compute-runtime/runtime/context/driver_diagnostics.cpp

74 lines
11 KiB
C++

/*
* Copyright (c) 2017, Intel Corporation
*
* Permission is hereby granted, free of charge, to any person obtaining a
* copy of this software and associated documentation files (the "Software"),
* to deal in the Software without restriction, including without limitation
* the rights to use, copy, modify, merge, publish, distribute, sublicense,
* and/or sell copies of the Software, and to permit persons to whom the
* Software is furnished to do so, subject to the following conditions:
*
* The above copyright notice and this permission notice shall be included
* in all copies or substantial portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
* OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
* THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
* OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
* ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
* OTHER DEALINGS IN THE SOFTWARE.
*/
#include "driver_diagnostics.h"
namespace OCLRT {
DriverDiagnostics::DriverDiagnostics(cl_diagnostics_verbose_level level) {
this->verboseLevel = level;
}
bool DriverDiagnostics::validFlags(cl_diagnostics_verbose_level flags) const {
return !!(verboseLevel & flags);
}
const char *DriverDiagnostics::hintFormat[] = {
"Performance hint: clCreateBuffer with pointer %p and size %u doesn't meet alignment restrictions. Size should be aligned to %u bytes and pointer should be aligned to %u. Buffer is not sharing the same physical memory with CPU.", //CL_BUFFER_DOESNT_MEET_ALIGNMENT_RESTRICTIONS
"Performance hint: clCreateBuffer with pointer %p and size %u meets alignment restrictions and buffer will share the same physical memory with CPU.", //CL_BUFFER_MEETS_ALIGNMENT_RESTRICTIONS
"Performance hint: clCreateBuffer needs to allocate memory for buffer. For subsequent operations the buffer will share the same physical memory with CPU.", //CL_BUFFER_NEEDS_ALLOCATE_MEMORY
"Performance hint: clCreateImage with pointer %p meets alignment restrictions and image will share the same physical memory with CPU.", //CL_IMAGE_MEETS_ALIGNMENT_RESTRICTIONS
"Performance hint: Driver calls internal clFlush on the command queue each time 1 command is enqueued.", //DRIVER_CALLS_INTERNAL_CL_FLUSH
"Performance hint: Profiling adds overhead on all enqueue commands with events.", //PROFILING_ENABLED
"Performance hint: Profiled kernels will be executed with disabled preemption.", //PROFILING_ENABLED_WITH_DISABLED_PREEMPTION
"Performance hint: Subbuffer created from buffer %p shares the same memory with buffer.", //SUBBUFFER_SHARES_MEMORY
"Performance hint: clSVMAlloc with pointer %p and size %u meets alignment restrictions.", //CL_SVM_ALLOC_MEETS_ALIGNMENT_RESTRICTIONS
"Performance hint: clEnqueueReadBuffer call on a buffer %p with pointer %p will require driver to copy the data.Consider using clEnqueueMapBuffer with buffer that shares the same physical memory with CPU.", //CL_ENQUEUE_READ_BUFFER_REQUIRES_COPY_DATA
"Performance hint: clEnqueueReadBuffer call on a buffer %p with pointer %p will not require any data copy as the buffer shares the same physical memory with CPU.", //CL_ENQUEUE_READ_BUFFER_DOESNT_REQUIRE_COPY_DATA
"Performance hint: Pointer %p and size %u passed to clEnqueueReadBuffer doesn't meet alignment restrictions. Size should be aligned to %u bytes and pointer should be aligned to %u. Driver needs to disable L3 caching.", //CL_ENQUEUE_READ_BUFFER_DOESNT_MEET_ALIGNMENT_RESTRICTIONS
"Performance hint: clEnqueueReadBufferRect call on a buffer %p with pointer %p will require driver to copy the data.Consider using clEnqueueMapBuffer with buffer that shares the same physical memory with CPU.", //CL_ENQUEUE_READ_BUFFER_RECT_REQUIRES_COPY_DATA
"Performance hint: clEnqueueReadBufferRect call on a buffer %p with pointer %p will not require any data copy as the buffer shares the same physical memory with CPU.", //CL_ENQUEUE_READ_BUFFER_RECT_DOESNT_REQUIRES_COPY_DATA
"Performance hint: Pointer %p and size %u passed to clEnqueueReadBufferRect doesn't meet alignment restrictions. Size should be aligned to %u bytes and pointer should be aligned to %u. Driver needs to disable L3 caching.", //CL_ENQUEUE_READ_BUFFER_RECT_DOESNT_MEET_ALIGNMENT_RESTRICTIONS
