ShuriZma
/
dolphin
mirror of https://github.com/dolphin-emu/dolphin.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

OpenCL: More windows support and some unfinished work 

git-svn-id: https://dolphin-emu.googlecode.com/svn/trunk@4380 8ced0084-cf51-0410-be5f-012b33b47a6e
This commit is contained in:
XTra.KrazzY 2009-10-07 21:30:36 +00:00
parent 5523f28b51
commit 98db1a0ca3
7 changed files with 132 additions and 119 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

34
Source/Core/Common/Src/OpenCL.cpp
View File

@ -27,8 +27,12 @@ namespace OpenCL {
cl_command_queue g_cmdq = NULL;
#endif
bool g_bInitialized = false;
bool Initialize() {
if(g_bInitialized)
return true;
#if defined(HAVE_OPENCL) && HAVE_OPENCL
if(g_context)
return false;
@ -41,7 +45,7 @@ bool Initialize() {
err = clGetDeviceIDs(NULL, gpu ? CL_DEVICE_TYPE_GPU : CL_DEVICE_TYPE_CPU, 1, &device_id, NULL);
if (err != CL_SUCCESS)
{
PanicAlert("Error: Failed to create a device group!\n");
PanicAlert("Error: Failed to create a device group!");
return false;
}
@ -50,7 +54,7 @@ bool Initialize() {
g_context = clCreateContext(0, 1, &device_id, NULL, NULL, &err);
if (!g_context)
{
PanicAlert("Error: Failed to create a compute context!\n");
PanicAlert("Error: Failed to create a compute context!");
return false;
}
@ -59,17 +63,18 @@ bool Initialize() {
g_cmdq = clCreateCommandQueue(g_context, device_id, 0, &err);
if (!g_cmdq)
{
PanicAlert("Error: Failed to create a command commands!\n");
PanicAlert("Error: Failed to create a command commands!");
return false;
}
printf("Initialized OpenCL fine!\n");
//PanicAlert("Initialized OpenCL fine!");
g_bInitialized = true;
return true;
#else
return false;
#endif
}
#if defined(HAVE_OPENCL) && HAVE_OPENCL
cl_context GetInstance() {
cl_context GetContext() {
return g_context;
}
@ -83,21 +88,18 @@ cl_program CompileProgram(const char *Kernel) {
program = clCreateProgramWithSource(OpenCL::g_context, 1, (const char **) & Kernel, NULL, &err);
if (!program)
{
printf("Error: Failed to create compute program!\n");
printf("Error: Failed to create compute program!");
return NULL;
}
// Build the program executable
//
err = clBuildProgram(program , 0, NULL, NULL, NULL, NULL);
if (err != CL_SUCCESS)
{
size_t len;
char buffer[2048];
printf("Error: Failed to build program executable!\n");
clGetProgramBuildInfo(program , OpenCL::device_id, CL_PROGRAM_BUILD_LOG, sizeof(buffer), buffer, &len);
printf("%s\n", buffer);
if(err != CL_SUCCESS) {
char *errors[16384] = {0};
err = clGetProgramBuildInfo(program, OpenCL::device_id, CL_PROGRAM_BUILD_LOG, sizeof(errors),
errors, NULL);
PanicAlert("Error log:\n%s\n", errors);
return NULL;
}
@ -112,7 +114,7 @@ cl_kernel CompileKernel(cl_program program, const char *Function)
cl_kernel kernel = clCreateKernel(program, Function, &err);
if (!kernel || err != CL_SUCCESS)
{
printf("Error: Failed to create compute kernel!\n");
PanicAlert("Error: Failed to create compute kernel!");
return NULL;
}
return kernel;

