Merge pull request #12 from ludovic111/collector-validation

Combine foreground terrain controls and compare Java collector settings
This commit is contained in:
Ludovic Marie
2026-10-04 01:11:18 +02:00
committed by GitHub
20 changed files with 225 additions and 13 deletions
+2
View File
@@ -45,6 +45,8 @@ jobs:
run: python tools/build_mod.py
- name: Verify streaming JFR analysis against the JDK exporter
run: python tools/test_jfr.py
- name: Verify ZGC pause and allocation event interpretation
run: python tools/test_jfr.py --gc zgc
- name: Build pinned portable packs
run: python tools/build_pack.py
- name: Measure isolated worker pacing on this platform
+9 -2
View File
@@ -6,7 +6,7 @@ Client-first performance research for Minecraft Java **26.3**, preserving vanill
## Play on your Mac
Download **horizon-mac-starter-0.10.0.zip** from [Releases](https://github.com/ludovic111/minecraft-optimization-lab/releases), extract it and open `Mac-Start.command`. It verifies and installs official universal Prism Launcher in `~/Applications` if needed, then opens the balanced pack import. Sign into Microsoft **inside Prism yourself**. This project needs no credentials.
Download **horizon-mac-starter-0.11.0.zip** from [Releases](https://github.com/ludovic111/minecraft-optimization-lab/releases), extract it and open `Mac-Start.command`. It verifies and installs official universal Prism Launcher in `~/Applications` if needed, then opens the balanced pack import. Sign into Microsoft **inside Prism yourself**. This project needs no credentials.
Use native **Java 25 ARM64**, maximum heap **6144 MiB**, minimum **512 MiB** on your 16 GB M2 Pro. Prism can download Java automatically. Start at 1280×800 or 1920×1200; full Retina resolution is a larger GPU workload. [START-HERE.txt](packs/START-HERE.txt) has details. The same `.mrpack` files import into Prism on Windows and Linux.
@@ -20,6 +20,8 @@ Use native **Java 25 ARM64**, maximum heap **6144 MiB**, minimum **512 MiB** on
| queue-wait (experimental) | 12 | 128 | 6144 MiB |
| adaptive-queue (experimental) | 12 | 128 | 6144 MiB |
| dispatch-retry (experimental) | 12 | 128 | 6144 MiB |
| foreground-first (combined experiment) | 12 | 128 | 6144 MiB |
| foreground-vulkan (combined experiment) | 12 | 128 | 6144 MiB |
| low-memory | 8 | none | 3072 MiB |
| far-distance | 12 | 256 | 8192 MiB, with sufficient physical RAM |
| cinematic | 12 | 128 | 6144 MiB |
@@ -31,6 +33,11 @@ Distant LOD is simplified terrain. Nearby blocks retain full detail. Initial gen
The optional `queue-wait` pack yields during pending distant-terrain debounce retries; `adaptive-queue` combines it with the existing controllers. [Queue experiment and measurement limits](docs/debounce-queue.md).
Version 0.11 adds combined foreground-first OpenGL/Vulkan profiles with a 120 FPS
adaptive goal, plus explicit collector experiments and GC measurement guidance.
See [combined controls](docs/foreground-first.md), [Vulkan on your Mac](docs/vulkan.md)
and [collector comparisons](docs/collectors.md). The frame goal is not a rendering cap.
Version 0.10 adds an isolated [rejected-dispatch retry experiment](docs/dispatch-retry.md).
It bounds retry passes when a DH worker rejects queued work, preserving original
requeue handling and allowing other executors to progress. Balanced keeps it off.
@@ -82,7 +89,7 @@ python3 tools/build_mod.py
python3 tools/build_release.py
```
Gitignored `dist/` contains twelve packs, our original instrumentation JAR, client/server starter ZIPs and SHA256SUMS. [mods.lock.json](packs/mods.lock.json) pins mod versions, exact dependency closure, SHA-1/SHA-512 and source/license; [fabric.lock.json](packs/fabric.lock.json) pins loader/library SHA-256. Packs use the [Modrinth format](https://support.modrinth.com/en/articles/8802351-modrinth-modpack-format-mrpack). They embed only original project code; third-party mods, game and assets download from their publishers.
