2026-09-07T12:42:10.377535Z  INFO retained operator-plan finalist finalist=0 values=11 operations=7 estimated_cycles=498794 estimated_exchange_cycles=217560 exchange_row_bytes=44628 peak_standard_bytes=351828 peak_interleaved_bytes=53248 peak_total_bytes=396884 maximum_standard_allocation_bytes=98304 active_tile_counts={192, 460, 1472} padding_group_counts={} conversion_sources=[]
2026-09-07T12:42:10.397622Z  INFO retained operator-plan finalist finalist=1 values=11 operations=7 estimated_cycles=505040 estimated_exchange_cycles=219000 exchange_row_bytes=44952 peak_standard_bytes=183448 peak_interleaved_bytes=53248 peak_total_bytes=220120 maximum_standard_allocation_bytes=98304 active_tile_counts={192, 480, 1472} padding_group_counts={16} conversion_sources=[]
2026-09-07T12:42:10.401140Z  INFO retained operator-plan finalist finalist=2 values=12 operations=8 estimated_cycles=560870 estimated_exchange_cycles=222240 exchange_row_bytes=45812 peak_standard_bytes=176884 peak_interleaved_bytes=196416 peak_total_bytes=356148 maximum_standard_allocation_bytes=98304 active_tile_counts={192, 480, 1472} padding_group_counts={16} conversion_sources=[Some(OperationId(2))]
2026-09-07T12:42:10.410564Z  INFO retained operator-plan finalist finalist=3 values=13 operations=9 estimated_cycles=595217 estimated_exchange_cycles=202440 exchange_row_bytes=41488 peak_standard_bytes=287248 peak_interleaved_bytes=180224 peak_total_bytes=467472 maximum_standard_allocation_bytes=98304 active_tile_counts={192, 460, 1380, 1472} padding_group_counts={} conversion_sources=[Some(OperationId(1)), Some(OperationId(2))]
2026-09-07T12:42:12.964605Z  INFO retained operator-plan finalist finalist=0 values=11 operations=7 estimated_cycles=498794 estimated_exchange_cycles=217560 exchange_row_bytes=44628 peak_standard_bytes=351828 peak_interleaved_bytes=53248 peak_total_bytes=396884 maximum_standard_allocation_bytes=98304 active_tile_counts={192, 460, 1472} padding_group_counts={} conversion_sources=[]
2026-09-07T12:42:12.977459Z  INFO retained operator-plan finalist finalist=1 values=11 operations=7 estimated_cycles=498794 estimated_exchange_cycles=217560 exchange_row_bytes=44628 peak_standard_bytes=339540 peak_interleaved_bytes=61056 peak_total_bytes=396884 maximum_standard_allocation_bytes=98304 active_tile_counts={192, 460, 1472} padding_group_counts={} conversion_sources=[]
2026-09-07T12:42:12.989487Z  INFO retained operator-plan finalist finalist=2 values=11 operations=7 estimated_cycles=504296 estimated_exchange_cycles=225240 exchange_row_bytes=46356 peak_standard_bytes=350484 peak_interleaved_bytes=53248 peak_total_bytes=392468 maximum_standard_allocation_bytes=98304 active_tile_counts={192, 368, 1472} padding_group_counts={} conversion_sources=[]
2026-09-07T12:42:13.001798Z  INFO retained operator-plan finalist finalist=3 values=11 operations=7 estimated_cycles=505040 estimated_exchange_cycles=219000 exchange_row_bytes=44952 peak_standard_bytes=183448 peak_interleaved_bytes=53248 peak_total_bytes=220120 maximum_standard_allocation_bytes=98304 active_tile_counts={192, 480, 1472} padding_group_counts={16} conversion_sources=[]
2026-09-07T12:42:13.065649Z  INFO ipu_codegen.package.phase{phase="lower_mid"}: package build phase finished phase="lower_mid" elapsed_ms=5922 success=true
2026-09-07T12:42:21.670638Z  INFO screened tile mappings by exchange resource load candidates=89 baseline_cycles=81248 candidate_cycles=81248 baseline_pressure=76526 candidate_pressure=76526
