项目文件夹

文件
wehub-resource-sync 59a0a3844c
PR Test AMD / cancel-on-close (push) Has been skipped
PR Test NVIDIA ARM / scan (push) Has been skipped
PR Test NVIDIA / cancel-on-close (push) Has been skipped
PR Test AMD / scan (push) Has been skipped
PR Test NVIDIA ARM / cancel-on-close (push) Has been skipped
PR Test NVIDIA / scan (push) Has been skipped
Release Docker Images / build (cu129-torch-2.11.0) (push) Has been skipped
Release Docker Images / build (cu130-torch-2.11.0) (push) Has been skipped
Release PyPI / publish (push) Has been skipped
Scheduler Python Test / test (push) Successful in 27m19s
Docs / build (push) Successful in 28m8s
Scheduler C++ Test / test (push) Successful in 28m19s
Scheduler C++ Test / test-flat (push) Successful in 28m18s
Docs / deploy (push) Has been cancelled
PR Test AMD / finish (push) Has been cancelled
PR Test NVIDIA / finish (push) Has been cancelled
PR Test NVIDIA ARM / finish (push) Has been cancelled
PR Test NVIDIA ARM / ${{ matrix.name }} (${{ matrix.runner }}) (push) Has been cancelled
PR Test AMD / ${{ matrix.name }} (${{ matrix.runner }}) (push) Has been cancelled
PR Test NVIDIA / ${{ matrix.name }} (${{ matrix.runner }}) (push) Has been cancelled
chore: import upstream snapshot with attribution
2026-07-13 12:32:31 +08:00

329 行
11 KiB
C++

// Copyright (c) 2026 LightSeek Foundation
//
// Permission is hereby granted, free of charge, to any person obtaining a copy
// of this software and associated documentation files (the "Software"), to deal
// in the Software without restriction, including without limitation the rights
// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
// copies of the Software, and to permit persons to whom the Software is
// furnished to do so, subject to the following conditions:
//
// The above copyright notice and this permission notice shall be included in
// all copies or substantial portions of the Software.
//
// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
// SOFTWARE.
#include <gtest/gtest.h>
#include <memory>
#include <numeric>
#include "resource/allocator/mamba_chunk_allocator.h"
#include "resource/allocator/mamba_host_allocator.h"
#include "resource/radix_tree/mamba_slot.h"
#include "resource/allocator/local_mamba_allocator.h"
#include "resource/radix_tree/tree_node.h"
#include "resource/allocator/page_allocator.h"
namespace tokenspeed::test {
TEST(MambaChunkAllocatorTest, AllocateReturnsSequentialSlots) {
MambaChunkAllocator allocator(4);
EXPECT_EQ(allocator.TotalSlots(), 4);
EXPECT_EQ(allocator.AvailableSlots(), 4);
auto slot0 = allocator.Allocate();
ASSERT_TRUE(slot0.has_value());
EXPECT_EQ(allocator.AvailableSlots(), 3);
auto slot1 = allocator.Allocate();
ASSERT_TRUE(slot1.has_value());
EXPECT_NE(slot0->Index(), slot1->Index());
}
TEST(MambaChunkAllocatorTest, ExhaustedPoolReturnsNullopt) {
MambaChunkAllocator allocator(2);
auto s0 = allocator.Allocate();
auto s1 = allocator.Allocate();
auto s2 = allocator.Allocate();
EXPECT_FALSE(s2.has_value());
}
TEST(MambaChunkAllocatorTest, AllocateReturnsSmallestFreeIndexRegardlessOfFreeOrder) {
// TP-determinism invariant: the index returned by Allocate() must depend
// only on the SET of free indices, not on the order in which Free() was
// called. Replays the same Free sequence in two different orders and
// expects the same Allocate result.
MambaChunkAllocator allocator_a(8);
MambaChunkAllocator allocator_b(8);
// Drain both pools.
std::vector<std::optional<MambaSlot>> slots_a, slots_b;
for (int i = 0; i < 8; ++i) {
slots_a.push_back(allocator_a.Allocate());
slots_b.push_back(allocator_b.Allocate());
}
// Free the same set of indices in opposite orders.
slots_a[3].reset();
slots_a[6].reset();
slots_a[1].reset();
slots_b[1].reset();
slots_b[6].reset();
slots_b[3].reset();
auto next_a = allocator_a.Allocate();
auto next_b = allocator_b.Allocate();
ASSERT_TRUE(next_a.has_value());
ASSERT_TRUE(next_b.has_value());
EXPECT_EQ(next_a->Index(), next_b->Index());
EXPECT_EQ(next_a->Index(), 1); // smallest free index after freeing {1,3,6}
}
TEST(MambaHostAllocatorTest, AllocateFreeAndDrainReleased) {
tokenspeed::MambaHostAllocator allocator(3);
auto slot = allocator.Allocate();
ASSERT_TRUE(slot.has_value());
std::int32_t idx = slot->Index();
EXPECT_EQ(allocator.AvailableSlots(), 2);
slot.reset();
EXPECT_EQ(allocator.AvailableSlots(), 3);
auto released = allocator.DrainReleased();
ASSERT_EQ(released.size(), 1);
EXPECT_EQ(released[0], idx);
EXPECT_TRUE(allocator.DrainReleased().empty());
}
TEST(MambaHostAllocatorTest, AllocateReturnsSmallestFreeIndexRegardlessOfFreeOrder) {
// Mirror of the MambaChunkAllocator determinism test for the host pool.