Gitignored `dist/` contains fourteen packs, our original instrumentation JAR, client/server starter ZIPs and SHA256SUMS. [mods.lock.json](packs/mods.lock.json) pins mod versions, exact dependency closure, SHA-1/SHA-512 and source/license; [fabric.lock.json](packs/fabric.lock.json) pins loader/library SHA-256. Packs use the [Modrinth format](https://support.modrinth.com/en/articles/8802351-modrinth-modpack-format-mrpack). They embed only original project code; third-party mods, game and assets download from their publishers.
The server starter uses Lithium/FerriteCore and `/horizonserver bench <label> <seconds>`. [Server instructions](docs/server-quickstart.txt) cover setup and a baseline. Files are prepared before you accept Minecraft's EULA locally. A 50-active-player workload remains to be measured.
+49
View File
@@ -0,0 +1,49 @@
# Compare Java collectors
Use the balanced pack first with the launcher's existing JVM policy. Java 25
HotSpot supports explicit G1 (`-XX:+UseG1GC`) and ZGC (`-XX:+UseZGC`) selections;
the [Java command reference](https://docs.oracle.com/en/java/javase/25/docs/specs/man/java.html)
documents both. A collector choice can change CPU, heap headroom, collection
behavior and frame pacing. Lower pauses alone do not establish lower RAM or
higher FPS. The packs do not automatically select a collector.
For a Mac comparison, duplicate one instance. In Edit > Settings > Java, enable
the Java Arguments override checkbox. Preserve existing non-collector arguments,
but use exactly one collector selector in each instance. Choose G1 in one and
ZGC in the other. Use the same native Java 25 ARM64, maximum/minimum heap, compact
header setting, world copy, resolution, shaders, distances and warmed camera route.
Alternate at least three 120-second runs rather than comparing unrelated captures.
Remove the explicit selector to restore the launcher's policy, then restart.
The viewer shows the actual collector names captured by the game. Check them:
setting a launcher's text field does not establish which flags reached the VM.
Prism's per-instance Java Arguments and Memory settings have separate override
checkboxes. A global collector selection may need removal or replacement when
creating an explicit per-instance experiment.
Record FPS, slowest-1% rate, frame spikes, heap and process RSS together. RSS does
not represent all shared GPU memory on Apple Silicon. `gc_time_ms` sums collector
MXBean counters; concurrent cycles and overlapping phases can contribute, so it
is not a measured pause total. Use JFR root pause events for pause analysis.
The streaming reader is checked against the JDK's event exporter with both
default-GC and ZGC fixtures in CI on Windows, Linux and ARM macOS.
Developer demo and server launchers accept `--gc default`, `--gc g1` or `--gc zgc`.
For example:
```sh
python3 tools/runtime.py demo --instance .local/runtime/collector-test --heap 6G --gc zgc --compact-headers
python3 tools/server.py run --heap 4G --gc zgc
python3 tools/test_jfr.py --gc zgc
```
Prepare demo mods/worlds before comparing. The server retains its EULA requirement.
Options affect only the launched process; they do not change global Java settings.
The collector and compact-header experiments can be tested independently.
A local 3 GiB, compact-header ZGC client completed a 120-second game recording.
That recording also encountered a distant-terrain rejection storm. Its workload
and cache history differed from G1 captures, so the larger memory footprint and
different frame rate cannot be attributed to the collector alone. The dispatcher
experiment addresses the identified retry loop separately. No M2 collector
performance winner is established.
+1 -1
View File
@@ -9,7 +9,7 @@ Native renderer buffers and GPU allocations are outside this optimization.
## Try on your Mac
Import the balanced pack normally. In Prism, edit the instance's Java settings
and add `-XX:+UseCompactObjectHeaders` to its JVM arguments, retaining any existing
check the Java Arguments override box, and add `-XX:+UseCompactObjectHeaders` to its JVM arguments, retaining any existing
arguments. Use Java 25 HotSpot matching your CPU architecture. Restart the game.
