Skip to content

Commit 783002b

Browse files
Maetveisvmustya
andcommitted
[OpenCL] Add decls for cl_intel_subgroup_local_block_io
This extension extends the subgroup block read and write functions defined by `cl_intel_subgroups` (and, when supported, `cl_intel_subgroups_char`, `cl_intel_subgroups_short`, and `cl_intel_subgroups_long`) to support reading from and writing to pointers to the `__local` memory address space in addition to pointers to the `__global` memory address space. It is already supported by the Intel OpenCL compiler. Co-authored-by: Victor Mustya <[email protected]>
1 parent 651c520 commit 783002b

File tree

1 file changed

+66
-0
lines changed

1 file changed

+66
-0
lines changed

clang/lib/Headers/opencl-c.h

Lines changed: 66 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -17651,6 +17651,72 @@ void __ovld __conv intel_sub_group_block_write_us4( __global ushort* p, u
1765117651
void __ovld __conv intel_sub_group_block_write_us8( __global ushort* p, ushort8 data );
1765217652
#endif // cl_intel_subgroups_short
1765317653

17654+
#if defined(cl_intel_subgroup_local_block_io)
17655+
uint __ovld __conv intel_sub_group_block_read( const __local uint* p );
17656+
uint2 __ovld __conv intel_sub_group_block_read2( const __local uint* p );
17657+
uint4 __ovld __conv intel_sub_group_block_read4( const __local uint* p );
17658+
uint8 __ovld __conv intel_sub_group_block_read8( const __local uint* p );
17659+
17660+
void __ovld __conv intel_sub_group_block_write( __local uint* p, uint data );
17661+
void __ovld __conv intel_sub_group_block_write2( __local uint* p, uint2 data );
17662+
void __ovld __conv intel_sub_group_block_write4( __local uint* p, uint4 data );
17663+
void __ovld __conv intel_sub_group_block_write8( __local uint* p, uint8 data );
17664+
17665+
#if defined(cl_intel_subgroups_char) || defined(cl_intel_subgroups_short) || \
17666+
defined(cl_intel_subgroups_long)
17667+
uint __ovld __conv intel_sub_group_block_read_ui( const __local uint* p );
17668+
uint2 __ovld __conv intel_sub_group_block_read_ui2( const __local uint* p );
17669+
uint4 __ovld __conv intel_sub_group_block_read_ui4( const __local uint* p );
17670+
uint8 __ovld __conv intel_sub_group_block_read_ui8( const __local uint* p );
17671+
17672+
void __ovld __conv intel_sub_group_block_write_ui( __local uint* p, uint data );
17673+
void __ovld __conv intel_sub_group_block_write_ui2( __local uint* p, uint2 data );
17674+
void __ovld __conv intel_sub_group_block_write_ui4( __local uint* p, uint4 data );
17675+
void __ovld __conv intel_sub_group_block_write_ui8( __local uint* p, uint8 data );
17676+
#endif // defined(cl_intel_subgroups_char) || defined(cl_intel_subgroups_short) ||
17677+
// defined(cl_intel_subgroups_long)
17678+
17679+
#if defined(cl_intel_subgroups_char)
17680+
uchar __ovld __conv intel_sub_group_block_read_uc( const __local uchar* p );
17681+
uchar2 __ovld __conv intel_sub_group_block_read_uc2( const __local uchar* p );
17682+
uchar4 __ovld __conv intel_sub_group_block_read_uc4( const __local uchar* p );
17683+
uchar8 __ovld __conv intel_sub_group_block_read_uc8( const __local uchar* p );
17684+
uchar16 __ovld __conv intel_sub_group_block_read_uc16( const __local uchar* p );
17685+
17686+
void __ovld __conv intel_sub_group_block_write_uc( __local uchar* p, uchar data );
17687+
void __ovld __conv intel_sub_group_block_write_uc2( __local uchar* p, uchar2 data );
17688+
void __ovld __conv intel_sub_group_block_write_uc4( __local uchar* p, uchar4 data );
17689+
void __ovld __conv intel_sub_group_block_write_uc8( __local uchar* p, uchar8 data );
17690+
void __ovld __conv intel_sub_group_block_write_uc16( __local uchar* p, uchar16 data );
17691+
#endif // defined(cl_intel_subgroups_char)
17692+
17693+
#if defined(cl_intel_subgroups_short)
17694+
ushort __ovld __conv intel_sub_group_block_read_us( const __local ushort* p );
17695+
ushort2 __ovld __conv intel_sub_group_block_read_us2( const __local ushort* p );
17696+
ushort4 __ovld __conv intel_sub_group_block_read_us4( const __local ushort* p );
17697+
ushort8 __ovld __conv intel_sub_group_block_read_us8( const __local ushort* p );
17698+
ushort16 __ovld __conv intel_sub_group_block_read_us16( const __local ushort* p );
17699+
17700+
void __ovld __conv intel_sub_group_block_write_us( __local ushort* p, ushort data );
17701+
void __ovld __conv intel_sub_group_block_write_us2( __local ushort* p, ushort2 data );
17702+
void __ovld __conv intel_sub_group_block_write_us4( __local ushort* p, ushort4 data );
17703+
void __ovld __conv intel_sub_group_block_write_us8( __local ushort* p, ushort8 data );
17704+
void __ovld __conv intel_sub_group_block_write_us16( __local ushort* p, ushort16 data );
17705+
#endif // defined(cl_intel_subgroups_short)
17706+
17707+
#if defined(cl_intel_subgroups_long)
17708+
ulong __ovld __conv intel_sub_group_block_read_ul( const __local ulong* p );
17709+
ulong2 __ovld __conv intel_sub_group_block_read_ul2( const __local ulong* p );
17710+
ulong4 __ovld __conv intel_sub_group_block_read_ul4( const __local ulong* p );
17711+
ulong8 __ovld __conv intel_sub_group_block_read_ul8( const __local ulong* p );
17712+
17713+
void __ovld __conv intel_sub_group_block_write_ul( __local ulong* p, ulong data );
17714+
void __ovld __conv intel_sub_group_block_write_ul2( __local ulong* p, ulong2 data );
17715+
void __ovld __conv intel_sub_group_block_write_ul4( __local ulong* p, ulong4 data );
17716+
void __ovld __conv intel_sub_group_block_write_ul8( __local ulong* p, ulong8 data );
17717+
#endif // defined(cl_intel_subgroups_long)
17718+
#endif // cl_intel_subgroup_local_block_io
17719+
1765417720
#ifdef cl_intel_device_side_avc_motion_estimation
1765517721
#pragma OPENCL EXTENSION cl_intel_device_side_avc_motion_estimation : begin
1765617722

0 commit comments

Comments
 (0)