Skip to main content
The hot command produces dual-ranked tables showing the hottest methods in your profile. It’s the go-to command for identifying performance bottlenecks.

​
Usage

ap-query hot <file> [flags]

​
What It Shows

Two ranked tables:
  1. RANK BY SELF TIME - Methods appearing at the top of stacks (doing actual work)
  2. RANK BY TOTAL TIME - Methods appearing anywhere in stacks (orchestrating work)
Each table shows:
  • Method name
  • SELF% - Percentage of samples where this is the leaf frame
  • TOTAL% - Percentage of samples containing this method anywhere
  • SAMPLES - Raw sample count for the ranking dimension

​
Flags

​
--top
int
default:"10"
Limit output rows in each table. Controls how many methods are shown.
​
--fqn
boolean
Show fully-qualified names (full package paths) instead of short names.
​
--assert-below
float
default:"0"
Exit with status 1 if the top method’s self% >= threshold. Useful for CI gates to catch performance regressions.Example: --assert-below 10.0 fails if any method consumes ≥10% self-time.

​
Shared Flags

​
-e, --event
string
default:"cpu"
Event type: cpu, wall, alloc, lock, or hardware counter name.
​
-t, --thread
string
Filter to threads containing this substring.
​
--from
duration
Start of time window (JFR only). Examples: 5s, 1m30s, 2m.
​
--to
duration
End of time window (JFR only). Examples: 10s, 2m, 1m30s.
​
--no-idle
boolean
Remove idle leaf frames from analysis.

​
Examples

​
Basic Usage

ap-query hot profile.jfr

=== RANK BY SELF TIME ===
METHOD                                              SELF%   TOTAL%   SAMPLES
HashMap.hash                                         8.3%   12.1%       998
ArrayList.grow                                       6.2%    6.2%       747
String.hashCode                                      5.1%    8.9%       614
UserService.validateInput                            4.8%    4.8%       578
JSONParser.parse                                     4.2%    9.3%       506
ByteBuffer.put                                       3.9%    3.9%       469
HashMap.resize                                       3.1%    3.1%       373
StringBuilder.append                                 2.8%    2.8%       337
Arrays.copyOf                                        2.3%    2.3%       277
MessageDigest.update                                 2.1%    2.1%       253

=== RANK BY TOTAL TIME ===
METHOD                                              SELF%   TOTAL%   SAMPLES
UserService.processRequest                           2.1%   45.3%      5456
RequestHandler.handle                                1.2%   38.7%      4661
HashMap.put                                          1.8%   18.4%      2215
HashMap.hash                                         8.3%   12.1%       998
JSONParser.parse                                     4.2%    9.3%       506
String.hashCode                                      5.1%    8.9%       614
ArrayList.add                                        0.9%    7.8%       939
HashMap.get                                          1.3%    6.4%       770
ArrayList.grow                                       6.2%    6.2%       747
UserService.validateInput                            4.8%    4.8%       578

​
Show More Methods

ap-query hot profile.jfr --top 25
Show top 25 methods in each table instead of the default 10.

​
Fully-Qualified Names

ap-query hot profile.jfr --fqn

=== RANK BY SELF TIME ===
METHOD                                              SELF%   TOTAL%   SAMPLES
java.util.HashMap.hash                               8.3%   12.1%       998
java.util.ArrayList.grow                             6.2%    6.2%       747
java.lang.String.hashCode                            5.1%    8.9%       614
com.example.UserService.validateInput                4.8%    4.8%       578
...

​
Allocation Profiling

ap-query hot profile.jfr --event alloc
Show methods allocating the most memory.

​
CI Performance Gate

ap-query hot profile.jfr --assert-below 5.0
Fails (exit code 1) if any method has self-time ≥ 5%: 
ERROR: ASSERT FAILED: HashMap.hash self=8.3% >= threshold 5.0%
Use in CI pipelines to detect performance regressions where a single method becomes too dominant.

​
Wall-Clock Time for Blocking Analysis

ap-query hot profile.jfr --event wall
Show where threads spend wall-clock time (includes blocking/waiting).

​
Filter to Specific Threads

ap-query hot profile.jfr --thread worker
Analyze only threads with “worker” in their name.

​
Time Window

ap-query hot profile.jfr --from 30s --to 1m
Analyze only the 30-60 second window of the recording.

​
Remove Idle Frames

ap-query hot profile.jfr --event wall --no-idle
Exclude idle/waiting frames (useful for wall-clock profiles).

​
Understanding Self vs Total

High Self Time

Method does significant work itself. Target for micro-optimization or algorithm improvement.

High Total Time

Method orchestrates expensive operations. May not need optimization itself, but investigate what it calls.

Low Self, High Total

Wrapper/orchestrator method. Use tree command to see what it calls.

High Self, High Total

Method both does work and calls expensive code. Prime optimization target.

​
Reading the Output

HashMap.hash    8.3%   12.1%    998
  • HashMap.hash appeared at the top of the stack in 8.3% of samples (doing work)
  • It appeared somewhere in the stack in 12.1% of samples (called by others too)
  • It was seen in 998 samples (absolute count)
SELF% measures direct cost. TOTAL% measures direct + indirect cost (including everything this method calls).

​
When to Use

Finding Bottlenecks

Identify which methods consume the most CPU or allocate the most memory

Before Deep-Dive

Get method names to investigate with tree, trace, or callers commands

Comparing Profiles

Run on multiple profiles to see ranking changes over time

CI Validation

Use --assert-below to catch performance regressions automatically

​
See Also

  • info - Overview including hot methods plus drill-down
  • tree - See what a hot method calls
  • trace - Find the hottest path from a method to leaf
  • callers - See who calls a hot method

Build docs developers (and LLMs) love

Get started for freeTalk to us