fumiama/base16384-sycl
1
0
Fork 0
mirror of https://github.com/fumiama/base16384-sycl.git synced 2026-06-05 08:40:34 +08:00
Code Activity

optimize(test): move test kernels into test class

Browse Source
This commit is contained in:
源文雨
2025-10-09 16:08:09 +08:00
parent cbe9cda397
commit 24ea4ca7bd
6 changed files with 44 additions and 28 deletions
Download Patch File Download Diff File Expand all files Collapse all files

15
include/test.hpp Normal file
View File

@@ -0,0 +1,15 @@
#ifndef _TEST_KERNELS_H_
#define _TEST_KERNELS_H_
#include <stdint.h>
namespace base16384 {
class test {
public:
// base16384_test_kernels_basic is a demo calculation that implements
// mod, bit, plus and mul calculations.
SYCL_EXTERNAL static uint8_t kernels_basic(uint8_t in);
};
} // namespace base16384
#endif

10
include/xeinfo.hpp
View File

@@ -1,6 +1,9 @@
#ifndef _XEINFO_HPP_
#define _XEINFO_HPP_
#include <stdint.h>
#include <iomanip>
#include <iostream>
#include <sstream>
#include <string>
@@ -59,7 +62,7 @@ class xeinfo {
const int num_subslices_per_slice;
const int num_eus_per_subslice;
const int num_threads_per_eu;
const int global_mem_size;
const uint64_t global_mem_size;
const int local_mem_size;
const int max_work_group_size;
const std::vector<unsigned long long> sub_group_sizes;
@@ -84,8 +87,9 @@ class xeinfo {
builder << " 每个 XeCore 的硬件线程数: " << num_thread_per_xecore << "\n";
builder << " 每个向量引擎的硬件线程数: " << num_threads_per_eu << "\n";
builder << " 硬件线程总数: " << total_hardware_threads << "\n";
builder << " GPU 内存大小: " << global_mem_size << " 字节\n";
builder << " 每个工作组的共享本地内存: " << local_mem_size << " 字节\n";
builder << " GPU 内存大小: " << global_mem_size << " B (" << std::fixed << std::setprecision(2)
<< (double)global_mem_size / 1024 / 1024 / 1024 << " GB)\n";
builder << " 每个工作组的共享本地内存: " << local_mem_size << " B\n";
builder << " 最大工作组大小: " << max_work_group_size << "\n";
builder << " 支持的子组大小:";
for (size_t i = 0; i < sub_group_sizes.size(); i++) builder << " " << sub_group_sizes[i];