Skip to content

Profiling

Quick profiling in your terminal

Note

This is only meant to be used for quick profiling or programmatically accessing the profiling results. For more detailed and GUI friendly profiling proceed to the next section.

Simply replace the use of Base.@time or Base.@timed with Reactant.Profiler.@time or Reactant.Profiler.@timed. We will automatically compile the function if it is not already a Reactant compiled function (with sync=true).

julia
using Reactant

x = Reactant.to_rarray(randn(Float32, 100, 2))
W = Reactant.to_rarray(randn(Float32, 10, 100))
b = Reactant.to_rarray(randn(Float32, 10))

linear(x, W, b) = (W * x) .+ b

Reactant.@time linear(x, W, b)
Debug: Profiling directory: /home/runner/.julia/scratchspaces/3c362404-f566-11ee-1572-e11a4b42c853/reactant_profiling/jl_Qc4Pp3
@ Reactant.Profiler ~/work/Reactant.jl/Reactant.jl/src/Profiler.jl:626
WARNING: All log messages before absl::InitializeLog() is called are written to STDERR
I0000 00:00:1789720092.939982    3980 profiler_session.cc:171] Profiler session initializing.
I0000 00:00:1789720092.940051    3980 profiler_session.cc:186] Profiler session started.
I0000 00:00:1789720092.940393    3980 profiler_session.cc:134] Profiler session collecting data.
I0000 00:00:1789720092.940993    3980 save_profile.cc:205] Collecting XSpace to repository: /home/runner/.julia/scratchspaces/3c362404-f566-11ee-1572-e11a4b42c853/reactant_profiling/jl_Qc4Pp3/plugins/profile/2026_09_18_08_28_12/runnervmlun5p.xplane.pb
I0000 00:00:1789720092.941164    3980 save_profile.cc:178] Creating directory: /home/runner/.julia/scratchspaces/3c362404-f566-11ee-1572-e11a4b42c853/reactant_profiling/jl_Qc4Pp3/plugins/profile/2026_09_18_08_28_12

