{ "status": "passed", "scope": "Three independent launches on one physical GPU; configured dtype/policy, timing arithmetic, sample errors and separately profiled kernel presence verified. Not a universal hardware ceiling, memory bandwidth measurement, model MFU or complete calibration of all GPUs.", "identities": [ { "repeat": 0, "device": "_CudaDeviceProperties(name='NVIDIA L20X', major=9, minor=0, total_memory=143166MB, multi_processor_count=132, uuid=006eb78c-23cb-f37d-eb7c-0ccb578b8f11, pci_bus_id=8, pci_device_id=0, pci_domain_id=0, L2_cache_size=60MB)", "cuda_visible_devices": "0", "script_sha256": "7f3223715774352b5b899e80609b797e8f0954d59acaf05343f4f9ad90440171" }, { "repeat": 1, "device": "_CudaDeviceProperties(name='NVIDIA L20X', major=9, minor=0, total_memory=143166MB, multi_processor_count=132, uuid=006eb78c-23cb-f37d-eb7c-0ccb578b8f11, pci_bus_id=8, pci_device_id=0, pci_domain_id=0, L2_cache_size=60MB)", "cuda_visible_devices": "0", "script_sha256": "7f3223715774352b5b899e80609b797e8f0954d59acaf05343f4f9ad90440171" }, { "repeat": 2, "device": "_CudaDeviceProperties(name='NVIDIA L20X', major=9, minor=0, total_memory=143166MB, multi_processor_count=132, uuid=006eb78c-23cb-f37d-eb7c-0ccb578b8f11, pci_bus_id=8, pci_device_id=0, pci_domain_id=0, L2_cache_size=60MB)", "cuda_visible_devices": "0", "script_sha256": "7f3223715774352b5b899e80609b797e8f0954d59acaf05343f4f9ad90440171" } ], "rows": [ { "repeat": 0, "mode": "bf16", "n": 4096, "median_ms": 0.1745920032262802, "tflops": 787.2007361864795, "sample_max_abs": 0.21205348521471024, "sample_relative_l2": 0.001489220105284778, "kernels": [ "nvjet_sm90_tst_256x128_64x4_1x2_h_bz_coopA_NNT" ], "checks": { "100_positive_samples": true, "median_recomputed": true, "flops_recomputed": true, "throughput_recomputed": true, "finite_saved_sample": true, "sample_error_recomputed": true, "gpu_kernel_recorded": true, "policy_matches": true }, "passed": true, "manifest": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_calibration_v1_r0/manifest.json", "sample_sha256": "41879a27c3bbc5af65d2045530858ba3ce440fa57ef39eb5460da5a993c2c8ac", "trace_sha256": "6b9ba77e181577140edf9fa6d756bcc0bc181ceaeb7feb30bdfceb19cd99cac8" }, { "repeat": 0, "mode": "bf16", "n": 8192, "median_ms": 1.583679974079132, "tflops": 694.2763978658862, "sample_max_abs": 0.7960862126201391, "sample_relative_l2": 0.0017052041160023565, "kernels": [ "nvjet_sm90_tst_320x128_64x3_1x2_h_bz_coopB_NNT" ], "checks": { "100_positive_samples": true, "median_recomputed": true, "flops_recomputed": true, "throughput_recomputed": true, "finite_saved_sample": true, "sample_error_recomputed": true, "gpu_kernel_recorded": true, "policy_matches": true }, "passed": true, "manifest": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_calibration_v1_r0/manifest.json", "sample_sha256": "c8ec82ddc420f3b09f36c67359d348a736834d7f8e1a34dc1a79ea9b037f14e5", "trace_sha256": "8317ee709c3dab12a77c4b0d37d25d5f1b9feaf64437bb8b4fcee89118cc4a49" }, { "repeat": 0, "mode": "bf16", "n": 12288, "median_ms": 5.245408058166504, "tflops": 707.4476766333986, "sample_max_abs": 0.23426778620341793, "sample_relative_l2": 0.000944183566314058, "kernels": [ "nvjet_sm90_tst_192x192_64x3_2x1_v_bz_coopB_NNN" ], "checks": { "100_positive_samples": true, "median_recomputed": true, "flops_recomputed": true, "throughput_recomputed": true, "finite_saved_sample": true, "sample_error_recomputed": true, "gpu_kernel_recorded": true, "policy_matches": true }, "passed": true, "manifest": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_calibration_v1_r0/manifest.json", "sample_sha256": "327de61023cd329fc3688eb8477b8f6e3295e279e9126aaf172f6099caeba72b", "trace_sha256": "4b0fc7fd2ac0cc8a10c18a5cd81e551bcc6b673c1fa9b92e5506e7f4412a3a62" }, { "repeat": 0, "mode": "fp32_ieee", "n": 4096, "median_ms": 2.6854079961776733, "tflops": 51.17991518146455, "sample_max_abs": 0.00017243917903897454, "sample_relative_l2": 1.2215468105051305e-06, "kernels": [ "sm80_xmma_gemm_f32f32_f32f32_f32_nn_n_tilesize256x128x8_stage3_warpsize4x2x1_ffma_aligna4_alignc4_execute_kernel__5x_cublas" ], "checks": { "100_positive_samples": true, "median_recomputed": true, "flops_recomputed": true, "throughput_recomputed": true, "finite_saved_sample": true, "sample_error_recomputed": true, "gpu_kernel_recorded": true, "policy_matches": true }, "passed": true, "manifest": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_calibration_v1_r0/manifest.json", "sample_sha256": "c5d4cd32738feb4dfe1e6fc068460cab1c361bce38965fcb581ef98f88c091be", "trace_sha256": "fa3511be84c3cf3a6a7e746ab473cc6447e3701d636d1d4d26802b0535943fc5" }, { "repeat": 0, "mode": "fp32_ieee", "n": 8192, "median_ms": 21.452272415161133, "tflops": 51.25385350779592, "sample_max_abs": 0.0005305199582608111, "sample_relative_l2": 1.7392301630633194e-06, "kernels": [ "sm80_xmma_gemm_f32f32_f32f32_f32_nn_n_tilesize256x128x8_stage3_warpsize4x2x1_ffma_aligna4_alignc4_execute_kernel__5x_cublas" ], "checks": { "100_positive_samples": true, "median_recomputed": true, "flops_recomputed": true, "throughput_recomputed": true, "finite_saved_sample": true, "sample_error_recomputed": true, "gpu_kernel_recorded": true, "policy_matches": true }, "passed": true, "manifest": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_calibration_v1_r0/manifest.json", "sample_sha256": "dde076b83db484998513f57f4c8668a77d9504ffd55427da1c7d685b6170593a", "trace_sha256": "3c701fe03b78610af74aecd9bf955fe7ba324c442f47b938b9e83bbd439a9f8f" }, { "repeat": 0, "mode": "fp32_ieee", "n": 12288, "median_ms": 73.74459075927734, "tflops": 50.32032458973488, "sample_max_abs": 0.0002790938920824715, "sample_relative_l2": 1.2136394292723015e-06, "kernels": [ "void cutlass::Kernel2(cutlass_80_simt_sgemm_256x128_8x4_nn_align1::Params)" ], "checks": { "100_positive_samples": true, "median_recomputed": true, "flops_recomputed": true, "throughput_recomputed": true, "finite_saved_sample": true, "sample_error_recomputed": true, "gpu_kernel_recorded": true, "policy_matches": true }, "passed": true, "manifest": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_calibration_v1_r0/manifest.json", "sample_sha256": "115241b05cd87f4ca3f52554ffb4fb65496254f03872e1d5dcf4f0ce17ac12bf", "trace_sha256": "6b82c0712092c5e781117ef82a153b2eed5a217f0316bea6c23c7204a98b6cad" }, { "repeat": 0, "mode": "fp32_tf32_allowed", "n": 4096, "median_ms": 0.34856000542640686, "tflops": 394.3050015272568, "sample_max_abs": 0.032902745670348565, "sample_relative_l2": 0.00031315381502780934, "kernels": [ "sm90_xmma_gemm_f32f32_tf32f32_f32_nn_n_tilesize128x256x32_warpgroupsize2x1x1_execute_segment_k_off_kernel__5x_cublas" ], "checks": { "100_positive_samples": true, "median_recomputed": true, "flops_recomputed": true, "throughput_recomputed": true, "finite_saved_sample": true, "sample_error_recomputed": true, "gpu_kernel_recorded": true, "policy_matches": true }, "passed": true, "manifest": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_calibration_v1_r0/manifest.json", "sample_sha256": "73404de058399df07678ec601175d3522e035357e611166665e22da9f51bd074", "trace_sha256": "40bc54b084c05e2ed56b8d4d604e566a1f1181b08fa88cdff557f42e441d92c9" }, { "repeat": 0, "mode": "fp32_tf32_allowed", "n": 8192, "median_ms": 3.0247679948806763, "tflops": 363.50279744988325, "sample_max_abs": 0.04182198168585671, "sample_relative_l2": 0.00017571443710893855, "kernels": [ "sm90_xmma_gemm_f32f32_tf32f32_f32_nn_n_tilesize128x256x32_warpgroupsize2x1x1_execute_segment_k_off_kernel__5x_cublas" ], "checks": { "100_positive_samples": true, "median_recomputed": true, "flops_recomputed": true, "throughput_recomputed": true, "finite_saved_sample": true, "sample_error_recomputed": true, "gpu_kernel_recorded": true, "policy_matches": true }, "passed": true, "manifest": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_calibration_v1_r0/manifest.json", "sample_sha256": "6854e7a3023501d559c7df3df704ab9f8f2dc4b7060d4aef535e9900242a3bd0", "trace_sha256": "559a72ca1708e2b4d4e0e4c029dea5251f90f46d0db03524f4c6fdd5bd87d40b" }, { "repeat": 0, "mode": "fp32_tf32_allowed", "n": 12288, "median_ms": 10.352112293243408, "tflops": 358.46324292347464, "sample_max_abs": 0.06719220805530313, "sample_relative_l2": 0.00031417125138381976, "kernels": [ "sm90_xmma_gemm_f32f32_tf32f32_f32_nn_n_tilesize128x256x32_warpgroupsize2x1x1_execute_segment_k_off_kernel__5x_cublas" ], "checks": { "100_positive_samples": true, "median_recomputed": true, "flops_recomputed": true, "throughput_recomputed": true, "finite_saved_sample": true, "sample_error_recomputed": true, "gpu_kernel_recorded": true, "policy_matches": true }, "passed": true, "manifest": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_calibration_v1_r0/manifest.json", "sample_sha256": "e88bb5e9531399c6aa9317ea83d886cd8d4d09e361049c00e5cac9fea7713da3", "trace_sha256": "2a17ed57b9c3a642187121df908c2dacc2838d931f15c0eb4659f40e6e6b4fcd" }, { "repeat": 1, "mode": "bf16", "n": 4096, "median_ms": 0.1746239960193634, "tflops": 787.0565134516788, "sample_max_abs": 0.21205348521471024, "sample_relative_l2": 0.001489220105284778, "kernels": [ "nvjet_sm90_tst_256x128_64x4_1x2_h_bz_coopA_NNT" ], "checks": { "100_positive_samples": true, "median_recomputed": true, "flops_recomputed": true, "throughput_recomputed": true, "finite_saved_sample": true, "sample_error_recomputed": true, "gpu_kernel_recorded": true, "policy_matches": true }, "passed": true, "manifest": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_calibration_v1_r1/manifest.json", "sample_sha256": "41879a27c3bbc5af65d2045530858ba3ce440fa57ef39eb5460da5a993c2c8ac", "trace_sha256": "7af5025c92e07be0b5b5bd0ce4d6a18c8eed48119dfae3fdb4d50f7d5a084921" }, { "repeat": 1, "mode": "bf16", "n": 8192, "median_ms": 1.584719955921173, "tflops": 693.8207748743032, "sample_max_abs": 0.7960862126201391, "sample_relative_l2": 0.0017052041160023565, "kernels": [ "nvjet_sm90_tst_320x128_64x3_1x2_h_bz_coopB_NNT" ], "checks": { "100_positive_samples": true, "median_recomputed": true, "flops_recomputed": true, "throughput_recomputed": true, "finite_saved_sample": true, "sample_error_recomputed": true, "gpu_kernel_recorded": true, "policy_matches": true }, "passed": true, "manifest": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_calibration_v1_r1/manifest.json", "sample_sha256": "c8ec82ddc420f3b09f36c67359d348a736834d7f8e1a34dc1a79ea9b037f14e5", "trace_sha256": "cb64a342a232cc8816672d999c22cfc5536ce17cda28823456ead4035ff8037c" }, { "repeat": 1, "mode": "bf16", "n": 12288, "median_ms": 5.248816013336182, "tflops": 706.988344479112, "sample_max_abs": 0.23426778620341793, "sample_relative_l2": 0.000944183566314058, "kernels": [ "nvjet_sm90_tst_192x192_64x3_2x1_v_bz_coopB_NNN" ], "checks": { "100_positive_samples": true, "median_recomputed": true, "flops_recomputed": true, "throughput_recomputed": true, "finite_saved_sample": true, "sample_error_recomputed": true, "gpu_kernel_recorded": true, "policy_matches": true }, "passed": true, "manifest": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_calibration_v1_r1/manifest.json", "sample_sha256": "327de61023cd329fc3688eb8477b8f6e3295e279e9126aaf172f6099caeba72b", "trace_sha256": "1d737401a6239f2760234a6d35c59c4ca3631f5b512f57066fd7c1955a6bb3ad" }, { "repeat": 1, "mode": "fp32_ieee", "n": 4096, "median_ms": 2.685728073120117, "tflops": 51.17381571408743, "sample_max_abs": 0.00017243917903897454, "sample_relative_l2": 1.2215468105051305e-06, "kernels": [ "sm80_xmma_gemm_f32f32_f32f32_f32_nn_n_tilesize256x128x8_stage3_warpsize4x2x1_ffma_aligna4_alignc4_execute_kernel__5x_cublas" ], "checks": { "100_positive_samples": true, "median_recomputed": true, "flops_recomputed": true, "throughput_recomputed": true, "finite_saved_sample": true, "sample_error_recomputed": true, "gpu_kernel_recorded": true, "policy_matches": true }, "passed": true, "manifest": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_calibration_v1_r1/manifest.json", "sample_sha256": "c5d4cd32738feb4dfe1e6fc068460cab1c361bce38965fcb581ef98f88c091be", "trace_sha256": "48089405d6f0cce45ca54795b0034aa349d85022b947806dc358d7aa9a063a31" }, { "repeat": 1, "mode": "fp32_ieee", "n": 8192, "median_ms": 21.44215965270996, "tflops": 51.27802635482376, "sample_max_abs": 0.0005305199582608111, "sample_relative_l2": 1.7392301630633194e-06, "kernels": [ "sm80_xmma_gemm_f32f32_f32f32_f32_nn_n_tilesize256x128x8_stage3_warpsize4x2x1_ffma_aligna4_alignc4_execute_kernel__5x_cublas" ], "checks": { "100_positive_samples": true, "median_recomputed": true, "flops_recomputed": true, "throughput_recomputed": true, "finite_saved_sample": true, "sample_error_recomputed": true, "gpu_kernel_recorded": true, "policy_matches": true }, "passed": true, "manifest": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_calibration_v1_r1/manifest.json", "sample_sha256": "dde076b83db484998513f57f4c8668a77d9504ffd55427da1c7d685b6170593a", "trace_sha256": "4249bc585d4ee04f942ee8712ab151493cf295d1fb2566e860c730d333d5ca38" }, { "repeat": 1, "mode": "fp32_ieee", "n": 12288, "median_ms": 73.70916748046875, "tflops": 50.344507618096365, "sample_max_abs": 0.0002790938920824715, "sample_relative_l2": 1.2136394292723015e-06, "kernels": [ "void cutlass::Kernel2(cutlass_80_simt_sgemm_256x128_8x4_nn_align1::Params)" ], "checks": { "100_positive_samples": true, "median_recomputed": true, "flops_recomputed": true, "throughput_recomputed": true, "finite_saved_sample": true, "sample_error_recomputed": true, "gpu_kernel_recorded": true, "policy_matches": true }, "passed": true, "manifest": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_calibration_v1_r1/manifest.json", "sample_sha256": "115241b05cd87f4ca3f52554ffb4fb65496254f03872e1d5dcf4f0ce17ac12bf", "trace_sha256": "67c07fd25d1aef172cf748c5fd5258eb58adca911aa621148572e89bd034d035" }, { "repeat": 1, "mode": "fp32_tf32_allowed", "n": 4096, "median_ms": 0.3470880091190338, "tflops": 395.9772445635404, "sample_max_abs": 0.032902745670348565, "sample_relative_l2": 0.00031315381502780934, "kernels": [ "sm90_xmma_gemm_f32f32_tf32f32_f32_nn_n_tilesize128x256x32_warpgroupsize2x1x1_execute_segment_k_off_kernel__5x_cublas" ], "checks": { "100_positive_samples": true, "median_recomputed": true, "flops_recomputed": true, "throughput_recomputed": true, "finite_saved_sample": true, "sample_error_recomputed": true, "gpu_kernel_recorded": true, "policy_matches": true }, "passed": true, "manifest": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_calibration_v1_r1/manifest.json", "sample_sha256": "73404de058399df07678ec601175d3522e035357e611166665e22da9f51bd074", "trace_sha256": "d207dae717731d8a52cd2782d6fa41d5d27dfbb3f738008234d0c661b97d0adb" }, { "repeat": 1, "mode": "fp32_tf32_allowed", "n": 8192, "median_ms": 3.048416018486023, "tflops": 360.6829320894547, "sample_max_abs": 0.04182198168585671, "sample_relative_l2": 0.00017571443710893855, "kernels": [ "sm90_xmma_gemm_f32f32_tf32f32_f32_nn_n_tilesize128x256x32_warpgroupsize2x1x1_execute_segment_k_off_kernel__5x_cublas" ], "checks": { "100_positive_samples": true, "median_recomputed": true, "flops_recomputed": true, "throughput_recomputed": true, "finite_saved_sample": true, "sample_error_recomputed": true, "gpu_kernel_recorded": true, "policy_matches": true }, "passed": true, "manifest": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_calibration_v1_r1/manifest.json", "sample_sha256": "6854e7a3023501d559c7df3df704ab9f8f2dc4b7060d4aef535e9900242a3bd0", "trace_sha256": "56815f0b45556a6754e6f0272a318c54bcb480dd06f92899591baf655f2c2c22" }, { "repeat": 1, "mode": "fp32_tf32_allowed", "n": 12288, "median_ms": 10.359295845031738, "tflops": 358.21467011425335, "sample_max_abs": 0.06719220805530313, "sample_relative_l2": 0.00031417125138381976, "kernels": [ "sm90_xmma_gemm_f32f32_tf32f32_f32_nn_n_tilesize128x256x32_warpgroupsize2x1x1_execute_segment_k_off_kernel__5x_cublas" ], "checks": { "100_positive_samples": true, "median_recomputed": true, "flops_recomputed": true, "throughput_recomputed": true, "finite_saved_sample": true, "sample_error_recomputed": true, "gpu_kernel_recorded": true, "policy_matches": true }, "passed": true, "manifest": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_calibration_v1_r1/manifest.json", "sample_sha256": "e88bb5e9531399c6aa9317ea83d886cd8d4d09e361049c00e5cac9fea7713da3", "trace_sha256": "418f3263215d4daab4fbc891c0bfc0076a669fda70ac43ef11f7199bcc6e0b58" }, { "repeat": 2, "mode": "bf16", "n": 4096, "median_ms": 0.1743519976735115, "tflops": 788.2843632762142, "sample_max_abs": 0.21205348521471024, "sample_relative_l2": 0.001489220105284778, "kernels": [ "nvjet_sm90_tst_256x128_64x4_1x2_h_bz_coopA_NNT" ], "checks": { "100_positive_samples": true, "median_recomputed": true, "flops_recomputed": true, "throughput_recomputed": true, "finite_saved_sample": true, "sample_error_recomputed": true, "gpu_kernel_recorded": true, "policy_matches": true }, "passed": true, "manifest": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_calibration_v1_r2/manifest.json", "sample_sha256": "41879a27c3bbc5af65d2045530858ba3ce440fa57ef39eb5460da5a993c2c8ac", "trace_sha256": "f034ee26d57049401d0d9f75b15f4f2e7772e04b8897e6d1c5c8448d1c176b66" }, { "repeat": 