16
Source/Core/Core/Src/HW/SystemTimers.cpp
View File

@ -238,11 +238,11 @@ void Init()
CPU_CORE_CLOCK = 729000000u;
SI_PERIOD = GetTicksPerSecond() / 60; // once a frame is good for controllers
if (!Core::GetStartupParameter().bDSPThread) {
DSP_PERIOD = 12000; // TO BE TWEAKED
} else {
//if (!Core::GetStartupParameter().bDSPThread) {
// DSP_PERIOD = 12000; // TO BE TWEAKED
//} else {
DSP_PERIOD = (int)(GetTicksPerSecond() * 0.003f);
}
//}
// This is the biggest question mark.
AI_PERIOD = GetTicksPerSecond() / 80;
@ -254,11 +254,11 @@ void Init()
CPU_CORE_CLOCK = 486000000;
SI_PERIOD = GetTicksPerSecond() / 60; // once a frame is good for controllers
if (!Core::GetStartupParameter().bDSPThread) {
DSP_PERIOD = 12000; // TO BE TWEAKED
} else {
//if (!Core::GetStartupParameter().bDSPThread) {
// DSP_PERIOD = 12000; // TO BE TWEAKED
//} else {
DSP_PERIOD = (int)(GetTicksPerSecond() * 0.005f);
}
//}
// This is the biggest question mark.
AI_PERIOD = GetTicksPerSecond() / 80;

191
Source/Core/VideoCommon/Src/OpenCL/TextureDecoder.cpp → Source/Core/VideoCommon/Src/OpenCL/OCLTextureDecoder.cpp
View File

