Skip to content

Fix arange error type for fractional step with integer output on XPU #3701

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
AKloniecki wants to merge 5 commits into
base: main
Choose a base branch
from
+7 −17
Open

Fix arange error type for fractional step with integer output on XPU #3701

Changes from all commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
8584d5b
Fix arange error type for fractional step with integer output on XPU
AKloniecki May 18, 2026
3e68faf
Scope arange int64 temporaries to constexpr branch
Copilot May 18, 2026
10a5471
Merge branch 'main' into aklonieckix/fix-arange
AKloniecki May 21, 2026
2035a8b
Merge branch 'main' into aklonieckix/fix-arange
AKloniecki May 22, 2026
cd57131
Merge branch 'main' into aklonieckix/fix-arange
AKloniecki May 25, 2026
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
24 changes: 7 additions & 17 deletions src/ATen/native/xpu/RangeFactories.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -19,6 +19,7 @@
#include <ATen/core/Tensor.h>
#include <ATen/native/DispatchStub.h>
#include <ATen/native/RangeFactories.h>
#include <ATen/native/RangeUtils.h>
#include <ATen/native/TensorIterator.h>
#include <ATen/native/xpu/sycl/RangeFactoriesKernel.h>
#include <comm/xpu_aten.h>
Expand All @@ -43,23 +44,7 @@ Tensor& arange_out_xpu(
out.scalar_type(),
"arange_xpu_preprocess",
[&]() {
using accscalar_t = at::acc_type_device<scalar_t, kXPU>;
auto xstart = start.to<accscalar_t>();
auto xend = end.to<accscalar_t>();
auto xstep = step.to<accscalar_t>();

TORCH_CHECK(xstep > 0 || xstep < 0, "step must be nonzero");
TORCH_CHECK(
std::isfinite(static_cast<double>(xstart)) &&
std::isfinite(static_cast<double>(xend)),
"unsupported range: ",
xstart,
" -> ",
xend);
TORCH_CHECK(
((xstep > 0) && (xend >= xstart)) ||
((xstep < 0) && (xend <= xstart)),
"upper bound and larger bound inconsistent with step sign");
arange_check_bounds(start, end, step);

// we use double precision for (start - end) / step
// to compute size_d for consistency across devices.
Expand All @@ -71,6 +56,11 @@ Tensor& arange_out_xpu(
// than double
double size_d;
if constexpr (std::is_same_v<scalar_t, int64_t>) {
using accscalar_t = at::acc_type_device<scalar_t, kXPU>;
auto xstart = start.to<accscalar_t>();
auto xend = end.to<accscalar_t>();
auto xstep = step.to<accscalar_t>();
TORCH_CHECK_VALUE(xstep != 0, "step must be nonzero");
int64_t sgn = (xstep > 0) - (xstep < 0);
size_d = std::ceil((xend - xstart + xstep - sgn) / xstep);
} else {
Expand Down
Loading