2026-09-22T15:53:26.729084Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:compile_runtime: close time.busy=278µs time.idle=3.80µs 2026-09-22T15:54:19.445351Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:construct_mid:gemm_candidates: close time.busy=52.7s time.idle=4.03µs 2026-09-22T15:55:59.236564Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:construct_mid:gemm_candidates: close time.busy=99.8s time.idle=4.40µs 2026-09-22T15:56:02.243087Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:construct_mid: close time.busy=156s time.idle=2.44µs 2026-09-22T15:56:02.305877Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}: wrote planner memory profile path=artifacts/kernel-cost-mlp-final-20260922/fp8-b2-n1-default/memory/baseline-1179116-0000.json scope="baseline" peak_step=2 effective_peak_bytes=246784 2026-09-22T15:56:03.129516Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:group_exchanges: close time.busy=82.4µs time.idle=3.02µs 2026-09-22T15:56:03.130745Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:merge_copies: close time.busy=1.04ms time.idle=888ns 2026-09-22T15:56:03.154846Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:eliminate: eliminated padding clears using kernel access contracts removed=0 2026-09-22T15:56:03.155895Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:eliminate: close time.busy=25.1ms time.idle=423ns 2026-09-22T15:56:03.189528Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:place_provisional_storage: close time.busy=33.1ms time.idle=970ns 2026-09-22T15:56:03.446022Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:schedule_provisional_exchanges: prepared large physical exchange phase phase=4 transfers=2520 maximum_identical_destinations=1 2026-09-22T15:56:03.446049Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:schedule_provisional_exchanges: optimized physical exchange schedule phase=4 initial_horizon=12660 selected_horizon=12660 endpoint_lower_bound=12288 lower_bound_gap=372 selected_kind="full-duplex" neighborhood_improvements=0 2026-09-22T15:56:03.764741Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:schedule_provisional_exchanges: prepared large physical exchange phase phase=1 transfers=22100 maximum_identical_destinations=1 2026-09-22T15:56:03.764783Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:schedule_provisional_exchanges: optimized physical exchange schedule phase=1 initial_horizon=11012 selected_horizon=11012 endpoint_lower_bound=9120 lower_bound_gap=1892 selected_kind="full-duplex" neighborhood_improvements=0 2026-09-22T15:56:04.245611Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:schedule_provisional_exchanges: retrying exchange schedule with incremental instruction-alignment validation transfers=104724 2026-09-22T15:56:05.007238Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:schedule_provisional_exchanges: retrying exchange schedule with incremental instruction-alignment validation transfers=16932 2026-09-22T15:56:06.118809Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:schedule_provisional_exchanges: finished incremental instruction-alignment validation transfers=104724 elapsed_ms=1873 success=true 2026-09-22T15:56:06.716424Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:schedule_provisional_exchanges: retrying exchange schedule with incremental instruction-alignment validation transfers=284688 2026-09-22T15:56:07.037376Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:schedule_provisional_exchanges: compared ordinary and paired exchange schedules phase=3 candidates=23559 ordinary_horizon=13859 paired_horizon=13055 use_paired=true 2026-09-22T15:56:07.095281Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:schedule_provisional_exchanges: prepared large physical exchange phase phase=3 transfers=24786 maximum_identical_destinations=1 2026-09-22T15:56:07.095302Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:schedule_provisional_exchanges: optimized physical exchange schedule phase=3 initial_horizon=13055 selected_horizon=13055 endpoint_lower_bound=6976 lower_bound_gap=6079 selected_kind="full-duplex" neighborhood_improvements=0 