Skip to content

[SYCL][CUDA][libclc] Add remaining atomics #5192

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
merged 5 commits into from
Jan 18, 2022
Merged

[SYCL][CUDA][libclc] Add remaining atomics #5192

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
7cd8c49
[SYCL][CUDA][libclc] implemented atomic min, max and sub
t4c1 Dec 9, 2021
585bfe0
Merge branch 'sycl' into remaining_ptx_atomics
t4c1 Dec 21, 2021
340196d
format
t4c1 Dec 21, 2021
b555211
Merge branch 'sycl' into remaining_ptx_atomics
t4c1 Jan 17, 2022
109a845
format new files
t4c1 Jan 17, 2022
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
1 change: 1 addition & 0 deletions libclc/ptx-nvidiacl/libspirv/SOURCES
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -95,6 +95,7 @@ atomic/atomic_dec.cl
atomic/atomic_inc.cl
atomic/atomic_max.cl
atomic/atomic_min.cl
atomic/atomic_sub.cl
atomic/atomic_xchg.cl
atomic/atomic_or.cl
atomic/atomic_xor.cl
Expand Down
2 changes: 1 addition & 1 deletion libclc/ptx-nvidiacl/libspirv/atomic/atomic_inc_dec_helpers.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -23,7 +23,7 @@
volatile ADDR_SPACE TYPE *pointer, enum Scope scope, \
enum MemorySemanticsMask semantics) { \
return _Z21__spirv_AtomicIAddEXTPU3##ADDR_SPACE_MANGLED##TYPE_MANGLED##N5__spv5Scope4FlagENS1_19MemorySemanticsMask4FlagE##TYPE_MANGLED( \
pointer, scope, semantics, 1); \
pointer, scope, semantics, VAL); \
}

#define __CLC_NVVM_ATOMIC_INCDEC(TYPE, TYPE_MANGLED, OP_MANGLED, VAL) \
Expand Down
57 changes: 57 additions & 0 deletions libclc/ptx-nvidiacl/libspirv/atomic/atomic_max.cl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -20,3 +20,60 @@ __CLC_NVVM_ATOMIC(unsigned long, m, unsigned long, ul, max,
#undef __CLC_NVVM_ATOMIC_TYPES
#undef __CLC_NVVM_ATOMIC
#undef __CLC_NVVM_ATOMIC_IMPL

#define __CLC_NVVM_ATOMIC_MAX_IMPL(TYPE, TYPE_MANGLED, TYPE_INT, \
TYPE_INT_MANGLED, OP_MANGLED, ADDR_SPACE, \
ADDR_SPACE_MANGLED) \
TYPE_INT \
_Z18__spirv_AtomicLoadPU3##ADDR_SPACE_MANGLED##K##TYPE_INT_MANGLED##N5__spv5Scope4FlagENS1_19MemorySemanticsMask4FlagE( \
volatile ADDR_SPACE const TYPE_INT *, enum Scope, \
enum MemorySemanticsMask); \
TYPE_INT \
_Z29__spirv_AtomicCompareExchange##PU3##ADDR_SPACE_MANGLED##TYPE_INT_MANGLED##N5__spv5Scope4FlagENS1_19MemorySemanticsMask4FlagES5_##TYPE_INT_MANGLED##TYPE_INT_MANGLED( \
volatile ADDR_SPACE TYPE_INT *, enum Scope, enum MemorySemanticsMask, \
enum MemorySemanticsMask, TYPE_INT, TYPE_INT); \
_CLC_DECL TYPE \
_Z21__spirv_Atomic##OP_MANGLED##PU3##ADDR_SPACE_MANGLED##TYPE_MANGLED##N5__spv5Scope4FlagENS1_19MemorySemanticsMask4FlagE##TYPE_MANGLED( \
volatile ADDR_SPACE TYPE *pointer, enum Scope scope, \
enum MemorySemanticsMask semantics, TYPE val) { \
enum MemorySemanticsMask load_order; \
switch (semantics) { \
case SequentiallyConsistent: \
load_order = SequentiallyConsistent; \
break; \
case Acquire: \
case AcquireRelease: \
load_order = Acquire; \
break; \
default: \
load_order = None; \
} \
volatile ADDR_SPACE TYPE_INT *pointer_int = \
(volatile ADDR_SPACE TYPE_INT *)pointer; \
TYPE_INT val_int = *(TYPE_INT *)&val; \
TYPE_INT old_int = \
_Z18__spirv_AtomicLoadPU3##ADDR_SPACE_MANGLED##K##TYPE_INT_MANGLED##N5__spv5Scope4FlagENS1_19MemorySemanticsMask4FlagE( \
pointer_int, scope, load_order); \
TYPE old = *(TYPE *)&old_int; \
while (val > old) { \
TYPE_INT tmp_int = \
_Z29__spirv_AtomicCompareExchange##PU3##ADDR_SPACE_MANGLED##TYPE_INT_MANGLED##N5__spv5Scope4FlagENS1_19MemorySemanticsMask4FlagES5_##TYPE_INT_MANGLED##TYPE_INT_MANGLED( \
pointer_int, scope, semantics, semantics, val_int, old_int); \
if (old_int == tmp_int) { \
return *(TYPE *)&tmp_int; \
} \
old_int = tmp_int; \
old = *(TYPE *)&old_int; \
} \
return old; \
}

