MCPcopy Create free account

hub / github.com/LaurentMazare/ug / functions

Functions526 in github.com/LaurentMazare/ug

⨍ Functions 526◇ Types & classes 143
↓ 1 callersMethodcompile_ptx
(&self, ptx_code: &str, func_name: &'static str)
ug-cuda/src/runtime.rs:188
↓ 1 callersMethodcopy_device_to_host
(&self, dst: &mut [DT])
ug-core/src/cpu_runtime.rs:396
↓ 1 callersMethodcopy_host_to_device
(&mut self, src: &[DT])
ug-cuda/src/runtime.rs:367
↓ 1 callersMethodcos
(&self)
ug-core/src/interpreter.rs:40
↓ 1 callersMethodcuda_code
(&self, name: &str)
ug-pyo3/src/lib.rs:396
↓ 1 callersMethoddata_type
(&self)
ug-metal/src/runtime.rs:323
↓ 1 callersMethoddims3
(&self)
ug-core/src/lazy_buffer.rs:217
↓ 1 callersMethoddiv
(self, rhs: usize)
ug-core/src/lower_op.rs:82
↓ 1 callersMethoddst
(&self)
ug-core/src/lang.rs:376
↓ 1 callersMethodelem_count
The total number of elements, this is the product of all dimension sizes.
ug-core/src/layout.rs:140
↓ 1 callersMethodeval
TODO: We currently use i32 for indexes, however this prevents numerous compile time optimizations and it would be better to use u32 and/or u64.
ug-core/src/lower_op.rs:92
↓ 1 callersFunctioneval_add
()
ug-core/examples/dot_prod.rs:3
↓ 1 callersFunctioneval_add
()
ug-cuda/examples/cuda_gen.rs:3
↓ 1 callersFunctioneval_dotprod
()
ug-core/examples/dot_prod.rs:14
↓ 1 callersFunctioneval_dotprod
()
ug-cuda/examples/cuda_gen.rs:11
↓ 1 callersFunctioneval_lower_add
()
ug-cuda/examples/cuda_gen.rs:33
↓ 1 callersFunctioneval_softmax
()
ug-cuda/examples/cuda_gen.rs:57
↓ 1 callersMethodexp
(&self)
ug-core/src/interpreter.rs:34
↓ 1 callersFunctionextract_const
Simple optimization that extract the constant bits that do not depend of the index on some specific dim so that these can be evaluated out of loop.
ug-core/src/lower_op.rs:254
↓ 1 callersMethodfinal_size
(self)
ug-core/src/display.rs:140
↓ 1 callersMethodflops_mem_per_thread
(&self)
ug-core/src/lang.rs:755
↓ 1 callersMethodfold_op
(&self)
ug-core/src/lower_op.rs:243
↓ 1 callersMethodfunction_constant_values
(&self)
ug-metal/src/runtime.rs:344
↓ 1 callersMethodfwd
(&self, xs: &LB<D>)
ug-llama/src/model.rs:125
↓ 1 callersFunctiongemm_reduced_precision_bf16
This bool controls whether reduced precision reductions (e.g., with fp16 accumulation type) are allowed with bf16 GEMMs.
ug-cuda/src/gemm.rs:120
↓ 1 callersFunctiongemm_reduced_precision_f16
This bool controls whether reduced precision reductions (e.g., with fp16 accumulation type) are allowed with f16 GEMMs.
ug-cuda/src/gemm.rs:108
↓ 1 callersFunctiongemm_reduced_precision_f32
This bool controls whether reduced precision reductions (e.g., with tf32 accumulation type) are allowed with f32 GEMMs.
