nihui
/
ncnn
Not watched
1
0
Fork 0
Code
Releases 46 Wiki evaluate Activity
Issues 0 Pull Requests 0 Datasets Model Cloudbrain HPC
Browse Source

cast layer and shader for fp32 fp16 conversion

tags/20190611
nihuini 7 years ago
parent
573c2bcd93
commit
2b8ff843e9
7 changed files with 676 additions and 0 deletions
Split View
Diff Options
Show Stats Download Patch File Download Diff File
  1. +5
    -0
      src/CMakeLists.txt
  2. +402
    -0
      src/layer/cast.cpp
  3. +57
    -0
      src/layer/cast.h
  4. +53
    -0
      src/layer/shader/cast_fp16_to_fp32.comp
  5. +53
    -0
      src/layer/shader/cast_fp16_to_fp32_pack4.comp
  6. +53
    -0
      src/layer/shader/cast_fp32_to_fp16.comp
  7. +53
    -0
      src/layer/shader/cast_fp32_to_fp16_pack4.comp

+ 5
- 0
src/CMakeLists.txt View File

@@ -105,6 +105,10 @@ macro(ncnn_add_layer class)

list(APPEND SHADER_SPV_HEX_FILES ${SHADER_SPV_HEX_FILE})

if("${name}" STREQUAL "cast")
continue()
endif()

# fp16 storage
set(SHADER_fp16s_SRC_NAME_WE "${SHADER_SRC_NAME_WE}_fp16s")

@@ -220,6 +224,7 @@ ncnn_add_layer(PSROIPooling)
ncnn_add_layer(ROIAlign OFF)
ncnn_add_layer(Packing)
ncnn_add_layer(Requantize)
ncnn_add_layer(Cast)

add_custom_target(generate-spirv DEPENDS ${SHADER_SPV_HEX_FILES})



+ 402
- 0
src/layer/cast.cpp View File

@@ -0,0 +1,402 @@
// Tencent is pleased to support the open source community by making ncnn available.
//
// Copyright (C) 2019 THL A29 Limited, a Tencent company. All rights reserved.
//
// Licensed under the BSD 3-Clause License (the "License"); you may not use this file except
// in compliance with the License. You may obtain a copy of the License at
//
// https://opensource.org/licenses/BSD-3-Clause
//
// Unless required by applicable law or agreed to in writing, software distributed
// under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR
// CONDITIONS OF ANY KIND, either express or implied. See the License for the
// specific language governing permissions and limitations under the License.

#include "cast.h"

namespace ncnn {

DEFINE_LAYER_CREATOR(Cast)

Cast::Cast()
{
one_blob_only = true;
support_inplace = false;
support_vulkan = true;

#if NCNN_VULKAN
pipeline_cast_fp32_to_fp16 = 0;
pipeline_cast_fp32_to_fp16_pack4 = 0;
pipeline_cast_fp16_to_fp32 = 0;
pipeline_cast_fp16_to_fp32_pack4 = 0;
#endif // NCNN_VULKAN
}

int Cast::load_param(const ParamDict& pd)
{
type_from = pd.get(0, 0);
type_to = pd.get(1, 0);

return 0;
}

// convert float to half precision floating point
static unsigned short float32_to_float16(float value)
{
// 1 : 8 : 23
union
{
unsigned int u;
float f;
} tmp;

tmp.f = value;

// 1 : 8 : 23
unsigned short sign = (tmp.u & 0x80000000) >> 31;
unsigned short exponent = (tmp.u & 0x7F800000) >> 23;
unsigned int significand = tmp.u & 0x7FFFFF;

// fprintf(stderr, "%d %d %d\n", sign, exponent, significand);