#define __CLC_NVVM_ATOMIC_MAX(TYPE, TYPE_MANGLED, TYPE_INT, TYPE_INT_MANGLED, \
OP_MANGLED) \
__CLC_NVVM_ATOMIC_MAX_IMPL(TYPE, TYPE_MANGLED, TYPE_INT, TYPE_INT_MANGLED, \
OP_MANGLED, __global, AS1) \
__CLC_NVVM_ATOMIC_MAX_IMPL(TYPE, TYPE_MANGLED, TYPE_INT, TYPE_INT_MANGLED, \
OP_MANGLED, __local, AS3)

__CLC_NVVM_ATOMIC_MAX(float, f, int, i, FMaxEXT)
__CLC_NVVM_ATOMIC_MAX(double, d, long, l, FMaxEXT)
57 changes: 57 additions & 0 deletions libclc/ptx-nvidiacl/libspirv/atomic/atomic_min.cl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -18,3 +18,60 @@ __CLC_NVVM_ATOMIC(ulong, m, ulong, ul, min, _Z18__spirv_AtomicUMin)
#undef __CLC_NVVM_ATOMIC_TYPES
#undef __CLC_NVVM_ATOMIC
#undef __CLC_NVVM_ATOMIC_IMPL

#define __CLC_NVVM_ATOMIC_MIN_IMPL(TYPE, TYPE_MANGLED, TYPE_INT, \
TYPE_INT_MANGLED, OP_MANGLED, ADDR_SPACE, \
ADDR_SPACE_MANGLED) \
TYPE_INT \
_Z18__spirv_AtomicLoadPU3##ADDR_SPACE_MANGLED##K##TYPE_INT_MANGLED##N5__spv5Scope4FlagENS1_19MemorySemanticsMask4FlagE( \
volatile ADDR_SPACE const TYPE_INT *, enum Scope, \
enum MemorySemanticsMask); \
TYPE_INT \
_Z29__spirv_AtomicCompareExchange##PU3##ADDR_SPACE_MANGLED##TYPE_INT_MANGLED##N5__spv5Scope4FlagENS1_19MemorySemanticsMask4FlagES5_##TYPE_INT_MANGLED##TYPE_INT_MANGLED( \
volatile ADDR_SPACE TYPE_INT *, enum Scope, enum MemorySemanticsMask, \
enum MemorySemanticsMask, TYPE_INT, TYPE_INT); \
_CLC_DECL TYPE \
_Z21__spirv_Atomic##OP_MANGLED##PU3##ADDR_SPACE_MANGLED##TYPE_MANGLED##N5__spv5Scope4FlagENS1_19MemorySemanticsMask4FlagE##TYPE_MANGLED( \
volatile ADDR_SPACE TYPE *pointer, enum Scope scope, \
enum MemorySemanticsMask semantics, TYPE val) { \
enum MemorySemanticsMask load_order; \
switch (semantics) { \
case SequentiallyConsistent: \
load_order = SequentiallyConsistent; \
break; \
case Acquire: \
case AcquireRelease: \
load_order = Acquire; \
break; \
default: \
load_order = None; \
} \
volatile ADDR_SPACE TYPE_INT *pointer_int = \
(volatile ADDR_SPACE TYPE_INT *)pointer; \
TYPE_INT val_int = *(TYPE_INT *)&val; \
TYPE_INT old_int = \
_Z18__spirv_AtomicLoadPU3##ADDR_SPACE_MANGLED##K##TYPE_INT_MANGLED##N5__spv5Scope4FlagENS1_19MemorySemanticsMask4FlagE( \
pointer_int, scope, load_order); \
TYPE old = *(TYPE *)&old_int; \
while (val < old) { \
TYPE_INT tmp_int = \
_Z29__spirv_AtomicCompareExchange##PU3##ADDR_SPACE_MANGLED##TYPE_INT_MANGLED##N5__spv5Scope4FlagENS1_19MemorySemanticsMask4FlagES5_##TYPE_INT_MANGLED##TYPE_INT_MANGLED( \
pointer_int, scope, semantics, semantics, val_int, old_int); \
if (old_int == tmp_int) { \
return *(TYPE *)&tmp_int; \
} \
old_int = tmp_int; \
old = *(TYPE *)&old_int; \
} \
return old; \
}

#define __CLC_NVVM_ATOMIC_MIN(TYPE, TYPE_MANGLED, TYPE_INT, TYPE_INT_MANGLED, \
OP_MANGLED) \
__CLC_NVVM_ATOMIC_MIN_IMPL(TYPE, TYPE_MANGLED, TYPE_INT, TYPE_INT_MANGLED, \
OP_MANGLED, __global, AS1) \
__CLC_NVVM_ATOMIC_MIN_IMPL(TYPE, TYPE_MANGLED, TYPE_INT, TYPE_INT_MANGLED, \
OP_MANGLED, __local, AS3)

__CLC_NVVM_ATOMIC_MIN(float, f, int, i, FMinEXT)
__CLC_NVVM_ATOMIC_MIN(double, d, long, l, FMinEXT)
Loading