Remove the argument to return to the launcher's default, or use
`-XX:-UseCompactObjectHeaders` to explicitly test the disabled mode. Do not supply
+30
View File
@@ -0,0 +1,30 @@
# Combined foreground-first experiments
`foreground-first` combines the adaptive background-terrain budget, precise
worker pacing, bounded/wakeable queue waits and rejected-dispatch retries.
`foreground-vulkan` uses the same combination with Vulkan preference and the
game's OpenGL fallback. Both retain balanced mods, 12 full-detail chunks,
128-chunk LOD radius, two DH workers and simulation distance 12.
The adaptive target is 120 FPS to suit the first M2 Pro/ProMotion testing target.
It is a heuristic frame-time goal, not an FPS cap or a guaranteed rate. The
controller reduces background duty cycle when frames fall behind that target
or heap pressure rises. This can delay distant-terrain completion. It cannot
identify whether GPU, CPU, disk, generation or other work caused a slow frame.
Uncapped game options do not override VSync, driver or display behavior.
Start with balanced, then compare the combined OpenGL profile at equal graphics,
heap and world settings. The individual profiles remain available for isolating
each hook. Compare `foreground-first` against `foreground-vulkan` only after the
isolated Vulkan pack works on the device. Each combined profile explicitly
enables all four controls; changing `/horizon adaptive off` disables only the
adaptive controller. Other hooks require their configuration flag and restart.
Change `targetFps` in `config/horizon-adaptive.properties` to evaluate a different
goal, then restart. The recorder captures the goal, actual applied hook flags,
backend and effective video settings. Java compact headers and collector changes
remain separate optional experiments. Normal simulation/tick rules are retained.
These profiles combine measured scheduler mechanisms, not independently measured
Minecraft FPS gains that can be added together. Their interaction, visible
terrain latency, memory and frame pacing need whole-game testing on real GPUs.
+36 -1
View File
@@ -22,6 +22,41 @@ Version 0.2.0 targets Minecraft 26.3 / Fabric 0.19.5 / Java 25. This is an initi
## Interpretation limits
The exact 0.11.0 JAR (SHA-256
`c2f59ca0b2f1e356f51b41cd0d9f1d98d6b86e240f5b5a1424f27de57f0e9c4a`)
ran `foreground-first` with an explicit G1 collector, compact headers and 3 GiB
heap. Its 120-second OpenGL world capture produced 640 frame intervals and
2,401 server ticks with matching CSV counts/means and 120 RSS samples each.
All four controls and the wakeable idle path were reported enabled, with adaptive
goal 120 FPS, uncapped rendering and actual distances 12/12. World saves, all
three DH database closures, worker shutdown and client quit completed. The offline
viewer displayed the actual combined state, goal/cap distinction and collector
accounting guidance. These software-rendered checks validate integration rather
than establishing M2 FPS, RAM, LOD-completion or real-GPU visual improvements.
The same exact JAR then ran `foreground-vulkan` with all controls enabled,
adaptive goal 120, uncapped rendering and distances 12/12. The actual Vulkan
backend and driver were captured. Its 120-second recording produced 618 frame
intervals and 2,413 server ticks with matching CSV counts/means and 120 RSS
samples each. This verifies the combination on Linux software Vulkan; it does
not validate MoltenVK, all distant-geometry visuals or GPU performance on the Mac.
The exact 0.10.0 JAR (SHA-256
`101790057c969333e2d9753640d4996649e8ced27e465897d8b6beab2474ed4c`)
ran the standalone dispatch-retry profile with 3 GiB heap and compact headers.
Its 120-second recording produced 568 frame intervals and 2,407 server ticks,
with matching CSV counts/means and 120/119 RSS samples. The actual frame report
retained custom graphics preset and 12/12 render/simulation distances, enabled
the dispatcher hook, and disabled adaptive/pacing/debounce/idle hooks. The world
saved, all three DH databases closed, worker threads stopped and the client quit.
The offline viewer read actual dispatch metadata, JVM settings, RSS and CSV.
