Revert "revert tg calib and opencl cleanup (#37113)" (#37115)

* Revert "revert tg calib and opencl cleanup (#37113)"

This reverts commit 51312afd3d.

* power draw is a lil higher

* just don't miss a cycle

* fix warp targets

* fix tinygrad dep
This commit is contained in:
Adeeb Shihadeh
2026-02-07 21:36:44 -08:00
committed by GitHub
parent c65cf18c75
commit 667f3bb32f
48 changed files with 309 additions and 5237 deletions
-1
View File
@@ -5,7 +5,6 @@ common_libs = [
'swaglog.cc',
'util.cc',
'ratekeeper.cc',
'clutil.cc',
]
_common = env.Library('common', common_libs, LIBS="json11")
-98
View File
@@ -1,98 +0,0 @@
#include "common/clutil.h"
#include <cassert>
#include <iostream>
#include <memory>
#include "common/util.h"
#include "common/swaglog.h"
namespace { // helper functions
template <typename Func, typename Id, typename Name>
std::string get_info(Func get_info_func, Id id, Name param_name) {
size_t size = 0;
CL_CHECK(get_info_func(id, param_name, 0, NULL, &size));
std::string info(size, '\0');
CL_CHECK(get_info_func(id, param_name, size, info.data(), NULL));
return info;
}
inline std::string get_platform_info(cl_platform_id id, cl_platform_info name) { return get_info(&clGetPlatformInfo, id, name); }
inline std::string get_device_info(cl_device_id id, cl_device_info name) { return get_info(&clGetDeviceInfo, id, name); }
void cl_print_info(cl_platform_id platform, cl_device_id device) {
size_t work_group_size = 0;
cl_device_type device_type = 0;
clGetDeviceInfo(device, CL_DEVICE_MAX_WORK_GROUP_SIZE, sizeof(work_group_size), &work_group_size, NULL);
clGetDeviceInfo(device, CL_DEVICE_TYPE, sizeof(device_type), &device_type, NULL);
const char *type_str = "Other...";
switch (device_type) {
case CL_DEVICE_TYPE_CPU: type_str ="CL_DEVICE_TYPE_CPU"; break;
case CL_DEVICE_TYPE_GPU: type_str = "CL_DEVICE_TYPE_GPU"; break;
case CL_DEVICE_TYPE_ACCELERATOR: type_str = "CL_DEVICE_TYPE_ACCELERATOR"; break;
}
LOGD("vendor: %s", get_platform_info(platform, CL_PLATFORM_VENDOR).c_str());
LOGD("platform version: %s", get_platform_info(platform, CL_PLATFORM_VERSION).c_str());
LOGD("profile: %s", get_platform_info(platform, CL_PLATFORM_PROFILE).c_str());
LOGD("extensions: %s", get_platform_info(platform, CL_PLATFORM_EXTENSIONS).c_str());
LOGD("name: %s", get_device_info(device, CL_DEVICE_NAME).c_str());
LOGD("device version: %s", get_device_info(device, CL_DEVICE_VERSION).c_str());
LOGD("max work group size: %zu", work_group_size);
LOGD("type = %d, %s", (int)device_type, type_str);
}
void cl_print_build_errors(cl_program program, cl_device_id device) {
cl_build_status status;
clGetProgramBuildInfo(program, device, CL_PROGRAM_BUILD_STATUS, sizeof(status), &status, NULL);
size_t log_size;
clGetProgramBuildInfo(program, device, CL_PROGRAM_BUILD_LOG, 0, NULL, &log_size);
std::string log(log_size, '\0');
clGetProgramBuildInfo(program, device, CL_PROGRAM_BUILD_LOG, log_size, &log[0], NULL);
LOGE("build failed; status=%d, log: %s", status, log.c_str());
}
} // namespace
cl_device_id cl_get_device_id(cl_device_type device_type) {
cl_uint num_platforms = 0;
CL_CHECK(clGetPlatformIDs(0, NULL, &num_platforms));
std::unique_ptr<cl_platform_id[]> platform_ids = std::make_unique<cl_platform_id[]>(num_platforms);
CL_CHECK(clGetPlatformIDs(num_platforms, &platform_ids[0], NULL));
for (size_t i = 0; i < num_platforms; ++i) {