I0000 00:00:1789720092.941260    3980 save_profile.cc:184] Dumped gzipped tool data for trace.json.gz to /home/runner/.julia/scratchspaces/3c362404-f566-11ee-1572-e11a4b42c853/reactant_profiling/jl_Qc4Pp3/plugins/profile/2026_09_18_08_28_12/runnervmlun5p.trace.json.gz
I0000 00:00:1789720092.941278    3980 profiler_session.cc:217] Profiler session tear down.
Debug: Starting XProf gRPC server...
@ Reactant.Profiler ~/work/Reactant.jl/Reactant.jl/src/Profiler.jl:598
Debug: Initializing XProf stubs for worker service at 0.0.0.0:37959
@ Reactant.Profiler ~/work/Reactant.jl/Reactant.jl/src/Profiler.jl:397
I0000 00:00:1789720092.956267    3980 stub_factory.cc:163] Created gRPC channel for address: 0.0.0.0:37959
Debug: Starting XProf gRPC server on port 37959
@ Reactant.Profiler ~/work/Reactant.jl/Reactant.jl/src/Profiler.jl:413
I0000 00:00:1789720092.956666    3980 grpc_server.cc:94] Server listening on 0.0.0.0:37959 with max_concurrent_requests 1
I0000 00:00:1789720092.966627    3980 xplane_to_tools_data_with_profile_processor.cc:142] serving tool: memory_profile with options: {} using ProfileProcessor session_id: /home/runner/.julia/scratchspaces/3c362404-f566-11ee-1572-e11a4b42c853/reactant_profiling/jl_Qc4Pp3/plugins/profile/2026_09_18_08_28_12
I0000 00:00:1789720092.966651    3980 xplane_to_tools_data_with_profile_processor.cc:165] Using local processing for tool: memory_profile
I0000 00:00:1789720092.966654    3980 memory_profile_processor.cc:47] Processing memory profile for host: runnervmlun5p
I0000 00:00:1789720092.966983    3980 xplane_to_tools_data_with_profile_processor.cc:170] Total time for tool memory_profile: 338.16us session_id: /home/runner/.julia/scratchspaces/3c362404-f566-11ee-1572-e11a4b42c853/reactant_profiling/jl_Qc4Pp3/plugins/profile/2026_09_18_08_28_12
I0000 00:00:1789720092.980300    3980 xplane_to_tools_data_with_profile_processor.cc:142] serving tool: op_profile with options: {} using ProfileProcessor session_id: /home/runner/.julia/scratchspaces/3c362404-f566-11ee-1572-e11a4b42c853/reactant_profiling/jl_Qc4Pp3/plugins/profile/2026_09_18_08_28_12
I0000 00:00:1789720092.980328    3980 xplane_to_tools_data_with_profile_processor.cc:165] Using local processing for tool: op_profile
I0000 00:00:1789720092.980331    3980 multi_xplanes_to_op_stats.cc:118] ConvertMultiXSpaceToCombinedOpStatsWithCache: Started
I0000 00:00:1789720092.980390    3980 multi_xplanes_to_op_stats.cc:134] ConvertMultiXSpaceToCombinedOpStatsWithCache: Cache miss, calling ConvertMultiXSpacesToCombinedOpStats
I0000 00:00:1789720092.980392    3980 multi_xplanes_to_op_stats.cc:45] ConvertMultiXSpacesToCombinedOpStats: Started. Number of XSpaces: 1
I0000 00:00:1789720092.980397    3980 multi_xplanes_to_op_stats.cc:55] ConvertMultiXSpacesToCombinedOpStats: Starting to process XSpace 0/1
I0000 00:00:1789720092.980559    3980 derived_timeline.cc:693] GenerateDerivedTimeLines: creating derived_timeline_trace_events XprofThreadPoolExecutor
I0000 00:00:1789720092.980566    3980 xprof_thread_pool_executor.cc:22] Creating derived_timeline_trace_events XprofThreadPoolExecutor with 4 threads.
I0000 00:00:1789720092.980773    3980 derived_timeline.cc:705] GenerateDerivedTimeLines: waiting for derived_timeline_trace_events threads to join
I0000 00:00:1789720092.980960    3980 derived_timeline.cc:709] GenerateDerivedTimeLines: derived_timeline_trace_events threads joined successfully
I0000 00:00:1789720092.981309    3980 derived_timeline.cc:758] GenerateDerivedTimeLines: creating ProcessTensorCorePlanes XprofThreadPoolExecutor
I0000 00:00:1789720092.981319    3980 xprof_thread_pool_executor.cc:22] Creating ProcessTensorCorePlanes XprofThreadPoolExecutor with 4 threads.
I0000 00:00:1789720092.981439    3980 derived_timeline.cc:769] GenerateDerivedTimeLines: waiting for ProcessTensorCorePlanes threads to join
I0000 00:00:1789720092.981661    3980 derived_timeline.cc:772] GenerateDerivedTimeLines: ProcessTensorCorePlanes threads joined successfully
I0000 00:00:1789720092.984762    3980 xplane_to_op_stats.cc:405] ConvertXSpaceToOpStats: creating op_stats_threads XprofThreadPoolExecutor
I0000 00:00:1789720092.984779    3980 xprof_thread_pool_executor.cc:22] Creating op_stats_threads XprofThreadPoolExecutor with 4 threads.
I0000 00:00:1789720092.984905    3980 xplane_to_op_stats.cc:461] ConvertXSpaceToOpStats: Scheduled 0 OpMetricsDb generation tasks.
I0000 00:00:1789720092.985145    3980 xplane_to_op_stats.cc:417] ConvertXSpaceToOpStats: Combining 0 op_metrics_dbs.
I0000 00:00:1789720092.985151    3980 xplane_to_op_stats.cc:422] ConvertXSpaceToOpStats: Finished combining op_metrics_dbs.
I0000 00:00:1789720092.985374    3980 xplane_to_op_stats.cc:687] ConvertXSpaceToOpStats: Final OpStats size: 221 bytes (0.000210762 MiB).
I0000 00:00:1789720092.985443    3980 multi_xplanes_to_op_stats.cc:67] ConvertMultiXSpacesToCombinedOpStats: Finished processing XSpace 0.
I0000 00:00:1789720092.985457    3980 multi_xplanes_to_op_stats.cc:72] ConvertMultiXSpacesToCombinedOpStats: Finished extracting all 1 OpStats. Time: 5.067371ms
I0000 00:00:1789720092.985464    3980 multi_xplanes_to_op_stats.cc:85] ConvertMultiXSpacesToCombinedOpStats: Starting ComputeStepIntersectionToMergeOpStats.
I0000 00:00:1789720092.985467    3980 multi_xplanes_to_op_stats.cc:94] ConvertMultiXSpacesToCombinedOpStats: Finished ComputeStepIntersectionToMergeOpStats in 1.422us
I0000 00:00:1789720092.985469    3980 multi_xplanes_to_op_stats.cc:99] ConvertMultiXSpacesToCombinedOpStats: Starting CombineAllOpStats.