Applied-hook fixtures verify exact task completion, independent-executor progress,
recovered slots and shutdown against the actual upstream rejection path in Linux
CI; Windows and ARM macOS compile the fixture. [Results and limits](dispatch-retry.md)
describe synthetic, caller-only failure-path measurements. An unchanged client
from the same saved world did not reproduce the earlier game allocation storm;
these variable captures do not prove a repeatable game-FPS or RAM improvement.
The exact 0.9.0 JAR (SHA-256
`2c5a731d4c93df427cc6054f120a52bf9456b3343b94ad9b79de1584b14d783f`)
ran the combined queue/adaptive/pacing profile with a 3 GiB heap and compact
@@ -73,7 +108,7 @@ continued updating, and worlds/databases/workers closed cleanly. The actual
publisher queue microbenchmark passes identity/priority/count checks on Linux,
Windows and Apple Silicon macOS; its raw results and timing overshoot are documented
in [the queue experiment](debounce-queue.md). These are separate from game-FPS,
block-update-latency or RAM-reduction claims. All twelve pack overrides retain
block-update-latency or RAM-reduction claims. All fourteen pack overrides retain
simulation distance 12 and explicitly set the intended experimental flags.
The 0.6.0 resident-memory sampler ran in the actual adaptive-pacing demo world.
+5
View File
@@ -37,6 +37,11 @@ when comparing peaks. Process CPU counters cover the whole process over the
recording wall time, including native work and the integrated server; they do
not identify which thread or mod caused CPU use.
`gc_time_ms` sums collector MXBean collection-time counters. With concurrent
collectors these can include cycle duration and overlapping phases; it is not a
direct stop-the-world pause measurement. The viewer calls it collection time.
Use [JFR pause events](profiling.md) when investigating GC-induced stalls.
The native-read test exercises the real library, overlapping recordings,
background execution, idle cancellation and unavailable-reader behavior on
the build platform. This validates measurement availability, not an FPS or
+5
View File
@@ -40,6 +40,11 @@ Reports capture the actual compact object header flag and collector names.
Compare against the same workload with the flag omitted; object allocation
changes do not establish a whole-server RAM/TPS gain.
Optional collector comparison: select --gc g1 or --gc zgc on the run command.
Match world, player activity, heap, Java and header mode; use one collector at a
time. gc_time_ms includes collector cycle accounting, not a measured pause total.
Use JFR pause events for pause analysis and compare RSS/CPU/TPS together.
Compare generated reports:
python3 tools/compare_benchmarks.py baseline-ticks.json optimized-ticks.json --strict
+32
View File
@@ -0,0 +1,32 @@
# Compare the Vulkan renderer on your Mac
The `vulkan` profile asks Minecraft 26.3 to use its existing Vulkan backend, with
its original OpenGL fallback. It uses balanced mods/distances and omits Iris.
The game renderer handles the backend; this project has not written a separate
Metal renderer. The benchmark records the actual backend and driver.
The checksum-verified 26.3 manifest includes LWJGL Vulkan 3.4.3 Mac ARM64 natives.
Inspecting that official native JAR confirmed `libMoltenVK.dylib`; its paired
LWJGL loader first tries MoltenVK on macOS. [MoltenVK](https://github.com/KhronosGroup/MoltenVK)
implements Vulkan functionality over Apple's Metal framework. This verifies a
bundled route to Metal, rather than proving that a particular Mac backend starts
or performs better. No separately installed Vulkan SDK is required for the bundled
library; driver/backend compatibility still needs a real Mac test.
Import `m2-balanced` and `vulkan` as separate Prism instances. Keep Java 25 ARM64,
heap, header/collector flags, world copies, resolution, view/simulation distance
and camera route equal. Warm terrain and shader/pipeline caches before comparing.
Run at least three alternating 120-second F8 recordings. Inspect nearby blocks,
distant terrain, transparency, water, lighting and reloads as well as timings.
Use the viewer's actual backend field: OpenGL fallback is an OpenGL result.
After the isolated test works, `foreground-first` and `foreground-vulkan` compare
the two backends with the same combined terrain controls. Their 120 FPS adaptive
goal controls background duty cycle; rendering remains uncapped in pack options.