// 1 : 5 : 10
unsigned short fp16;
if (exponent == 0)
{
// zero or denormal, always underflow
fp16 = (sign << 15) | (0x00 << 10) | 0x00;
}
else if (exponent == 0xFF)
{
// infinity or NaN
fp16 = (sign << 15) | (0x1F << 10) | (significand ? 0x200 : 0x00);
}
else
{
// normalized
short newexp = exponent + (- 127 + 15);
if (newexp >= 31)
{
// overflow, return infinity
fp16 = (sign << 15) | (0x1F << 10) | 0x00;
}
else if (newexp <= 0)
{
// underflow
if (newexp >= -10)
{
// denormal half-precision
unsigned short sig = (significand | 0x800000) >> (14 - newexp);
fp16 = (sign << 15) | (0x00 << 10) | sig;
}
else
{
// underflow
fp16 = (sign << 15) | (0x00 << 10) | 0x00;
}
}
else
{
fp16 = (sign << 15) | (newexp << 10) | (significand >> 13);
}
}

return fp16;
}

// convert half precision floating point to float
static float float16_to_float32(unsigned short value)
{
// 1 : 5 : 10
unsigned short sign = (value & 0x8000) >> 15;
unsigned short exponent = (value & 0x7c00) >> 10;
unsigned short significand = value & 0x03FF;

// fprintf(stderr, "%d %d %d\n", sign, exponent, significand);

// 1 : 8 : 23
union
{
unsigned int u;
float f;
} tmp;
if (exponent == 0)
{
if (significand == 0)
{
// zero
tmp.u = (sign << 31);
}
else
{
// denormal
exponent = 0;
// find non-zero bit
while ((significand & 0x200) == 0)
{
significand <<= 1;
exponent++;
}
significand <<= 1;
significand &= 0x3FF;
tmp.u = (sign << 31) | ((-exponent + (-15 + 127)) << 23) | (significand << 13);
}
}
else if (exponent == 0x1F)
{
// infinity or NaN
tmp.u = (sign << 31) | (0xFF << 23) | (significand << 13);
}
else
{
// normalized
tmp.u = (sign << 31) | ((exponent + (-15 + 127)) << 23) | (significand << 13);
}

return tmp.f;
}

// round to nearest
static signed char float32_to_int8(float value)
{
float tmp;
if (value >= 0.f) tmp = value + 0.5;
else tmp = value - 0.5;

if (tmp > 127)
return 127;
if (tmp < -128)
return -128;

return tmp;
}

int Cast::forward(const Mat& bottom_blob, Mat& top_blob, const Option& opt) const
{
if (type_from == type_to)
{
top_blob = bottom_blob;
return 0;
}

int w = bottom_blob.w;
int h = bottom_blob.h;
int channels = bottom_blob.c;
int dims = bottom_blob.dims;
size_t elemsize = bottom_blob.elemsize;
int packing = bottom_blob.packing;

size_t out_elemsize = elemsize;
if (type_to == 1)
{
// float32
out_elemsize = 4 * packing;
}
else if (type_to == 2)
{
// float16
out_elemsize = 2 * packing;
}
else if (type_to == 3)
{
// int8
out_elemsize = packing;
}

if (dims == 1)
{
top_blob.create(w, out_elemsize, packing, opt.blob_allocator);
}
else if (dims == 2)
{
top_blob.create(w, h, out_elemsize, packing, opt.blob_allocator);
}
else if (dims == 3)
{
top_blob.create(w, h, channels, out_elemsize, packing, opt.blob_allocator);
}
if (top_blob.empty())
return -100;

int size = w * h * packing;

if (type_from == 1 && type_to == 2)
{
#pragma omp parallel for num_threads(opt.num_threads)
for (int q=0; q<channels; q++)
{
const float* ptr = bottom_blob.channel(q);
unsigned short* outptr = top_blob.channel(q);

for (int i=0; i<size; i++)
{
outptr[i] = float32_to_float16(ptr[i]);
}
}
}

if (type_from == 2 && type_to == 1)
{
#pragma omp parallel for num_threads(opt.num_threads)
for (int q=0; q<channels; q++)
{
const unsigned short* ptr = bottom_blob.channel(q);
float* outptr = top_blob.channel(q);

for (int i=0; i<size; i++)
{
outptr[i] = float16_to_float32(ptr[i]);
}
}
}

// TODO more cast type

return 0;
}

#if NCNN_VULKAN
int Cast::create_pipeline()
{
std::vector<vk_specialization_type> specializations;

if (type_from == 1 && type_to == 2)
{
// pack1
{
pipeline_cast_fp32_to_fp16 = new Pipeline(vkdev);
pipeline_cast_fp32_to_fp16->set_optimal_local_size_xyz();
pipeline_cast_fp32_to_fp16->create("cast_fp32_to_fp16", specializations, 2, 10);
}