I0000 00:00:1789720092.985474    3980 multi_xplanes_to_op_stats.cc:106] ConvertMultiXSpacesToCombinedOpStats: Finished CombineAllOpStats in 3.185us
I0000 00:00:1789720092.985475    3980 multi_xplanes_to_op_stats.cc:109] ConvertMultiXSpacesToCombinedOpStats: Overall Finished in 5.083946ms
I0000 00:00:1789720092.985478    3980 multi_xplanes_to_op_stats.cc:138] ConvertMultiXSpaceToCombinedOpStatsWithCache: Starting to write cache file.
I0000 00:00:1789720092.985579    3980 multi_xplanes_to_op_stats.cc:145] ConvertMultiXSpaceToCombinedOpStatsWithCache: Finished writing cache file.
I0000 00:00:1789720092.985581    3980 multi_xplanes_to_op_stats.cc:149] ConvertMultiXSpaceToCombinedOpStatsWithCache: Overall Finished in 5.250587ms
I0000 00:00:1789720092.985683    3980 xplane_to_tools_data_with_profile_processor.cc:170] Total time for tool op_profile: 5.360152ms session_id: /home/runner/.julia/scratchspaces/3c362404-f566-11ee-1572-e11a4b42c853/reactant_profiling/jl_Qc4Pp3/plugins/profile/2026_09_18_08_28_12
Debug: `op_profile` data missing keys for metrics
  data_available_keys =
   KeySet for a JSON.Object{String, Any} with 4 entries. Keys:
     "byProgram"
     "deviceType"
     "byProgramExcludeIdle"
     "aggDvfsTimeScaleMultiplier"
  by_program_available_keys =
   KeySet for a JSON.Object{String, Any} with 3 entries. Keys:
     "name"
     "children"
     "numChildren"
@ Reactant.Profiler ~/work/Reactant.jl/Reactant.jl/src/Profiler.jl:828
I0000 00:00:1789720093.233020    3980 xplane_to_tools_data_with_profile_processor.cc:142] serving tool: overview_page with options: {} using ProfileProcessor session_id: /home/runner/.julia/scratchspaces/3c362404-f566-11ee-1572-e11a4b42c853/reactant_profiling/jl_Qc4Pp3/plugins/profile/2026_09_18_08_28_12
I0000 00:00:1789720093.233045    3980 xplane_to_tools_data_with_profile_processor.cc:165] Using local processing for tool: overview_page
I0000 00:00:1789720093.233048    3980 overview_page_processor.cc:84] OverviewPageProcessor::ProcessSession: Started
I0000 00:00:1789720093.233050    3980 overview_page_processor.cc:86] OverviewPageProcessor::ProcessSession: Starting ConvertMultiXSpaceToCombinedOpStatsWithCache
I0000 00:00:1789720093.233052    3980 multi_xplanes_to_op_stats.cc:118] ConvertMultiXSpaceToCombinedOpStatsWithCache: Started
I0000 00:00:1789720093.233114    3980 multi_xplanes_to_op_stats.cc:126] ConvertMultiXSpaceToCombinedOpStatsWithCache: Cache hit, reading binary proto
I0000 00:00:1789720093.233165    3980 multi_xplanes_to_op_stats.cc:131] ConvertMultiXSpaceToCombinedOpStatsWithCache: Finished reading cache file.
I0000 00:00:1789720093.233168    3980 multi_xplanes_to_op_stats.cc:149] ConvertMultiXSpaceToCombinedOpStatsWithCache: Overall Finished in 117.227us
I0000 00:00:1789720093.233173    3980 overview_page_processor.cc:90] OverviewPageProcessor::ProcessSession: Starting ConvertOpStatsToOverviewPage
I0000 00:00:1789720093.233182    3980 op_stats_to_overview_page.cc:388] ConvertOpStatsToOverviewPage: Starting ComputeRunEnvironment
I0000 00:00:1789720093.233189    3980 op_stats_to_overview_page.cc:393] ConvertOpStatsToOverviewPage: Starting ComputeAnalysisResult
I0000 00:00:1789720093.233193    3980 op_stats_to_overview_page.cc:396] ConvertOpStatsToOverviewPage: Starting ConvertOpStatsToInputPipelineAnalysis
I0000 00:00:1789720093.233365    3980 op_stats_to_overview_page.cc:401] ConvertOpStatsToOverviewPage: Starting ComputeBottleneckAnalysis
I0000 00:00:1789720093.233374    3980 op_stats_to_overview_page.cc:407] ConvertOpStatsToOverviewPage: Starting ComputeGenericRecommendation
I0000 00:00:1789720093.233477    3980 op_stats_to_overview_page.cc:412] ConvertOpStatsToOverviewPage: Starting SetCommonRecommendation
I0000 00:00:1789720093.233491    3980 op_stats_to_overview_page.cc:425] ConvertOpStatsToOverviewPage: Starting PopulateOverviewDiagnostics
I0000 00:00:1789720093.233493    3980 op_stats_to_overview_page.cc:429] ConvertOpStatsToOverviewPage: Starting setting utilizations
I0000 00:00:1789720093.233494    3980 op_stats_to_overview_page.cc:435] ConvertOpStatsToOverviewPage: Overall Finished in 312.832us
I0000 00:00:1789720093.233496    3980 overview_page_processor.cc:94] OverviewPageProcessor::ProcessSession: Not a training run, Starting to convert inference stats.
I0000 00:00:1789720093.233504    3980 xprof_thread_pool_executor.cc:22] Creating ConvertMultiXSpaceToInferenceStats XprofThreadPoolExecutor with 1 threads.
I0000 00:00:1789720093.233853    3980 overview_page_processor.cc:99] OverviewPageProcessor::ProcessSession: Starting to compute InferenceLatency
I0000 00:00:1789720093.233858    3980 overview_page_processor.cc:104] OverviewPageProcessor::ProcessSession: Starting to serialize OverviewPage toJson
I0000 00:00:1789720093.234123    3980 overview_page_processor.cc:107] OverviewPageProcessor::ProcessSession: Starting to set Output
I0000 00:00:1789720093.234132    3980 overview_page_processor.cc:109] OverviewPageProcessor::ProcessSession: Overall Finished in 1.085197ms
I0000 00:00:1789720093.234143    3980 xplane_to_tools_data_with_profile_processor.cc:170] Total time for tool overview_page: 1.105297ms session_id: /home/runner/.julia/scratchspaces/3c362404-f566-11ee-1572-e11a4b42c853/reactant_profiling/jl_Qc4Pp3/plugins/profile/2026_09_18_08_28_12
  runtime: 0.00027752s
  compile time: 0.58021831s
julia
Reactant.@timed nrepeat=100 linear(x, W, b)
AggregateProfilingResult(
    runtime = 0.00001718s, 
    compile_time = 0.10460205s, )