Do not add Iris or shaders to one side of an isolated backend comparison.
Linux software rendering checks reached the actual Vulkan backend, rendered the
demo world and recorded matching CSV/JSON frame/tick samples. An unavailable
Vulkan-driver test reached the menu using OpenGL fallback. These tests do not
validate MoltenVK, real GPU speeds, all visual details or macOS 27. The user's
M2 Pro remains the priority device for those checks.
@@ -17,6 +17,7 @@ import net.fabricmc.fabric.api.client.keymapping.v1.KeyMappingHelper;
import net.fabricmc.fabric.api.client.event.lifecycle.v1.ClientTickEvents;
import net.fabricmc.loader.api.FabricLoader;
import net.minecraft.client.Minecraft;
import net.minecraft.client.Options;
import net.minecraft.client.KeyMapping;
import net.minecraft.network.chat.Component;
import static net.fabricmc.fabric.api.client.command.v2.ClientCommands.argument;
@@ -100,6 +101,7 @@ public final class HorizonClient implements ClientModInitializer {
context.put("simulation_distance", client.options.simulationDistance().get());
context.put("vsync", client.options.enableVsync().get());
context.put("fps_limit", client.options.framerateLimit().get());
context.put("fps_limit_unlimited", client.options.framerateLimit().get() >= Options.UNLIMITED_FRAMERATE_CUTOFF);
context.put("framebuffer_width", client.getWindow().getWidth());
context.put("framebuffer_height", client.getWindow().getHeight());
var device = RenderSystem.getDevice().getDeviceInfo();
@@ -1,7 +1,7 @@
{
"schemaVersion": 1,
"id": "horizonlab",
"version": "0.10.0",
"version": "0.11.0",
"name": "Horizon Lab",
"description": "Client frame pacing and server tick measurements for reproducible optimization experiments.",
"authors": [
+8
View File
@@ -58,6 +58,14 @@ PROFILES
This is a separate experiment: adaptive, worker-pacing and queue-wait are off.
Set enabled=false in config/horizon-dispatch-retry.properties and restart
to disable it. A synthetic applied-hook test checks exact task completion.
- foreground-first: combines adaptive terrain, precise pacing, wakeable queue
waits and bounded rejected dispatch retries. Uses OpenGL and a 120 FPS
adaptive goal. This is not an FPS cap; terrain may take longer to appear.
- foreground-vulkan: the same combined controls with Vulkan preference and
OpenGL fallback. On Apple Silicon the game bundles a MoltenVK library.
Test the isolated vulkan pack first, then compare both combined profiles
using equal worlds, heap, resolution and camera routes. Check the actual
backend in the benchmark viewer; a fallback is not a Vulkan result.
- low-memory: 8 full-detail chunks, no LOD. Set maximum memory 3072 MiB.
For a matched comparison set both instances to the same render distance.
- far-distance: 12 full-detail plus 256 LOD chunks. Start with 8192 MiB ONLY on
+1 -1
View File
@@ -3,7 +3,7 @@
"minecraft": "26.3",
"java_major": 25,
"fabric_loader": "0.19.5",
"pack_version": "0.10.0",
"pack_version": "0.11.0",
"options_data_version": 5023,
"projects": {
"sodium": {
+14 -1
View File
@@ -54,7 +54,7 @@ class PackTests(unittest.TestCase):
for name, profile in build_pack.PROFILES.items():
files = build_pack.overrides(name, b"original fixture")
dispatch = files["config/horizon-dispatch-retry.properties"].decode()
self.assertEqual(dispatch, "enabled=true\n" if name == "dispatch-retry" else "enabled=false\n")
self.assertEqual(dispatch, "enabled=true\n" if name in {"dispatch-retry", "foreground-first", "foreground-vulkan"} else "enabled=false\n")
if name == "dispatch-retry":
for setting in ["worker-pacing", "queue-wait"]:
self.assertEqual(files[f"config/horizon-{setting}.properties"], b"enabled=false\n")
@@ -62,6 +62,19 @@ class PackTests(unittest.TestCase):
self.assertEqual(files["config/DistantHorizons.toml"],
build_pack.overrides("m2-balanced", b"original fixture")["config/DistantHorizons.toml"])
def test_combined_profiles_enable_all_controls_without_changing_simulation(self):
for name, backend in [("foreground-first", "opengl"), ("foreground-vulkan", "vulkan")]:
files = build_pack.overrides(name, b"original fixture")
self.assertEqual(files["config/horizon-adaptive.properties"], b"enabled=true\ntargetFps=120\n")
for key in ["worker-pacing", "queue-wait", "dispatch-retry"]:
self.assertEqual(files[f"config/horizon-{key}.properties"], b"enabled=true\n")
options = dict(line.split(":", 1) for line in files["options.txt"].decode().splitlines())
self.assertEqual(options["simulationDistance"], "12")
self.assertEqual(options["renderDistance"], "12")
self.assertEqual(options["maxFps"], "260")
self.assertEqual(json.loads(options["preferredGraphicsBackend"]), backend)
self.assertNotIn("config/iris.properties", files)
def test_wrong_game_or_download_path_is_rejected(self):
for field, value in [("game_versions", ["1.21"]), ("filename", "../bad.jar"), ("url", "http://unsafe.invalid")]:
lock = copy.deepcopy(build_pack.read_lock())
+5
View File
@@ -91,6 +91,7 @@ function compareReports(a,b){
rows.push({label:'GC collection time / second',key:'gc_collection_ms_per_second',unit:'ms/s',baseline:av,candidate:bv,change_percent:av?100*(bv/av-1):null});
}
warnings.push('Memory peaks are sampled and can miss short spikes. RSS includes resident JVM and native pages, but is not a complete GPU/unified-memory measurement. Heap/nonheap measure JVM pools only. CPU includes the whole process and integrated server.');
warnings.push('GC collection time comes from collector MXBean counters. It may include concurrent cycles and overlapping collector phases; it is not a measured stop-the-world pause total. Use JFR pause events for that comparison.');
for(const [label,r] of [['Baseline',a],['Candidate',b]])if(number(r.process_rss_last_sample_age_ms)&&r.process_rss_last_sample_age_ms>2500)warnings.push(label+' resident-memory sample is '+format(r.process_rss_last_sample_age_ms/1000)+' seconds old. The background reader may be slow or unavailable.');
return{kind:a.kind,differences,warnings,metrics:rows};
}
@@ -100,6 +101,10 @@ function renderDetails(side){
const r=entry.report,fields=[['Profile',r.profile||r.label||'Unlabeled'],['Recording',format(r.elapsed_seconds)+' s · '+format(r.stats.count)+' samples'],['Platform',[r.os,r.architecture].filter(Boolean).join(' ')||'Not captured']];
if(r.kind==='frames')fields.push(['Graphics',[r.gpu,r.graphics_backend].filter(Boolean).join(' · ')||'Not captured'],['Window',number(r.framebuffer_width)?r.framebuffer_width+' × '+r.framebuffer_height:'Not captured'],['Experiments','Adaptive: '+(r.adaptive_budget_enabled?'on':'off')+' · precise pacing: '+(r.precise_worker_pacing?'on':'off')+' · queue wait: '+(r.debounce_queue_wait?'on':'off')]);
fields.push(['JVM', 'Compact headers: '+(r.compact_object_headers===true?'on':r.compact_object_headers===false?'off':'not captured')+' · GC: '+(Array.isArray(r.garbage_collectors)?r.garbage_collectors.join(', '):'not captured')]);
if(r.kind==='frames'){
fields.push(['Frame limit',r.fps_limit_unlimited===true||(r.minecraft==='26.3'&&number(r.fps_limit)&&r.fps_limit>=260)?'Uncapped':number(r.fps_limit)?r.fps_limit+' FPS':'not captured']);