2026-09-07T12:42:21.691446Z  INFO modelled expanded operator plan finalist=6 modelled_cycles=341881 estimated_cycles=394337 estimated_exchange_cycles=145104
2026-09-07T12:42:21.722050Z  INFO screened tile mappings by exchange resource load candidates=89 baseline_cycles=82176 candidate_cycles=82176 baseline_pressure=77475 candidate_pressure=77475
2026-09-07T12:42:21.743887Z  INFO modelled expanded operator plan finalist=1 modelled_cycles=367839 estimated_cycles=439847 estimated_exchange_cycles=165584
2026-09-07T12:42:21.974718Z  INFO screened tile mappings by exchange resource load candidates=89 baseline_cycles=82176 candidate_cycles=82176 baseline_pressure=77475 candidate_pressure=77475
2026-09-07T12:42:21.994824Z  INFO modelled expanded operator plan finalist=7 modelled_cycles=360207 estimated_cycles=432215 estimated_exchange_cycles=165584
2026-09-07T12:42:22.515332Z  INFO screened tile mappings by exchange resource load candidates=65 baseline_cycles=84640 candidate_cycles=81184 baseline_pressure=77412 candidate_pressure=73956
2026-09-07T12:42:22.524009Z  INFO screened tile mappings by exchange resource load candidates=65 baseline_cycles=85792 candidate_cycles=82336 baseline_pressure=78558 candidate_pressure=75102
2026-09-07T12:42:22.537774Z  INFO modelled expanded operator plan finalist=8 modelled_cycles=348449 estimated_cycles=406953 estimated_exchange_cycles=151088
2026-09-07T12:42:22.546018Z  INFO modelled expanded operator plan finalist=9 modelled_cycles=370777 estimated_cycles=442849 estimated_exchange_cycles=165808
2026-09-07T12:42:22.636243Z  INFO screened tile mappings by exchange resource load candidates=89 baseline_cycles=81248 candidate_cycles=81248 baseline_pressure=76526 candidate_pressure=76526
2026-09-07T12:42:22.669555Z  INFO modelled expanded operator plan finalist=0 modelled_cycles=349513 estimated_cycles=401969 estimated_exchange_cycles=145104
2026-09-07T12:42:23.137239Z  INFO screened tile mappings by exchange resource load candidates=93 baseline_cycles=87392 candidate_cycles=87392 baseline_pressure=72804 candidate_pressure=72804
2026-09-07T12:42:23.140960Z  INFO screened tile mappings by exchange resource load candidates=93 baseline_cycles=87648 candidate_cycles=87648 baseline_pressure=73156 candidate_pressure=73156
2026-09-07T12:42:23.171991Z  INFO modelled expanded operator plan finalist=4 modelled_cycles=477125 estimated_cycles=567581 estimated_exchange_cycles=188048
2026-09-07T12:42:23.174596Z  INFO modelled expanded operator plan finalist=5 modelled_cycles=483029 estimated_cycles=578349 estimated_exchange_cycles=193168
2026-09-07T12:42:24.194730Z  INFO screened tile mappings by exchange resource load candidates=123 baseline_cycles=94976 candidate_cycles=94976 baseline_pressure=77346 candidate_pressure=77346
2026-09-07T12:42:24.228499Z  INFO modelled expanded operator plan finalist=3 modelled_cycles=631597 estimated_cycles=963261 estimated_exchange_cycles=437440
2026-09-07T12:42:24.418597Z  INFO screened tile mappings by exchange resource load candidates=123 baseline_cycles=82368 candidate_cycles=82368 baseline_pressure=77549 candidate_pressure=77549
2026-09-07T12:42:24.445011Z  INFO modelled expanded operator plan finalist=2 modelled_cycles=665231 estimated_cycles=1102823 estimated_exchange_cycles=531360
2026-09-07T12:42:24.474906Z  INFO ipu_codegen.package.phase{phase="select_finalist"}: prepared large physical exchange phase phase=1 transfers=1334 maximum_identical_destinations=1