Note that the information returned depends on the backend. Specifically CUDA and TPU backends provide more detailed information regarding memory usage and allocation (something like the following will be displayed on GPUs):

julia
AggregateProfilingResult(
    runtime = 0.00003829s, 
    compile_time = 2.18053260s,  # time spent compiling by Reactant
    GPU_0_bfc = MemoryProfileSummary(
        peak_bytes_usage_lifetime = 64.010 MiB,  # peak memory usage over the entire program (lifetime of memory allocator)
        peak_stats = MemoryAggregationStats(
            stack_reserved_bytes = 0 bytes,  # memory usage by stack reservation
            heap_allocated_bytes = 30.750 KiB,  # memory usage by heap allocation
            free_memory_bytes = 23.518 GiB,  # free memory available for allocation or reservation
            fragmentation = 0.514931,  # fragmentation of memory within [0, 1]
            peak_bytes_in_use = 30.750 KiB # The peak memory usage over the entire program
        )
        peak_stats_time = 0.04975365s, 
        memory_capacity = 23.518 GiB # memory capacity of the allocator
    )
    flops = FlopsSummary(
        Flops = 2.8369974648038653e-9,  # [flops / (peak flops * program time)], capped at 1.0
        UncappedFlops = 2.8369974648038653e-9, 
        RawFlops = 4060.0,  # Total FLOPs performed
        BF16Flops = 4060.0,  # Total FLOPs Normalized to the bf16 (default) devices peak bandwidth
        RawTime = 0.00040298422s,  # Raw time in seconds
        RawFlopsRate = 1.0074836180930361e7,  # Raw FLOPs rate in FLOPs/seconds
        BF16FlopsRate = 1.0074836180930361e7,  # BF16 FLOPs rate in FLOPs/seconds
    )
)

Additionally for GPUs and TPUs, we can use the Reactant.@profile macro to profile the function and get information regarding each of the kernels executed.

julia
Reactant.@profile linear(x, W, b)
Debug: Profiling directory: /home/runner/.julia/scratchspaces/3c362404-f566-11ee-1572-e11a4b42c853/reactant_profiling/jl_Fictiv
@ Reactant.Profiler ~/work/Reactant.jl/Reactant.jl/src/Profiler.jl:626
I0000 00:00:1789720093.854239    3980 profiler_session.cc:171] Profiler session initializing.
I0000 00:00:1789720093.854493    3980 profiler_session.cc:186] Profiler session started.
I0000 00:00:1789720093.854581    3980 profiler_session.cc:134] Profiler session collecting data.
I0000 00:00:1789720093.855091    3980 save_profile.cc:205] Collecting XSpace to repository: /home/runner/.julia/scratchspaces/3c362404-f566-11ee-1572-e11a4b42c853/reactant_profiling/jl_Fictiv/plugins/profile/2026_09_18_08_28_13/runnervmlun5p.xplane.pb
I0000 00:00:1789720093.855241    3980 save_profile.cc:178] Creating directory: /home/runner/.julia/scratchspaces/3c362404-f566-11ee-1572-e11a4b42c853/reactant_profiling/jl_Fictiv/plugins/profile/2026_09_18_08_28_13

