Sign in
Join now
raven
stepbrobd.com
/
Star
Starred
Star
0
Something went wrong. Try again.
Fork
0
Eye filled
Watch
Watching
Watch
forked from raven-ml/raven
Star
Starred
Star
0
Something went wrong. Try again.
Fork
0
Eye filled
Watch
Watching
Watch
Overview
Overview
Issues
Issues
Pulls
Pulls
Pipelines
Pipelines
raven
/
dev
/
nx2
/
lib
/
cpu
1 folder
·
42 files
at
master
Thibaut Mattio
·
perf(nx2): Run a batch of small matrices a matrix a unit
·
2w ago
7feec3b7
probe
feat(nx2): Add nx2 with its CPU backend
3w ago
amx.h
feat(nx2): Add nx2 with its CPU backend
3w ago
apply.c
perf(nx2): Price a kind in the dtype it computes in
2w ago
assemble.c
perf(nx2): Join a pad's whole inner axes into its rows
2w ago
convert.c
refactor(nx2): Keep only what varies in nx.cpu's target table
2w ago
copy.c
perf(nx2): Price a kind in the dtype it computes in
2w ago
cpu.h
perf(nx2): Price a kind in the dtype it computes in
2w ago
dune
refactor(nx2): Give each derived contraction its own file
2w ago
fft_body.h
refactor(nx2): Give a transform one line cost for its call and jobs
2w ago
fft.c
refactor(nx2): Give a transform one line cost for its call and jobs
2w ago
gemm_amx.c
feat(nx2): Add nx2 with its CPU backend
3w ago
gemm_avx2.c
refactor(nx2): Give each derived contraction its own file
2w ago
gemm_avx2.h
feat(nx2): Add nx2 with its CPU backend
3w ago
gemm_complex.c
refactor(nx2): Give a contraction one cost formula at a known speed
2w ago
gemm_generic.c
feat(nx2): Require x86-64-v3 in nx.cpu
2w ago
gemm_lanes.h
refactor(nx2): Define the lane and block trees once in cpu.h
2w ago
gemm_neon.c
feat(nx2): Add nx2 with its CPU backend
3w ago
gemm_quads.c
refactor(nx2): Give a contraction one cost formula at a known speed
2w ago
gemm_rounded.c
perf(nx2): Price a kind in the dtype it computes in
2w ago
gemm_runs.c
perf(nx2): Price a kind in the dtype it computes in
2w ago
gemm_scales.c
refactor(nx2): Give a contraction one cost formula at a known speed
2w ago
gemm_segments.c
refactor(nx2): Give each derived contraction its own file
2w ago
gemm_view.c
refactor(nx2): Rename nx.cpu's view.c to gemm_view.c
2w ago
gemm_vnni.c
refactor(nx2): Give each derived contraction its own file
2w ago
gemm.c
refactor(nx2): Give a contraction one cost formula at a known speed
2w ago
gemm.h
refactor(nx2): Give a contraction one cost formula at a known speed
2w ago
index.c
refactor(nx2): Size a scatter's jobs by its updates' cost
2w ago
job.c
refactor(nx2): Size a walk's threads by its call's cost
2w ago
linalg_body.h
perf(nx2): Run a batch of small matrices a matrix a unit
2w ago
linalg.c
perf(nx2): Run a batch of small matrices a matrix a unit
2w ago
map.c
perf(nx2): Price a kind in the dtype it computes in
2w ago
nx_cpu.ml
refactor(nx2): Drop nx.cpu's OCaml re-checks of plain loads
2w ago
nx_cpu.mli
feat(nx2): Compute host loops of any width
2w ago
reduce_body.h
refactor(nx2): Write reduce's float-only layers as plain C
2w ago
reduce_float.h
refactor(nx2): Write reduce's float-only layers as plain C
2w ago
reduce_network.h
refactor(nx2): Write reduce's float-only layers as plain C
2w ago
reduce.c
perf(nx2): Price a kind in the dtype it computes in
2w ago
rows.c
perf(nx2): Price a kind in the dtype it computes in
2w ago
runs.h
refactor(nx2): Move nx.cpu's copies and runs out of nx_array.h
3w ago
sort.c
fix(nx2): Release a sort's lock before counting its threads
2w ago
stage.c
refactor(nx2): Keep only what varies in nx.cpu's target table
2w ago
target.c
refactor(nx2): Move nx.cpu's kinds' costs into rows.c
2w ago
walk.c
refactor(nx2): Size a walk's threads by its call's cost
2w ago