2026-09-07T12:42:24.474962Z  INFO ipu_codegen.package.phase{phase="select_finalist"}: optimized physical exchange schedule phase=1 initial_horizon=2913 selected_horizon=2913 endpoint_lower_bound=2784 lower_bound_gap=129 selected_kind="full-duplex" neighborhood_improvements=0
2026-09-07T12:42:24.475747Z  INFO ipu_codegen.package.phase{phase="select_finalist"}: finished large physical exchange phase phase=1 tile=367 row_words=11 horizon=2913
2026-09-07T12:42:24.526294Z  INFO ipu_codegen.package.phase{phase="select_finalist"}: prepared large physical exchange phase phase=5 transfers=1334 maximum_identical_destinations=1
2026-09-07T12:42:24.526343Z  INFO ipu_codegen.package.phase{phase="select_finalist"}: optimized physical exchange schedule phase=5 initial_horizon=2897 selected_horizon=2897 endpoint_lower_bound=2784 lower_bound_gap=113 selected_kind="full-duplex" neighborhood_improvements=0
2026-09-07T12:42:24.527357Z  INFO ipu_codegen.package.phase{phase="select_finalist"}: finished large physical exchange phase phase=5 tile=1291 row_words=11 horizon=2897
2026-09-07T12:42:24.937117Z  INFO ipu_codegen.package.phase{phase="select_finalist"}: prepared large physical exchange phase phase=3 transfers=1288 maximum_identical_destinations=1
2026-09-07T12:42:24.937183Z  INFO ipu_codegen.package.phase{phase="select_finalist"}: optimized physical exchange schedule phase=3 initial_horizon=2887 selected_horizon=2887 endpoint_lower_bound=2784 lower_bound_gap=103 selected_kind="full-duplex" neighborhood_improvements=0
2026-09-07T12:42:24.938241Z  INFO ipu_codegen.package.phase{phase="select_finalist"}: finished large physical exchange phase phase=3 tile=919 row_words=11 horizon=2887
2026-09-07T12:42:37.527812Z  INFO ipu_codegen.package.phase{phase="select_finalist"}: prepared large physical exchange phase phase=6 transfers=84731 maximum_identical_destinations=1
2026-09-07T12:42:37.527884Z  INFO ipu_codegen.package.phase{phase="select_finalist"}: optimized physical exchange schedule phase=6 initial_horizon=6389 selected_horizon=6389 endpoint_lower_bound=1312 lower_bound_gap=5077 selected_kind="full-duplex" neighborhood_improvements=0
2026-09-07T12:42:37.581580Z  INFO ipu_codegen.package.phase{phase="select_finalist"}: finished large physical exchange phase phase=6 tile=532 row_words=619 horizon=6389
2026-09-07T12:42:37.676242Z  INFO ipu_codegen.package.phase{phase="select_finalist"}: scheduled operator-plan finalist finalist=6 mapped=false analytical_cycles=394337 analytical_exchange_cycles=145104 scheduled_exchange_cycles=100169 refined_cycles=349402
2026-09-07T12:42:38.672393Z  INFO ipu_codegen.package.phase{phase="select_finalist"}: selected physically scheduled operator-plan finalist selected=6 mapped=false
2026-09-07T12:42:38.672449Z  INFO ipu_codegen.package.phase{phase="select_finalist"}: package build phase finished phase="select_finalist" elapsed_ms=25606 success=true
2026-09-07T12:42:38.672464Z  INFO lowered graph for package construction logical_shards=136188 exchange_phases=19
2026-09-07T12:42:38.672798Z  INFO ipu_codegen.package.phase{phase="compile_runtime"}: package build phase finished phase="compile_runtime" elapsed_ms=0 success=true
2026-09-07T12:42:38.689024Z  INFO ipu_codegen.package.phase{phase="plan_kernels"}: package build phase finished phase="plan_kernels" elapsed_ms=16 success=true