tokenspeed::MambaHostAllocator allocator_a(8);
tokenspeed::MambaHostAllocator allocator_b(8);
std::vector<std::optional<tokenspeed::MambaSlot>> slots_a, slots_b;
for (int i = 0; i < 8; ++i) {
slots_a.push_back(allocator_a.Allocate());
slots_b.push_back(allocator_b.Allocate());
}
slots_a[5].reset();
slots_a[2].reset();
slots_a[7].reset();
slots_b[7].reset();
slots_b[2].reset();
slots_b[5].reset();
auto next_a = allocator_a.Allocate();
auto next_b = allocator_b.Allocate();
ASSERT_TRUE(next_a.has_value());
ASSERT_TRUE(next_b.has_value());
EXPECT_EQ(next_a->Index(), next_b->Index());
EXPECT_EQ(next_a->Index(), 2);
}
TEST(MambaHostAllocatorTest, DrainReleasedReturnsSortedRegardlessOfFreeOrder) {
// The drain output must be order-independent across TP ranks too — the
// consumer treats it as a SET of released indices.
tokenspeed::MambaHostAllocator allocator(8);
std::vector<std::optional<tokenspeed::MambaSlot>> slots;
for (int i = 0; i < 8; ++i) {
slots.push_back(allocator.Allocate());
}
slots[5].reset();
slots[1].reset();
slots[7].reset();
slots[3].reset();
auto released = allocator.DrainReleased();
ASSERT_EQ(released.size(), 4);
EXPECT_EQ(released[0], 1);
EXPECT_EQ(released[1], 3);
EXPECT_EQ(released[2], 5);
EXPECT_EQ(released[3], 7);
}
TEST(MambaSlotTest, CustomReleaserRunsOnDestruction) {
std::vector<std::int32_t> released;
{
tokenspeed::MambaSlot slot(7, [&released](std::int32_t idx) { released.push_back(idx); });
EXPECT_EQ(slot.Index(), 7);
}
ASSERT_EQ(released.size(), 1);
EXPECT_EQ(released[0], 7);
}
TEST(MambaSlotTest, RAIIFreesOnDestruction) {
MambaChunkAllocator allocator(2);
{
auto slot = allocator.Allocate();
ASSERT_TRUE(slot.has_value());
EXPECT_EQ(allocator.AvailableSlots(), 1);
}
EXPECT_EQ(allocator.AvailableSlots(), 2);
}
TEST(MambaSlotTest, MoveTransfersOwnership) {
MambaChunkAllocator allocator(2);
auto slot = allocator.Allocate();
ASSERT_TRUE(slot.has_value());
std::int32_t idx = slot->Index();
MambaSlot moved = std::move(*slot);
EXPECT_EQ(moved.Index(), idx);
EXPECT_EQ(allocator.AvailableSlots(), 1);
}
TEST(MambaSlotTest, UniquePtrOwnership) {
MambaChunkAllocator allocator(2);
{
auto slot = allocator.Allocate();
auto ptr = std::make_unique<MambaSlot>(std::move(*slot));
EXPECT_EQ(allocator.AvailableSlots(), 1);
}
EXPECT_EQ(allocator.AvailableSlots(), 2);
}
TEST(TreeNodeMambaTest, DefaultHasNoMamba) {
tokenspeed::TreeNode node;
EXPECT_FALSE(node.HasMamba());
}
TEST(TreeNodeMambaTest, AttachAndDetachMamba) {
tokenspeed::MambaChunkAllocator allocator(4);
tokenspeed::TreeNode node;
auto slot = allocator.Allocate();
ASSERT_TRUE(slot.has_value());
std::int32_t idx = slot->Index();
node.AttachMamba(std::make_unique<tokenspeed::MambaSlot>(std::move(*slot)));
EXPECT_TRUE(node.HasMamba());
EXPECT_EQ(node.MambaSlotIndex(), idx);
auto detached = node.DetachMamba();
EXPECT_FALSE(node.HasMamba());
EXPECT_EQ(detached->Index(), idx);
EXPECT_EQ(allocator.AvailableSlots(), 3);
}
TEST(TreeNodeMambaTest, DestructorFreesMambaSlot) {
tokenspeed::MambaChunkAllocator allocator(4);
{
tokenspeed::TreeNode node;
auto slot = allocator.Allocate();
node.AttachMamba(std::make_unique<tokenspeed::MambaSlot>(std::move(*slot)));
EXPECT_EQ(allocator.AvailableSlots(), 3);
}
EXPECT_EQ(allocator.AvailableSlots(), 4);
}
TEST(TreeNodeMambaTest, AttachAndDetachMambaHost) {
tokenspeed::MambaHostAllocator allocator(4);
