Skip to content

Commit

Permalink
Extra resources (#17229)
Browse files Browse the repository at this point in the history 
This recreates a [closed PR](#13996) to implement extra resources which we're hoping to use in lowRISC/opentitan#16436

Fixes:#16817

Closes #16785.

PiperOrigin-RevId: 498557024
Change-Id: I60d8f8f4a4a02748147cabb4cd60a2a9b95a2c68

Co-authored-by: kshyanashree <109167932+kshyanashree@users.noreply.github.com>
  • Loading branch information
@ShreeM01
Drew Macrae and ShreeM01 authored Jan 19, 2023
1 parent 0849b3b commit bcfc7f7
Show file tree
Hide file tree
Showing 10 changed files with 338 additions and 24 deletions.
24 changes: 24 additions & 0 deletions src/main/java/com/google/devtools/build/lib/actions/ExecutionRequirements.java
View file
Original file line number Diff line number Diff line change
Expand Up @@ -154,6 +154,30 @@ public String parseIfMatches(String tag) throws ValidationException {
return null;
});

/** How many extra resources an action requires for execution. */
public static final ParseableRequirement RESOURCES =
ParseableRequirement.create(
"resources:<str>:<float>",
Pattern.compile("resources:(.+:.+)"),
s -> {
Preconditions.checkNotNull(s);

int splitIndex = s.indexOf(":");
String resourceCount = s.substring(splitIndex + 1);
float value;
try {
value = Float.parseFloat(resourceCount);
} catch (NumberFormatException e) {
return "can't be parsed as a float";
}

if (value < 0) {
return "can't be negative";
}

return null;
});

/** If an action supports running in persistent worker mode. */
public static final String SUPPORTS_WORKERS = "supports-workers";

Expand Down
88 changes: 82 additions & 6 deletions src/main/java/com/google/devtools/build/lib/actions/ResourceManager.java
View file
Original file line number Diff line number Diff line change
Expand Up @@ -27,8 +27,13 @@
import com.google.devtools.build.lib.worker.WorkerPool;
import java.io.IOException;
import java.util.Deque;
import java.util.HashMap;
import java.util.HashSet;
import java.util.Iterator;
import java.util.LinkedList;
import java.util.Map;
import java.util.NoSuchElementException;
import java.util.Set;
import java.util.concurrent.CountDownLatch;
import javax.annotation.Nullable;

Expand Down Expand Up @@ -171,14 +176,16 @@ public static ResourceManager instance() {
// definition in the ResourceSet class.
private double usedRam;

// Used amount of extra resources. Corresponds to the extra resource
// definition in the ResourceSet class.
private Map<String, Float> usedExtraResources;

// Used local test count. Corresponds to the local test count definition in the ResourceSet class.
private int usedLocalTestCount;

/** If set, local-only actions are given priority over dynamically run actions. */
private boolean prioritizeLocalActions;

private ResourceManager() {}

@VisibleForTesting
public static ResourceManager instanceForTestingOnly() {
return new ResourceManager();
Expand All @@ -192,6 +199,7 @@ public static ResourceManager instanceForTestingOnly() {
public synchronized void resetResourceUsage() {
usedCpu = 0;
usedRam = 0;
usedExtraResources = new HashMap<>();
usedLocalTestCount = 0;
for (Pair<ResourceSet, LatchWithWorker> request : localRequests) {
request.second.latch.countDown();
Expand Down Expand Up @@ -286,6 +294,20 @@ private Worker incrementResources(ResourceSet resources)
throws IOException, InterruptedException {
usedCpu += resources.getCpuUsage();
usedRam += resources.getMemoryMb();

resources
.getExtraResourceUsage()
.entrySet()
.forEach(
resource -> {
String key = (String) resource.getKey();
float value = resource.getValue();
if (usedExtraResources.containsKey(key)) {
value += (float) usedExtraResources.get(key);
}
usedExtraResources.put(key, value);
});

usedLocalTestCount += resources.getLocalTestCount();

if (resources.getWorkerKey() != null) {
Expand All @@ -298,6 +320,7 @@ private Worker incrementResources(ResourceSet resources)
public synchronized boolean inUse() {
return usedCpu != 0.0
|| usedRam != 0.0
|| !usedExtraResources.isEmpty()
|| usedLocalTestCount != 0
|| !localRequests.isEmpty()
|| !dynamicWorkerRequests.isEmpty()
Expand Down Expand Up @@ -357,7 +380,7 @@ public void acquireResourceOwnership() {
* wait.
*/
private synchronized LatchWithWorker acquire(ResourceSet resources, ResourcePriority priority)
throws IOException, InterruptedException {
throws IOException, InterruptedException, NoSuchElementException {
if (areResourcesAvailable(resources)) {
Worker worker = incrementResources(resources);
return new LatchWithWorker(/* latch= */ null, worker);
Expand Down Expand Up @@ -405,6 +428,7 @@ private boolean release(ResourceSet resources, @Nullable Worker worker)
private synchronized void releaseResourcesOnly(ResourceSet resources) {
usedCpu -= resources.getCpuUsage();
usedRam -= resources.getMemoryMb();

usedLocalTestCount -= resources.getLocalTestCount();

