lightseekorg--tokenspeed
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
162 行
6.8 KiB
C++
162 行
6.8 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 <cstdint>
|
|
#include <string>
|
|
#include <vector>
|
|
|
|
#include "cache/block_pool.h"
|
|
#include "cache/full_attn_manager.h"
|
|
#include "cache/swa_manager.h"
|
|
|
|
namespace tokenspeed::test {
|
|
namespace {
|
|
|
|
// The unified Match with a raised floor (begin_blocks > 0 over a non-device pool) is the
|
|
// host-tier lookup: slots below the floor are device-valid, holes come back as the queried
|
|
// pool's null block.
|
|
|
|
// Publish a host page for `key` (the scheduler's store path minus the D2H write):
|
|
// allocate -> hash -> free leaves it cached-and-evictable, exactly like a committed store.
|
|
CacheBlock* Put(BlockPool& host_pool, const std::string& key) {
|
|
CacheBlock* block = host_pool.AllocateBlocks(1).front();
|
|
host_pool.CacheFullBlock(block, key);
|
|
host_pool.FreeBlocks({block});
|
|
return block;
|
|
}
|
|
|
|
TEST(HostTierMatchTest, FullWalksContiguousRunFromBegin) {
|
|
BlockPool pool(4);
|
|
FullAttnManager mgr(/*block_size=*/4);
|
|
EXPECT_TRUE(mgr.MatchIsPrefixClosed());
|
|
BlockPool host_pool(9);
|
|
std::vector<std::string> keys{"k0", "k1", "k2", "k3", "k4"};
|
|
std::vector<CacheBlock*> put;
|
|
for (std::size_t j = 1; j <= 4; ++j) {
|
|
put.push_back(Put(host_pool, keys[j]));
|
|
}
|
|
// Slots below begin=1 are device-valid; the run covers all extension slots, no holes.
|
|
PrefixMatch m = mgr.Match(host_pool, keys, /*begin_blocks=*/1, /*max_blocks=*/5);
|
|
EXPECT_EQ(m.blocks, put);
|
|
EXPECT_EQ(m.num_hit_blocks, 4);
|
|
}
|
|
|
|
TEST(HostTierMatchTest, FullStopsAtFirstMiss) {
|
|
BlockPool pool(4);
|
|
FullAttnManager mgr(4);
|
|
BlockPool host_pool(9);
|
|
std::vector<std::string> keys{"k0", "k1", "k2", "k3"};
|
|
CacheBlock* p0 = Put(host_pool, keys[0]);
|
|
CacheBlock* p1 = Put(host_pool, keys[1]);
|
|
(void)Put(host_pool, keys[3]); // beyond the gap at k2: unreachable
|
|
EXPECT_EQ(mgr.Match(host_pool, keys, 0, 4).blocks, (std::vector<CacheBlock*>{p0, p1}));
|
|
}
|
|
|
|
TEST(HostTierMatchTest, FullEmptyOnBeginMissOrEmptyRange) {
|
|
BlockPool pool(4);
|
|
FullAttnManager mgr(4);
|
|
BlockPool host_pool(9);
|
|
std::vector<std::string> keys{"k0", "k1"};
|
|
(void)Put(host_pool, keys[1]);
|
|
EXPECT_TRUE(mgr.Match(host_pool, keys, 0, 2).blocks.empty()); // miss right at begin
|
|
EXPECT_TRUE(mgr.Match(host_pool, keys, 2, 2).blocks.empty()); // empty extension range
|
|
}
|
|
|
|
TEST(HostTierMatchTest, SwaTrailingRunAtEnd) {
|
|
BlockPool pool(4);
|
|
// block_size 4, window 10 -> pages_needed = ceil(9/4) = 3.
|
|
SwaManager mgr(4, /*sliding_window=*/10);
|
|
EXPECT_FALSE(mgr.MatchIsPrefixClosed());
|
|
BlockPool host_pool(9);
|
|
std::vector<std::string> keys{"k0", "k1", "k2", "k3", "k4"};
|
|
CacheBlock* p2 = Put(host_pool, keys[2]);
|
|
CacheBlock* p3 = Put(host_pool, keys[3]);
|
|
CacheBlock* p4 = Put(host_pool, keys[4]);
|
|
CacheBlock* hole = host_pool.NullBlock();
|
|
// Trailing run [2, 5) covers the window at boundary 5; slots below stay holes.
