Skip to content

linked sub mul div tanh sigmoid from oss #6619

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
wants to merge 2 commits into
base: main
Choose a base branch
from
Open

linked sub mul div tanh sigmoid from oss #6619

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
887579a
port add ops, create new buck target, add op_add kernel modification
Nov 1, 2024
4ed5dbc
linked sub mul div tanh sigmoid from oss (#6619)
zonglinpeng Nov 1, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
119 changes: 119 additions & 0 deletions backends/cadence/hifi/operators/targets.bzl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -31,3 +31,122 @@ def define_common_targets():
"@EXECUTORCH_CLIENTS",
],
)

runtime.cxx_library(
name = "op_add",
srcs = glob([
"op_add.cpp",
]),
platforms = CXX,
deps = [
"//executorch/kernels/portable/cpu/util:all_deps",
"//executorch/kernels/portable/cpu/pattern:all_deps",
"//executorch/runtime/kernel:kernel_includes",
"//executorch/kernels/portable/cpu:scalar_utils",
"//executorch/backends/cadence/hifi/kernels:kernels",
"//executorch/backends/cadence/hifi/third-party/nnlib:nnlib-extensions"
],
visibility = [
"//executorch/backends/cadence/...",
"@EXECUTORCH_CLIENTS",
],
)


runtime.cxx_library(
name = "op_mul",
srcs = glob([
"op_mul.cpp",
]),
platforms = CXX,
deps = [
"//executorch/kernels/portable/cpu/util:all_deps",
"//executorch/kernels/portable/cpu/pattern:all_deps",
"//executorch/runtime/kernel:kernel_includes",
"//executorch/kernels/portable/cpu:scalar_utils",
"//executorch/backends/cadence/hifi/kernels:kernels",
"//executorch/backends/cadence/hifi/third-party/nnlib:nnlib-extensions"
],
visibility = [
"//executorch/backends/cadence/...",
"@EXECUTORCH_CLIENTS",
],
)

runtime.cxx_library(
name = "op_sub",
srcs = glob([
"op_sub.cpp",
]),
platforms = CXX,
deps = [
"//executorch/kernels/portable/cpu/util:all_deps",
"//executorch/kernels/portable/cpu/pattern:all_deps",
"//executorch/runtime/kernel:kernel_includes",
"//executorch/kernels/portable/cpu:scalar_utils",
"//executorch/backends/cadence/hifi/kernels:kernels",
"//executorch/backends/cadence/hifi/third-party/nnlib:nnlib-extensions"
],
visibility = [
"//executorch/backends/cadence/...",
"@EXECUTORCH_CLIENTS",
],
)

runtime.cxx_library(
name = "op_div",
srcs = glob([
"op_div.cpp",
]),
platforms = CXX,
deps = [
"//executorch/kernels/portable/cpu/util:all_deps",
"//executorch/kernels/portable/cpu/pattern:all_deps",
"//executorch/runtime/kernel:kernel_includes",
"//executorch/kernels/portable/cpu:scalar_utils",
"//executorch/backends/cadence/hifi/kernels:kernels",
"//executorch/backends/cadence/hifi/third-party/nnlib:nnlib-extensions"
],
visibility = [
"//executorch/backends/cadence/...",
"@EXECUTORCH_CLIENTS",
],
)

runtime.cxx_library(
name = "op_sigmoid",
srcs = glob([
"op_sigmoid.cpp",
]),
platforms = CXX,
deps = [
"//executorch/kernels/portable/cpu/util:all_deps",
"//executorch/kernels/portable/cpu/pattern:all_deps",
"//executorch/runtime/kernel:kernel_includes",
"//executorch/backends/cadence/hifi/kernels:kernels",
"//executorch/backends/cadence/hifi/third-party/nnlib:nnlib-extensions"
],
visibility = [
"//executorch/backends/cadence/...",
"@EXECUTORCH_CLIENTS",
],
)

runtime.cxx_library(
name = "op_tanh",
srcs = glob([
"op_tanh.cpp",
]),
platforms = CXX,
deps = [
"//executorch/kernels/portable/cpu/util:all_deps",
"//executorch/kernels/portable/cpu/pattern:all_deps",
"//executorch/runtime/kernel:kernel_includes",
"//executorch/backends/cadence/hifi/kernels:kernels",
"//executorch/backends/cadence/hifi/third-party/nnlib:nnlib-extensions"
],
visibility = [
"//executorch/backends/cadence/...",
"@EXECUTORCH_CLIENTS",
],
)
5 changes: 5 additions & 0 deletions backends/cadence/hifi/third-party/nnlib/TARGETS
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,5 @@
load("targets.bzl", "define_common_targets")

oncall("odai_jarvis")

define_common_targets()
18 changes: 18 additions & 0 deletions backends/cadence/hifi/third-party/nnlib/targets.bzl
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,18 @@
# Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved.