// TODO(bazel-team): (2010) rounding error can accumulate and value below can end up being
Expand All @@ -416,6 +440,19 @@ private synchronized void releaseResourcesOnly(ResourceSet resources) {
if (usedRam < epsilon) {
usedRam = 0;
}

Set<String> toRemove = new HashSet<>();
for (Map.Entry<String, Float> resource : resources.getExtraResourceUsage().entrySet()) {
String key = (String) resource.getKey();
float value = (float) usedExtraResources.get(key) - resource.getValue();
usedExtraResources.put(key, value);
if (value < epsilon) {
toRemove.add(key);
}
}
for (String key : toRemove) {
usedExtraResources.remove(key);
}
}

private synchronized boolean processAllWaitingThreads() throws IOException, InterruptedException {
Expand Down Expand Up @@ -454,9 +491,35 @@ private synchronized void processWaitingThreads(Deque<Pair<ResourceSet, LatchWit
}
}

/** Throws an exception if requested extra resource isn't being tracked */
private void assertExtraResourcesTracked(ResourceSet resources) throws NoSuchElementException {
for (Map.Entry<String, Float> resource : resources.getExtraResourceUsage().entrySet()) {
String key = (String) resource.getKey();
if (!availableResources.getExtraResourceUsage().containsKey(key)) {
throw new NoSuchElementException(
"Resource " + key + " is not tracked in this resource set.");
}
}
}

/** Return true iff all requested extra resources are considered to be available. */
private boolean areExtraResourcesAvailable(ResourceSet resources) throws NoSuchElementException {
for (Map.Entry<String, Float> resource : resources.getExtraResourceUsage().entrySet()) {
String key = (String) resource.getKey();
float used = (float) usedExtraResources.getOrDefault(key, 0f);
float requested = resource.getValue();
float available = availableResources.getExtraResourceUsage().get(key);
float epsilon = 0.0001f; // Account for possible rounding errors.
if (requested != 0.0 && used != 0.0 && requested + used > available + epsilon) {
return false;
}
}
return true;
}

// Method will return true if all requested resources are considered to be available.
@VisibleForTesting
boolean areResourcesAvailable(ResourceSet resources) {
boolean areResourcesAvailable(ResourceSet resources) throws NoSuchElementException {
Preconditions.checkNotNull(availableResources);
// Comparison below is robust, since any calculation errors will be fixed
// by the release() method.
Expand All @@ -472,7 +535,15 @@ boolean areResourcesAvailable(ResourceSet resources) {
workerKey == null
|| (activeWorkers < availableWorkers && workerPool.couldBeBorrowed(workerKey));

if (usedCpu == 0.0 && usedRam == 0.0 && usedLocalTestCount == 0 && workerIsAvailable) {
// We test for tracking of extra resources whenever acquired and throw an
// exception before acquiring any untracked resource.
assertExtraResourcesTracked(resources);

if (usedCpu == 0.0
&& usedRam == 0.0
&& usedExtraResources.isEmpty()
&& usedLocalTestCount == 0
&& workerIsAvailable) {
return true;
}
// Use only MIN_NECESSARY_???_RATIO of the resource value to check for
Expand Down Expand Up @@ -503,7 +574,12 @@ boolean areResourcesAvailable(ResourceSet resources) {
localTestCount == 0
|| usedLocalTestCount == 0
|| usedLocalTestCount + localTestCount <= availableLocalTestCount;
return cpuIsAvailable && ramIsAvailable && localTestCountIsAvailable && workerIsAvailable;
boolean extraResourcesIsAvailable = areExtraResourcesAvailable(resources);
return cpuIsAvailable
&& ramIsAvailable
&& extraResourcesIsAvailable
&& localTestCountIsAvailable
&& workerIsAvailable;
}

@VisibleForTesting
Expand Down
63 changes: 58 additions & 5 deletions src/main/java/com/google/devtools/build/lib/actions/ResourceSet.java
View file
Original file line number Diff line number Diff line change
Expand Up @@ -14,7 +14,9 @@

package com.google.devtools.build.lib.actions;

import com.google.common.base.Joiner;
import com.google.common.base.Splitter;
import com.google.common.collect.ImmutableMap;
import com.google.common.primitives.Doubles;
import com.google.devtools.build.lib.concurrent.ThreadSafety.Immutable;
import com.google.devtools.build.lib.util.OS;
Expand Down Expand Up @@ -43,6 +45,12 @@ public class ResourceSet implements ResourceSetOrBuilder {
/** The number of CPUs, or fractions thereof. */
private final double cpuUsage;