2026-09-07T12:42:38.690702Z  INFO ipu_codegen.package.phase{phase="compile_kernels"}: compiling kernel source=/srv/home/gc-sdk/ipu-stack/crates/ipu-tests/../../device/attention_softmax_f16.S name="attention_softmax_d72_p64_tail" target=ipu21
2026-09-07T12:42:38.752140Z  INFO ipu_codegen.package.phase{phase="compile_kernels"}: kernel artifact created object=/srv/home/.cache/ipu-stack/kernels/d1aeff246c16eee775962d5664dfc77d79facb795f565495484d61f4893c48ba/kernel.o metadata=/srv/home/.cache/ipu-stack/kernels/d1aeff246c16eee775962d5664dfc77d79facb795f565495484d61f4893c48ba/kernel.json
2026-09-07T12:42:38.752814Z  INFO ipu_codegen.package.phase{phase="compile_kernels"}: compiling kernel source=/srv/home/gc-sdk/ipu-stack/crates/ipu-tests/../../device/attention_softmax_f16.S name="attention_softmax_d72_p64_full" target=ipu21
2026-09-07T12:42:38.812427Z  INFO ipu_codegen.package.phase{phase="compile_kernels"}: kernel artifact created object=/srv/home/.cache/ipu-stack/kernels/52cced871f4bcb4740a48d8e795cf5d54ec677212d9b13f9fd6852637690bf38/kernel.o metadata=/srv/home/.cache/ipu-stack/kernels/52cced871f4bcb4740a48d8e795cf5d54ec677212d9b13f9fd6852637690bf38/kernel.json
2026-09-07T12:42:38.812914Z  INFO ipu_codegen.package.phase{phase="compile_kernels"}: package build phase finished phase="compile_kernels" elapsed_ms=123 success=true
2026-09-07T12:42:38.814979Z  INFO ipu_codegen.package.phase{phase="link_runtime"}: package build phase finished phase="link_runtime" elapsed_ms=0 success=true
2026-09-07T12:42:40.234951Z  INFO ipu_codegen.package.phase{phase="size_tile_code"}: package build phase finished phase="size_tile_code" elapsed_ms=21 success=true
2026-09-07T12:42:40.235052Z  INFO allocated package support memory linked_end=325244 profile_bytes=0 exchange_table_bytes=4424 host_code_bytes=4540 generated_code_bytes=3204 code_address=397824 standard_ranges=[(409600, 524288)]
2026-09-07T12:42:40.411120Z  INFO ipu_codegen.package.phase{phase="place_storage"}: package build phase finished phase="place_storage" elapsed_ms=176 success=true
2026-09-07T12:42:40.417505Z  INFO ipu_codegen.package.phase{phase="lower_exchanges"}: reused exchange optimization after validating relocated rows phase=4
2026-09-07T12:42:40.422998Z  INFO ipu_codegen.package.phase{phase="lower_exchanges"}: reused exchange optimization after validating relocated rows phase=3
2026-09-07T12:42:40.423562Z  INFO ipu_codegen.package.phase{phase="lower_exchanges"}: prepared large physical exchange phase phase=3 transfers=1288 maximum_identical_destinations=1
2026-09-07T12:42:40.423581Z  INFO ipu_codegen.package.phase{phase="lower_exchanges"}: optimized physical exchange schedule phase=3 initial_horizon=2887 selected_horizon=2887 endpoint_lower_bound=2784 lower_bound_gap=103 selected_kind="reused" neighborhood_improvements=0
2026-09-07T12:42:40.424121Z  INFO ipu_codegen.package.phase{phase="lower_exchanges"}: reused exchange optimization after validating relocated rows phase=5
2026-09-07T12:42:40.424466Z  INFO ipu_codegen.package.phase{phase="lower_exchanges"}: finished large physical exchange phase phase=3 tile=919 row_words=11 horizon=2887
2026-09-07T12:42:40.424688Z  INFO ipu_codegen.package.phase{phase="lower_exchanges"}: prepared large physical exchange phase phase=5 transfers=1334 maximum_identical_destinations=1
2026-09-07T12:42:40.424705Z  INFO ipu_codegen.package.phase{phase="lower_exchanges"}: optimized physical exchange schedule phase=5 initial_horizon=2897 selected_horizon=2897 endpoint_lower_bound=2784 lower_bound_gap=113 selected_kind="reused" neighborhood_improvements=0