if(r.adaptive_budget_enabled&&number(r.adaptive_target_fps))fields.push(['Adaptive frame goal',r.adaptive_target_fps+' FPS (not a cap)']);
}
fields.push(['Rejected dispatch retry', r.dispatch_retry_backoff===true?'on':r.dispatch_retry_backoff===false?'off':'not captured']);
fields.push(['Wakeable queue idle', r.wakeable_idle_wait===true?'on':r.wakeable_idle_wait===false?'off':'not captured']);
for(const [label,value]of fields)target.append(text('dt',label),text('dd',value));
+2
View File
@@ -31,6 +31,8 @@ def test_metrics():
subprocess.run(["java", "-cp", str(destination), "dev.horizonlab.metrics.MetricsTest"], check=True)
subprocess.run(["java", "-XX:+UseCompactObjectHeaders", "-cp", str(destination),
"dev.horizonlab.metrics.MetricsTest"], check=True)
subprocess.run(["java", "-Xms32M", "-Xmx256M", "-XX:+UseZGC", "-XX:+UseCompactObjectHeaders", "-cp", str(destination),
"dev.horizonlab.metrics.MetricsTest"], check=True)
budget_sources = [SOURCE / "main/java/dev/horizonlab/budget/TerrainBudget.java",
SOURCE / "test/java/dev/horizonlab/budget/TerrainBudgetTest.java"]
subprocess.run(["javac", "--release", "25", "-d", str(destination), *map(str, budget_sources)], check=True)
+3 -1
View File
@@ -24,6 +24,8 @@ PROFILES = {
"queue-wait": {"title": "Debounce queue wait experiment", "mods": CORE + ["distanthorizons"], "render": 12, "lod": 128, "heap": 6144, "queue_wait": True},
"adaptive-queue": {"title": "Adaptive pacing and queue wait experiment", "mods": CORE + ["distanthorizons"], "render": 12, "lod": 128, "heap": 6144, "adaptive": True, "pacing": True, "queue_wait": True},
"dispatch-retry": {"title": "Rejected dispatch retry experiment", "mods": CORE + ["distanthorizons"], "render": 12, "lod": 128, "heap": 6144, "dispatch_retry": True},
"foreground-first": {"title": "Foreground-first combined experiment", "mods": CORE + ["distanthorizons"], "render": 12, "lod": 128, "heap": 6144, "adaptive": True, "pacing": True, "queue_wait": True, "dispatch_retry": True, "target_fps": 120},
"foreground-vulkan": {"title": "Foreground-first Vulkan experiment", "mods": CORE + ["distanthorizons"], "render": 12, "lod": 128, "heap": 6144, "adaptive": True, "pacing": True, "queue_wait": True, "dispatch_retry": True, "target_fps": 120, "backend": "vulkan"},
"low-memory": {"title": "Low memory", "mods": CORE, "render": 8, "lod": 0, "heap": 3072},
"far-distance": {"title": "Far distance", "mods": CORE + ["distanthorizons"], "render": 12, "lod": 256, "heap": 8192},
"cinematic": {"title": "Cinematic (optional)", "mods": CORE + ["distanthorizons", "iris", "complementary-reimagined"], "render": 12, "lod": 128, "heap": 6144},
@@ -80,7 +82,7 @@ def overrides(profile_name, mod_bytes):
experiment = any(profile.get(key, False) for key in ["adaptive", "pacing", "queue_wait", "dispatch_retry"])
config = ROOT / "packs/config" / f"{'m2-balanced' if experiment else profile_name}.toml"
files["config/horizon-adaptive.properties"] = (
f"enabled={str(profile.get('adaptive', False)).lower()}\ntargetFps=60\n").encode()
f"enabled={str(profile.get('adaptive', False)).lower()}\ntargetFps={profile.get('target_fps', 60)}\n").encode()
files["config/horizon-worker-pacing.properties"] = (
f"enabled={str(profile.get('pacing', False)).lower()}\n").encode()
files["config/horizon-queue-wait.properties"] = (
+11 -1
View File
@@ -42,6 +42,15 @@ def install_original_mod(source, mod_dir, filename=None):
CACHE = ROOT / ".local" / "runtime"
def jvm_options(heap, compact_headers=False, collector="default"):
options = ["-Xms512M", f"-Xmx{heap}"]
if compact_headers:
options.append("-XX:+UseCompactObjectHeaders")
