This repository was archived by the owner on Apr 18, 2026. It is now read-only.

Add CUDA & OpenCL support#227

Open

crazyks wants to merge 190 commits intogoogle:masterfrom

ianhuang-777:opencl

Commits on Jun 2, 2017

Commits on Jun 3, 2017

调整cu编译
strongtu
committed
CUDA编译支持宏开关
strongtu
committed

Commits on Jun 4, 2017

Commits on Jun 5, 2017

完成CUDA并行优化，计算结果正常
strongtu
committed
修正命令行提示，Max Thread Per MP和SP是不一样的概念
strongtu
committed
调整参数试试性能情况
strongtu
committed

Commits on Jun 6, 2017

修正64、32位判断的宏
strongtu
committed
优化
strongtu
committed
恢复factor=2的支持，性能差别不大，但是编译时间变长了
strongtu
committed
优化编译和Test脚本
strongtu
committed
减少kernel中一些冗余的数据copy
strongtu
committed
Merge branch 'master' of https://github.com/ianhuang-777/guetzli
strongtu
committed
优化clDiffmapOpsinDynamicsImageEx
strongtu
committed
增加一些调试信息
strongtu
committed

Commits on Jun 7, 2017

kernel运算用float替代double，节省运算时间
strongtu
committed
修正数组长度
strongtu
committed
我也不知道为什么，删除掉这个空行计算结果就正确了
strongtu
committed
修正编译配置
strongtu
committed
修正warning
strongtu
committed
换一组编译参数
strongtu
committed
Merge branch 'googleMaster'
strongtu
committed

Commits on Jul 17, 2017

fix build

strongtu
authored and
kornelski
committed
add sample picture

strongtu
authored and
kornelski
committed
Merge branch 'master' of https://github.com/ianhuang-777/guetzli

strongtu
authored and
kornelski
committed
float is enough

strongtu
authored and
kornelski
committed
Add OpenCL Support

strongtu
authored and
kornelski
committed
MinSquareVal with OpenCL

strongtu
authored and
kornelski
committed
OpenCL 优化卷积

strongtu
authored and
kornelski
committed
fix setupopencl

strongtu
authored and
kornelski
committed
Merge branch 'master' of https://github.com/ianhuang-777/guetzli

strongtu
authored and
kornelski
committed
Add comment for understanding.

ianhuang-777
authored and
kornelski
committed
尝试看一下全OpenCL化Blur函数，不过目前计算误差有些大，是否有Bug？

strongtu
authored and
kornelski
committed
add opencl process line

strongtu
authored and
kornelski
committed
Merge branch 'master' of https://github.com/ianhuang-777/guetzli

strongtu
authored and
kornelski
committed
add function

strongtu
authored and
kornelski
committed
搭建 clDiffmapOpsinDynamicsImage 的计算流程

strongtu
authored and
kornelski
committed
Convert OpsinDynamicsImage to opencl

ianhuang-777
authored and
kornelski
committed
fix opencl compile error

strongtu
authored and
kornelski
committed
open cl compiler error fix

ianhuang-777
authored and
kornelski
committed
Implement clConvolutionEx

ianhuang-777
authored and
kornelski
committed
Remove useless code

ianhuang-777
authored and
kornelski
committed
Implement clUpsampleEx

ianhuang-777
authored and
kornelski
committed
Implement clMinSquareValEx

ianhuang-777
authored and
kornelski
committed
Implement clMaskEx

ianhuang-777
authored and
kornelski
committed
Implement clScaleImageEx

ianhuang-777
authored and
kornelski
committed
验证clOpinDynamicImage的效果

strongtu
authored and
kornelski
committed
尝试双精度运算支持

strongtu
authored and
kornelski
committed
Print More DeviceInfo

strongtu
authored and
kornelski
committed
Merge branch 'master' of https://github.com/ianhuang-777/guetzli

strongtu
authored and
kornelski
committed
Implement clMaskHighIntensityChangeEx

ianhuang-777
authored and
kornelski
committed
Implement clDiffPrecomputeEx

ianhuang-777
authored and
kornelski
committed
Implement clDiffPrecomputeEx

ianhuang-777
authored and
kornelski
committed
Merge branch 'master' of https://github.com/ianhuang-777/guetzli

