Age | Commit message (Collapse) | Author | Files | Lines |
|
for this case 1.0f/src, 2.0f/src can be converted,
but 3.0f/src and i/src cant
Signed-off-by: rander.wang <rander.wang@intel.com>
Reviewed-by: Yang Rong <rong.r.yang@intel.com>
|
|
Check the device supported subgroup sizes, and use
intel_reqd_sub_group_size to build kernels in these size. Then check if
there is spill for each kernel.
V2: Fix memory leak
Signed-off-by: Pan Xiuli <xiuli.pan@intel.com>
Reviewed-by: Yang Rong <rong.r.yang@intel.com>
|
|
the negative Add is like:
exp -a
llvm transfer it to:
add x -a, 0
exp x
Signed-off-by: rander.wang <rander.wang@intel.com>
Reviewed-by: Pan Xiuli <xiuli.pan@intel.com>
|
|
Signed-off-by: rander.wang <rander.wang@intel.com>
Reviewed-by: Yang Rong <rong.r.yang@intel.com>
|
|
the test case modify src as well as dst.
and it introduce cross workitem memory dependency in dst[id] = src[id+1];
The compiler may order 'then' and 'else' block not as written.
If compiler order the else block first. src[3+1] will be modified in else part.
And the utest will get wrong result. If user want to get the old behaviour,
It should use two if-then and order them as required.
Signed-off-by: Ruiling Song <ruiling.song@intel.com>
Reviewed-by: Yang Rong <rong.r.yang@intel.com>
|
|
After clang check the image qualifier, can't use default qualifier
to write_image.
Signed-off-by: Yang Rong <rong.r.yang@intel.com>
Reviewed-by: Pan Xiuli <xiuli.pan@intel.com>
|
|
Signed-off-by: Luo Xionghu <xionghu.luo@intel.com>
Reviewed-by: Yang Rong <rong.r.yang@intel.com>
|
|
v2: remove printf code in kernel; disable DEBUG_OUTPUT.
Signed-off-by: Luo Xionghu <xionghu.luo@intel.com>
Reviewed-by: Pan Xiuli <xiuli.pan@intel.com>
|
|
Signed-off-by: Luo Xionghu <xionghu.luo@intel.com>
Reviewed-by: Pan Xiuli <xiuli.pan@intel.com>
|
|
v2: use read_only/write_only instead of read_write to run on OpenCL-1.2
platform; fix local size issue on IVB platform;
Signed-off-by: Luo Xionghu <xionghu.luo@intel.com>
Reviewed-by: Yang Rong <rong.r.yang@intel.com>
|
|
Signed-off-by: Chuanbo Weng <chuanbo.weng@intel.com>
Reviewed-by: Yang Rong <rong.r.yang@intel.com>
|
|
Signed-off-by: Yang Rong <rong.r.yang@intel.com>
Reviewed-by: Pan Xiuli <xiuli.pan@intel.com>
|
|
although we have eliminate ConstantExpr in llvm instructions,
but in program scope variable, we still meet ConstantExpr.
So, we handle it here. also enhance the test case to hit it.
Signed-off-by: Ruiling Song <ruiling.song@intel.com>
Reviewed-by: Pan Xiuli <xiuli.pan@intel.com>
|
|
Add compiler_ceil64.spir for load_program_from_spir utest.
Signed-off-by: Pan Xiuli <xiuli.pan@intel.com>
Reviewed-by: Yang Rong <rong.r.yang@intel.com>
|
|
Add test case for builtin with user struct type and int type and
runtime tset for creatPipe and pipe query.
Signed-off-by: Pan Xiuli <xiuli.pan@intel.com>
Reviewed-by: Yang Rong <rong.r.yang@intel.com>
|
|
Signed-off-by: Luo Xionghu <xionghu.luo@intel.com>
Reviewed-by: Ruiling Song <ruiling.song@intel.com>
|
|
only dword is supported so far.
Signed-off-by: Luo Xionghu <xionghu.luo@intel.com>
Reviewed-by: Ruiling Song <ruiling.song@intel.com>
|
|
Check all type of ctz function and 0 num bound case.
V2: Fix type warning
Signed-off-by: Pan Xiuli <xiuli.pan@intel.com>
Reviewed-by: Yang Rong <rong.r.yang@intel.com>
|
|
Different from ocl 1.2, which gather all constant into one surface.
ocl2 only gather program global/constant into one surface. But
keep other constant passed through kernel argument in separate buffer.
Signed-off-by: Ruiling Song <ruiling.song@intel.com>
Reviewed-by: Yang Rong <rong.r.yang@intel.com>
|
|
Signed-off-by: Ruiling Song <ruiling.song@intel.com>
Reviewed-by: Yang Rong <rong.r.yang@intel.com>
|
|
Add ushort block read/write for buffer and image.
Refine uint block read/write with suffix _ui.