if collector != "default":
options.append({"g1": "-XX:+UseG1GC", "zgc": "-XX:+UseZGC"}[collector])
return options
def fetch(url, path, checksum, algorithm="sha1"):
for attempt in range(3):
try:
@@ -138,6 +147,7 @@ def main():
parser.add_argument("--instance", type=Path, default=CACHE / "demo")
parser.add_argument("--heap", default="4G")
parser.add_argument("--compact-headers", action="store_true", help="Enable Java 25 compact object headers for this demo process")
parser.add_argument("--gc", choices=["default", "g1", "zgc"], default="default", help="Select a collector for this process; default leaves JVM policy in use")
args = parser.parse_args()
if not re.fullmatch(r"[1-9][0-9]*[MGmg]", args.heap):
parser.error("--heap must look like 4G")
@@ -146,7 +156,7 @@ def main():
return
instance = args.instance.resolve()
instance.mkdir(parents=True, exist_ok=True)
command = ["java", "-Xms512M", f"-Xmx{args.heap}", *(["-XX:+UseCompactObjectHeaders"] if args.compact_headers else []),
command = ["java", *jvm_options(args.heap, args.compact_headers, args.gc),
"--enable-native-access=ALL-UNNAMED",
f"-Djava.library.path={natives}", f"-Dorg.lwjgl.system.SharedLibraryExtractPath={natives}",
"-Dfabric.log.disableAnsi=true", "-cp", os.pathsep.join(map(str, jars)),
+2 -1
View File
@@ -51,6 +51,7 @@ def main():
parser.add_argument("--heap", default="4G")
parser.add_argument("--baseline", action="store_true", help="Use a separate instance for the baseline")
parser.add_argument("--compact-headers", action="store_true", help="Enable Java 25 compact object headers for this server process")
parser.add_argument("--gc", choices=["default", "g1", "zgc"], default="default", help="Select a collector for this process; default leaves JVM policy in use")
args = parser.parse_args()
if not re.fullmatch(r"[1-9][0-9]*[MGmg]", args.heap):
parser.error("--heap must look like 4G")
@@ -67,7 +68,7 @@ def main():
eula.write_text("# Read https://www.minecraft.net/eula then change this value yourself.\neula=false\n")
print(f"Server prepared. Read Minecraft's EULA and set eula=true in {eula} before running.")
return
subprocess.run(["java", "-Xms512M", f"-Xmx{args.heap}", *(["-XX:+UseCompactObjectHeaders"] if args.compact_headers else []),
subprocess.run(["java", *runtime.jvm_options(args.heap, args.compact_headers, args.gc),
"--enable-native-access=ALL-UNNAMED",
"-cp", os.pathsep.join(map(str, libraries)), "net.fabricmc.loader.impl.launch.knot.KnotServer",
"--nogui"], cwd=instance, check=True)
+7 -3
View File
@@ -1,6 +1,7 @@
#!/usr/bin/env python3
"""Verify the streaming JFR reader against the JDK's own JSON exporter."""
import argparse
import json
from pathlib import Path
import subprocess
@@ -9,10 +10,11 @@ ROOT = Path(__file__).resolve().parents[1]
BUILD = ROOT / ".local/jfr-tests"
def main():
def main(collector="default"):
BUILD.mkdir(parents=True, exist_ok=True)
fixture = BUILD / "fixture.jfr"
subprocess.run(["java", "-Xmx128M", str(ROOT / "tests/java/JfrFixture.java"), str(fixture)], check=True)
gc_flags = ["-XX:+UseZGC"] if collector == "zgc" else []
subprocess.run(["java", "-Xmx128M", *gc_flags, str(ROOT / "tests/java/JfrFixture.java"), str(fixture)], check=True)
summary = json.loads(subprocess.check_output(
["java", str(ROOT / "tools/JfrSummary.java"), str(fixture)], text=True))
events = json.loads(subprocess.check_output(
@@ -65,4 +67,6 @@ def main():
if __name__ == "__main__":
main()
parser = argparse.ArgumentParser(description=__doc__)
parser.add_argument("--gc", choices=["default", "zgc"], default="default")
main(parser.parse_args().gc)