2026-09-22T15:56:07.935152Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:schedule_provisional_exchanges: prepared large physical exchange phase phase=6 transfers=104724 maximum_identical_destinations=1 2026-09-22T15:56:07.935192Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:schedule_provisional_exchanges: optimized physical exchange schedule phase=6 initial_horizon=936 selected_horizon=936 endpoint_lower_bound=672 lower_bound_gap=264 selected_kind="full-duplex" neighborhood_improvements=0 2026-09-22T15:56:08.348653Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:schedule_provisional_exchanges: finished incremental instruction-alignment validation transfers=16932 elapsed_ms=3341 success=true 2026-09-22T15:56:11.516039Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:schedule_provisional_exchanges: finished incremental instruction-alignment validation transfers=284688 elapsed_ms=4799 success=true 2026-09-22T15:56:11.737626Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:schedule_provisional_exchanges: retrying exchange schedule with incremental instruction-alignment validation transfers=16932 2026-09-22T15:56:14.839927Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:schedule_provisional_exchanges: retrying exchange schedule with incremental instruction-alignment validation transfers=385393 2026-09-22T15:56:15.060884Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:schedule_provisional_exchanges: finished incremental instruction-alignment validation transfers=16932 elapsed_ms=3323 success=true 2026-09-22T15:56:16.501060Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:schedule_provisional_exchanges: prepared large physical exchange phase phase=5 transfers=284688 maximum_identical_destinations=1 2026-09-22T15:56:16.501084Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:schedule_provisional_exchanges: optimized physical exchange schedule phase=5 initial_horizon=12196 selected_horizon=12196 endpoint_lower_bound=10752 lower_bound_gap=1444 selected_kind="full-duplex" neighborhood_improvements=0 2026-09-22T15:56:16.679613Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:schedule_provisional_exchanges: compared ordinary and paired exchange schedules phase=0 candidates=3644 ordinary_horizon=28574 paired_horizon=29853 use_paired=false 2026-09-22T15:56:16.848504Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:schedule_provisional_exchanges: prepared large physical exchange phase phase=0 transfers=16932 maximum_identical_destinations=1 2026-09-22T15:56:16.848529Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:schedule_provisional_exchanges: optimized physical exchange schedule phase=0 initial_horizon=28574 selected_horizon=28574 endpoint_lower_bound=27144 lower_bound_gap=1430 selected_kind="full-duplex" neighborhood_improvements=0 2026-09-22T15:56:28.187808Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:schedule_provisional_exchanges: finished incremental instruction-alignment validation transfers=385393 elapsed_ms=13347 success=true 2026-09-22T15:56:36.274673Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:schedule_provisional_exchanges: prepared large physical exchange phase phase=2 transfers=385393 maximum_identical_destinations=1 2026-09-22T15:56:36.274701Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:schedule_provisional_exchanges: optimized physical exchange schedule phase=2 initial_horizon=3775 selected_horizon=3775 endpoint_lower_bound=2280 lower_bound_gap=1495 selected_kind="full-duplex" neighborhood_improvements=0 2026-09-22T15:56:36.727071Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:schedule_provisional_exchanges: close time.busy=33.5s time.idle=4.08µs 2026-09-22T15:56:36.768394Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:plan_kernels: close time.busy=41.2ms time.idle=3.11µs 2026-09-22T15:56:36.768560Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:compile_kernels: compiling kernel source=/srv/home/gc-sdk/ipu-stack/crates/ipu-tests/../../device/rearrange_f8.cpp name="rearrange_row_major_to_amp_f8_o0_r57_p57_c288_p288_codelet" target="ipu21" 