List of all items

Structs

• buffers::BufferArena
• buffers::BufferSizes
• buffers::MoeFp8Scratch
• buffers::decode_meta::DecodeMetaLayout
• cuda_backend::LedgerProbe
• cuda_backend::MetraleCudaBackend
• cuda_backend::tensormap::TensorMap
• cuda_host::CudaHost
• gpu::DevicePtr
• gpu::GraphHandle
• gpu::KernelHandle
• kernel::KernelModule
• kernel_args::KernelLaunch
• mlx_int8::MlxInt8Weight
• mlx_int8::MlxQuantConfig
• op_cache::OpCache
• registry::MetraleRegistry
• registry::RawCudaFunc
• stream::MetraleStream
• timing::GpuSpans
• timing::KernelSpan

Enums

• gpu::KernelArg

Traits

• gpu::GpuBackend

Functions

• buffers::q12_batched_scratch_bytes
• buffers::q12_batched_scratch_bytes_varlen
• buffers::q2_dequant_scratch_bytes
• buffers::ssm_rowwise_w_bf16_bytes
• buffers::ssm_rowwise_w_bf16_bytes_for
• buffers::ssm_rowwise_w_bf16_layer_bytes
• cublaslt::bf16_gemm_act_weight_t
• cublaslt::bf16_gemm_act_weight_t_f32_out
• cublaslt::fp8_gemm_act_weight_t_blkscaled
• cublaslt::fp8_gemm_act_weight_t_blkscaled_ldc
• cublaslt::fp8_gemm_act_weight_t_rowwise
• cublaslt::prewarm
• cublaslt::scale_layout::act_scale_rowmajor_to_kmajor
• cublaslt::scale_layout::blk128x128_stride_ok
• cublaslt::scale_layout::k_groups
• cublaslt::scale_layout::rowmajor_index
• cublaslt::scale_layout::vec128_b_elems
• cublaslt::scale_layout::vec128_b_index
• cuda_backend::arch_preflight::check_arch
• cuda_backend::arch_preflight::check_sm_count
• cuda_backend::arch_preflight::device_compute_capability
• cuda_backend::arch_preflight::device_compute_capability_of
• cuda_backend::arch_preflight::device_sm_count_of
• cuda_backend::arch_preflight::preflight_arch
• cuda_backend::arch_preflight::preflight_device_arch
• cuda_backend::cuda_free_memory_bytes
• cuda_backend::spawn_oom_watchdog
• cuda_host::host
• cuda_host::release
• cutlass::bf16_gemm_act_weight_t
• cutlass::nvfp4_gemm_bf16_act_weight_t
• cutlass::nvfp4_grouped_down
• cutlass::nvfp4_grouped_gate_up
• cutlass::nvfp4_grouped_gate_up_fused
• cutlass::pack_bf16_weight_to_nvfp4_t
• cutlass::pack_weight_sfb
• cutlass::transpose_nvfp4_packed_kton
• flashinfer::available
• flashinfer::ragged_prefill_bf16_hd128
• flashinfer::ragged_prefill_bf16_hd256
• gpu::baseline_free_bytes
• gpu::pack_kernel_args
• gpu::set_baseline_free_bytes
• hermetic_enabled
• host_heap::alloc_zeroed
• host_heap::free
• kernel::launch_config
• kernel::launch_vector_add
• kernel_args::div_ceil
• mlx_int8::gemv_gate_up
• pinned_hosts::is_pinned
• pinned_hosts::live_count
• pinned_hosts::register
• pinned_hosts::unregister
• registry::cuda_error_text
• registry::is_teardown_noop
• set_hermetic
• set_ssm_tail_midchunk
• ssm_tail_boundary
• ssm_tail_ckpt_enabled
• ssm_tail_midchunk_enabled
• timing::harvest
• timing::install_for_stream
• timing::kernel_begin
• timing::kernel_end
• timing::lane_begin
• timing::lane_end

Constants

• buffers::GATEUP_FUSED_MAX_M
• buffers::Q12_SIZING_STREAMS
• buffers::VERIFY_ROW_CAP
• buffers::decode_meta::DECODE_META_MAX_ROWS
• buffers::decode_meta::DECODE_META_MIN_ROWS
• registry::CUDA_ERROR_DEINITIALIZED
• timing::SPAN_SLOTS