Files
ktransformers/archive/csrc/balance_serve/kvc2/test/kvc2test/flush-back.cpp
Jiaqi Liao 57d14d22bc Refactor: restructure repository to focus on kt-kernel and KT-SFT modulesq recon (#1581)
* refactor: move legacy code to archive/ directory

  - Moved ktransformers, csrc, third_party, merge_tensors to archive/
  - Moved build scripts and configurations to archive/
  - Kept kt-kernel, KT-SFT, doc, and README files in root
  - Preserved complete git history for all moved files

* refactor: restructure repository to focus on kt-kernel and KT-SFT modules

* fix README

* fix README

* fix README

* fix README

* docs: add performance benchmarks to kt-kernel section

Add comprehensive performance data for kt-kernel to match KT-SFT's presentation:
- AMX kernel optimization: 21.3 TFLOPS (3.9× faster than PyTorch)
- Prefill phase: up to 20× speedup vs baseline
- Decode phase: up to 4× speedup
- NUMA optimization: up to 63% throughput improvement
- Multi-GPU (8×L20): 227.85 tokens/s total throughput with DeepSeek-R1 FP8

Source: https://lmsys.org/blog/2025-10-22-KTransformers/

This provides users with concrete performance metrics for both core modules,
making it easier to understand the capabilities of each component.

* refactor: improve kt-kernel performance data with specific hardware and models

Replace generic performance descriptions with concrete benchmarks:
- Specify exact hardware: 8×L20 GPU + Xeon Gold 6454S, Single/Dual-socket Xeon + AMX
- Include specific models: DeepSeek-R1-0528 (FP8), DeepSeek-V3 (671B)
- Show detailed metrics: total throughput, output throughput, concurrency details
- Match KT-SFT presentation style for consistency

This provides users with actionable performance data they can use to evaluate
hardware requirements and expected performance for their use cases.

* fix README

* docs: clean up performance table and improve formatting

* add pic for README

* refactor: simplify .gitmodules and backup legacy submodules

- Remove 7 legacy submodules from root .gitmodules (archive/third_party/*)
- Keep only 2 active submodules for kt-kernel (llama.cpp, pybind11)
- Backup complete .gitmodules to archive/.gitmodules
- Add documentation in archive/README.md for researchers who need legacy submodules

This reduces initial clone size by ~500MB and avoids downloading unused dependencies.

* refactor: move doc/ back to root directory

Keep documentation in root for easier access and maintenance.

* refactor: consolidate all images to doc/assets/

- Move kt-kernel/assets/heterogeneous_computing.png to doc/assets/
- Remove KT-SFT/assets/ (images already in doc/assets/)
- Update KT-SFT/README.md image references to ../doc/assets/
- Eliminates ~7.9MB image duplication
- Centralizes all documentation assets in one location

* fix pic path for README
2025-11-10 17:42:26 +08:00

58 lines
2.3 KiB
C++

#include <future>
#include "common.hpp"
int main(int argc, char* argv[]) {
init(argc, argv);
spdlog::set_level(spdlog::level::debug);
config.gpu_cache_config->total_kvcache_pages = 12;
auto kvc2 = kvc2::create_kvc2(config);
// #pragma omp parallel for
for (size_t ti = 0; ti < 2; ti++) {
SPDLOG_WARN("Test {}", ti);
auto [kcache, vcache] = kvc2->get_kvcache();
std::mt19937 gen(ti + 123);
size_t total_page = 10;
TokenLength total_length = total_page * config.num_token_per_page;
auto tokens = random_ids(total_length, gen);
TokenLength prompt_length = 3 * config.num_token_per_page;
auto k1 = random_kvcache(total_page, gen);
auto v1 = random_kvcache(total_page, gen);
{
std::promise<std::shared_ptr<DoubleCacheHandleInterface>> p;
kvc2->lookup_to_gpu_async(test_model_name, test_quant_type, tokens.data(), prompt_length, total_length,
[&p](std::shared_ptr<DoubleCacheHandleInterface> h) { p.set_value(h); });
auto fut = p.get_future();
fut.wait();
auto h = fut.get();
assert(h->matched_length() % config.num_token_per_page == 0);
size_t matched_block = h->matched_length() / config.num_token_per_page;
auto block_idx = h->get_gpu_block_idx();
cmp_handle_gpu(block_idx, kcache, vcache, k1, v1, matched_block);
for (size_t at = matched_block; at < block_idx.size(); at++) {
copy_cpu_gpu(block_idx, kcache, vcache, k1, v1, at);
}
h->append_tokens(tokens.data(), total_length);
cmp_handle_gpu(block_idx, kcache, vcache, k1, v1, total_page);
}
{
std::promise<std::shared_ptr<DoubleCacheHandleInterface>> p;
kvc2->lookup_to_gpu_async(test_model_name, test_quant_type, tokens.data(), total_length, total_length,
[&p](std::shared_ptr<DoubleCacheHandleInterface> h) { p.set_value(h); });
auto fut = p.get_future();
fut.wait();
auto h = fut.get();
assert(h->matched_length() == total_length);
size_t matched_block = h->matched_length() / config.num_token_per_page;
auto block_idx = h->get_gpu_block_idx();
cmp_handle_gpu(block_idx, kcache, vcache, k1, v1, matched_block);
}
}
kvc2->save();
SPDLOG_CRITICAL("All Test Passed: {}", argv[0]);
return 0;
}