@ -15,127 +15,136 @@
// Official SVN repository and contact information can be found at
// http://code.google.com/p/dolphin-emu/
#include "TextureDecoder.h"
#include "OCLTextureDecoder.h"
#include "OpenCL.h"
#include <CL/cl.h>
#include <fcntl.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <math.h>
#include <unistd.h>
#include <sys/types.h>
#include <sys/stat.h>
struct sDecoders
{
cl_program program; // compute program
cl_kernel kernel; // compute kernel
const char **cKernel;
cl_mem src, dst; // texture buffer memory objects
};
const char *Kernel = " \n \
__kernel void Decode(__global unsigned char *dst, \n \
const __global unsigned char *src, \n \
const __global int width, const __global int height) \n \
{ \n \
int x = get_global_id(0) % width, y = get_global_id(0) / width; \n \
if((y % 4) == 0 && (x % 8) == 0) \n \
{ \n \
int srcOffset = (x * 4) + (y * width); \n \
// for (int y = 0; y < height; y += 4) \n \
// for (int x = 0; x < width; x += 8) \n \
for (int iy = 0; iy < 4; iy++, srcOffset += 8) \n\
{ \n \
dst[(y + iy)*width + x] = src[srcOffset]; \n \
dst[(y + iy)*width + x + 1] = src[srcOffset + 1]; \n \
dst[(y + iy)*width + x + 2] = src[srcOffset + 2]; \n \
dst[(y + iy)*width + x + 3] = src[srcOffset + 3]; \n \
dst[(y + iy)*width + x + 4] = src[srcOffset + 4]; \n \
dst[(y + iy)*width + x + 5] = src[srcOffset + 5]; \n \
dst[(y + iy)*width + x + 6] = src[srcOffset + 6]; \n \
dst[(y + iy)*width + x + 7] = src[srcOffset + 7]; \n \
} \n \
} \n \
// XK's neatly aligned kernel
const char *XKernel = " \n\
kernel void DecodeI8(global uchar *dst, \n\
const global uchar *src, int width) \n\
{ \n\
int x = get_global_id(0) * 8, y = get_global_id(1) * 4; \n\
int srcOffset = 0; \n\
for (int iy = 0; iy < 4; iy++) \n\
{ \n\
dst[(y + iy)*width + x] = src[srcOffset]; \n\
dst[(y + iy)*width + x + 1] = src[srcOffset + 1]; \n\
dst[(y + iy)*width + x + 2] = src[srcOffset + 2]; \n\
dst[(y + iy)*width + x + 3] = src[srcOffset + 3]; \n\
dst[(y + iy)*width + x + 4] = src[srcOffset + 4]; \n\
dst[(y + iy)*width + x + 5] = src[srcOffset + 5]; \n\
dst[(y + iy)*width + x + 6] = src[srcOffset + 6]; \n\
dst[(y + iy)*width + x + 7] = src[srcOffset + 7]; \n\
srcOffset += 8; \n\
} \n\
}\n";
// memcpy(dst + (y + iy)*width+x, src, 8);
const char *KernelOld = " \n \
__kernel void Decode(__global uchar *dst, \n \
const __global uchar *src, \n \
int width, int height) \n \
{ \n \
dst[get_global_id(0)] = 0xFF; \n \
} \n ";
sDecoders Decoders[] = { {NULL, NULL, &Kernel}, };
const char *Kernel = " \n \
__kernel void DecodeI8(__global unsigned char *dst, \n \
const __global unsigned char *src, \n \
const __global int width) \n \
{ \n \
int x = get_global_id(0) % width, y = get_global_id(0) / width; \n \
if((y % 4) == 0 && (x % 8) == 0) \n \
{ \n \
int srcOffset = (x * 4) + (y * width); \n \
// for (int y = 0; y < height; y += 4) \n \
// for (int x = 0; x < width; x += 8) \n \
for (int iy = 0; iy < 4; iy++, srcOffset += 8) \n\
{ \n \
dst[(y + iy)*width + x] = src[srcOffset]; \n \
dst[(y + iy)*width + x + 1] = src[srcOffset + 1]; \n \
dst[(y + iy)*width + x + 2] = src[srcOffset + 2]; \n \
dst[(y + iy)*width + x + 3] = src[srcOffset + 3]; \n \
dst[(y + iy)*width + x + 4] = src[srcOffset + 4]; \n \
dst[(y + iy)*width + x + 5] = src[srcOffset + 5]; \n \
dst[(y + iy)*width + x + 6] = src[srcOffset + 6]; \n \
dst[(y + iy)*width + x + 7] = src[srcOffset + 7]; \n \
} \n \
} \n \
}\n";
sDecoders Decoders[] = { {NULL, NULL, NULL, NULL}, };
bool g_Inited = false;
PC_TexFormat TexDecoder_Decode_OpenCL(u8 *dst, const u8 *src, int width, int height, int texformat, int tlutaddr, int tlutfmt)
{
int err;
#if defined(HAVE_OPENCL) && HAVE_OPENCL
cl_int err;
if(!g_Inited)
{
g_Inited = true;
if(!OpenCL::Initialize())
return PC_TEX_FMT_NONE;
#if defined(HAVE_OPENCL) && HAVE_OPENCL
// TODO: Switch this over to the OpenCl.h backend
// Create the compute program from the source buffer
//
Decoders[0].program = clCreateProgramWithSource(OpenCL::g_context, 1, (const char **) & Kernel, NULL, &err);
if (!Decoders[0].program)
{
printf("Error: Failed to create compute program!\n");
return PC_TEX_FMT_NONE;
}
Decoders[0].program = OpenCL::CompileProgram(Kernel);
Decoders[0].kernel = OpenCL::CompileKernel(Decoders[0].program, "DecodeI8");
// Build the program executable
//
err = clBuildProgram(Decoders[0].program , 0, NULL, NULL, NULL, NULL);
if (err != CL_SUCCESS)
{
size_t len;
char buffer[2048];
printf("Error: Failed to build program executable!\n");
clGetProgramBuildInfo(Decoders[0].program , OpenCL::device_id, CL_PROGRAM_BUILD_LOG, sizeof(buffer), buffer, &len);
printf("%s\n", buffer);
exit(1);
}
// Create the compute kernel in the program we wish to run
//
Decoders[0].kernel = clCreateKernel(Decoders[0].program, "Decode", &err);
if (!Decoders[0].kernel || err != CL_SUCCESS)
{
printf("Error: Failed to create compute kernel!\n");
exit(1);
}
#endif
g_Inited = true;
}
/*switch(texformat)
{
case GX_TF_I8:
switch(texformat)
{
case GX_TF_I8:
{
int srcOffset = 0;
for (int y = 0; y < height; y += 4)
for (int x = 0; x < width; x += 8)
for (int iy = 0; iy < 4; iy++, srcOffset += 8)
{
printf("x: %d y: %d offset: %d\n", x, y, srcOffset);
memcpy(dst + (y + iy)*width+x, src + srcOffset, 8);
}
return PC_TEX_FMT_I8;
// TODO: Optimize
//PanicAlert("Really calling the OCL version");
Decoders[0].src = clCreateBuffer(OpenCL::GetContext(), CL_MEM_READ_ONLY | CL_MEM_COPY_HOST_PTR, width * height * sizeof(u8), (void *)src, NULL);
Decoders[0].dst = clCreateBuffer(OpenCL::GetContext(), CL_MEM_WRITE_ONLY, width * height * sizeof(u8), NULL, NULL);
clSetKernelArg(Decoders[0].kernel, 0, sizeof(cl_mem), &Decoders[0].dst);
clSetKernelArg(Decoders[0].kernel, 1, sizeof(cl_mem), &Decoders[0].src);
clSetKernelArg(Decoders[0].kernel, 2, sizeof(cl_int), &width);
//size_t global[] = { width, height, 0 }, local[3] = {0}; // Later on we'll get the max work-group size and actually use them
size_t global = width * height, local;
clGetKernelWorkGroupInfo(Decoders[0].kernel, OpenCL::device_id, CL_KERNEL_WORK_GROUP_SIZE, sizeof(local), &local, NULL);
err = clEnqueueNDRangeKernel(OpenCL::GetCommandQueue(), Decoders[0].kernel, 1 /* 2 */, NULL, &global, &local, 0, NULL, NULL);
if(err)
PanicAlert("Error queueing kernel");
clFinish(OpenCL::GetCommandQueue());
clEnqueueReadBuffer(OpenCL::GetCommandQueue(), Decoders[0].dst, CL_TRUE, 0, width * height * sizeof(u8), dst, 0, NULL, NULL);
clReleaseMemObject(Decoders[0].src);
clReleaseMemObject(Decoders[0].dst);
/*
for (int y = 0; y < height; y += 4)
for (int x = 0; x < width; x += 8)
for (int iy = 0; iy < 4; iy++, src += 8)
memcpy(dst + (y + iy)*width+x, src, 8);
*/
return PC_TEX_FMT_I8;
}
break;
default:
default:
return PC_TEX_FMT_NONE;
}
//return PC_TEX_FMT_NONE;*/
}
#else
return PC_TEX_FMT_NONE;
#endif
/* OLD CODE
switch(texformat)
{
case GX_TF_I8:
@ -211,9 +220,9 @@ PC_TexFormat TexDecoder_Decode_OpenCL(u8 *dst, const u8 *src, int width, int hei
return PC_TEX_FMT_NONE;
}
// TODO: clEnqueueNDRangeKernel
*/
/*switch (texformat)
/* switch (texformat)
{
case GX_TF_C4:
if (tlutfmt == 2)