ug-cuda/src/gemm.rs:96
↓ 1 callersFunctiongemm_strided_batched_bf16
( stream: &Arc<cudarc::driver::CudaStream>, cublas: &cudarc::cublas::CudaBlas, cfg: StridedBatched
ug-cuda/src/gemm.rs:234
↓ 1 callersFunctiongemm_strided_batched_f16
( stream: &Arc<cudarc::driver::CudaStream>, cublas: &cudarc::cublas::CudaBlas, cfg: StridedBatched
ug-cuda/src/gemm.rs:176
↓ 1 callersFunctiongemm_strided_batched_f32
( stream: &Arc<cudarc::driver::CudaStream>, cublas: &cudarc::cublas::CudaBlas, cfg: StridedBatched
ug-cuda/src/gemm.rs:130
↓ 1 callersFunctionget_num_threads
()
ug-core/src/utils.rs:14
↓ 1 callersMethodget_on_dim
(&self, dim: D1, index: usize)
ug-core/src/lazy_buffer.rs:529
↓ 1 callersMethodget_ssa
(&self, kernel: &Ssa)
ug-core/src/cache.rs:94
↓ 1 callersFunctionget_summarized_data
(t: &LB<D>, edge_items: usize)
ug-core/src/display.rs:372
↓ 1 callersFunctionid_cnts
Return the number of uses for each buffer that is reachable from b. The number of uses can be either 1 or 2 for the case where the buffer is used twic
ug-core/src/schedule.rs:480
↓ 1 callersMethodin_place_op
(&self)
ug-core/src/lazy_buffer.rs:130
↓ 1 callersFunctionindex_select
(src: &LB<D>, ids: &LB<D>)
ug-llama/src/model.rs:32
↓ 1 callersMethodinit_value
(&self, dtype: DType)
ug-core/src/lower_op.rs:235
↓ 1 callersMethodinsert_ssa
(&mut self, kernel: Ssa, func: std::sync::Arc<D::Func>)
ug-core/src/cache.rs:98
↓ 1 callersFunctionlayout
(op: LayoutOp, arg: Ast, shape: S)
ug-core/src/lang.rs:479
↓ 1 callersMethodlayout_op
( &self, op: &crate::lang::op::LayoutOp, shape: &Shape, arg_shape: &Shape,
ug-core/src/lower_op.rs:127
↓ 1 callersFunctionload
(arg: Arg, shape: Vec<usize>)
ug-pyo3/src/lib.rs:206
↓ 1 callersMethodload_slice_with_cast
( &self, name: &str, dst_dtype: DType, device: &D, )
ug-core/src/safetensors.rs:120
↓ 1 callersFunctionlower_add
()
ug-core/examples/dot_prod.rs:25
↓ 1 callersMethodlower_b
(&self, opts: &Opts)
ug-core/src/lower_op.rs:409
↓ 1 callersMethodlower_b
(&self)
ug-core/src/lower.rs:143
↓ 1 callersFunctionmatmul
( blas: &cudarc::cublas::CudaBlas, dst: &mut Slice, lhs: &Slice, rhs: &Slice, bmnk: (usize
ug-cuda/src/gemm.rs:292
↓ 1 callersMethodmax
(&self, axis: usize)
ug-pyo3/src/lib.rs:194
↓ 1 callersMethodmod_
(self, rhs: usize)
ug-core/src/lower_op.rs:86
↓ 1 callersMethodname
(&self)
ug-core/src/lang.rs:601
↓ 1 callersMethodoffset
(&self)
ug-core/src/layout.rs:656
↓ 1 callersMethodoptimize
(mut self)
ug-core/src/lower_op.rs:493
↓ 1 callersMethodprogram_id
()
ug-core/src/lang.rs:219
↓ 1 callersFunctionpy_file
(module, origin)
stub.py:145
↓ 1 callersFunctionpyi_file
(obj, indent="")
stub.py:85
↓ 1 callersFunctionrms_norm
(src: &LB<D>, alpha: &LB<D>, eps: f32)
ug-llama/src/model.rs:65
↓ 1 callersMethodrun3
(&self, v1: &mut [T], v2: &mut [T], v3: &mut [T])
ug-core/src/cpu_runtime.rs:450
↓ 1 callersFunctionrun_one
(args: &Args, n_cols: usize)
ug-metal/examples/metal_bench.rs:30
↓ 1 callersFunctionrun_one
(args: &Args, n_cols: usize)
ug-cuda/examples/benchmark.rs:27
↓ 1 callersMethodset
Sets self using data from values.