load("@fbsource//tools/build_defs:platform_defs.bzl", "CXX")
load("@fbsource//xplat/executorch/build:runtime_wrapper.bzl", "runtime")

def define_common_targets():
runtime.cxx_library(
name = "nnlib-extensions",
srcs = native.glob(["*.c", "*.cpp"]),
exported_headers = glob(["*.h"]),
visibility = [
"//executorch/backends/cadence/...",
"@EXECUTORCH_CLIENTS",
],
deps = [
"fbsource//third-party/nnlib-hifi4/xa_nnlib:libxa_nnlib",
],
)
2 changes: 0 additions & 2 deletions backends/cadence/hifi/third-party/nnlib/xa_nn_elm_add_f32_broadcast.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -25,7 +25,6 @@
#include "xa_nnlib_err_chk.h"
#include "xa_nnlib_kernels_api.h"


#if HAVE_VFPU
static void internal_elm_add_broadcast_2D_f32xf32_f32(FLOAT32 * __restrict__ p_out,
const FLOAT32 * __restrict__ p_inp1,
Expand Down Expand Up @@ -425,4 +424,3 @@ WORD32 xa_nn_elm_add_broadcast_4D_f32xf32_f32(FLOAT32 * __restrict__ p_out,
return 0;

}

56 changes: 28 additions & 28 deletions backends/cadence/hifi/third-party/nnlib/xa_nn_elm_div_mode_f32_broadcast.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -54,7 +54,7 @@ WORD32 xa_nn_elm_div_mode_f32xf32_f32(FLOAT32 * __restrict__ p_out,
XT_LASX2IP(x1, inp1_a, inp1);
XT_LASX2IP(x2, inp2_a, inp2);
y = XT_DIV_SX2(x1, x2);
y = FITRUNC_SX2(y);
y = XT_FITRUNC_SX2(y);
XT_SASX2IP(y, out_a, out);
}
}
Expand All @@ -66,7 +66,7 @@ WORD32 xa_nn_elm_div_mode_f32xf32_f32(FLOAT32 * __restrict__ p_out,
XT_LASX2IP(x1, inp1_a, inp1);
XT_LASX2IP(x2, inp2_a, inp2);
y = XT_DIV_SX2(x1, x2);
y = FIFLOOR_SX2(y);
y = XT_FIFLOOR_SX2(y);
XT_SASX2IP(y, out_a, out);
}
}
Expand All @@ -80,9 +80,9 @@ WORD32 xa_nn_elm_div_mode_f32xf32_f32(FLOAT32 * __restrict__ p_out,
XT_LSIP(a2, (xtfloat *)inp2, 0);
a = XT_DIV_S(a1, a2);
if(mode == 0)
a = FITRUNC_S(a);
a = XT_FITRUNC_S(a);
else
a = FIFLOOR_S(a);
a = XT_FIFLOOR_S(a);
XT_SSI(a, (xtfloat *)out, 0);
}