tokenspeed::TreeNode node;
auto slot = allocator.Allocate();
ASSERT_TRUE(slot.has_value());
std::int32_t idx = slot->Index();
node.AttachMambaHost(std::make_unique<tokenspeed::MambaSlot>(std::move(*slot)));
EXPECT_TRUE(node.HasMambaOnHost());
EXPECT_EQ(node.MambaHostSlotIndex(), idx);
auto detached = node.DetachMambaHost();
EXPECT_FALSE(node.HasMambaOnHost());
EXPECT_EQ(detached->Index(), idx);
EXPECT_EQ(allocator.AvailableSlots(), 3);
}
TEST(TreeNodeMambaTest, DestructorFreesMambaHostSlot) {
tokenspeed::MambaHostAllocator allocator(4);
{
tokenspeed::TreeNode node;
auto slot = allocator.Allocate();
node.AttachMambaHost(std::make_unique<tokenspeed::MambaSlot>(std::move(*slot)));
EXPECT_EQ(allocator.AvailableSlots(), 3);
}
EXPECT_EQ(allocator.AvailableSlots(), 4);
EXPECT_EQ(allocator.DrainReleased().size(), 1);
}
TEST(TreeNodeMambaTest, SplitKeepsMambaOnSuffix) {
tokenspeed::MambaChunkAllocator mamba_alloc(4);
tokenspeed::PageAllocator page_alloc(2, 8);
tokenspeed::TreeNode root;
auto tokens = tokenspeed::token_vec_t(8);
std::iota(tokens.begin(), tokens.end(), 1);
auto child_ptr = std::make_unique<tokenspeed::TreeNode>(tokens);
auto pages = page_alloc.Allocate(4);
child_ptr->AttachResource<tokenspeed::ResourceType::Device>(
std::make_unique<tokenspeed::DeviceResource>(std::move(pages)));
auto slot = mamba_alloc.Allocate();
std::int32_t idx = slot->Index();
child_ptr->AttachMamba(std::make_unique<tokenspeed::MambaSlot>(std::move(*slot)));
tokenspeed::TreeNode* child = child_ptr.get();
root.AddChild(tokens, std::move(child_ptr));
tokenspeed::TreeNode prefix;
child->SplitSelfInto(prefix, 2, 2);
EXPECT_TRUE(child->HasMamba());
EXPECT_EQ(child->MambaSlotIndex(), idx);
EXPECT_FALSE(prefix.HasMamba());
}
TEST(LocalMambaAllocatorTest, AllocateWorkingAndCheckpoint) {
tokenspeed::MambaChunkAllocator allocator(8);
tokenspeed::LocalMambaAllocator local(&allocator);
EXPECT_FALSE(local.HasWorking());
EXPECT_TRUE(local.AllocateWorking());
EXPECT_TRUE(local.HasWorking());
EXPECT_GE(local.WorkingIndex(), 0);
EXPECT_FALSE(local.HasCheckpoint());
EXPECT_TRUE(local.AllocateCheckpoint());
EXPECT_TRUE(local.HasCheckpoint());
EXPECT_GE(local.CheckpointIndex(), 0);
EXPECT_NE(local.WorkingIndex(), local.CheckpointIndex());
EXPECT_EQ(allocator.AvailableSlots(), 6);
}
TEST(LocalMambaAllocatorTest, DetachCheckpointTransfersOwnership) {
tokenspeed::MambaChunkAllocator allocator(8);
tokenspeed::LocalMambaAllocator local(&allocator);
local.AllocateWorking();
local.AllocateCheckpoint();
std::int32_t cp_idx = local.CheckpointIndex();
auto detached = local.DetachCheckpoint();
EXPECT_FALSE(local.HasCheckpoint());
EXPECT_EQ(detached->Index(), cp_idx);
EXPECT_EQ(allocator.AvailableSlots(), 6);
}
TEST(LocalMambaAllocatorTest, DetachWorkingTransfersOwnership) {
tokenspeed::MambaChunkAllocator allocator(8);
tokenspeed::LocalMambaAllocator local(&allocator);
local.AllocateWorking();
std::int32_t w_idx = local.WorkingIndex();
auto detached = local.DetachWorking();
EXPECT_FALSE(local.HasWorking());
EXPECT_EQ(detached->Index(), w_idx);
}
TEST(LocalMambaAllocatorTest, DestructorFreesAllSlots) {
tokenspeed::MambaChunkAllocator allocator(8);
{
tokenspeed::LocalMambaAllocator local(&allocator);
local.AllocateWorking();
local.AllocateCheckpoint();
EXPECT_EQ(allocator.AvailableSlots(), 6);
}
EXPECT_EQ(allocator.AvailableSlots(), 8);
}
} // namespace tokenspeed::test