// pack4
{
pipeline_cast_fp32_to_fp16_pack4 = new Pipeline(vkdev);
pipeline_cast_fp32_to_fp16_pack4->set_optimal_local_size_xyz();
pipeline_cast_fp32_to_fp16_pack4->create("cast_fp32_to_fp16_pack4", specializations, 2, 10);
}
}

if (type_from == 2 && type_to == 1)
{
// pack1
{
pipeline_cast_fp16_to_fp32 = new Pipeline(vkdev);
pipeline_cast_fp16_to_fp32->set_optimal_local_size_xyz();
pipeline_cast_fp16_to_fp32->create("cast_fp16_to_fp32", specializations, 2, 10);
}

// pack4
{
pipeline_cast_fp16_to_fp32_pack4 = new Pipeline(vkdev);
pipeline_cast_fp16_to_fp32_pack4->set_optimal_local_size_xyz();
pipeline_cast_fp16_to_fp32_pack4->create("cast_fp16_to_fp32_pack4", specializations, 2, 10);
}
}

return 0;
}

int Cast::destroy_pipeline()
{
delete pipeline_cast_fp32_to_fp16;
pipeline_cast_fp32_to_fp16 = 0;

delete pipeline_cast_fp32_to_fp16_pack4;
pipeline_cast_fp32_to_fp16_pack4 = 0;

delete pipeline_cast_fp16_to_fp32;
pipeline_cast_fp16_to_fp32 = 0;

delete pipeline_cast_fp16_to_fp32_pack4;
pipeline_cast_fp16_to_fp32_pack4 = 0;

return 0;
}

int Cast::forward(const VkMat& bottom_blob, VkMat& top_blob, VkCompute& cmd, const Option& opt) const
{
if (type_from == type_to)
{
top_blob = bottom_blob;
return 0;
}

int w = bottom_blob.w;
int h = bottom_blob.h;
int channels = bottom_blob.c;
int dims = bottom_blob.dims;
size_t elemsize = bottom_blob.elemsize;
int packing = bottom_blob.packing;

size_t out_elemsize = elemsize;
if (type_to == 1)
{
// float32
out_elemsize = 4 * packing;
}
else if (type_to == 2)
{
// float16
out_elemsize = 2 * packing;
}
else if (type_to == 3)
{
// int8
out_elemsize = packing;
}

if (dims == 1)
{
top_blob.create(w, out_elemsize, packing, opt.blob_vkallocator, opt.staging_vkallocator);
}
else if (dims == 2)
{
top_blob.create(w, h, out_elemsize, packing, opt.blob_vkallocator, opt.staging_vkallocator);
}
else if (dims == 3)
{
top_blob.create(w, h, channels, out_elemsize, packing, opt.blob_vkallocator, opt.staging_vkallocator);
}
if (top_blob.empty())
return -100;

// fprintf(stderr, "Cast::forward %p %p\n", bottom_blob.buffer(), top_blob.buffer());

std::vector<VkMat> bindings(2);
bindings[0] = bottom_blob;
bindings[1] = top_blob;

std::vector<vk_constant_type> constants(10);
constants[0].i = bottom_blob.dims;
constants[1].i = bottom_blob.w;
constants[2].i = bottom_blob.h;
constants[3].i = bottom_blob.c;
constants[4].i = bottom_blob.cstep;
constants[5].i = top_blob.dims;
constants[6].i = top_blob.w;
constants[7].i = top_blob.h;
constants[8].i = top_blob.c;
constants[9].i = top_blob.cstep;

const Pipeline* pipeline = 0;

if (type_from == 1 && type_to == 2)
{
pipeline = packing == 4 ? pipeline_cast_fp32_to_fp16_pack4 : pipeline_cast_fp32_to_fp16;
}
if (type_from == 2 && type_to == 1)
{
pipeline = packing == 4 ? pipeline_cast_fp16_to_fp32_pack4 : pipeline_cast_fp16_to_fp32;
}

// TODO more cast type

// record
cmd.record_pipeline(pipeline, bindings, constants, top_blob);

return 0;
}
#endif // NCNN_VULKAN

} // namespace ncnn

+ 57
- 0
src/layer/cast.h View File

@@ -0,0 +1,57 @@
// Tencent is pleased to support the open source community by making ncnn available.
//
// Copyright (C) 2019 THL A29 Limited, a Tencent company. All rights reserved.
//
// Licensed under the BSD 3-Clause License (the "License"); you may not use this file except
// in compliance with the License. You may obtain a copy of the License at
//
// https://opensource.org/licenses/BSD-3-Clause
//
// Unless required by applicable law or agreed to in writing, software distributed
// under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR
// CONDITIONS OF ANY KIND, either express or implied. See the License for the
// specific language governing permissions and limitations under the License.

