Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
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
2 changes: 1 addition & 1 deletion .ci/docker/ci_commit_pins/pytorch.txt
Original file line number Diff line number Diff line change
@@ -1 +1 @@
59d5cf083b4f860dea76fe8936076177f9367f10
01f1cc44cbbfdf6307aa01b803a4ee22f9ade946
6 changes: 5 additions & 1 deletion backends/xnnpack/cmake/Dependencies.cmake
Original file line number Diff line number Diff line change
Expand Up @@ -35,7 +35,11 @@ set(XNNPACK_BUILD_TESTS
set(XNNPACK_ENABLE_AVXVNNI
OFF
CACHE BOOL ""
)
)
# Work around observed failure: https://github.com/pytorch/executorch/pull/10362#issuecomment-2906391232
set(XNNPACK_ENABLE_AVX512VNNIGFNI
OFF
CACHE BOOL "")

if(EXECUTORCH_XNNPACK_ENABLE_KLEIDI)
set(XNNPACK_ENABLE_KLEIDIAI
Expand Down
2 changes: 1 addition & 1 deletion install_requirements.py
Original file line number Diff line number Diff line change
Expand Up @@ -71,7 +71,7 @@ def python_is_compatible():
#
# NOTE: If you're changing, make the corresponding change in .ci/docker/ci_commit_pins/pytorch.txt
# by picking the hash from the same date in https://hud.pytorch.org/hud/pytorch/pytorch/nightly/
NIGHTLY_VERSION = "dev20250524"
NIGHTLY_VERSION = "dev20250422"


def install_requirements(use_pytorch_nightly):
Expand Down
10 changes: 10 additions & 0 deletions runtime/core/portable_type/c10/c10/macros/Macros.h
Original file line number Diff line number Diff line change
Expand Up @@ -508,4 +508,14 @@ __host__ __device__

#endif

// This macro is used to find older C++ compilers
// that don't support move optimization for return values.

#if (defined(__GNUC__) && __GNUC__ < 13) || \
(defined(__clang_major__) && __clang_major__ < 13)
#define C10_RETURN_MOVE_IF_OLD_COMPILER 1
#else
#define C10_RETURN_MOVE_IF_OLD_COMPILER 0
#endif

#endif // C10_MACROS_MACROS_H_
8 changes: 4 additions & 4 deletions runtime/core/portable_type/c10/c10/util/BFloat16.h
Original file line number Diff line number Diff line change
Expand Up @@ -31,7 +31,7 @@ inline C10_HOST_DEVICE float f32_from_bits(uint16_t src) {
uint32_t tmp = src;
tmp <<= 16;

#if defined(USE_ROCM)
#if defined(USE_ROCM) && defined(__HIPCC__)
float* tempRes;

// We should be using memcpy in order to respect the strict aliasing rule
Expand All @@ -48,7 +48,7 @@ inline C10_HOST_DEVICE float f32_from_bits(uint16_t src) {
inline C10_HOST_DEVICE uint16_t bits_from_f32(float src) {
uint32_t res = 0;

#if defined(USE_ROCM)
#if defined(USE_ROCM) && defined(__HIPCC__)
// We should be using memcpy in order to respect the strict aliasing rule
// but it fails in the HIP environment.
uint32_t* tempRes = reinterpret_cast<uint32_t*>(&src);
Expand All @@ -61,7 +61,7 @@ inline C10_HOST_DEVICE uint16_t bits_from_f32(float src) {
}

inline C10_HOST_DEVICE uint16_t round_to_nearest_even(float src) {
#if defined(USE_ROCM)
#if defined(USE_ROCM) && defined(__HIPCC__)
if (src != src) {
#elif defined(_MSC_VER)
if (isnan(src)) {
Expand All @@ -87,7 +87,7 @@ struct alignas(2) BFloat16 {
uint16_t x;

// HIP wants __host__ __device__ tag, CUDA does not
#if defined(USE_ROCM)
#if defined(USE_ROCM) && defined(__HIPCC__)
C10_HOST_DEVICE BFloat16() = default;
#else
BFloat16() = default;
Expand Down
Loading