Signed-off-by: Pan Xiuli <xiuli.pan@intel.com>
Reviewed-by: Yang Rong <rong.r.yang@intel.com>
|
|
Signed-off-by: Pan Xiuli <xiuli.pan@intel.com>
Reviewed-by: Yang Rong <rong.r.yang@intel.com>
|
|
Signed-off-by: Pan Xiuli <xiuli.pan@intel.com>
Reviewed-by: Yang Rong <rong.r.yang@intel.com>
|
|
Signed-off-by: Pan Xiuli <xiuli.pan@intel.com>
Reviewed-by: Yang Rong <rong.r.yang@intel.com>
|
|
Signed-off-by: Pan Xiuli <xiuli.pan@intel.com>
Reviewed-by: Ruiling Song <ruiling.song@intel.com>
|
|
Sub group broadcast case will fail in simd 16, for the kernel doesn't
fit the test case.
Signed-off-by: Pan Xiuli <xiuli.pan@intel.com>
Reviewed-by: Yang Rong <rong.r.yang@intel.com>
|
|
In simd8 mode, the sub group size is 8. We should use some index less
than 8.
Signed-off-by: Pan Xiuli <xiuli.pan@intel.com>
Reviewed-by: Yang Rong <rong.r.yang@intel.com>
|
|
Check if device support subgroup and half first, use build options
to hide code for unsported device.
V2: Fix half part test case for utest multithread.
Signed-off-by: Pan Xiuli <xiuli.pan@intel.com>
Reviewed-by: Yang Rong <rong.r.yang@intel.com>
|
|
Mad now can support half type, add a test.
Signed-off-by: Pan Xiuli <xiuli.pan@intel.com>
Reviewed-by: Yang Rong <rong.r.yang@intel.com>
|
|
V2: Half program is different with normal program, reorder the test case
order.
Signed-off-by: Pan Xiuli <xiuli.pan@intel.com>
Signed-off-by: Pan Xiuli <xiuli.pan@intel.com>
Reviewed-by: Yang Rong <rong.r.yang@intel.com>
|
|
Functions to benchmark math functions on intervals.
Tests: sin, cos, exp2, exp, exp10, log2, log, log10
Signed-off-by: Grigore Lupescu <grigore.lupescu at intel.com>
Reviewed-by: Ruiling Song <ruiling.song@intel.com>
|
|
V2:Add subgroups extension check.
Signed-off-by: Pan Xiuli <xiuli.pan@intel.com>
Reviewed-by: Yang Rong <rong.r.yang@intel.com>
|
|
the default access qualifier is __read_only, which is not correct
for the image to be write in kernel, set the qualifer explictly.
Signed-off-by: Guo Yejun <yejun.guo@intel.com>
Reviewed-by: Yang Rong <rong.r.yang@intel.com>
|
|
Signed-off-by: Pan Xiuli <xiuli.pan@intel.com>
Reviewed-by: Yang Rong <rong.r.yang@intel.com>
|
|
V2: Rename test case to buffer block read/write test
Signed-off-by: Pan Xiuli <xiuli.pan@intel.com>
Reviewed-by: Yang Rong <rong.r.yang@intel.com>
|
|
Long type need to be fixed before gen8, so hide them now.
Signed-off-by: Pan Xiuli <xiuli.pan@intel.com>
Reviewed-by: Yang Rong <rong.r.yang@intel.com>
|
|
This utest does not follow spec, so just remove them.
Signed-off-by: Pan Xiuli <xiuli.pan@intel.com>
Reviewed-by: Yang Rong <rong.r.yang@intel.com>
|
|
Signed-off-by: Pan Xiuli <xiuli.pan@intel.com>
Reviewed-by: Yang Rong <rong.r.yang@intel.com>
|
|
^ has lower precedence than <, add parentheses.
Signed-off-by: Luo Xionghu <xionghu.luo@intel.com>
Reviewed-by: Yang Rong <rong.r.yang@intel.com>
|
|
HSW and IVB does not support long type, now hide these tests.
V2: Remove some unsupport kernel.
Signed-off-by: Pan Xiuli <xiuli.pan@intel.com>
Reviewed-by: Yang Rong <rong.r.yang@intel.com>
|
|
Fix the problem with AND implemention, use predicate simd width to
get in-thread all and any result.
V2: Fix a typo in utest.