#ifndef LAYER_CAST_H
#define LAYER_CAST_H

#include "layer.h"

namespace ncnn {

class Cast : public Layer
{
public:
Cast();

virtual int load_param(const ParamDict& pd);

virtual int forward(const Mat& bottom_blob, Mat& top_blob, const Option& opt) const;

#if NCNN_VULKAN
virtual int create_pipeline();
virtual int destroy_pipeline();

virtual int forward(const VkMat& bottom_blob, VkMat& top_blob, VkCompute& cmd, const Option& opt) const;
#endif // NCNN_VULKAN

public:
// element type
// 0 = auto
// 1 = float32
// 2 = float16
// 3 = int8
int type_from;
int type_to;

#if NCNN_VULKAN
Pipeline* pipeline_cast_fp32_to_fp16;
Pipeline* pipeline_cast_fp32_to_fp16_pack4;
Pipeline* pipeline_cast_fp16_to_fp32;
Pipeline* pipeline_cast_fp16_to_fp32_pack4;
#endif // NCNN_VULKAN
};

} // namespace ncnn

#endif // LAYER_CAST_H

+ 53
- 0
src/layer/shader/cast_fp16_to_fp32.comp View File

@@ -0,0 +1,53 @@
// Tencent is pleased to support the open source community by making ncnn available.
//
// Copyright (C) 2019 THL A29 Limited, a Tencent company. All rights reserved.
//
// Licensed under the BSD 3-Clause License (the "License"); you may not use this file except
// in compliance with the License. You may obtain a copy of the License at
//
// https://opensource.org/licenses/BSD-3-Clause
//
// Unless required by applicable law or agreed to in writing, software distributed
// under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR
// CONDITIONS OF ANY KIND, either express or implied. See the License for the
// specific language governing permissions and limitations under the License.

#version 450

#extension GL_AMD_gpu_shader_half_float: require

layout (local_size_x_id = 233) in;
layout (local_size_y_id = 234) in;
layout (local_size_z_id = 235) in;

layout (binding = 0) readonly buffer bottom_blob { float16_t bottom_blob_data[]; };
layout (binding = 1) writeonly buffer top_blob { float top_blob_data[]; };

layout (push_constant) uniform parameter
{
int dims;
int w;
int h;
int c;
int cstep;

int outdims;
int outw;
int outh;
int outc;
int outcstep;
} p;

void main()
{
int gx = int(gl_GlobalInvocationID.x);
int gy = int(gl_GlobalInvocationID.y);
int gz = int(gl_GlobalInvocationID.z);

if (gx >= p.w || gy >= p.h || gz >= p.c)
return;

float16_t v = bottom_blob_data[gz * p.cstep + gy * p.w + gx];

top_blob_data[gz * p.outcstep + gy * p.outw + gx] = float(v);
}

+ 53
- 0
src/layer/shader/cast_fp16_to_fp32_pack4.comp View File

@@ -0,0 +1,53 @@
// Tencent is pleased to support the open source community by making ncnn available.
//
// Copyright (C) 2019 THL A29 Limited, a Tencent company. All rights reserved.
//
// Licensed under the BSD 3-Clause License (the "License"); you may not use this file except
// in compliance with the License. You may obtain a copy of the License at
//
// https://opensource.org/licenses/BSD-3-Clause
//
// Unless required by applicable law or agreed to in writing, software distributed
// under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR
// CONDITIONS OF ANY KIND, either express or implied. See the License for the
// specific language governing permissions and limitations under the License.

