Skip to content

Add ROCm support for xobjects. #166

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

Merged
szymonlopaciuk merged 7 commits into from
Mar 4, 2026
Merged

Add ROCm support for xobjects. #166

Show file tree
Hide file tree
Changes from 4 commits
Commits
Show all changes
7 commits
Select commit Hold shift + click to select a range
22e2b29
Add support for ROCm in ContextCupy
ekatralis Dec 10, 2025
4f87acb
Add definition for NULL, missing for ROCm
ekatralis Dec 10, 2025
19258bd
Clean up and add comments
ekatralis Dec 15, 2025
b6e3d27
Merge branch 'xsuite:main' into RocmSupportMain
ekatralis Jan 5, 2026
947e0f2
Fix headers on HIPRTC
ekatralis Jan 16, 2026
ebc7fa4
Fix formatting for black in CI
ekatralis Jan 19, 2026
901dde3
Pin black version in CI
ekatralis Jan 19, 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
2 changes: 1 addition & 1 deletion tests/test_shared_memory.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -47,7 +47,7 @@ class TestElement(xo.HybridClass):
unsigned int gid = blockIdx.x*blockDim.x + threadIdx.x; // global thread ID: 0,1,2,3

// init shared memory with chunk of input array
extern __shared__ double sdata[2];
extern __shared__ double sdata[];
sdata[tid] = input_arr[gid];
__syncthreads();

Expand Down
11 changes: 9 additions & 2 deletions xobjects/context_cupy.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -352,15 +352,22 @@ def __invert__(self):

cudaheader: List[SourceType] = [
"""\
typedef signed long long int64_t; //only_for_context cuda
typedef signed int int32_t; //only_for_context cuda
typedef signed short int16_t; //only_for_context cuda
typedef signed char int8_t; //only_for_context cuda
typedef unsigned long long uint64_t; //only_for_context cuda
typedef unsigned int uint32_t; //only_for_context cuda
typedef unsigned short uint16_t; //only_for_context cuda
typedef unsigned char uint8_t; //only_for_context cuda

#if defined(__CUDACC__) && !defined(__HIPCC__)
typedef signed long long int64_t;
typedef unsigned long long uint64_t;
#endif

#ifndef NULL
#define NULL nullptr
#endif

"""
]

Expand Down
10 changes: 9 additions & 1 deletion xobjects/headers/atomicadd.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -101,7 +101,7 @@ DEF_ATOMIC_ADD(double , f64)
// -------------------------------------------
#if defined(XO_CONTEXT_CUDA)
// CUDA compiler may not have <stdint.h>, so define the types if needed.
#ifdef __CUDACC_RTC__
#if defined(__CUDACC_RTC__) && !defined(__HIPCC__)
// NVRTC (CuPy RawModule default) can’t see <stdint.h>, so detect it via __CUDACC_RTC__
typedef signed char int8_t;
typedef short int16_t;
Expand All @@ -111,6 +111,14 @@ DEF_ATOMIC_ADD(double , f64)
typedef unsigned short uint16_t;
typedef unsigned int uint32_t;
typedef unsigned long long uint64_t;
#elif defined(__HIPCC__) && !defined(__CUDACC_RTC__)
// ROCm appears to have definitions for 64-bit int types
typedef signed char int8_t;
typedef short int16_t;
typedef int int32_t;
typedef unsigned char uint8_t;
typedef unsigned short uint16_t;
typedef unsigned int uint32_t;
#else
// Alternatively, NVCC path is fine with host headers
#include <stdint.h>
Expand Down