strongtu
authored and
kornelski
committed
test for clDiffmapOpsinDynamicsImage

strongtu
authored and
kornelski
committed
fix runtime bug

strongtu
authored and
kornelski
committed
remove useless code

strongtu
authored and
kornelski
committed
添加测试用例

strongtu
authored and
kornelski
committed
增加测试用例

strongtu
authored and
kornelski
committed
修改测试用例框架

strongtu
authored and
kornelski
committed
测试用例分工

strongtu
authored and
kornelski
committed
MapBuffer之后要进行Unmap

strongtu
authored and
kornelski
committed
先排查>100*100的计算精度问题

strongtu
authored and
kornelski
committed
Remove _constant for opencl 1.2

ianhuang-777
authored and
kornelski
committed
Remove _constant for opencl 2.0

ianhuang-777
authored and
kornelski
committed
修复nVidia显卡的问题

strongtu
authored and
kornelski
committed
fixed n卡 __constant的问题

strongtu
authored and
kornelski
committed
Fix __constant error for nvidia device

ianhuang-777
authored and
kornelski
committed
Optimize clDoMask

ianhuang-777
authored and
kornelski
committed
32位平台编译配置

strongtu
authored and
kornelski
committed
Move some local constant array to __constant

ianhuang-777
authored and
kornelski
committed
for test

strongtu
authored and
kornelski
committed
测试卷积函数，节省一块中间缓存的使用

strongtu
authored and
kornelski
committed
Merge branch 'master' of https://github.com/ianhuang-777/guetzli

strongtu
authored and
kornelski
committed
修正blockDiffMap计算

strongtu
authored and
kornelski
committed
Merge remote-tracking branch 'origin/master'

crazyks
authored and
kornelski
committed
add clMinSquareVal test

crazyks
authored and
kornelski
committed
Merge branch 'master' of https://github.com/ianhuang-777/guetzli

strongtu
authored and
kornelski
committed
修正OpsinDynamicsImage运算结果

strongtu
authored and
kornelski
committed
remove redundant parameter

crazyks
authored and
kornelski
committed
Merge branch 'master' of https://github.com/ianhuang-777/guetzli

ianhuang-777
authored and
kornelski
committed
Add tclAverage55

ianhuang-777
authored and
kornelski
committed
修正计算结果+增加comparator子类

strongtu
authored and
kornelski
committed
fix-mapbuffer长度和需要的不符

strongtu
authored and
kornelski
committed
添加 clButteraugliComparator，避免对第三方库代码破坏太大

strongtu
authored and
kornelski
committed
规范kernel函数名以cl开头

strongtu
authored and
kornelski
committed
修正n卡上的编译问题

strongtu
authored and
kornelski
committed
Merge branch 'master' of https://github.com/ianhuang-777/guetzli

strongtu
authored and
kornelski
committed
增加 SelectFrequencyMaskingBatch 化处理

strongtu
authored and
kornelski
committed
建立cl端的批量化ComputeZeroingOrder都有哦

strongtu
authored and
kornelski
committed
调整分工

strongtu
authored and
kornelski
committed
分配工作

strongtu
authored and
kornelski
committed
clComputeBlockZeroingOrder

strongtu
authored and
kornelski
committed
修正n卡编译兼容问题

strongtu
authored and
kornelski
committed
Implement part of BlurEx

ianhuang-777
authored and
kornelski
committed
Merge remote-tracking branch 'origin/master'

crazyks
authored and
kornelski
committed
Fix BlurEx

ianhuang-777
authored and
kornelski
committed
fix data type of coeff_t

crazyks
authored and
kornelski
committed
modify MakeInputOrder

crazyks
authored and
kornelski
committed
Add BlockToImage

crazyks
authored and
kornelski
committed
Add MaskHighIntensityChangeBlock

crazyks
authored and
kornelski
committed
SelectFrequencyMaskingBatch 计算流程修正，终于可以正常跑起来了

strongtu
authored and
kornelski
committed
Merge branch 'master' of https://github.com/ianhuang-777/guetzli

strongtu
authored and
kornelski
committed
对于8x8的块，暂时不做check，否则速度太慢了

strongtu
authored and
kornelski
committed
Merge branch 'master' of https://github.com/ianhuang-777/guetzli

strongtu
authored and
kornelski
committed
增加访问接口，主要用于数据校验

strongtu
authored and
kornelski
committed
增加校验原图数据变化的代码，to be delete

strongtu
authored and
kornelski
committed
增加factor_x = factor_y = 2时的batch化原型

strongtu
authored and
kornelski
committed
翻译ComputeBlockEx2为OpenCL

strongtu
authored and
kornelski
committed
clComputeBlockZeroingOrderFactor调试

strongtu
authored and
kornelski
committed
精简代码

strongtu
authored and
kornelski
committed
factor 2支持完成

strongtu
authored and
kornelski
committed
合并类型声明，在opencl中include

strongtu
authored and
kornelski
committed
修正 clEnqueueUnmapMemObject 参数传递bug

strongtu
authored and
kornelski
committed
精简代码

strongtu
authored and
kornelski
committed
清理代码

strongtu
authored and
kornelski
committed
清理代码

strongtu
authored and
kornelski
committed
清理代码

strongtu
authored and
kornelski
committed
清理代码

strongtu
authored and
kornelski
committed
去掉编译事件

strongtu
authored and
kornelski
committed
精简代码

strongtu
authored and
kornelski
committed
精简代码

strongtu
authored and
kornelski
committed
const 控制

strongtu
authored and
kornelski
committed
Fix Average5x5

ianhuang-777
authored and
kornelski
committed
Inline ScaleIamge in kernel Average5x5