I0000 00:00:1789720093.855334    3980 save_profile.cc:184] Dumped gzipped tool data for trace.json.gz to /home/runner/.julia/scratchspaces/3c362404-f566-11ee-1572-e11a4b42c853/reactant_profiling/jl_Fictiv/plugins/profile/2026_09_18_08_28_13/runnervmlun5p.trace.json.gz
I0000 00:00:1789720093.855348    3980 profiler_session.cc:217] Profiler session tear down.
I0000 00:00:1789720093.855406    3980 xplane_to_tools_data_with_profile_processor.cc:142] serving tool: memory_profile with options: {} using ProfileProcessor session_id: /home/runner/.julia/scratchspaces/3c362404-f566-11ee-1572-e11a4b42c853/reactant_profiling/jl_Fictiv/plugins/profile/2026_09_18_08_28_13
I0000 00:00:1789720093.855411    3980 xplane_to_tools_data_with_profile_processor.cc:165] Using local processing for tool: memory_profile
I0000 00:00:1789720093.855413    3980 memory_profile_processor.cc:47] Processing memory profile for host: runnervmlun5p
I0000 00:00:1789720093.855553    3980 xplane_to_tools_data_with_profile_processor.cc:170] Total time for tool memory_profile: 146.38us session_id: /home/runner/.julia/scratchspaces/3c362404-f566-11ee-1572-e11a4b42c853/reactant_profiling/jl_Fictiv/plugins/profile/2026_09_18_08_28_13
I0000 00:00:1789720093.855577    3980 xplane_to_tools_data_with_profile_processor.cc:142] serving tool: op_profile with options: {} using ProfileProcessor session_id: /home/runner/.julia/scratchspaces/3c362404-f566-11ee-1572-e11a4b42c853/reactant_profiling/jl_Fictiv/plugins/profile/2026_09_18_08_28_13
I0000 00:00:1789720093.855580    3980 xplane_to_tools_data_with_profile_processor.cc:165] Using local processing for tool: op_profile
I0000 00:00:1789720093.855582    3980 multi_xplanes_to_op_stats.cc:118] ConvertMultiXSpaceToCombinedOpStatsWithCache: Started
I0000 00:00:1789720093.855604    3980 multi_xplanes_to_op_stats.cc:134] ConvertMultiXSpaceToCombinedOpStatsWithCache: Cache miss, calling ConvertMultiXSpacesToCombinedOpStats
I0000 00:00:1789720093.855605    3980 multi_xplanes_to_op_stats.cc:45] ConvertMultiXSpacesToCombinedOpStats: Started. Number of XSpaces: 1
I0000 00:00:1789720093.855608    3980 multi_xplanes_to_op_stats.cc:55] ConvertMultiXSpacesToCombinedOpStats: Starting to process XSpace 0/1
I0000 00:00:1789720093.855670    3980 derived_timeline.cc:693] GenerateDerivedTimeLines: creating derived_timeline_trace_events XprofThreadPoolExecutor
I0000 00:00:1789720093.855676    3980 xprof_thread_pool_executor.cc:22] Creating derived_timeline_trace_events XprofThreadPoolExecutor with 4 threads.
I0000 00:00:1789720093.855844    3980 derived_timeline.cc:705] GenerateDerivedTimeLines: waiting for derived_timeline_trace_events threads to join
I0000 00:00:1789720093.856056    3980 derived_timeline.cc:709] GenerateDerivedTimeLines: derived_timeline_trace_events threads joined successfully
I0000 00:00:1789720093.856467    3980 derived_timeline.cc:758] GenerateDerivedTimeLines: creating ProcessTensorCorePlanes XprofThreadPoolExecutor
I0000 00:00:1789720093.856486    3980 xprof_thread_pool_executor.cc:22] Creating ProcessTensorCorePlanes XprofThreadPoolExecutor with 4 threads.
I0000 00:00:1789720093.856608    3980 derived_timeline.cc:769] GenerateDerivedTimeLines: waiting for ProcessTensorCorePlanes threads to join
I0000 00:00:1789720093.856800    3980 derived_timeline.cc:772] GenerateDerivedTimeLines: ProcessTensorCorePlanes threads joined successfully
I0000 00:00:1789720093.857236    3980 xplane_to_op_stats.cc:405] ConvertXSpaceToOpStats: creating op_stats_threads XprofThreadPoolExecutor
I0000 00:00:1789720093.857261    3980 xprof_thread_pool_executor.cc:22] Creating op_stats_threads XprofThreadPoolExecutor with 4 threads.
I0000 00:00:1789720093.857382    3980 xplane_to_op_stats.cc:461] ConvertXSpaceToOpStats: Scheduled 0 OpMetricsDb generation tasks.
I0000 00:00:1789720093.857570    3980 xplane_to_op_stats.cc:417] ConvertXSpaceToOpStats: Combining 0 op_metrics_dbs.
I0000 00:00:1789720093.857575    3980 xplane_to_op_stats.cc:422] ConvertXSpaceToOpStats: Finished combining op_metrics_dbs.
I0000 00:00:1789720093.857747    3980 xplane_to_op_stats.cc:687] ConvertXSpaceToOpStats: Final OpStats size: 262 bytes (0.000249863 MiB).
I0000 00:00:1789720093.857815    3980 multi_xplanes_to_op_stats.cc:67] ConvertMultiXSpacesToCombinedOpStats: Finished processing XSpace 0.
I0000 00:00:1789720093.857844    3980 multi_xplanes_to_op_stats.cc:72] ConvertMultiXSpacesToCombinedOpStats: Finished extracting all 1 OpStats. Time: 2.2413ms
I0000 00:00:1789720093.857850    3980 multi_xplanes_to_op_stats.cc:85] ConvertMultiXSpacesToCombinedOpStats: Starting ComputeStepIntersectionToMergeOpStats.
I0000 00:00:1789720093.857852    3980 multi_xplanes_to_op_stats.cc:94] ConvertMultiXSpacesToCombinedOpStats: Finished ComputeStepIntersectionToMergeOpStats in 1.462us
I0000 00:00:1789720093.857854    3980 multi_xplanes_to_op_stats.cc:99] ConvertMultiXSpacesToCombinedOpStats: Starting CombineAllOpStats.
I0000 00:00:1789720093.857859    3980 multi_xplanes_to_op_stats.cc:106] ConvertMultiXSpacesToCombinedOpStats: Finished CombineAllOpStats in 3.185us
I0000 00:00:1789720093.857860    3980 multi_xplanes_to_op_stats.cc:109] ConvertMultiXSpacesToCombinedOpStats: Overall Finished in 2.255341ms
I0000 00:00:1789720093.857863    3980 multi_xplanes_to_op_stats.cc:138] ConvertMultiXSpaceToCombinedOpStatsWithCache: Starting to write cache file.
I0000 00:00:1789720093.857930    3980 multi_xplanes_to_op_stats.cc:145] ConvertMultiXSpaceToCombinedOpStatsWithCache: Finished writing cache file.
I0000 00:00:1789720093.857932    3980 multi_xplanes_to_op_stats.cc:149] ConvertMultiXSpaceToCombinedOpStatsWithCache: Overall Finished in 2.350274ms
I0000 00:00:1789720093.857955    3980 xplane_to_tools_data_with_profile_processor.cc:170] Total time for tool op_profile: 2.377745ms session_id: /home/runner/.julia/scratchspaces/3c362404-f566-11ee-1572-e11a4b42c853/reactant_profiling/jl_Fictiv/plugins/profile/2026_09_18_08_28_13
Debug: `op_profile` data missing keys for metrics
  data_available_keys =
   KeySet for a JSON.Object{String, Any} with 4 entries. Keys:
     "byProgram"
     "deviceType"
     "byProgramExcludeIdle"
     "aggDvfsTimeScaleMultiplier"
  by_program_available_keys =
   KeySet for a JSON.Object{String, Any} with 3 entries. Keys:
     "name"
     "children"
     "numChildren"
@ Reactant.Profiler ~/work/Reactant.jl/Reactant.jl/src/Profiler.jl:828