2026-09-07T12:42:40.425645Z  INFO ipu_codegen.package.phase{phase="lower_exchanges"}: finished large physical exchange phase phase=5 tile=1291 row_words=11 horizon=2897
2026-09-07T12:42:40.549722Z  INFO ipu_codegen.package.phase{phase="lower_exchanges"}: reused exchange optimization after validating relocated rows phase=13
2026-09-07T12:42:40.550264Z  INFO ipu_codegen.package.phase{phase="lower_exchanges"}: reused exchange optimization after validating relocated rows phase=7
2026-09-07T12:42:40.552022Z  INFO ipu_codegen.package.phase{phase="lower_exchanges"}: reused exchange optimization after validating relocated rows phase=12
2026-09-07T12:42:40.553864Z  INFO ipu_codegen.package.phase{phase="lower_exchanges"}: reused exchange optimization after validating relocated rows phase=11
2026-09-07T12:42:40.554989Z  INFO ipu_codegen.package.phase{phase="lower_exchanges"}: reused exchange optimization after validating relocated rows phase=8
2026-09-07T12:42:40.558868Z  INFO ipu_codegen.package.phase{phase="lower_exchanges"}: reused exchange optimization after validating relocated rows phase=15
2026-09-07T12:42:40.558911Z  INFO ipu_codegen.package.phase{phase="lower_exchanges"}: reused exchange optimization after validating relocated rows phase=10
2026-09-07T12:42:40.577044Z  INFO ipu_codegen.package.phase{phase="lower_exchanges"}: reused exchange optimization after validating relocated rows phase=2
2026-09-07T12:42:40.577728Z  INFO ipu_codegen.package.phase{phase="lower_exchanges"}: reused exchange optimization after validating relocated rows phase=1
2026-09-07T12:42:40.578452Z  INFO ipu_codegen.package.phase{phase="lower_exchanges"}: prepared large physical exchange phase phase=1 transfers=1334 maximum_identical_destinations=1
2026-09-07T12:42:40.578471Z  INFO ipu_codegen.package.phase{phase="lower_exchanges"}: optimized physical exchange schedule phase=1 initial_horizon=2913 selected_horizon=2913 endpoint_lower_bound=2784 lower_bound_gap=129 selected_kind="reused" neighborhood_improvements=0
2026-09-07T12:42:40.579344Z  INFO ipu_codegen.package.phase{phase="lower_exchanges"}: finished large physical exchange phase phase=1 tile=367 row_words=11 horizon=2913
2026-09-07T12:42:40.651448Z  INFO ipu_codegen.package.phase{phase="lower_exchanges"}: reused exchange optimization after validating relocated rows phase=9
2026-09-07T12:42:40.654657Z  INFO ipu_codegen.package.phase{phase="lower_exchanges"}: reused exchange optimization after validating relocated rows phase=18
2026-09-07T12:42:40.655751Z  INFO ipu_codegen.package.phase{phase="lower_exchanges"}: reused exchange optimization after validating relocated rows phase=14
2026-09-07T12:42:40.657829Z  INFO ipu_codegen.package.phase{phase="lower_exchanges"}: reused exchange optimization after validating relocated rows phase=17
2026-09-07T12:42:40.659779Z  INFO ipu_codegen.package.phase{phase="lower_exchanges"}: reused exchange optimization after validating relocated rows phase=16
2026-09-07T12:42:40.680081Z  INFO ipu_codegen.package.phase{phase="lower_exchanges"}: reused exchange optimization after validating relocated rows phase=0
2026-09-07T12:42:46.022423Z  INFO ipu_codegen.package.phase{phase="lower_exchanges"}: reused exchange optimization after validating relocated rows phase=6