Signed-off-by: Pan Xiuli <xiuli.pan@intel.com>
Reviewed-by: Yang Rong <rong.r.yang@intel.com>
|
|
Added the following performance tests:
benchmark_workgroup_broadcast_1D_int
benchmark_workgroup_broadcast_1D_long
benchmark_workgroup_broadcast_2D_int
benchmark_workgroup_broadcast_2D_long
Signed-off-by: Grigore Lupescu <grigore.lupescu at intel.com>
Reviewed-by: Pan Xiuli <xiuli.pan@intel.com>
|
|
Added the following performance tests:
benchmark_workgroup_reduce_add_int
benchmark_workgroup_reduce_add_long
benchmark_workgroup_reduce_min_int
benchmark_workgroup_reduce_min_long
benchmark_workgroup_scan_inclusive_add_int
benchmark_workgroup_scan_inclusive_add_long
benchmark_workgroup_scan_inclusive_min_int
benchmark_workgroup_scan_inclusive_min_long
Signed-off-by: Grigore Lupescu <grigore.lupescu at intel.com>
Reviewed-by: Pan Xiuli <xiuli.pan@intel.com>
|
|
Added the following unit tests:
compiler_workgroup_broadcast_1D_int
compiler_workgroup_broadcast_1D_long
compiler_workgroup_broadcast_2D_int
compiler_workgroup_broadcast_2D_long
compiler_workgroup_broadcast_3D_int
compiler_workgroup_broadcast_3D_long
Signed-off-by: Grigore Lupescu <grigore.lupescu at intel.com>
Reviewed-by: Pan Xiuli <xiuli.pan@intel.com>
|
|
Added the following unit tests:
compiler_workgroup_scan_inclusive_add_int
compiler_workgroup_scan_inclusive_add_uint
compiler_workgroup_scan_inclusive_add_long
compiler_workgroup_scan_inclusive_add_ulong
compiler_workgroup_scan_inclusive_add_float
compiler_workgroup_scan_inclusive_max_int
compiler_workgroup_scan_inclusive_max_uint
compiler_workgroup_scan_inclusive_max_long
compiler_workgroup_scan_inclusive_max_ulong
compiler_workgroup_scan_inclusive_max_float
compiler_workgroup_scan_inclusive_min_int
compiler_workgroup_scan_inclusive_min_uint
compiler_workgroup_scan_inclusive_min_long
compiler_workgroup_scan_inclusive_min_ulong
compiler_workgroup_scan_inclusive_min_float
Signed-off-by: Grigore Lupescu <grigore.lupescu at intel.com>
Reviewed-by: Pan Xiuli <xiuli.pan@intel.com>
|
|
Added the following unit tests:
compiler_workgroup_scan_exclusive_add_int
compiler_workgroup_scan_exclusive_add_uint
compiler_workgroup_scan_exclusive_add_long
compiler_workgroup_scan_exclusive_add_ulong
compiler_workgroup_scan_exclusive_add_float
compiler_workgroup_scan_exclusive_max_int
compiler_workgroup_scan_exclusive_max_uint
compiler_workgroup_scan_exclusive_max_long
compiler_workgroup_scan_exclusive_max_ulong
compiler_workgroup_scan_exclusive_max_float
compiler_workgroup_scan_exclusive_min_int
compiler_workgroup_scan_exclusive_min_uint
compiler_workgroup_scan_exclusive_min_long
compiler_workgroup_scan_exclusive_min_ulong
compiler_workgroup_scan_exclusive_min_float
Signed-off-by: Grigore Lupescu <grigore.lupescu at intel.com>
Reviewed-by: Pan Xiuli <xiuli.pan@intel.com>
|
|
Added the following unit tests:
compiler_workgroup_any
compiler_workgroup_all
compiler_workgroup_reduce_add_int
compiler_workgroup_reduce_add_uint
compiler_workgroup_reduce_add_long
compiler_workgroup_reduce_add_ulong
compiler_workgroup_reduce_add_float
compiler_workgroup_reduce_max_int
compiler_workgroup_reduce_max_uint
compiler_workgroup_reduce_max_long
compiler_workgroup_reduce_max_ulong
compiler_workgroup_reduce_max_float
compiler_workgroup_reduce_min_int
compiler_workgroup_reduce_min_uint
compiler_workgroup_reduce_min_long
compiler_workgroup_reduce_min_ulong
compiler_workgroup_reduce_min_float
Signed-off-by: Grigore Lupescu <grigore.lupescu at intel.com>
Reviewed-by: Pan Xiuli <xiuli.pan@intel.com>
|
|
v2: add uniform dest test coverage.
Signed-off-by: Luo Xionghu <xionghu.luo@intel.com>
Reviewed-by: Guo Yejun <yejun.guo@intel.com>
|
|
1. use clEnqueueMapBuffer/Image instead of clEnqueueReadBuffer/Image;
2. add sanity check for clEnqueueMapImage;
v2: disable OpenCL 2.0 specific builtin cases for stand alone utest.
v3: don't hide failed cases. fix utest build warnings.
Signed-off-by: Luo Xionghu <xionghu.luo@intel.com>
Reviewed-by: Yang Rong <rong.r.yang@intel.com>
|
|
Signed-off-by: Junyan He <junyan.he@linux.intel.com>
Reviewed-by: Yan Wang <yan.wang@linux.intel.com>
Reviewed-by: Junyan He <junyan.he@linux.intel.com>
|