ug-core/src/lazy_buffer.rs:332
↓ 1 callersMethodset_offset
(&mut self, offset: usize)
ug-core/src/layout.rs:660
↓ 1 callersFunctionset_param
(encoder: &ComputeCommandEncoderRef, position: u64, data: P)
ug-metal/src/utils.rs:4
↓ 1 callersFunctionsilu
(src: &LB<D>)
ug-llama/src/model.rs:103
↓ 1 callersMethodsin
(&self)
ug-core/src/interpreter.rs:37
↓ 1 callersMethodslice_from_values
(&self, data: &[T])
ug-metal/src/runtime.rs:94
↓ 1 callersFunctionsoftmax
()
ug-core/tests/basic_tests.rs:41
↓ 1 callersFunctionsoftmax
()
ug-core/examples/dot_prod.rs:37
↓ 1 callersFunctionsoftmax
(dim1, dim2)
py-examples/softmax.py:3
↓ 1 callersFunctionsoftmax_barrier
(_dim1: usize, dim2: usize)
ug-core/src/samples.rs:101
↓ 1 callersFunctionsoftmax_reduce
(_dim1: usize, dim2: usize)
ug-core/src/samples.rs:160
↓ 1 callersMethodsqrt
(&self)
ug-core/src/interpreter.rs:31
↓ 1 callersMethodsqueeze
Contrary to pytorch, this returns an error if the target dimension does not have a size 1.
ug-core/src/lazy_buffer.rs:519
↓ 1 callersMethodsrc
(&self)
ug-core/src/lang.rs:370
↓ 1 callersMethodstride_contiguous
The strides given in number of elements for a contiguous n-dimensional arrays using this shape.
ug-core/src/layout.rs:146
↓ 1 callersMethodsum
(&self, axis: usize)
ug-pyo3/src/lib.rs:184
↓ 1 callersMethodto_index_plus_one
(&self, shape: &Shape, op: &'static str)
ug-core/src/layout.rs:240
↓ 1 callersMethodto_indexes_internal
(self, _: &Shape, _: &'static str)
ug-core/src/layout.rs:327
↓ 1 callersMethodto_vec
(&self)
ug-pyo3/src/lib.rs:59
↓ 1 callersMethodwalk
(&mut self, b: &LazyBuffer<D>)
ug-core/src/schedule.rs:327
↓ 1 callersMethodwith_global_axis
(mut self, axis: usize, block_dim: usize)
ug-core/src/lower_op.rs:21
Method__add__
(&self, rhs: &Self)
ug-pyo3/src/lib.rs:149
Method__mul__
(&self, rhs: &Self)
ug-pyo3/src/lib.rs:159
Method__neg__
(&self)
ug-pyo3/src/lib.rs:169
Method__str__
(&self)
ug-pyo3/src/lib.rs:138
Method__sub__
(&self, rhs: &Self)
ug-pyo3/src/lib.rs:154
Method__truediv__
(&self, rhs: &Self)
ug-pyo3/src/lib.rs:164
Function_index_select
(src: &LB, ids: &LB)
ug-llama/src/custom.rs:51
Function_transpose
(src: &LB, dim1: usize, dim2: usize)
ug-llama/src/custom.rs:6
Methodadd
(&mut self, src_id: Id, v: i32)
ug-core/src/block.rs:68
Functionaffine
()
ug-core/tests/basic_tests.rs:119
Methodall_args
(&self, args: &mut std::collections::BTreeSet<Arg>)
ug-core/src/lang.rs:338
Methodalloc_uninit
TODO: Should this be marked as unsafe?
ug-core/src/lazy_buffer.rs:279
Methodallocate_uninit
(&self, dtype: ug::DType, len: usize)
ug-metal/src/runtime.rs:176
Methodas_usize
(&self)
ug-core/src/lazy_buffer.rs:45
Methodast
(&self)
ug-core/src/schedule.rs:19
Methodbf16
()
ug-pyo3/src/lib.rs:88
Methodbinary
(&self, op: crate::lang::BinaryOp, rhs: Self)
ug-core/src/lazy_buffer.rs:259
Methodbinary
( &mut self, op: lang::BinaryOp, lhs: I1, rhs: I2, dtype: lang::DType,
ug-core/src/block.rs:121
Methodbinary
(op: &str, lhs: usize, rhs: usize, dtype: DType)
ug-pyo3/src/lib.rs:359
Methodblas
(&self)
ug-cuda/src/runtime.rs:175
← previousnext →201–300 of 526, ranked by callers