/**
* Map of extra resources (for example: GPUs, embedded boards, ...) mapping name of the resource
* to a value.
*/
private final ImmutableMap<String, Float> extraResourceUsage;

/** The number of local tests. */
private final int localTestCount;

Expand All @@ -51,8 +59,18 @@ public class ResourceSet implements ResourceSetOrBuilder {

private ResourceSet(
double memoryMb, double cpuUsage, int localTestCount, @Nullable WorkerKey workerKey) {
this(memoryMb, cpuUsage, ImmutableMap.of(), localTestCount, workerKey);
}

private ResourceSet(
double memoryMb,
double cpuUsage,
@Nullable ImmutableMap<String, Float> extraResourceUsage,
int localTestCount,
@Nullable WorkerKey workerKey) {
this.memoryMb = memoryMb;
this.cpuUsage = cpuUsage;
this.extraResourceUsage = extraResourceUsage;
this.localTestCount = localTestCount;
this.workerKey = workerKey;
}
Expand Down Expand Up @@ -83,21 +101,51 @@ public static ResourceSet createWithLocalTestCount(int localTestCount) {
}

/**
* Returns a new ResourceSet with the provided values for memoryMb, cpuUsage, ioUsage, and
* localTestCount. Most action resource definitions should use {@link #createWithRamCpu} or {@link
* Returns a new ResourceSet with the provided values for memoryMb, cpuUsage, and localTestCount.
* Most action resource definitions should use {@link #createWithRamCpu} or {@link
* #createWithLocalTestCount(int)}. Use this method primarily when constructing ResourceSets that
* represent available resources.
*/
public static ResourceSet create(double memoryMb, double cpuUsage, int localTestCount) {
return createWithWorkerKey(memoryMb, cpuUsage, localTestCount, /* workerKey= */ null);
return ResourceSet.createWithWorkerKey(
memoryMb, cpuUsage, ImmutableMap.of(), localTestCount, /* workerKey= */ null);
}

/**
* Returns a new ResourceSet with the provided values for memoryMb, cpuUsage, extraResources, and
* localTestCount. Most action resource definitions should use {@link #createWithRamCpu} or {@link
* #createWithLocalTestCount(int)}. Use this method primarily when constructing ResourceSets that
* represent available resources.
*/
public static ResourceSet create(
double memoryMb,
double cpuUsage,
ImmutableMap<String, Float> extraResourceUsage,
int localTestCount) {
return createWithWorkerKey(
memoryMb, cpuUsage, extraResourceUsage, localTestCount, /* workerKey= */ null);
}

public static ResourceSet createWithWorkerKey(
double memoryMb, double cpuUsage, int localTestCount, WorkerKey workerKey) {
if (memoryMb == 0 && cpuUsage == 0 && localTestCount == 0 && workerKey == null) {
return ResourceSet.createWithWorkerKey(
memoryMb, cpuUsage, /* extraResourceUsage= */ ImmutableMap.of(), localTestCount, workerKey);
}

public static ResourceSet createWithWorkerKey(
double memoryMb,
double cpuUsage,
ImmutableMap<String, Float> extraResourceUsage,
int localTestCount,
WorkerKey workerKey) {
if (memoryMb == 0
&& cpuUsage == 0
&& extraResourceUsage.size() == 0
&& localTestCount == 0
&& workerKey == null) {
return ZERO;
}
return new ResourceSet(memoryMb, cpuUsage, localTestCount, workerKey);
return new ResourceSet(memoryMb, cpuUsage, extraResourceUsage, localTestCount, workerKey);
}

/** Returns the amount of real memory (resident set size) used in MB. */
Expand All @@ -124,6 +172,10 @@ public double getCpuUsage() {
return cpuUsage;
}

public ImmutableMap<String, Float> getExtraResourceUsage() {
return extraResourceUsage;
}

/** Returns the local test count used. */
public int getLocalTestCount() {
return localTestCount;
Expand All @@ -138,6 +190,7 @@ public String toString() {
+ "CPU: "
+ cpuUsage
+ "\n"
+ Joiner.on("\n").withKeyValueSeparator(": ").join(extraResourceUsage.entrySet())
+ "Local tests: "
+ localTestCount
+ "\n";
Expand Down
Loading

0 comments on commit bcfc7f7

Please sign in to comment.