I0000 00:00:1789720093.858375    3980 xplane_to_tools_data_with_profile_processor.cc:142] serving tool: overview_page with options: {} using ProfileProcessor session_id: /home/runner/.julia/scratchspaces/3c362404-f566-11ee-1572-e11a4b42c853/reactant_profiling/jl_Fictiv/plugins/profile/2026_09_18_08_28_13
I0000 00:00:1789720093.858382    3980 xplane_to_tools_data_with_profile_processor.cc:165] Using local processing for tool: overview_page
I0000 00:00:1789720093.858384    3980 overview_page_processor.cc:84] OverviewPageProcessor::ProcessSession: Started
I0000 00:00:1789720093.858385    3980 overview_page_processor.cc:86] OverviewPageProcessor::ProcessSession: Starting ConvertMultiXSpaceToCombinedOpStatsWithCache
I0000 00:00:1789720093.858387    3980 multi_xplanes_to_op_stats.cc:118] ConvertMultiXSpaceToCombinedOpStatsWithCache: Started
I0000 00:00:1789720093.858416    3980 multi_xplanes_to_op_stats.cc:126] ConvertMultiXSpaceToCombinedOpStatsWithCache: Cache hit, reading binary proto
I0000 00:00:1789720093.858449    3980 multi_xplanes_to_op_stats.cc:131] ConvertMultiXSpaceToCombinedOpStatsWithCache: Finished reading cache file.
I0000 00:00:1789720093.858452    3980 multi_xplanes_to_op_stats.cc:149] ConvertMultiXSpaceToCombinedOpStatsWithCache: Overall Finished in 65.459us
I0000 00:00:1789720093.858454    3980 overview_page_processor.cc:90] OverviewPageProcessor::ProcessSession: Starting ConvertOpStatsToOverviewPage
I0000 00:00:1789720093.858456    3980 op_stats_to_overview_page.cc:388] ConvertOpStatsToOverviewPage: Starting ComputeRunEnvironment
I0000 00:00:1789720093.858460    3980 op_stats_to_overview_page.cc:393] ConvertOpStatsToOverviewPage: Starting ComputeAnalysisResult
I0000 00:00:1789720093.858463    3980 op_stats_to_overview_page.cc:396] ConvertOpStatsToOverviewPage: Starting ConvertOpStatsToInputPipelineAnalysis
I0000 00:00:1789720093.858496    3980 op_stats_to_overview_page.cc:401] ConvertOpStatsToOverviewPage: Starting ComputeBottleneckAnalysis
I0000 00:00:1789720093.858500    3980 op_stats_to_overview_page.cc:407] ConvertOpStatsToOverviewPage: Starting ComputeGenericRecommendation
I0000 00:00:1789720093.858505    3980 op_stats_to_overview_page.cc:412] ConvertOpStatsToOverviewPage: Starting SetCommonRecommendation
I0000 00:00:1789720093.858513    3980 op_stats_to_overview_page.cc:425] ConvertOpStatsToOverviewPage: Starting PopulateOverviewDiagnostics
I0000 00:00:1789720093.858515    3980 op_stats_to_overview_page.cc:429] ConvertOpStatsToOverviewPage: Starting setting utilizations
I0000 00:00:1789720093.858517    3980 op_stats_to_overview_page.cc:435] ConvertOpStatsToOverviewPage: Overall Finished in 61.002us
I0000 00:00:1789720093.858518    3980 overview_page_processor.cc:94] OverviewPageProcessor::ProcessSession: Not a training run, Starting to convert inference stats.
I0000 00:00:1789720093.858522    3980 xprof_thread_pool_executor.cc:22] Creating ConvertMultiXSpaceToInferenceStats XprofThreadPoolExecutor with 1 threads.
I0000 00:00:1789720093.858760    3980 overview_page_processor.cc:99] OverviewPageProcessor::ProcessSession: Starting to compute InferenceLatency
I0000 00:00:1789720093.858765    3980 overview_page_processor.cc:104] OverviewPageProcessor::ProcessSession: Starting to serialize OverviewPage toJson
I0000 00:00:1789720093.858965    3980 overview_page_processor.cc:107] OverviewPageProcessor::ProcessSession: Starting to set Output
I0000 00:00:1789720093.858974    3980 overview_page_processor.cc:109] OverviewPageProcessor::ProcessSession: Overall Finished in 590.891us
I0000 00:00:1789720093.858983    3980 xplane_to_tools_data_with_profile_processor.cc:170] Total time for tool overview_page: 605.233us session_id: /home/runner/.julia/scratchspaces/3c362404-f566-11ee-1572-e11a4b42c853/reactant_profiling/jl_Fictiv/plugins/profile/2026_09_18_08_28_13
I0000 00:00:1789720093.940742    3980 xplane_to_tools_data_with_profile_processor.cc:142] serving tool: kernel_stats with options: {} using ProfileProcessor session_id: /home/runner/.julia/scratchspaces/3c362404-f566-11ee-1572-e11a4b42c853/reactant_profiling/jl_Fictiv/plugins/profile/2026_09_18_08_28_13
I0000 00:00:1789720093.940775    3980 xplane_to_tools_data_with_profile_processor.cc:165] Using local processing for tool: kernel_stats
I0000 00:00:1789720093.940778    3980 multi_xplanes_to_op_stats.cc:118] ConvertMultiXSpaceToCombinedOpStatsWithCache: Started
I0000 00:00:1789720093.940840    3980 multi_xplanes_to_op_stats.cc:126] ConvertMultiXSpaceToCombinedOpStatsWithCache: Cache hit, reading binary proto
I0000 00:00:1789720093.940898    3980 multi_xplanes_to_op_stats.cc:131] ConvertMultiXSpaceToCombinedOpStatsWithCache: Finished reading cache file.
I0000 00:00:1789720093.940901    3980 multi_xplanes_to_op_stats.cc:149] ConvertMultiXSpaceToCombinedOpStatsWithCache: Overall Finished in 123.186us
I0000 00:00:1789720093.940973    3980 xplane_to_tools_data_with_profile_processor.cc:170] Total time for tool kernel_stats: 203.828us session_id: /home/runner/.julia/scratchspaces/3c362404-f566-11ee-1572-e11a4b42c853/reactant_profiling/jl_Fictiv/plugins/profile/2026_09_18_08_28_13
I0000 00:00:1789720094.056439    3980 xplane_to_tools_data_with_profile_processor.cc:142] serving tool: framework_op_stats with options: {} using ProfileProcessor session_id: /home/runner/.julia/scratchspaces/3c362404-f566-11ee-1572-e11a4b42c853/reactant_profiling/jl_Fictiv/plugins/profile/2026_09_18_08_28_13
I0000 00:00:1789720094.056473    3980 xplane_to_tools_data_with_profile_processor.cc:165] Using local processing for tool: framework_op_stats
I0000 00:00:1789720094.056477    3980 multi_xplanes_to_op_stats.cc:118] ConvertMultiXSpaceToCombinedOpStatsWithCache: Started
I0000 00:00:1789720094.056538    3980 multi_xplanes_to_op_stats.cc:126] ConvertMultiXSpaceToCombinedOpStatsWithCache: Cache hit, reading binary proto
I0000 00:00:1789720094.056588    3980 multi_xplanes_to_op_stats.cc:131] ConvertMultiXSpaceToCombinedOpStatsWithCache: Finished reading cache file.
I0000 00:00:1789720094.056591    3980 multi_xplanes_to_op_stats.cc:149] ConvertMultiXSpaceToCombinedOpStatsWithCache: Overall Finished in 115.074us
I0000 00:00:1789720094.056754    3980 xplane_to_tools_data_with_profile_processor.cc:170] Total time for tool framework_op_stats: 284.018us session_id: /home/runner/.julia/scratchspaces/3c362404-f566-11ee-1572-e11a4b42c853/reactant_profiling/jl_Fictiv/plugins/profile/2026_09_18_08_28_13

