Renderer-cropped offsets; rounds can overlap; occupancy is measured compute coverage. epoch=0 next=0 compute=0..2592 span=2592 occupancy=2.94% tiles=149 work=112260 exposed=N/A operation.0 copy_u64 tiles=114 calls=167 max=594 cycles operation.0 static_copy_u32 tiles=83 calls=92 max=2148 cycles operation.0 fill_zero_u64 tiles=24 calls=24 max=1002 cycles barrier=0..2592 exchangeEnd=53298 scheduled=50443 lateTile=1025 unattributed=0 exchangeOccupancy=91.56% late tile: operation.0 static_copy_u32 0..2592 late tile: operation.0 copy_u64 102..444 epoch=0 next=1 compute=50754..254016 span=203262 occupancy=95.79% tiles=1440 work=286619040 exposed=200718 operation.0 gemm_f8_init_rows_interleaved_k384_c432_r61 tiles=1080 calls=1080 max=130170 cycles operation.0 rearrange_row_major_to_amp_f8_o0_r61_p61_c384_p384 tiles=1080 calls=1080 max=70548 cycles operation.0 gemm_f8_init_rows_interleaved_k384_c432_r60 tiles=360 calls=360 max=129840 cycles operation.0 rearrange_row_major_to_amp_f8_o0_r60_p60_c384_p384 tiles=360 calls=360 max=64170 cycles barrier=7746..254016 exchangeEnd=269556 scheduled=15233 lateTile=1360 unattributed=0 exchangeOccupancy=84.26% late tile: operation.0 gemm_f8_init_rows_interleaved_k384_c432_r61 123846..254016 late tile: operation.0 rearrange_row_major_to_amp_f8_o0_r61_p61_c384_p384 53298..123846 epoch=0 next=2 compute=263706..292908 span=29202 occupancy=64.84% tiles=1440 work=27872856 exposed=23352 operation.0 reduce_sum_f16 tiles=1440 calls=1440 max=18852 cycles operation.0 copy_strided_u64 tiles=144 calls=144 max=6222 cycles operation.0 copy_u64 tiles=48 calls=48 max=4638 cycles barrier=254274..292908 exchangeEnd=299910 scheduled=6715 lateTile=1059 unattributed=0 exchangeOccupancy=74.51% late tile: operation.0 reduce_sum_f16 274056..292908 late tile: operation.0 copy_strided_u64 267834..274056 epoch=0 next=3 compute=298068..343536 span=45468 occupancy=94.21% tiles=1458 work=63053676 exposed=43626 operation.1 gelu_tanh_approx_f16 tiles=1458 calls=1458 max=42150 cycles operation.1 copy_strided_u64 tiles=1002 calls=1002 max=834 cycles operation.2 copy_u64 tiles=1230 calls=1230 max=642 cycles barrier=293166..343536 exchangeEnd=358470 scheduled=14660 lateTile=31 unattributed=0 exchangeOccupancy=88.94% late tile: operation.1 gelu_tanh_approx_f16 300744..342894 late tile: operation.1 copy_strided_u64 299910..300744 late tile: operation.2 copy_u64 342894..343536 epoch=0 next=4 compute=357294..418392 span=61098 occupancy=56.51% tiles=1440 work=50825664 exposed=59922 operation.2 cast_f16_f8 tiles=1440 calls=1440 max=12342 cycles operation.2 rearrange_row_major_to_amp_f16_o0_r121_p121_c224_p224 tiles=684 calls=684 max=21654 cycles operation.2 rearrange_row_major_to_amp_f16_o0_r122_p122_c224_p224 tiles=684 calls=684 max=21654 cycles operation.2 rearrange_row_major_to_amp_f16_o0_r121_p121_c48_p224 tiles=36 calls=36 max=34254 cycles operation.2 rearrange_row_major_to_amp_f16_o0_r122_p122_c48_p224 tiles=36 calls=36 max=34254 cycles operation.2 fill_zero_u64 tiles=72 calls=1728 max=558 cycles barrier=343776..418392 exchangeEnd=440472 scheduled=21802 lateTile=951 unattributed=0 exchangeOccupancy=93.25% late tile: operation.2 rearrange_row_major_to_amp_f16_o0_r122_p122_c48_p224 358404..392658 late tile: operation.2 fill_zero_u64 405000..418392 late tile: operation.2 cast_f16_f8 392658..405000 epoch=0 next=5 compute=438390..549018 span=110628 occupancy=95.89% tiles=1440 work=156150720 exposed=108546 operation.2 gemm_f8_init_rows_interleaved_k224_c384_r122 tiles=720 calls=720 max=108546 cycles operation.2 gemm_f8_init_rows_interleaved_k224_c384_r121 tiles=720 calls=720 max=108330 cycles barrier=418632..549018 exchangeEnd=577734 scheduled=28426 lateTile=1122 unattributed=0 exchangeOccupancy=80.52% late tile: operation.2 gemm_f8_init_rows_interleaved_k224_c384_r122 440472..549018 epoch=0 next=6 compute=571794..605406 span=33612 occupancy=61.00% tiles=1440 work=30178584 exposed=27672 operation.2 reduce_sum_f16 tiles=1440 calls=1440 max=25890 cycles operation.2 copy_strided_u64 tiles=1428 calls=1428 max=1782 cycles barrier=549282..605406 exchangeEnd=607320 scheduled=1672 lateTile=157 unattributed=0 exchangeOccupancy=75.74% late tile: operation.2 reduce_sum_f16 579516..605406 late tile: operation.2 copy_strided_u64 577734..579516 epoch=0 next=end compute=606822..608454 span=1632 occupancy=0.52% tiles=12 work=12420 exposed=N/A copy_strided_u64 tiles=12 calls=12 max=1170 cycles