_DEFAULT
• FA_SPW_DEFAULT
• FA_SP_GEMMA
• FA_V4_MAX_DEFAULT
• FA_VEC_MIN_DEFAULT
• FUSED_MR1_DEFAULT
• KV_FP8_FORCE
• MMQ_SK_FORCE
• PRIME_NANOS
• RMS_BLOCK_DEFAULT
• ROUTER_W8_DEFAULT
• cache::KV_FP8_FORCE
• pp::PRIME_PIPE_OVERLAPS
• pp::PRIME_SPLIT_CHUNKS
• pp::STEP35_PRIME_BATCHES
• pp::STEP35_PRIME_BATCH_SPLITS

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_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::PRIME_CHUNK_MAX_TOKENS
• decode_batch::BATCH_PHASE_NAMES
• hybrid_forward::PRIME_MIN_T
• moe_cache::PROJ_DOWN
• moe_cache::PROJ_GATE
• moe_cache::PROJ_UP
• spec::SPEC_TELEM_POS