|
|
PrefixMatch m = mgr.Match(host_pool, keys, 0, 5);
|
|
EXPECT_EQ(m.blocks, (std::vector<CacheBlock*>{hole, hole, p2, p3, p4}));
|
|
EXPECT_EQ(m.num_hit_blocks, 3);
|
|
}
|
|
|
|
TEST(HostTierMatchTest, SwaInteriorBoundaryShrink) {
|
|
BlockPool pool(4);
|
|
SwaManager mgr(4, 10); // pages_needed = 3
|
|
BlockPool host_pool(9);
|
|
std::vector<std::string> keys{"k0", "k1", "k2", "k3", "k4"};
|
|
CacheBlock* p1 = Put(host_pool, keys[1]);
|
|
CacheBlock* p2 = Put(host_pool, keys[2]);
|
|
CacheBlock* p3 = Put(host_pool, keys[3]);
|
|
// Miss at 4 invalidates boundary 5; boundary 4 needs [1, 4), which hits.
|
|
EXPECT_EQ(mgr.Match(host_pool, keys, 0, 5).blocks, (std::vector<CacheBlock*>{host_pool.NullBlock(), p1, p2, p3}));
|
|
}
|
|
|
|
TEST(HostTierMatchTest, SwaShortRunAtBottomSuffices) {
|
|
BlockPool pool(4);
|
|
SwaManager mgr(4, 10); // pages_needed = 3, but only 2 extension slots exist
|
|
BlockPool host_pool(9);
|
|
std::vector<std::string> keys{"k0", "k1"};
|
|
CacheBlock* p0 = Put(host_pool, keys[0]);
|
|
CacheBlock* p1 = Put(host_pool, keys[1]);
|
|
// The window clamps to begin: a full 2-run from the bottom is a valid boundary 2.
|
|
EXPECT_EQ(mgr.Match(host_pool, keys, 0, 2).blocks, (std::vector<CacheBlock*>{p0, p1}));
|
|
}
|
|
|
|
TEST(HostTierMatchTest, SwaBeginAboveZeroInteriorBoundary) {
|
|
BlockPool pool(4);
|
|
SwaManager mgr(4, /*sliding_window=*/9); // pages_needed = ceil(8/4) = 2
|
|
BlockPool host_pool(9);
|
|
std::vector<std::string> keys{"k0", "k1", "k2", "k3", "k4", "k5", "k6"};
|
|
CacheBlock* p3 = Put(host_pool, keys[3]);
|
|
CacheBlock* p4 = Put(host_pool, keys[4]);
|
|
CacheBlock* p5 = Put(host_pool, keys[5]);
|
|
(void)p3; // hit at slot 3 sits below the winning run's window and stays a hole
|
|
// Miss at 6 invalidates boundary 7; boundary 6 needs [4, 6), which hits -> vector
|
|
// covers [3, 6): hole at slot 3, pages for 4 and 5.
|
|
PrefixMatch m = mgr.Match(host_pool, keys, /*begin_blocks=*/3, /*max_blocks=*/7);
|
|
EXPECT_EQ(m.blocks, (std::vector<CacheBlock*>{host_pool.NullBlock(), p4, p5}));
|
|
EXPECT_EQ(m.num_hit_blocks, 2);
|
|
}
|
|
|
|
TEST(HostTierMatchTest, SwaAllMissReturnsEmpty) {
|
|
BlockPool pool(4);
|
|
SwaManager mgr(4, 10);
|
|
BlockPool host_pool(9);
|
|
std::vector<std::string> keys{"k0", "k1", "k2", "k3", "k4"};
|
|
EXPECT_TRUE(mgr.Match(host_pool, keys, 1, 5).blocks.empty());
|
|
}
|
|
|
|
TEST(HostTierMatchTest, SwaZeroNeededWindowAcceptsAllAsHoles) {
|
|
BlockPool pool(4);
|
|
SwaManager mgr(4, /*sliding_window=*/1); // pages_needed = 0
|
|
BlockPool host_pool(9);
|
|
std::vector<std::string> keys{"k0", "k1", "k2"};
|
|
CacheBlock* hole = host_pool.NullBlock();
|
|
// Zero needed pages: every boundary is resumable with no host page at all.
|
|
PrefixMatch m = mgr.Match(host_pool, keys, 1, 3);
|
|
EXPECT_EQ(m.blocks, (std::vector<CacheBlock*>{hole, hole}));
|
|
EXPECT_EQ(m.num_hit_blocks, 0);
|
|
}
|
|
|
|
} // namespace
|
|
} // namespace tokenspeed::test
|