"Performance hint: clEnqueueWriteBuffer call on a buffer %p require driver to copy the data. Consider using clEnqueueMapBuffer with buffer that shares the same physical memory with CPU.", //CL_ENQUEUE_WRITE_BUFFER_REQUIRES_COPY_DATA
"Performance hint: clEnqueueWriteBuffer call on a buffer %p with pointer %p will not require any data copy as the buffer shares the same physical memory with CPU.", //CL_ENQUEUE_WRITE_BUFFER_DOESNT_REQUIRE_COPY_DATA
"Performance hint: clEnqueueWriteBufferRect call on a buffer %p require driver to copy the data. Consider using clEnqueueMapBuffer with buffer that shares the same physical memory with CPU.", //CL_ENQUEUE_WRITE_BUFFER_RECT_REQUIRES_COPY_DATA
"Performance hint: clEnqueueWriteBufferRect call on a buffer %p will not require any data copy as the buffer shares the same physical memory with CPU.", //CL_ENQUEUE_WRITE_BUFFER_RECT_DOESNT_REQUIRE_COPY_DATA
"Performance hint: Pointer %p and size %u passed to clEnqueueReadImage doesn't meet alignment restrictions. Size should be aligned to %u bytes and pointer should be aligned to %u. Driver needs to disable L3 caching.", //CL_ENQUEUE_READ_IMAGE_DOESNT_MEET_ALIGNMENT_RESTRICTIONS
"Performance hint: clEnqueueReadImage call on an image %p will not require any data copy as the image shares the same physical memory with CPU.", //CL_ENQUEUE_READ_IMAGE_DOESNT_REQUIRES_COPY_DATA
"Performance hint: clEnqueueWriteImage call on an image %p require driver to copy the data.", //CL_ENQUEUE_WRITE_IMAGE_REQUIRES_COPY_DATA
"Performance hint: clEnqueueWriteImage call on an image %p will not require any data copy as the image shares the same physical memory with CPU.", //CL_ENQUEUE_WRITE_IMAGE_DOESNT_REQUIRES_COPY_DATA
"Performance hint: clEnqueueMapBuffer call on a buffer %p will require driver to make a copy as buffer is not sharing the same physical memory with CPU.", //CL_ENQUEUE_MAP_BUFFER_REQUIRES_COPY_DATA
"Performance hint: clEnqueueMapBuffer call on a buffer %p will not require any data copy as buffer shares the same physical memory with CPU.", //CL_ENQUEUE_MAP_BUFFER_DOESNT_REQUIRE_COPY_DATA
"Performance hint: clEnqueueMapImage call on an image %p will require driver to make a copy, as image is not sharing the same physical memory with CPU.", //CL_ENQUEUE_MAP_IMAGE_REQUIRES_COPY_DATA
"Performance hint: clEnqueueMapImage call on an image %p will not require any data copy as image shares the same physical memory with CPU.", //CL_ENQUEUE_MAP_IMAGE_DOESNT_REQUIRE_COPY_DATA
"Performance hint: clEnqueueUnmapMemObject call with pointer %p will not require any data copy.", //CL_ENQUEUE_UNMAP_MEM_OBJ_DOESNT_REQUIRE_COPY_DATA
"Performance hint: clEnqueueUnmapMemObject call with pointer %p will require driver to copy the data to memory object %p.", //CL_ENQUEUE_UNMAP_MEM_OBJ_REQUIRES_COPY_DATA
"Performance hint: clEnqueueSVMMap call with pointer %p will not require any data copy.", //CL_ENQUEUE_SVM_MAP_DOESNT_REQUIRE_COPY_DATA
"Performance hint: Printf detected in kernel %s, it may cause overhead.", //PRINTF_DETECTED_IN_KERNEL
"Performance hint: Null local workgroup size detected ( kernel name: %s ); following sizes will be used for execution : { %u, %u, %u }.", //NULL_LOCAL_WORKGROUP_SIZE
"Performance hint: Local workgroup sizes { %u, %u, %u } selected for this workload ( kernel name: %s ) may not be optimal, consider using following local workgroup size: { %u, %u, %u }.", //BAD_LOCAL_WORKGROUP_SIZE
"Performance hint: Kernel %s register pressure is too high, spill fills will be generated, additional surface needs to be allocated of size %u, consider simplifying your kernel.", //REGISTER_PRESSURE_TOO_HIGH
"Performance hint: Kernel %s private memory usage is too high and exhausts register space, additional surface needs to be allocated of size %u, consider reducing amount of private memory used, avoid using private memory arrays.", //PRIVATE_MEMORY_USAGE_TOO_HIGH
"Performance hint: Kernel %s submission requires coherency with CPU; this will impact performance." //KERNEL_REQUIRES_COHERENCY
};
} // namespace OCLRT