2026-09-07T12:42:46.072076Z  INFO ipu_codegen.package.phase{phase="lower_exchanges"}: prepared large physical exchange phase phase=6 transfers=84731 maximum_identical_destinations=1
2026-09-07T12:42:46.072121Z  INFO ipu_codegen.package.phase{phase="lower_exchanges"}: optimized physical exchange schedule phase=6 initial_horizon=6389 selected_horizon=6389 endpoint_lower_bound=1312 lower_bound_gap=5077 selected_kind="reused" neighborhood_improvements=0
2026-09-07T12:42:46.132530Z  INFO ipu_codegen.package.phase{phase="lower_exchanges"}: finished large physical exchange phase phase=6 tile=532 row_words=619 horizon=6389
2026-09-07T12:42:46.165723Z  INFO ipu_codegen.package.phase{phase="lower_exchanges"}: package build phase finished phase="lower_exchanges" elapsed_ms=5754 success=true
2026-09-07T12:42:48.320466Z  INFO ipu_codegen.package.phase{phase="optimize_exchange_placement"}: package build phase finished phase="optimize_exchange_placement" elapsed_ms=2154 success=true
2026-09-07T12:42:48.340351Z  INFO costed final placed program final_cycles=349402 final_exchange=100169
2026-09-07T12:42:49.075457Z  INFO ipu_codegen.package.phase{phase="prepare_tile_code"}: package build phase finished phase="prepare_tile_code" elapsed_ms=27 success=true
2026-09-07T12:42:49.110639Z  INFO ipu_codegen.package.phase{phase="emit_tile_code"}: package build phase finished phase="emit_tile_code" elapsed_ms=35 success=true
2026-09-07T12:42:49.355014Z  INFO ipu_codegen.package.phase{phase="build_tile_images"}: package build phase finished phase="build_tile_images" elapsed_ms=221 success=true
2026-09-07T12:42:49.601997Z  INFO writing application package tiles=1472
2026-09-07T12:42:49.662178Z  INFO application package written
2026-09-07T12:42:49.662250Z  INFO reading application package
2026-09-07T12:42:49.767136Z  INFO application package read tiles=1472 compiler=0.1.0
2026-09-07T12:42:49.770456Z  INFO waiting for hardware ownership path=artifacts/layout-sweep/device.lock
2026-09-07T12:42:49.770479Z  INFO acquired hardware ownership path=artifacts/layout-sweep/device.lock
2026-09-07T12:42:49.770486Z  INFO opening IPU device path="/dev/ipu0"
2026-09-07T12:42:49.770533Z  INFO IPU device attached path=/dev/ipu0
2026-09-07T12:42:49.770539Z  INFO initializing IPU device
2026-09-07T12:42:49.778368Z  INFO IPU device initialized
2026-09-07T12:42:49.778448Z  INFO replaying device configuration register_writes=2003
2026-09-07T12:42:49.804352Z  INFO loading application tile_count=1472 final_mark=23
2026-09-07T12:42:50.064139Z  INFO application loaded and initial synchronization completed tile_count=1472
2026-09-07T12:43:00.389165Z  INFO starting host exchange session startup_mark=23
2026-09-07T12:43:00.389255Z  INFO attaching host exchange pages pages=2
2026-09-07T12:43:00.389900Z  INFO host exchange session started
2026-09-07T12:43:00.391994Z  INFO invoking streaming host exchange call call="initialize" command=0 phases=132
2026-09-07T12:43:00.612963Z  INFO invoking streaming host exchange call call="run" command=0 phases=7
checkpoint=0 operation=0 tensors=1 samples=256 maximumAbsoluteError=0.000732
checkpoint=1 operation=1 tensors=1 samples=256 maximumAbsoluteError=0.000732
checkpoint=2 operation=2 tensors=1 samples=256 maximumAbsoluteError=0.000549
checkpoint=3 operation=6 tensors=1 samples=256 maximumAbsoluteError=0.000028
diagnosticCheckpoints=4 sampleLimit=256 numericalTest=PASS
package=artifacts/softmax-upgrade/flash-diagnostic/model.ipuexe tiles=1472 hardwareTest=PASS