2026-09-22T15:56:37.067678Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:compile_kernels: kernel artifact created object=/srv/home/.cache/ipu-stack/kernels/eadee33c7b291eb28057df512b36c5356f5f86a46f95e9f4e7b0485a6c6f0ae8/kernel.o metadata=/srv/home/.cache/ipu-stack/kernels/eadee33c7b291eb28057df512b36c5356f5f86a46f95e9f4e7b0485a6c6f0ae8/kernel.json 2026-09-22T15:56:37.068020Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:compile_kernels: compiling kernel source=/srv/home/gc-sdk/ipu-stack/crates/ipu-tests/../../device/worker_call.S name="rearrange_row_major_to_amp_f8_o0_r57_p57_c288_p288_wrapper" target="ipu21" 2026-09-22T15:56:37.114067Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:compile_kernels: kernel artifact created object=/srv/home/.cache/ipu-stack/kernels/b098327ea621799b4d3ba42a1d330df36537d798723c4c419fde43e8f5f01c4d/kernel.o metadata=/srv/home/.cache/ipu-stack/kernels/b098327ea621799b4d3ba42a1d330df36537d798723c4c419fde43e8f5f01c4d/kernel.json 2026-09-22T15:56:37.114266Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:compile_kernels: compiling kernel source=/srv/home/gc-sdk/ipu-stack/crates/ipu-tests/../../device/rearrange_f8.cpp name="rearrange_row_major_to_amp_f8_o0_r56_p56_c288_p288_codelet" target="ipu21" 2026-09-22T15:56:37.405415Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:compile_kernels: kernel artifact created object=/srv/home/.cache/ipu-stack/kernels/ed2b6548df05689bb8c141c57f6d7d759e096f4a56f6b82bde5f12a51bffcf8d/kernel.o metadata=/srv/home/.cache/ipu-stack/kernels/ed2b6548df05689bb8c141c57f6d7d759e096f4a56f6b82bde5f12a51bffcf8d/kernel.json 2026-09-22T15:56:37.405577Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:compile_kernels: compiling kernel source=/srv/home/gc-sdk/ipu-stack/crates/ipu-tests/../../device/worker_call.S name="rearrange_row_major_to_amp_f8_o0_r56_p56_c288_p288_wrapper" target="ipu21" 2026-09-22T15:56:37.444360Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:compile_kernels: kernel artifact created object=/srv/home/.cache/ipu-stack/kernels/8ae97c812f2b0b52b8bd9a0befbb84b48e0cb1b6cb579075040a49af67791b73/kernel.o metadata=/srv/home/.cache/ipu-stack/kernels/8ae97c812f2b0b52b8bd9a0befbb84b48e0cb1b6cb579075040a49af67791b73/kernel.json 2026-09-22T15:56:37.445405Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:compile_kernels: compiling kernel source=/srv/home/gc-sdk/ipu-stack/crates/ipu-tests/../../device/gemm_f16_amp.S name="gemm_f8_interleaved_packed0_k288_c320_r57" target="ipu21" 2026-09-22T15:56:37.488966Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:compile_kernels: kernel artifact created object=/srv/home/.cache/ipu-stack/kernels/c7df0070da57764fe064905b4d52c32cf0482820291cf548ca250352948056a1/kernel.o metadata=/srv/home/.cache/ipu-stack/kernels/c7df0070da57764fe064905b4d52c32cf0482820291cf548ca250352948056a1/kernel.json 2026-09-22T15:56:37.489467Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:compile_kernels: compiling kernel source=/srv/home/gc-sdk/ipu-stack/crates/ipu-tests/../../device/gemm_f16_amp.S name="gemm_f8_interleaved_packed0_k288_c320_r56" target="ipu21" 2026-09-22T15:56:37.531548Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:compile_kernels: kernel artifact created object=/srv/home/.cache/ipu-stack/kernels/f04cc46adef0738da821f8d0ddc382d101c2499acc96c6022a09615b1d34db4e/kernel.o metadata=/srv/home/.cache/ipu-stack/kernels/f04cc46adef0738da821f8d0ddc382d101c2499acc96c6022a09615b1d34db4e/kernel.json 2026-09-22T15:56:37.532334Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:compile_kernels: compiling kernel source=/srv/home/gc-sdk/ipu-stack/crates/ipu-tests/../../device/rearrange_amp_left_f16.S name="rearrange_row_major_to_amp_f16_o0_r105_p105_c256_p256" target="ipu21" 2026-09-22T15:56:37.567076Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:compile_kernels: kernel artifact created object=/srv/home/.cache/ipu-stack/kernels/09e916ee5fc756880b87ae8178f2aa890865f6289539481f225fcfe81bad22c1/kernel.o metadata=/srv/home/.cache/ipu-stack/kernels/09e916ee5fc756880b87ae8178f2aa890865f6289539481f225fcfe81bad22c1/kernel.json 