2, "mode": "bf16", "n": 8192, "median_ms": 1.5845440030097961, "tflops": 693.8978189860989, "sample_max_abs": 0.7960862126201391, "sample_relative_l2": 0.0017052041160023565, "kernels": [ "nvjet_sm90_tst_320x128_64x3_1x2_h_bz_coopB_NNT" ], "checks": { "100_positive_samples": true, "median_recomputed": true, "flops_recomputed": true, "throughput_recomputed": true, "finite_saved_sample": true, "sample_error_recomputed": true, "gpu_kernel_recorded": true, "policy_matches": true }, "passed": true, "manifest": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_calibration_v1_r2/manifest.json", "sample_sha256": "c8ec82ddc420f3b09f36c67359d348a736834d7f8e1a34dc1a79ea9b037f14e5", "trace_sha256": "36ac9c82fd0412ab1e76d96692a587ba929330ca02eda3245467ec5d94cd37c2" }, { "repeat": 2, "mode": "bf16", "n": 12288, "median_ms": 5.242448091506958, "tflops": 707.8471124503408, "sample_max_abs": 0.23426778620341793, "sample_relative_l2": 0.000944183566314058, "kernels": [ "nvjet_sm90_tst_192x192_64x3_2x1_v_bz_coopB_NNN" ], "checks": { "100_positive_samples": true, "median_recomputed": true, "flops_recomputed": true, "throughput_recomputed": true, "finite_saved_sample": true, "sample_error_recomputed": true, "gpu_kernel_recorded": true, "policy_matches": true }, "passed": true, "manifest": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_calibration_v1_r2/manifest.json", "sample_sha256": "327de61023cd329fc3688eb8477b8f6e3295e279e9126aaf172f6099caeba72b", "trace_sha256": "3fe025a8b846c275f2167527a5eca99dd7ea3893b1ad7786bba2544671753b6e" }, { "repeat": 2, "mode": "fp32_ieee", "n": 4096, "median_ms": 2.6843039989471436, "tflops": 51.20096439371519, "sample_max_abs": 0.00017243917903897454, "sample_relative_l2": 1.2215468105051305e-06, "kernels": [ "sm80_xmma_gemm_f32f32_f32f32_f32_nn_n_tilesize256x128x8_stage3_warpsize4x2x1_ffma_aligna4_alignc4_execute_kernel__5x_cublas" ], "checks": { "100_positive_samples": true, "median_recomputed": true, "flops_recomputed": true, "throughput_recomputed": true, "finite_saved_sample": true, "sample_error_recomputed": true, "gpu_kernel_recorded": true, "policy_matches": true }, "passed": true, "manifest": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_calibration_v1_r2/manifest.json", "sample_sha256": "c5d4cd32738feb4dfe1e6fc068460cab1c361bce38965fcb581ef98f88c091be", "trace_sha256": "1521061b56898a9c6080524e2be8f7d03821b107330ea541d2206e9b41e548e5" }, { "repeat": 2, "mode": "fp32_ieee", "n": 8192, "median_ms": 21.43668842315674, "tflops": 51.29111391049866, "sample_max_abs": 0.0005305199582608111, "sample_relative_l2": 1.7392301630633194e-06, "kernels": [ "sm80_xmma_gemm_f32f32_f32f32_f32_nn_n_tilesize256x128x8_stage3_warpsize4x2x1_ffma_aligna4_alignc4_execute_kernel__5x_cublas" ], "checks": { "100_positive_samples": true, "median_recomputed": true, "flops_recomputed": true, "throughput_recomputed": true, "finite_saved_sample": true, "sample_error_recomputed": true, "gpu_kernel_recorded": true, "policy_matches": true }, "passed": true, "manifest": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_calibration_v1_r2/manifest.json", "sample_sha256": "dde076b83db484998513f57f4c8668a77d9504ffd55427da1c7d685b6170593a", "trace_sha256": "4646b6cdd11c3979f955e69a852aa61906ff13d6bdef59899528e8f190704feb" }, { "repeat": 