ianhuang-777
authored and
kornelski
committed
Avoid const value computing in work item

ianhuang-777
authored and
kornelski
committed
Fix tclCalculateDiffmap

ianhuang-777
authored and
kornelski
committed
Merge branch 'master' of https://github.com/ianhuang-777/guetzli

strongtu
authored and
kornelski
committed
const control

strongtu
authored and
kornelski
committed
精简代码

strongtu
authored and
kornelski
committed
调整参数顺序

strongtu
authored and
kornelski
committed
调整参数规则

strongtu
authored and
kornelski
committed
调整参数规范

strongtu
authored and
kornelski
committed
Merge branch 'master' of https://github.com/ianhuang-777/guetzli

strongtu
authored and
kornelski
committed
调整代码，修正参数传递规则

strongtu
authored and
kornelski
committed
Merge branch 'master' of https://github.com/ianhuang-777/guetzli

strongtu
authored and
kornelski
committed
增加CUDA编译，请小心更新，没安装cuda会无法编译

strongtu
authored and
kornelski
committed
support cuda opt

strongtu
authored and
kornelski
committed
运行期编译.cu

strongtu
authored and
kornelski
committed
兼容CUDA编译，编译器语法检查

strongtu
authored and
kornelski
committed
cuScaleImage跑通

strongtu
authored and
kornelski
committed
cuOpsinDynamicsImage 完成

strongtu
authored and
kornelski
committed
增加剩余的cu入口函数

strongtu
authored and
kornelski
committed
fix linux build

crazyks
authored and
kornelski
committed
merge google的改动之后，每次compuare StartBlockComparisons都会重新计算原始图片的opsin

strongtu
authored and
kornelski
committed
修复处理png时的crash

strongtu
authored and
kornelski
committed
节省clComputeBlockZeroingOrderEx过程中的冗余计算

strongtu
authored and
kornelski
committed
静态库编译

strongtu
authored and
kornelski
committed
编译参数

strongtu
authored and
kornelski
committed
调整测试脚本，支持目录批量优化

strongtu
authored and
kornelski
committed
Merge branch 'master' of https://github.com/ianhuang-777/guetzli

strongtu
authored and
kornelski
committed
c优化选项

strongtu
authored and
kornelski
committed
优化c代码

strongtu
authored and
kornelski
committed
不优化c了

strongtu
authored and
kornelski
committed
优化c版本

strongtu
authored and
kornelski
committed
Modify the flag for creating CUDA context

crazyks
authored and
kornelski
committed
Add macro for opencl version

crazyks
authored and
kornelski
committed
Add simple cuda memory pool

ianhuang-777
authored and
kornelski
committed
Add missing files

ianhuang-777
authored and
kornelski
committed
Clean code

ianhuang-777
authored and
kornelski
committed
Modify makefile

strongtu
authored and
kornelski
committed
默认开启CUDA OPENCL

strongtu
authored and
kornelski
committed
移除tcmalloc，对性能没什么影响

strongtu
authored and
kornelski
committed
Change memory block status to enum

ianhuang-777
authored and
kornelski
committed
Remove tcmalloc

ianhuang-777
authored and
kornelski
committed
支持非主流JPEG格式

zhantong
authored and
kornelski
committed
Mofidy makefile

strongtu
authored and
kornelski
committed
修复libjpeg库在debug和32位下编译不成功的问题

zhantong
authored and
kornelski
committed
Translate the comment.

ianhuang-777
authored and
kornelski
committed
Remove some redundant files

crazyks
authored and
kornelski
committed
Modify makefile

strongtu
authored and
kornelski
committed
Disable CUDA & OpenCL by default

crazyks
authored and
kornelski
committed
Add netpbm

crazyks
authored and
kornelski
committed
Fix type cast error on Mac

crazyks
authored and
kornelski
committed
Update bazel version to 0.5.2

crazyks
authored and
kornelski
committed
Add oracle-java8-installer

crazyks
authored and
kornelski
committed
Try to fix Bazel build

crazyks
authored and
kornelski
committed
Add author information

crazyks
authored and
kornelski
committed
Update ReadMe

crazyks
authored and
kornelski
committed

Commits on Jul 18, 2017

Update ReadMe & fix some mistakes
crazyks
committed

Commits on Jul 19, 2017

Update appveyor.xml
crazyks
committed