╔================================================================================╗
║ SUMMARY                                                                        ║
╚================================================================================╝

AggregateProfilingResult(
    runtime = 0.00006202s,
    compile_time = 0.09712100s,  # time spent compiling by Reactant
)

On GPUs this would look something like the following:

julia
================================================================================
║ KERNEL STATISTICS                                                              ║
================================================================================

┌───────────────────┬─────────────┬────────────────┬──────────────┬──────────────┬──────────────┬──────────────┬───────────┬──────────┬────────────┬─────────────┐
│       Kernel Name │ Occurrences │ Total Duration │ Avg Duration │ Min Duration │ Max Duration │ Static Shmem │ Block Dim │ Grid Dim │ TensorCore │ Occupancy %
├───────────────────┼─────────────┼────────────────┼──────────────┼──────────────┼──────────────┼──────────────┼───────────┼──────────┼────────────┼─────────────┤
│ gemm_fusion_dot_1 │           10.00000250s │  0.00000250s │  0.00000250s │  0.00000250s │    2.000 KiB │    64,1,11,1,1 │          ✗ │      100.0%
│   loop_add_fusion │           10.00000131s │  0.00000131s │  0.00000131s │  0.00000131s │      0 bytes │    20,1,11,1,1 │          ✗ │       31.2%
└───────────────────┴─────────────┴────────────────┴──────────────┴──────────────┴──────────────┴──────────────┴───────────┴──────────┴────────────┴─────────────┘

================================================================================
║ FRAMEWORK OP STATISTICS                                                        ║
================================================================================