2026-09-22T15:56:37.567462Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:compile_kernels: compiling kernel source=/srv/home/gc-sdk/ipu-stack/crates/ipu-tests/../../device/rearrange_amp_left_f16.S name="rearrange_row_major_to_amp_f16_o0_r105_p105_c208_p256" target="ipu21" 2026-09-22T15:56:37.601647Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:compile_kernels: kernel artifact created object=/srv/home/.cache/ipu-stack/kernels/58e0bcbd9d738753676ee6d96e36f3f64ba2297228b6105c8b69e789c042c14e/kernel.o metadata=/srv/home/.cache/ipu-stack/kernels/58e0bcbd9d738753676ee6d96e36f3f64ba2297228b6105c8b69e789c042c14e/kernel.json 2026-09-22T15:56:37.602004Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:compile_kernels: compiling kernel source=/srv/home/gc-sdk/ipu-stack/crates/ipu-tests/../../device/rearrange_amp_left_f16.S name="rearrange_row_major_to_amp_f16_o0_r104_p104_c256_p256" target="ipu21" 2026-09-22T15:56:37.643364Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:compile_kernels: kernel artifact created object=/srv/home/.cache/ipu-stack/kernels/41d0f2d8a8d124534923b1866bfccb4fb720073da0e2a3b615469ca1381ff667/kernel.o metadata=/srv/home/.cache/ipu-stack/kernels/41d0f2d8a8d124534923b1866bfccb4fb720073da0e2a3b615469ca1381ff667/kernel.json 2026-09-22T15:56:37.643716Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:compile_kernels: compiling kernel source=/srv/home/gc-sdk/ipu-stack/crates/ipu-tests/../../device/rearrange_amp_left_f16.S name="rearrange_row_major_to_amp_f16_o0_r104_p104_c208_p256" target="ipu21" 2026-09-22T15:56:37.675915Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:compile_kernels: kernel artifact created object=/srv/home/.cache/ipu-stack/kernels/decb2c7a49e1de794a2e8ecb4a1972fdd2a87807d6b4a552f4c709adf865af1e/kernel.o metadata=/srv/home/.cache/ipu-stack/kernels/decb2c7a49e1de794a2e8ecb4a1972fdd2a87807d6b4a552f4c709adf865af1e/kernel.json 2026-09-22T15:56:37.677111Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:compile_kernels: compiling kernel source=/srv/home/gc-sdk/ipu-stack/crates/ipu-tests/../../device/gemm_f16_amp.S name="gemm_f8_interleaved_packed0_k256_c192_r105" target="ipu21" 2026-09-22T15:56:37.715356Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:compile_kernels: kernel artifact created object=/srv/home/.cache/ipu-stack/kernels/1bd32a300fe28e348a02baf7aec623fd109888ed65e432193a73c524d3f10a6f/kernel.o metadata=/srv/home/.cache/ipu-stack/kernels/1bd32a300fe28e348a02baf7aec623fd109888ed65e432193a73c524d3f10a6f/kernel.json 2026-09-22T15:56:37.716004Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:compile_kernels: compiling kernel source=/srv/home/gc-sdk/ipu-stack/crates/ipu-tests/../../device/gemm_f16_amp.S name="gemm_f8_interleaved_packed0_k256_c192_r104" target="ipu21" 2026-09-22T15:56:37.747517Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:compile_kernels: kernel artifact created object=/srv/home/.cache/ipu-stack/kernels/9e42f53db99522711c6ee30db1c4ab1e8a9ae1f43d3f48aef9aa8d719c68006f/kernel.o metadata=/srv/home/.cache/ipu-stack/kernels/9e42f53db99522711c6ee30db1c4ab1e8a9ae1f43d3f48aef9aa8d719c68006f/kernel.json 2026-09-22T15:56:37.747744Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:compile_kernels: close time.busy=979ms time.idle=4.71µs 2026-09-22T15:56:37.748386Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:link_runtime: close time.busy=456µs time.idle=2.60µs 2026-09-22T15:56:37.941074Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}: exchange table and repeat patch storage exchange_table_bytes=14688 maximum_uncompressed_repeat_patch_bytes=0 maximum_elided_arithmetic_patch_bytes=0 2026-09-22T15:56:38.265768Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:size_tile_code: close time.busy=26.2ms time.idle=5.27µs 2026-09-22T15:56:38.265825Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}: placing generated tile programs linked_end=319744 host_code_base=376832 host_code_bytes=572 generated_code_bytes=1432 exchange_table_bytes=14688 executable_ranges=[(377472, 524288)] 2026-09-22T15:56:38.265880Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}: allocated package support memory linked_end=319744 profile_bytes=136 exchange_table_bytes=14688 host_code_bytes=572 host_data_bytes=392 generated_code_bytes=1432 code_address=377472 available_ranges=[(327680, 360448), (362600, 376832), (409600, 949168)] 2026-09-22T15:56:38.337864Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:place_storage: close time.busy=59.5ms time.idle=2.61µs 2026-09-22T15:56:38.507238Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:lower_exchanges: reused exchange optimization after validating relocated rows phase=4 2026-09-22T15:56:38.514484Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:lower_exchanges: prepared large physical exchange phase phase=4 transfers=2520 maximum_identical_destinations=1 2026-09-22T15:56:38.514498Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:lower_exchanges: optimized physical exchange schedule phase=4 initial_horizon=12660 selected_horizon=12660 endpoint_lower_bound=12288 lower_bound_gap=372 selected_kind="reused" neighborhood_improvements=0 2026-09-22T15:56:38.518668Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:lower_exchanges: reused exchange optimization after validating relocated rows phase=1 2026-09-22T15:56:38.525123Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:lower_exchanges: prepared large physical exchange phase phase=1 transfers=22100 maximum_identical_destinations=1 2026-09-22T15:56:38.525137Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:lower_exchanges: optimized physical exchange schedule phase=1 initial_horizon=11012 selected_horizon=11012 endpoint_lower_bound=9120 lower_bound_gap=1892 selected_kind="reused" neighborhood_improvements=0 2026-09-22T15:56:39.101550Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:lower_exchanges: reused exchange optimization after validating relocated rows phase=3 2026-09-22T15:56:39.124314Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:lower_exchanges: prepared large physical exchange phase phase=3 transfers=24786 maximum_identical_destinations=1 2026-09-22T15:56:39.124330Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:lower_exchanges: optimized physical exchange schedule phase=3 initial_horizon=13055 selected_horizon=13055 endpoint_lower_bound=6976 lower_bound_gap=6079 selected_kind="reused" neighborhood_improvements=0 2026-09-22T15:56:40.479503Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:lower_exchanges: reused exchange optimization after validating relocated rows phase=6 2026-09-22T15:56:40.510600Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:lower_exchanges: prepared large physical exchange phase phase=6 transfers=104724 maximum_identical_destinations=1 2026-09-22T15:56:40.510621Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:lower_exchanges: optimized physical exchange schedule phase=6 initial_horizon=936 selected_horizon=936 endpoint_lower_bound=672 lower_bound_gap=264 selected_kind="reused" neighborhood_improvements=0 2026-09-22T15:56:44.704588Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:lower_exchanges: reused exchange optimization after validating relocated rows phase=5 2026-09-22T15:56:44.800587Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:lower_exchanges: prepared large physical exchange phase phase=5 transfers=284688 maximum_identical_destinations=1 2026-09-22T15:56:44.800618Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:lower_exchanges: optimized physical exchange schedule phase=5 initial_horizon=12196 selected_horizon=12196 endpoint_lower_bound=10752 lower_bound_gap=1444 selected_kind="reused" neighborhood_improvements=0 2026-09-22T15:56:48.804382Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:lower_exchanges: reused exchange optimization after validating relocated rows phase=0 2026-09-22T15:56:48.881404Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:lower_exchanges: prepared large physical exchange phase phase=0 transfers=16932 maximum_identical_destinations=1 2026-09-22T15:56:48.881423Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:lower_exchanges: optimized physical exchange schedule phase=0 