Expand Down Expand Up @@ -138,7 +138,7 @@ static void internal_elm_div_mode_broadcast_2D_f32xf32_f32(FLOAT32 * __restrict_
XT_LSX2IP(x1, p_a, 2 * sizeof(FLOAT32));
XT_LSX2IP(x2, p_b, 2 * sizeof(FLOAT32));
y = XT_DIV_SX2(x2, x1);
y = FITRUNC_SX2(y);
y = XT_FITRUNC_SX2(y);
XT_SSX2IP(y, p_c, 2 * sizeof(FLOAT32));
}
}
Expand All @@ -149,7 +149,7 @@ static void internal_elm_div_mode_broadcast_2D_f32xf32_f32(FLOAT32 * __restrict_
XT_LSX2IP(x1, p_a, 2 * sizeof(FLOAT32));
XT_LSX2IP(x2, p_b, 2 * sizeof(FLOAT32));
y = XT_DIV_SX2(x2, x1);
y = FIFLOOR_SX2(y);
y = XT_FIFLOOR_SX2(y);
XT_SSX2IP(y, p_c, 2 * sizeof(FLOAT32));
}
}
Expand All @@ -166,7 +166,7 @@ static void internal_elm_div_mode_broadcast_2D_f32xf32_f32(FLOAT32 * __restrict_
XT_LASX2IP(x1, vinp1, p_a);
XT_LASX2IP(x2, vinp2, p_b);
y = XT_DIV_SX2(x2, x1);
y = FITRUNC_SX2(y);
y = XT_FITRUNC_SX2(y);
XT_SASX2IP(y, out_a, p_c);
}
}
Expand All @@ -177,7 +177,7 @@ static void internal_elm_div_mode_broadcast_2D_f32xf32_f32(FLOAT32 * __restrict_
XT_LASX2IP(x1, vinp1, p_a);
XT_LASX2IP(x2, vinp2, p_b);
y = XT_DIV_SX2(x2, x1);
y = FIFLOOR_SX2(y);
y = XT_FIFLOOR_SX2(y);
XT_SASX2IP(y, out_a, p_c);
}
}
Expand All @@ -189,9 +189,9 @@ static void internal_elm_div_mode_broadcast_2D_f32xf32_f32(FLOAT32 * __restrict_
XT_LSIP(b0, (xtfloat *)p_b, sizeof(FLOAT32));
c0 = XT_DIV_S(b0, a0);
if(mode == 0)
c0 = FITRUNC_S(c0);
c0 = XT_FITRUNC_S(c0);
else
c0 = FIFLOOR_S(c0);
c0 = XT_FIFLOOR_S(c0);
XT_SSI(c0, (xtfloat *)p_c, 0);
}
}
Expand All @@ -213,7 +213,7 @@ static void internal_elm_div_mode_broadcast_2D_f32xf32_f32(FLOAT32 * __restrict_
XT_LSX2IP(x1, p_a, 2 * sizeof(FLOAT32));
XT_LSX2IP(x2, p_b, 2 * sizeof(FLOAT32));
y = XT_DIV_SX2(x1, x2);
y = FITRUNC_SX2(y);
y = XT_FITRUNC_SX2(y);
XT_SSX2IP(y, p_c, 2 * sizeof(FLOAT32));
}
}
Expand All @@ -224,7 +224,7 @@ static void internal_elm_div_mode_broadcast_2D_f32xf32_f32(FLOAT32 * __restrict_
XT_LSX2IP(x1, p_a, 2 * sizeof(FLOAT32));
XT_LSX2IP(x2, p_b, 2 * sizeof(FLOAT32));
y = XT_DIV_SX2(x1, x2);
y = FIFLOOR_SX2(y);
y = XT_FIFLOOR_SX2(y);
XT_SSX2IP(y, p_c, 2 * sizeof(FLOAT32));
}
}
Expand All @@ -241,7 +241,7 @@ static void internal_elm_div_mode_broadcast_2D_f32xf32_f32(FLOAT32 * __restrict_
XT_LASX2IP(x1, vinp1, p_a);
XT_LASX2IP(x2, vinp2, p_b);
y = XT_DIV_SX2(x1, x2);
y = FITRUNC_SX2(y);
y = XT_FITRUNC_SX2(y);
XT_SASX2IP(y, out_a, p_c);
}
}
Expand All @@ -252,7 +252,7 @@ static void internal_elm_div_mode_broadcast_2D_f32xf32_f32(FLOAT32 * __restrict_
XT_LASX2IP(x1, vinp1, p_a);
XT_LASX2IP(x2, vinp2, p_b);
y = XT_DIV_SX2(x1, x2);
y = FIFLOOR_SX2(y);
y = XT_FIFLOOR_SX2(y);
XT_SASX2IP(y, out_a, p_c);
}
}
Expand All @@ -264,9 +264,9 @@ static void internal_elm_div_mode_broadcast_2D_f32xf32_f32(FLOAT32 * __restrict_
XT_LSIP(b0, (xtfloat *)p_b, sizeof(FLOAT32));
c0 = XT_DIV_S(a0, b0);
if(mode == 0)
c0 = FITRUNC_S(c0);
c0 = XT_FITRUNC_S(c0);
else
c0 = FIFLOOR_S(c0);
c0 = XT_FIFLOOR_S(c0);
XT_SSI(c0, (xtfloat *)p_c, 0);
}
}
Expand Down Expand Up @@ -302,7 +302,7 @@ static void internal_elm_div_mode_broadcast_f32xf32_f32(FLOAT32 * __restrict__ p
{
XT_LSX2IP(x1, p_a, 2 * sizeof(FLOAT32));
y = XT_DIV_SX2(x2, x1);
y = FITRUNC_SX2(y);
y = XT_FITRUNC_SX2(y);
XT_SSX2IP(y, p_c, 2 * sizeof(FLOAT32));
}
}
