Skip to content

[Exeuctorch][Optimized] Use portable path for fallback in op_mul #6733

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: gh/kimishpatel/137/base
Choose a base branch
from
Open

[Exeuctorch][Optimized] Use portable path for fallback in op_mul #6733

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
87886a5
[Exeuctorch][Optimized] Use portable path for fallback in op_mul
kimishpatel Nov 8, 2024
d9c2e24
Update on "[Exeuctorch][Optimized] Use portable path for fallback in …
kimishpatel Nov 8, 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
54 changes: 3 additions & 51 deletions kernels/optimized/cpu/op_mul.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -9,6 +9,7 @@
#include <executorch/kernels/optimized/cpu/binary_ops.h>
#include <executorch/kernels/optimized/vec/functional.h>
#include <executorch/kernels/optimized/vec/vec.h>
#include <executorch/kernels/portable/cpu/op_mul_impl.h>
#include <executorch/kernels/portable/cpu/scalar_utils.h>
#include <executorch/kernels/portable/cpu/util/broadcast_util.h>
#include <executorch/runtime/core/exec_aten/util/tensor_util.h> // IWYU pragma: export
Expand Down Expand Up @@ -240,36 +241,7 @@ Tensor& opt_mul_out(
} else if (selected_optimized_path != ElementwiseOptimizedPath::kNone) {
return handle_broadcast_mul(ctx, a, b, out, selected_optimized_path);
} else {
ScalarType common_type =
promoteTypes(a_type, b_type, /*half_to_float*/ true);
ET_KERNEL_CHECK(ctx, canCast(common_type, out_type), InvalidArgument, out);

ET_KERNEL_CHECK(
ctx,
resize_to_broadcast_target_size(a, b, out) == Error::Ok,
InvalidArgument,
out);

ET_SWITCH_REALHBBF16_TYPES(a_type, ctx, "mul.out", CTYPE_A, [&]() {
ET_SWITCH_REALHBBF16_TYPES(b_type, ctx, "mul.out", CTYPE_B, [&]() {
using CTYPE_IN = typename torch::executor::
promote_types<CTYPE_A, CTYPE_B, /*half_to_float*/ true>::type;
ET_DCHECK(CppTypeToScalarType<CTYPE_IN>::value == common_type);
ET_SWITCH_REALHBBF16_TYPES(out_type, ctx, "mul.out", CTYPE_OUT, [&]() {
apply_binary_elementwise_fn<CTYPE_A, CTYPE_B, CTYPE_OUT>(
[](const CTYPE_A val_a, const CTYPE_B val_b) {
CTYPE_IN a_casted = static_cast<CTYPE_IN>(val_a);
CTYPE_IN b_casted = static_cast<CTYPE_IN>(val_b);
CTYPE_IN value = a_casted * b_casted;

return static_cast<CTYPE_OUT>(value);
},
a,
b,
out);
});
});
});
mul_out_impl(ctx, a, b, out);
}

return out;
Expand Down Expand Up @@ -315,27 +287,7 @@ Tensor& opt_mul_scalar_out(
});
});
} else {
ET_SWITCH_REALHBBF16_TYPES(a_type, ctx, "mul.Scalar_out", CTYPE_A, [&]() {
ET_SWITCH_SCALAR_OBJ_TYPES(b_type, ctx, "mul.Scalar_out", CTYPE_B, [&]() {
ET_SWITCH_REALB_TYPES(
common_type, ctx, "mul.Scalar_out", CTYPE_IN, [&]() {
ET_SWITCH_REALHBBF16_TYPES(
out_type, ctx, "mul.Scalar_out", CTYPE_OUT, [&]() {
CTYPE_B b_val;
ET_EXTRACT_SCALAR(b, b_val);
CTYPE_IN b_casted = static_cast<CTYPE_IN>(b_val);

const size_t n = a.numel();
const CTYPE_A* a_data = a.const_data_ptr<CTYPE_A>();
CTYPE_OUT* out_data = out.mutable_data_ptr<CTYPE_OUT>();
for (auto i = 0; i < n; ++i) {
out_data[i] = static_cast<CTYPE_OUT>(
static_cast<CTYPE_IN>(a_data[i]) * b_casted);
}
});
});
});
});
mul_scalar_out_impl(ctx, a, b, out);
}

return out;
Expand Down
1 change: 1 addition & 0 deletions kernels/optimized/cpu/targets.bzl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -69,6 +69,7 @@ _OPTIMIZED_ATEN_OPS = (
name = "op_mul",
deps = [
":binary_ops",
"//executorch/kernels/portable/cpu:op_mul_impl",
"//executorch/kernels/portable/cpu:scalar_utils",
"//executorch/kernels/portable/cpu/util:broadcast_util",
"//executorch/runtime/core/exec_aten/util:tensor_util",
Expand Down
Loading