LOGD("platform[%zu] CL_PLATFORM_NAME: %s", i, get_platform_info(platform_ids[i], CL_PLATFORM_NAME).c_str());
// Get first device
if (cl_device_id device_id = NULL; clGetDeviceIDs(platform_ids[i], device_type, 1, &device_id, NULL) == 0 && device_id) {
cl_print_info(platform_ids[i], device_id);
return device_id;
}
}
LOGE("No valid openCL platform found");
assert(0);
return nullptr;
}
cl_context cl_create_context(cl_device_id device_id) {
return CL_CHECK_ERR(clCreateContext(NULL, 1, &device_id, NULL, NULL, &err));
}
void cl_release_context(cl_context context) {
clReleaseContext(context);
}
cl_program cl_program_from_file(cl_context ctx, cl_device_id device_id, const char* path, const char* args) {
return cl_program_from_source(ctx, device_id, util::read_file(path), args);
}
cl_program cl_program_from_source(cl_context ctx, cl_device_id device_id, const std::string& src, const char* args) {
const char *csrc = src.c_str();
cl_program prg = CL_CHECK_ERR(clCreateProgramWithSource(ctx, 1, &csrc, NULL, &err));
if (int err = clBuildProgram(prg, 1, &device_id, args, NULL, NULL); err != 0) {
cl_print_build_errors(prg, device_id);
assert(0);
}
return prg;
}
-28
View File
@@ -1,28 +0,0 @@
#pragma once
#ifdef __APPLE__
#include <OpenCL/cl.h>
#else
#include <CL/cl.h>
#endif
#include <string>
#define CL_CHECK(_expr) \
do { \
assert(CL_SUCCESS == (_expr)); \
} while (0)
#define CL_CHECK_ERR(_expr) \
({ \
cl_int err = CL_INVALID_VALUE; \
__typeof__(_expr) _ret = _expr; \
assert(_ret&& err == CL_SUCCESS); \
_ret; \
})
cl_device_id cl_get_device_id(cl_device_type device_type);
cl_context cl_create_context(cl_device_id device_id);
void cl_release_context(cl_context context);
cl_program cl_program_from_source(cl_context ctx, cl_device_id device_id, const std::string& src, const char* args = nullptr);
cl_program cl_program_from_file(cl_context ctx, cl_device_id device_id, const char* path, const char* args);
-85
View File
@@ -1,85 +0,0 @@
#pragma once
typedef struct vec3 {
float v[3];
} vec3;
typedef struct vec4 {
float v[4];
} vec4;
typedef struct mat3 {
float v[3*3];
} mat3;
typedef struct mat4 {
float v[4*4];
} mat4;
static inline mat3 matmul3(const mat3 &a, const mat3 &b) {
mat3 ret = {{0.0}};
for (int r=0; r<3; r++) {
for (int c=0; c<3; c++) {
float v = 0.0;
for (int k=0; k<3; k++) {
v += a.v[r*3+k] * b.v[k*3+c];
}
ret.v[r*3+c] = v;
}
}
return ret;
}
static inline vec3 matvecmul3(const mat3 &a, const vec3 &b) {
vec3 ret = {{0.0}};
for (int r=0; r<3; r++) {
for (int c=0; c<3; c++) {
ret.v[r] += a.v[r*3+c] * b.v[c];
}
}
return ret;
}
static inline mat4 matmul(const mat4 &a, const mat4 &b) {
mat4 ret = {{0.0}};
for (int r=0; r<4; r++) {
for (int c=0; c<4; c++) {
float v = 0.0;
for (int k=0; k<4; k++) {
v += a.v[r*4+k] * b.v[k*4+c];
}
ret.v[r*4+c] = v;
}
}
return ret;
}
static inline vec4 matvecmul(const mat4 &a, const vec4 &b) {
vec4 ret = {{0.0}};
for (int r=0; r<4; r++) {
for (int c=0; c<4; c++) {
ret.v[r] += a.v[r*4+c] * b.v[c];
}
}
return ret;
}
// scales the input and output space of a transformation matrix
// that assumes pixel-center origin.
static inline mat3 transform_scale_buffer(const mat3 &in, float s) {
// in_pt = ( transform(out_pt/s + 0.5) - 0.5) * s
mat3 transform_out = {{
1.0f/s, 0.0f, 0.5f,
0.0f, 1.0f/s, 0.5f,
0.0f, 0.0f, 1.0f,
}};
mat3 transform_in = {{
s, 0.0f, -0.5f*s,
0.0f, s, -0.5f*s,
0.0f, 0.0f, 1.0f,
}};
return matmul3(transform_in, matmul3(in, transform_out));
}