Expand All @@ -312,7 +312,7 @@ static void internal_elm_div_mode_broadcast_f32xf32_f32(FLOAT32 * __restrict__ p
{
XT_LSX2IP(x1, p_a, 2 * sizeof(FLOAT32));
y = XT_DIV_SX2(x2, x1);
y = FIFLOOR_SX2(y);
y = XT_FIFLOOR_SX2(y);
XT_SSX2IP(y, p_c, 2 * sizeof(FLOAT32));
}
}
Expand All @@ -328,7 +328,7 @@ static void internal_elm_div_mode_broadcast_f32xf32_f32(FLOAT32 * __restrict__ p
{
XT_LASX2IP(x1, inp1_a, p_a);
y = XT_DIV_SX2(x2, x1);
y = FITRUNC_SX2(y);
y = XT_FITRUNC_SX2(y);
XT_SASX2IP(y, out_a, p_c);
}
}
Expand All @@ -338,7 +338,7 @@ static void internal_elm_div_mode_broadcast_f32xf32_f32(FLOAT32 * __restrict__ p
{
XT_LASX2IP(x1, inp1_a, p_a);
y = XT_DIV_SX2(x2, x1);
y = FIFLOOR_SX2(y);
y = XT_FIFLOOR_SX2(y);
XT_SASX2IP(y, out_a, p_c);
}
}
Expand All @@ -349,9 +349,9 @@ static void internal_elm_div_mode_broadcast_f32xf32_f32(FLOAT32 * __restrict__ p
XT_LSIP(a0_7, (xtfloat *)p_a, sizeof(FLOAT32));
out = XT_DIV_S(x2, a0_7);
if(mode == 0)
out = FITRUNC_S(out);
out = XT_FITRUNC_S(out);
else
out = FIFLOOR_S(out);
out = XT_FIFLOOR_S(out);
XT_SSI(out, (xtfloat *)p_c, 0);
}
}
Expand All @@ -366,7 +366,7 @@ static void internal_elm_div_mode_broadcast_f32xf32_f32(FLOAT32 * __restrict__ p
{
XT_LSX2IP(x1, p_a, 2 * sizeof(FLOAT32));
y = XT_DIV_SX2(x1, x2);
y = FITRUNC_SX2(y);
y = XT_FITRUNC_SX2(y);
XT_SSX2IP(y, p_c, 2 * sizeof(FLOAT32));
}
}
Expand All @@ -376,7 +376,7 @@ static void internal_elm_div_mode_broadcast_f32xf32_f32(FLOAT32 * __restrict__ p
{
XT_LSX2IP(x1, p_a, 2 * sizeof(FLOAT32));
y = XT_DIV_SX2(x1, x2);
y = FIFLOOR_SX2(y);
y = XT_FIFLOOR_SX2(y);
XT_SSX2IP(y, p_c, 2 * sizeof(FLOAT32));
}
}
Expand All @@ -392,7 +392,7 @@ static void internal_elm_div_mode_broadcast_f32xf32_f32(FLOAT32 * __restrict__ p
{
XT_LASX2IP(x1, inp1_a, p_a);
y = XT_DIV_SX2(x1, x2);
y = FITRUNC_SX2(y);
y = XT_FITRUNC_SX2(y);
XT_SASX2IP(y, out_a, p_c);
}
}
Expand All @@ -402,7 +402,7 @@ static void internal_elm_div_mode_broadcast_f32xf32_f32(FLOAT32 * __restrict__ p
{
XT_LASX2IP(x1, inp1_a, p_a);
y = XT_DIV_SX2(x1, x2);
y = FIFLOOR_SX2(y);
y = XT_FIFLOOR_SX2(y);
XT_SASX2IP(y, out_a, p_c);
}
}
Expand All @@ -413,9 +413,9 @@ static void internal_elm_div_mode_broadcast_f32xf32_f32(FLOAT32 * __restrict__ p
XT_LSIP(a0_7, (xtfloat *)p_a, sizeof(FLOAT32));
out = XT_DIV_S(a0_7, x2);
if(mode == 0)
out = FITRUNC_S(out);
out = XT_FITRUNC_S(out);
else
out = FIFLOOR_S(out);
out = XT_FIFLOOR_S(out);
XT_SSI(out, (xtfloat *)p_c, 0);
}
}
Expand Down
9 changes: 4 additions & 5 deletions backends/cadence/hifi/third-party/nnlib/xa_nn_elm_mul_f32_broadcast.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -20,11 +20,10 @@

******************************************************************************/
#include "xa_type_def.h"
#include "nnlib-hifi4/xa_nnlib/algo/common/include/xa_nnlib_common_fpu.h"
#include "nnlib-hifi4/xa_nnlib/algo/common/include/xa_nn_common.h"
#include "nnlib-hifi4/xa_nnlib/algo/common/include/xa_nnlib_err_chk.h"
#include "nnlib-hifi4/xa_nnlib/algo/kernels/basic/hifi4/xa_nn_basic_state.h"
#include "nnlib-hifi4/xa_nnlib/include/nnlib/xa_nnlib_kernels_api.h"
#include "xa_nnlib_common_fpu.h"
#include "xa_nn_common.h"
#include "xa_nnlib_err_chk.h"
#include "xa_nnlib_kernels_api.h"

#if HAVE_VFPU
static void internal_elm_mul_broadcast_2D_f32xf32_f32(FLOAT32 * __restrict__ p_out,
Expand Down
Loading