0
Source/Core/VideoCommon/Src/OpenCL/TextureDecoder.h → Source/Core/VideoCommon/Src/OpenCL/OCLTextureDecoder.h
View File

2
Source/Core/VideoCommon/Src/SConscript
View File

@ -41,7 +41,7 @@ env_vcommon = env.Clone()
if env_vcommon['HAVE_OPENCL']:
files += [
'OpenCL/TextureDecoder.cpp',
'OpenCL/OCLTextureDecoder.cpp',
]
env_vcommon.Append(CXXFLAGS = [ '-fPIC' ])

4
Source/Core/VideoCommon/Src/TextureDecoder.cpp
View File

@ -21,8 +21,10 @@
#include "CPUDetect.h"
#include "TextureDecoder.h"
#include "OpenCL.h"
#if defined(HAVE_OPENCL) && HAVE_OPENCL
#include "OpenCL/TextureDecoder.h"
#include "OpenCL/OCLTextureDecoder.h"
#endif
#include "LookUpTables.h"

4
Source/Core/VideoCommon/VideoCommon.vcproj
View File

@ -692,11 +692,11 @@
Name="OpenCL"
>
<File
RelativePath=".\Src\OpenCL\TextureDecoder.cpp"
RelativePath=".\Src\OpenCL\OCLTextureDecoder.cpp"
>
</File>
<File
RelativePath=".\Src\OpenCL\TextureDecoder.h"
RelativePath=".\Src\OpenCL\OCLTextureDecoder.h"
>
</File>
</Filter>