2, "mode": "fp32_ieee", "n": 12288, "median_ms": 73.71353530883789, "tflops": 50.341524500169875, "sample_max_abs": 0.0002790938920824715, "sample_relative_l2": 1.2136394292723015e-06, "kernels": [ "void cutlass::Kernel2(cutlass_80_simt_sgemm_256x128_8x4_nn_align1::Params)" ], "checks": { "100_positive_samples": true, "median_recomputed": true, "flops_recomputed": true, "throughput_recomputed": true, "finite_saved_sample": true, "sample_error_recomputed": true, "gpu_kernel_recorded": true, "policy_matches": true }, "passed": true, "manifest": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_calibration_v1_r2/manifest.json", "sample_sha256": "115241b05cd87f4ca3f52554ffb4fb65496254f03872e1d5dcf4f0ce17ac12bf", "trace_sha256": "71734778a01673efa8309d37f5bf18f52499a1bb1dbb9c1037dcfe3bd845efd8" }, { "repeat": 2, "mode": "fp32_tf32_allowed", "n": 4096, "median_ms": 0.3472000062465668, "tflops": 395.84951324683055, "sample_max_abs": 0.032902745670348565, "sample_relative_l2": 0.00031315381502780934, "kernels": [ "sm90_xmma_gemm_f32f32_tf32f32_f32_nn_n_tilesize128x256x32_warpgroupsize2x1x1_execute_segment_k_off_kernel__5x_cublas" ], "checks": { "100_positive_samples": true, "median_recomputed": true, "flops_recomputed": true, "throughput_recomputed": true, "finite_saved_sample": true, "sample_error_recomputed": true, "gpu_kernel_recorded": true, "policy_matches": true }, "passed": true, "manifest": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_calibration_v1_r2/manifest.json", "sample_sha256": "73404de058399df07678ec601175d3522e035357e611166665e22da9f51bd074", "trace_sha256": "ce63818a93960f283f3fc97a9f11e2958bbdd8d267ce5a3e4156ef5d694b813b" }, { "repeat": 2, "mode": "fp32_tf32_allowed", "n": 8192, "median_ms": 3.028656005859375, "tflops": 363.03615387446945, "sample_max_abs": 0.04182198168585671, "sample_relative_l2": 0.00017571443710893855, "kernels": [ "sm90_xmma_gemm_f32f32_tf32f32_f32_nn_n_tilesize128x256x32_warpgroupsize2x1x1_execute_segment_k_off_kernel__5x_cublas" ], "checks": { "100_positive_samples": true, "median_recomputed": true, "flops_recomputed": true, "throughput_recomputed": true, "finite_saved_sample": true, "sample_error_recomputed": true, "gpu_kernel_recorded": true, "policy_matches": true }, "passed": true, "manifest": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_calibration_v1_r2/manifest.json", "sample_sha256": "6854e7a3023501d559c7df3df704ab9f8f2dc4b7060d4aef535e9900242a3bd0", "trace_sha256": "10bbd72518e052597f9ef63bc8f3f7dface5fbfc0d3cab46114f81b2bdc647ab" }, { "repeat": 2, "mode": "fp32_tf32_allowed", "n": 12288, "median_ms": 10.37553596496582, "tflops": 357.65398108339787, "sample_max_abs": 0.06719220805530313, "sample_relative_l2": 0.00031417125138381976, "kernels": [ "sm90_xmma_gemm_f32f32_tf32f32_f32_nn_n_tilesize128x256x32_warpgroupsize2x1x1_execute_segment_k_off_kernel__5x_cublas" ], "checks": { "100_positive_samples": true, "median_recomputed": true, "flops_recomputed": true, "throughput_recomputed": true, "finite_saved_sample": true, "sample_error_recomputed": true, "gpu_kernel_recorded": true, "policy_matches": true }, "passed": true, "manifest": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_calibration_v1_r2/manifest.json", "sample_sha256": "e88bb5e9531399c6aa9317ea83d886cd8d4d09e361049c00e5cac9fea7713da3", "trace_sha256": "463a28e6a3bba02552b13c297296ad336e872e6df9b7922cdd6b59ddcdbb3bb7" } ] }