initial_horizon=28574 selected_horizon=28574 endpoint_lower_bound=27144 lower_bound_gap=1430 selected_kind="reused" neighborhood_improvements=0 2026-09-22T15:56:49.027004Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:lower_exchanges: reused exchange optimization after validating relocated rows phase=2 2026-09-22T15:56:49.158458Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:lower_exchanges: prepared large physical exchange phase phase=2 transfers=385393 maximum_identical_destinations=1 2026-09-22T15:56:49.158484Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:lower_exchanges: optimized physical exchange schedule phase=2 initial_horizon=3775 selected_horizon=3775 endpoint_lower_bound=2280 lower_bound_gap=1495 selected_kind="reused" neighborhood_improvements=0 2026-09-22T15:56:49.805188Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:lower_exchanges: close time.busy=11.5s time.idle=4.24µs 2026-09-22T15:56:49.808461Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}: costed final placed program final_cycles=359427 final_exchange=86408 2026-09-22T15:56:50.167595Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:prepare_tile_code: close time.busy=52.3ms time.idle=3.98µs 2026-09-22T15:56:50.186577Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:emit_tile_code: close time.busy=18.9ms time.idle=2.19µs 2026-09-22T15:56:50.346124Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}:build_tile_images: close time.busy=154ms time.idle=1.68µs 2026-09-22T15:56:50.528377Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}: wrote exact tile placement profile path=artifacts/kernel-cost-mlp-final-20260922/fp8-b2-n1-default/memory/placement-1179116.html 2026-09-22T15:56:50.608511Z INFO ipu_codegen.package.build{tile_count=1472 operations=3}: close time.busy=204s time.idle=3.98µs 2026-09-22T15:56:50.612552Z INFO writing application package tiles=1472 2026-09-22T15:56:50.957108Z INFO application package written 2026-09-22T15:56:50.957154Z INFO reading application package 2026-09-22T15:56:51.138959Z INFO application package read tiles=1472 compiler=0.1.0 2026-09-22T15:56:51.139024Z INFO waiting for hardware ownership path=/tmp/ipu-stack-device.lock 2026-09-22T15:56:51.139031Z INFO acquired hardware ownership path=/tmp/ipu-stack-device.lock 2026-09-22T15:56:51.139035Z INFO opening IPU device path="/dev/ipu0" 2026-09-22T15:56:51.139066Z INFO IPU device attached path=/dev/ipu0 2026-09-22T15:56:51.139069Z INFO initializing IPU device 2026-09-22T15:56:51.146862Z INFO IPU device initialized 2026-09-22T15:56:51.146869Z INFO replaying device configuration register_writes=2003 2026-09-22T15:56:51.152335Z INFO loading application tile_count=1472 final_mark=23 2026-09-22T15:56:51.338009Z INFO application loaded and initial synchronization completed tile_count=1472 2026-09-22T15:56:52.179856Z INFO prepared reference inputs elapsed_ms=838 2026-09-22T15:56:52.574453Z INFO configured host reference BLAS threads=4 referencePrecision=device 2026-09-22T15:56:53.391908Z INFO evaluated host reference elapsed_ms=1212 2026-09-22T15:56:53.394644Z INFO starting host exchange session startup_mark=23 2026-09-22T15:56:53.394668Z INFO attaching host exchange pages pages=2 2026-09-22T15:56:53.395458Z INFO host exchange session started 2026-09-22T15:56:53.395479Z INFO invoking host exchange call call="initialize" command=0 phases=4 2026-09-22T15:56:53.399774Z INFO invoking host exchange call call="run" command=0 phases=11 referenceInference=1 numericalTest=PASS profile=artifacts/kernel-cost-mlp-final-20260922/fp8-b2-n1-default/profile.capnp tiles=1472 samples=28646 workload=siglip-mlp-f16-b2-t729-d1152-h4304-n1 benchmark=siglip-mlp-f16 batch=2 tokens=729 rows=1458 dimension=1152 hiddenDimension=4304 blocks=1 biases=false activeTiles=1458 executionTiles=1472 inputBytes=1679616 weightBytes=9934848 cycles=357462 minimumTileCycles=350508 deviceMicroseconds=238.308 effectiveGemmTflops=121.340 peakTflops=282.624 efficiencyPercent=42.93 maximumAbsoluteError=0.005402 package=artifacts/kernel-cost-mlp-final-20260922/fp8-b2-n1-default/model.ipuexe tiles=1472 hardwareTest=PASS