Welcome to
mirror list
, hosted at
ThFree Co
, Russian Federation.
github.com/marian-nmt/intgemm.git - Unnamed repository; edit this file 'description' to name the repository.
index
:
github.com/marian-nmt/intgemm.git
4bit
absolute_std
add-postprocess-sigmoid
add127
benchmark_refactor
big-refactoring
compile_with_marian
debug_add127
hacky_nonmult8
log4-unstable
mac_support
marian-ssru
master
multiply-tiling
multiply-tiling-8x
multiply-tiling-manymatrices
overb
rearrangement-b
static
static-callbacks
static-multiply1x16
static-selectcolumnsb-colmajor
static-unquantize-and-add-bias
tests_for_kpu
unsigned
working
Unnamed repository; edit this file 'description' to name the repository.
www-data
summary
refs
log
tree
commit
diff
log msg
author
committer
range
path:
root
/
multiply.h
Age
Commit message (
Expand
)
Author
2019-03-15
Get rid of underscores
Nikolay Bogoychev
2019-03-15
Change tuple to struct
Kenneth Heafield
2019-03-15
First attempt at decoupling
Nikolay Bogoychev
2019-03-13
Separate part of the common output
Nikolay Bogoychev
2019-02-25
Genericize index type
Kenneth Heafield
2018-06-30
Fix maximum absolute value
Kenneth Heafield
2018-06-29
max absolute value
Kenneth Heafield
2018-06-28
Liberally apply static to inline functions
Kenneth Heafield
2018-06-28
Delete commented C version since there is now a C version
Kenneth Heafield
2018-06-28
Untested SelectColumnsB function
Kenneth Heafield
2018-06-26
Should fix SSSE3 illegal instruction?
Kenneth Heafield
2018-06-26
Extract inner loop from multiply
Kenneth Heafield
2018-06-26
Oops had the wrong width of sign instruction
Kenneth Heafield
2018-06-26
Merge branch 'master' of github.com:kpu/intgemm
Kenneth Heafield
2018-06-26
Sadly the 8-bit is not SSSE3, but AVX due to VEX
Kenneth Heafield
2018-06-26
8-bit initialize first round
kpu
2018-06-24
16-bit initialize sum instead of zeroing
kpu
2018-06-23
inline the template multiply functions
kpu
2018-06-23
AVX512 8-bit
kpu
2018-06-23
AVX512 16-bit multiply
kpu
2018-06-21
Fix headers and AVX512F
kpu
2018-06-19
Extract common SSE2 and AVX2 GEMM
Kenneth Heafield