#version 450

#extension GL_AMD_gpu_shader_half_float: require

layout (local_size_x_id = 233) in;
layout (local_size_y_id = 234) in;
layout (local_size_z_id = 235) in;

layout (binding = 0) readonly buffer bottom_blob { f16vec4 bottom_blob_data[]; };
layout (binding = 1) writeonly buffer top_blob { vec4 top_blob_data[]; };

layout (push_constant) uniform parameter
{
int dims;
int w;
int h;
int c;
int cstep;

int outdims;
int outw;
int outh;
int outc;
int outcstep;
} p;

void main()
{
int gx = int(gl_GlobalInvocationID.x);
int gy = int(gl_GlobalInvocationID.y);
int gz = int(gl_GlobalInvocationID.z);

if (gx >= p.w || gy >= p.h || gz >= p.c)
return;

f16vec4 v = bottom_blob_data[gz * p.cstep + gy * p.w + gx];

top_blob_data[gz * p.outcstep + gy * p.outw + gx] = vec4(v);
}

+ 53
- 0
src/layer/shader/cast_fp32_to_fp16.comp View File

@@ -0,0 +1,53 @@
// Tencent is pleased to support the open source community by making ncnn available.
//
// Copyright (C) 2019 THL A29 Limited, a Tencent company. All rights reserved.
//
// Licensed under the BSD 3-Clause License (the "License"); you may not use this file except
// in compliance with the License. You may obtain a copy of the License at
//
// https://opensource.org/licenses/BSD-3-Clause
//
// Unless required by applicable law or agreed to in writing, software distributed
// under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR
// CONDITIONS OF ANY KIND, either express or implied. See the License for the
// specific language governing permissions and limitations under the License.

#version 450

#extension GL_AMD_gpu_shader_half_float: require

layout (local_size_x_id = 233) in;
layout (local_size_y_id = 234) in;
layout (local_size_z_id = 235) in;

layout (binding = 0) readonly buffer bottom_blob { float bottom_blob_data[]; };
layout (binding = 1) writeonly buffer top_blob { float16_t top_blob_data[]; };

layout (push_constant) uniform parameter
{
int dims;
int w;
int h;
int c;
int cstep;

int outdims;
int outw;
int outh;
int outc;
int outcstep;
} p;

void main()
{
int gx = int(gl_GlobalInvocationID.x);
int gy = int(gl_GlobalInvocationID.y);
int gz = int(gl_GlobalInvocationID.z);

if (gx >= p.w || gy >= p.h || gz >= p.c)
return;

float v = bottom_blob_data[gz * p.cstep + gy * p.w + gx];

top_blob_data[gz * p.outcstep + gy * p.outw + gx] = float16_t(v);
}

+ 53
- 0
src/layer/shader/cast_fp32_to_fp16_pack4.comp View File

@@ -0,0 +1,53 @@
// Tencent is pleased to support the open source community by making ncnn available.
//
// Copyright (C) 2019 THL A29 Limited, a Tencent company. All rights reserved.
//
// Licensed under the BSD 3-Clause License (the "License"); you may not use this file except
// in compliance with the License. You may obtain a copy of the License at
//
// https://opensource.org/licenses/BSD-3-Clause
//
// Unless required by applicable law or agreed to in writing, software distributed
// under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR
// CONDITIONS OF ANY KIND, either express or implied. See the License for the
// specific language governing permissions and limitations under the License.

#version 450

#extension GL_AMD_gpu_shader_half_float: require

layout (local_size_x_id = 233) in;
layout (local_size_y_id = 234) in;
layout (local_size_z_id = 235) in;

layout (binding = 0) readonly buffer bottom_blob { vec4 bottom_blob_data[]; };
layout (binding = 1) writeonly buffer top_blob { f16vec4 top_blob_data[]; };

layout (push_constant) uniform parameter
{
int dims;
int w;
int h;
int c;
int cstep;

int outdims;
int outw;
int outh;
int outc;
int outcstep;
} p;

void main()
{
int gx = int(gl_GlobalInvocationID.x);
int gy = int(gl_GlobalInvocationID.y);
int gz = int(gl_GlobalInvocationID.z);

if (gx >= p.w || gy >= p.h || gz >= p.c)
return;

vec4 v = bottom_blob_data[gz * p.cstep + gy * p.w + gx];

top_blob_data[gz * p.outcstep + gy * p.outw + gx] = f16vec4(v);
}

Write Preview
Loading…
Cancel
Save