┌───────────────────┬─────────┬─────────────┬─────────────┬─────────────────┬───────────────┬──────────┬───────────┬──────────────┬──────────┐
│         Operation │    Type │ Host/Device │ Occurrences │ Total Self-Time │ Avg Self-Time │ Device % │ Memory BW │    FLOP Rate │ Bound By │
├───────────────────┼─────────┼─────────────┼─────────────┼─────────────────┼───────────────┼──────────┼───────────┼──────────────┼──────────┤
│ gemm_fusion_dot.1 │ Unknown │      Device │           10.00000250s │   0.00000250s │   65.55%1.82 GB/s │  1.6 GFLOP/s │      HBM │
+/add │     add │      Device │           10.00000131s │   0.00000131s │   34.45%0.14 GB/s │ 0.05 GFLOP/s │      HBM │
└───────────────────┴─────────┴─────────────┴─────────────┴─────────────────┴───────────────┴──────────┴───────────┴──────────────┴──────────┘

================================================================================
║ SUMMARY                                                                        ║
================================================================================

AggregateProfilingResult(
    runtime = 0.00005622s, 
    compile_time = 2.32802137s,  # time spent compiling by Reactant
    GPU_0_bfc = MemoryProfileSummary(
        peak_bytes_usage_lifetime = 64.010 MiB,  # peak memory usage over the entire program (lifetime of memory allocator)
        peak_stats = MemoryAggregationStats(
            stack_reserved_bytes = 0 bytes,  # memory usage by stack reservation
            heap_allocated_bytes = 81.750 KiB,  # memory usage by heap allocation
            free_memory_bytes = 23.518 GiB,  # free memory available for allocation or reservation
            fragmentation = 0.514564,  # fragmentation of memory within [0, 1]
            peak_bytes_in_use = 81.750 KiB # The peak memory usage over the entire program
        )
        peak_stats_time = 0.00608052s, 
        memory_capacity = 23.518 GiB # memory capacity of the allocator
    )
    flops = FlopsSummary(
        Flops = 2.033375207640664e-8,  # [flops / (peak flops * program time)], capped at 1.0
        UncappedFlops = 2.033375207640664e-8, 
        RawFlops = 4060.0,  # Total FLOPs performed
        BF16Flops = 4060.0,  # Total FLOPs Normalized to the bf16 (default) devices peak bandwidth
        RawTime = 0.00005622s,  # Raw time in seconds
        RawFlopsRate = 7.220987105380169e7,  # Raw FLOPs rate in FLOPs/seconds
        BF16FlopsRate = 7.220987105380169e7,  # BF16 FLOPs rate in FLOPs/seconds
    )
)

Capturing traces

When running Reactant, it is possible to capture traces using the XLA profiler. These traces can provide information about where the XLA specific parts of program spend time during compilation or execution. Note that tracing and compilation happen on the CPU even though the final execution is aimed to run on another device such as GPU or TPU. Therefore, including tracing and compilation in a trace will create annotations on the CPU.

Let's setup a simple function which we can then profile

julia
using Reactant

x = Reactant.to_rarray(randn(Float32, 100, 2))
W = Reactant.to_rarray(randn(Float32, 10, 100))
b = Reactant.to_rarray(randn(Float32, 10))

linear(x, W, b) = (W * x) .+ b
linear (generic function with 1 method)

The profiler can be accessed using the Reactant.with_profiler function.

julia
Reactant.with_profiler("./") do
    mylinear = Reactant.@compile linear(x, W, b)
    mylinear(x, W, b)
end
10×2 ConcretePJRTArray{Float32,2}:
 -14.4167   -3.19252
   5.07385  -7.00612
  20.7905   -4.6425
  -8.22616  19.9759
 -22.842    -3.31033
 -19.0995    0.244569
  14.1243    7.46187
 -15.8923   27.0745
   7.2197   -5.29203
  -4.71639   3.45019

Running this function should create a folder called plugins in the folder provided to Reactant.with_profiler which will contain the trace files. The traces can then be visualized in different ways.

Note

For more insights about the current state of Reactant, it is possible to fetch device information about allocations using the Reactant.XLA.allocatorstats function.

Perfetto UI

The first and easiest way to visualize a captured trace is to use the online perfetto.dev tool. Reactant.with_profiler has a keyword parameter called create_perfetto_link which will create a usable perfetto URL for the generated trace. The function will block execution until the URL has been clicked and the trace is visualized. The URL only works once.

julia
Reactant.with_profiler("./"; create_perfetto_link=true) do
    mylinear = Reactant.@compile linear(x, W, b)
    mylinear(x, W, b)
end

Note

It is recommended to use the Chrome browser to open the perfetto URL.

XProf

XProf is a complete web UI to analyze the log files captured by Reactant. It can be installed in the following manner:

bash
pip install xprof # or xprof-nightly

Launching xprof is then as simple as:

bash
xprof --logdir=./

which will then make the xprof interface available on port :8791 by default.

Tensorboard

Another option to visualize the generated trace files is to use the tensorboard profiler plugin. The tensorboard viewer can offer more details than the timeline view such as visualization for compute graphs.

First install tensorboard and its profiler plugin:

bash
pip install tensorboard tensorboard-plugin-profile

And then run the following in the folder where the plugins folder was generated:

bash
tensorboard --logdir ./

Adding Custom Annotations

By default, the traces contain only information captured from within XLA. The Reactant.Profiler.annotate function can be used to annotate traces for Julia code evaluated during tracing.

julia
Reactant.Profiler.annotate("my_annotation") do
    # Do things...
end

The added annotations will be captured in the traces and can be seen in the different viewers along with the default XLA annotations. When the profiler is not activated, then the custom annotations have no effect and can therefore always be activated.