{ "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": [ { "pid": 791028, "uuid": "GPU-46890c79-bf10-5b49-3d4a-a6b78f956122", "repeat": 0, "device": "_CudaDeviceProperties(name='NVIDIA L20X', major=9, minor=0, total_memory=143166MB, multi_processor_count=132, uuid=46890c79-bf10-5b49-3d4a-a6b78f956122, pci_bus_id=163, pci_device_id=0, pci_domain_id=1, L2_cache_size=60MB)", "cuda_visible_devices": "6", "script_sha256": "e2df72bbcb703640baf64ba38392d6f28ceb06a23c0dead33048204693c11cfd" }, { "pid": 795747, "uuid": "GPU-46890c79-bf10-5b49-3d4a-a6b78f956122", "repeat": 1, "device": "_CudaDeviceProperties(name='NVIDIA L20X', major=9, minor=0, total_memory=143166MB, multi_processor_count=132, uuid=46890c79-bf10-5b49-3d4a-a6b78f956122, pci_bus_id=163, pci_device_id=0, pci_domain_id=1, L2_cache_size=60MB)", "cuda_visible_devices": "6", "script_sha256": "e2df72bbcb703640baf64ba38392d6f28ceb06a23c0dead33048204693c11cfd" }, { "pid": 812254, "uuid": "GPU-46890c79-bf10-5b49-3d4a-a6b78f956122", "repeat": 2, "device": "_CudaDeviceProperties(name='NVIDIA L20X', major=9, minor=0, total_memory=143166MB, multi_processor_count=132, uuid=46890c79-bf10-5b49-3d4a-a6b78f956122, pci_bus_id=163, pci_device_id=0, pci_domain_id=1, L2_cache_size=60MB)", "cuda_visible_devices": "6", "script_sha256": "e2df72bbcb703640baf64ba38392d6f28ceb06a23c0dead33048204693c11cfd" } ], "rows": [ { "repeat": 0, "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, "sample_tolerance": true, "policy_matches": true }, "passed": true, "manifest": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_same_gpu_v1_r0/manifest.json", "sample_sha256": "41879a27c3bbc5af65d2045530858ba3ce440fa57ef39eb5460da5a993c2c8ac", "trace_sha256": "5948d5e2f15cb1b55219c219e35fc210e6c9f39ce8b41df99b2dbb37099ef0c4", "full_tensor_sha256": "504c586c2e66ef7e60ba02090192e4106ab5003333dde3303d24639b2614f193", "full_tensor_file": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_same_gpu_v1_r0/bf16_4096_full.pt" }, { "repeat": 0, "mode": "bf16", "n": 8192, "median_ms": 1.6212000250816345, "tflops": 678.2084941804974, "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, "sample_tolerance": true, "policy_matches": true }, "passed": true, "manifest": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_same_gpu_v1_r0/manifest.json", "sample_sha256": "c8ec82ddc420f3b09f36c67359d348a736834d7f8e1a34dc1a79ea9b037f14e5", "trace_sha256": "6559cef3d47b4e8c302029fd1dd854967c6af05bda3b1a81d49f1b977501bc80", "full_tensor_sha256": "703679569fc56438886833c0346a15f8a936d55771592b3346ea22ef53f492ad", "full_tensor_file": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_same_gpu_v1_r0/bf16_8192_full.pt" }, { "repeat": 0, "mode": "bf16", "n": 12288, "median_ms": 5.295056104660034, "tflops": 700.8144333878126, "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, "sample_tolerance": true, "policy_matches": true }, "passed": true, "manifest": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_same_gpu_v1_r0/manifest.json", "sample_sha256": "327de61023cd329fc3688eb8477b8f6e3295e279e9126aaf172f6099caeba72b", "trace_sha256": "5683c6099965401d47ffaebf20ce31ba257c70aab4e000761318aebbc53ec752", "full_tensor_sha256": "5e3f51868bec24da111e733f28f3b8602ba9cbea12a714b0383f2d9848bccd1b", "full_tensor_file": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_same_gpu_v1_r0/bf16_12288_full.pt" }, { "repeat": 0, "mode": "fp32_ieee", "n": 4096, "median_ms": 2.6847360134124756, "tflops": 51.19272538729276, "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, "sample_tolerance": true, "policy_matches": true }, "passed": true, "manifest": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_same_gpu_v1_r0/manifest.json", "sample_sha256": "c5d4cd32738feb4dfe1e6fc068460cab1c361bce38965fcb581ef98f88c091be", "trace_sha256": "58b62c971097a7900d127d36f7893ac1903649612c84761971cbd98bc7c61067", "full_tensor_sha256": "50b1a19971a7be37e9f9933391a08ecb9f9c51e10aeb4bc6ca35e3755e12598a", "full_tensor_file": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_same_gpu_v1_r0/fp32_ieee_4096_full.pt" }, { "repeat": 0, "mode": "fp32_ieee", "n": 8192, "median_ms": 21.438496589660645, "tflops": 51.286787913396516, "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, "sample_tolerance": true, "policy_matches": true }, "passed": true, "manifest": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_same_gpu_v1_r0/manifest.json", "sample_sha256": "dde076b83db484998513f57f4c8668a77d9504ffd55427da1c7d685b6170593a", "trace_sha256": "2960d7504d5eef60444ce9f4a44f2928b1c14924278f052dfaaeea23a939a1a4", "full_tensor_sha256": "96e46136de03fef076f36f5738fa726ee3268b8398388fd7b1df799ac1da0a7e", "full_tensor_file": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_same_gpu_v1_r0/fp32_ieee_8192_full.pt" }, { "repeat": 0, "mode": "fp32_ieee", "n": 12288, "median_ms": 73.77207946777344, "tflops": 50.30157439665296, "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, "sample_tolerance": true, "policy_matches": true }, "passed": true, "manifest": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_same_gpu_v1_r0/manifest.json", "sample_sha256": "115241b05cd87f4ca3f52554ffb4fb65496254f03872e1d5dcf4f0ce17ac12bf", "trace_sha256": "de4294b5dffb197f1ed634ff663c60cc7114e2e7b2f195cd76cd7cfa7b57cfa4", "full_tensor_sha256": "0ea2d93d684eec26c335e982cac6203cddc962ae88aafba6cc168166eae0c580", "full_tensor_file": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_same_gpu_v1_r0/fp32_ieee_12288_full.pt" }, { "repeat": 0, "mode": "fp32_tf32_allowed", "n": 4096, "median_ms": 0.34545600414276123, "tflops": 397.8479222355699, "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, "sample_tolerance": true, "policy_matches": true }, "passed": true, "manifest": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_same_gpu_v1_r0/manifest.json", "sample_sha256": "73404de058399df07678ec601175d3522e035357e611166665e22da9f51bd074", "trace_sha256": "d7de3d3a54d3359a18657380afb8332fc5569eca63153f48ef372328420c0d78", "full_tensor_sha256": "47bf60c3f95c189fd9b05b89fa86ad660f0e06e6b777219311e1d47c77f19f51", "full_tensor_file": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_same_gpu_v1_r0/fp32_tf32_allowed_4096_full.pt" }, { "repeat": 0, "mode": "fp32_tf32_allowed", "n": 8192, "median_ms": 3.0584640502929688, "tflops": 359.49797339310834, "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, "sample_tolerance": true, "policy_matches": true }, "passed": true, "manifest": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_same_gpu_v1_r0/manifest.json", "sample_sha256": "6854e7a3023501d559c7df3df704ab9f8f2dc4b7060d4aef535e9900242a3bd0", "trace_sha256": "ac7d3cab0e8c34bb5e72a26737ccbad95009bcb979eefe46385712cfbb184843", "full_tensor_sha256": "d5014779ef629267d03a9b2796cbde33ce3f875133ce47d2f081222a80c4e4ce", "full_tensor_file": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_same_gpu_v1_r0/fp32_tf32_allowed_8192_full.pt" }, { "repeat": 0, "mode": "fp32_tf32_allowed", "n": 12288, "median_ms": 10.263247966766357, "tflops": 361.56699670125755, "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, "sample_tolerance": true, "policy_matches": true }, "passed": true, "manifest": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_same_gpu_v1_r0/manifest.json", "sample_sha256": "e88bb5e9531399c6aa9317ea83d886cd8d4d09e361049c00e5cac9fea7713da3", "trace_sha256": "5387f6568df14c7a1960d0277d1c90326ffaf84c0332fa99ea44866cec308776", "full_tensor_sha256": "d2bff3f5249b9bd4da2ba7fa4769d2d6c440c0da4ffd1f124bebfcd0e70490c3", "full_tensor_file": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_same_gpu_v1_r0/fp32_tf32_allowed_12288_full.pt" }, { "repeat": 1, "mode": "bf16", "n": 4096, "median_ms": 0.17447999864816666, "tflops": 787.7060668090745, "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, "sample_tolerance": true, "policy_matches": true }, "passed": true, "manifest": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_same_gpu_v1_r1/manifest.json", "sample_sha256": "41879a27c3bbc5af65d2045530858ba3ce440fa57ef39eb5460da5a993c2c8ac", "trace_sha256": "4296ceaf4a84c93bc62fc47fa47d3ff3687d45a2d7dd51e4cb7ac795b29df1b6", "full_tensor_sha256": "504c586c2e66ef7e60ba02090192e4106ab5003333dde3303d24639b2614f193", "full_tensor_file": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_same_gpu_v1_r1/bf16_4096_full.pt" }, { "repeat": 1, "mode": "bf16", "n": 8192, "median_ms": 1.600767970085144, "tflops": 686.8650849613874, "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, "sample_tolerance": true, "policy_matches": true }, "passed": true, "manifest": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_same_gpu_v1_r1/manifest.json", "sample_sha256": "c8ec82ddc420f3b09f36c67359d348a736834d7f8e1a34dc1a79ea9b037f14e5", "trace_sha256": "b3d12bb6762e5f6f31257709fdeeb08647a71a939fcaa71bf40b965e3eb98bcf", "full_tensor_sha256": "703679569fc56438886833c0346a15f8a936d55771592b3346ea22ef53f492ad", "full_tensor_file": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_same_gpu_v1_r1/bf16_8192_full.pt" }, { "repeat": 1, "mode": "bf16", "n": 12288, "median_ms": 5.295552015304565, "tflops": 700.7488044720068, "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, "sample_tolerance": true, "policy_matches": true }, "passed": true, "manifest": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_same_gpu_v1_r1/manifest.json", "sample_sha256": "327de61023cd329fc3688eb8477b8f6e3295e279e9126aaf172f6099caeba72b", "trace_sha256": "9e9596d7d8b2fa16fe9704f860eca1060b057b42e1c1b692657c3b70c594df6f", "full_tensor_sha256": "5e3f51868bec24da111e733f28f3b8602ba9cbea12a714b0383f2d9848bccd1b", "full_tensor_file": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_same_gpu_v1_r1/bf16_12288_full.pt" }, { "repeat": 1, "mode": "fp32_ieee", "n": 4096, "median_ms": 2.683583974838257, "tflops": 51.21470196597206, "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, "sample_tolerance": true, "policy_matches": true }, "passed": true, "manifest": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_same_gpu_v1_r1/manifest.json", "sample_sha256": "c5d4cd32738feb4dfe1e6fc068460cab1c361bce38965fcb581ef98f88c091be", "trace_sha256": "7a94330c9c9951763ed0463544eb341397d9a204895e0eba5244221885cc9b7a", "full_tensor_sha256": "50b1a19971a7be37e9f9933391a08ecb9f9c51e10aeb4bc6ca35e3755e12598a", "full_tensor_file": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_same_gpu_v1_r1/fp32_ieee_4096_full.pt" }, { "repeat": 1, "mode": "fp32_ieee", "n": 8192, "median_ms": 21.443135261535645, "tflops": 51.275693333347874, "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, "sample_tolerance": true, "policy_matches": true }, "passed": true, "manifest": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_same_gpu_v1_r1/manifest.json", "sample_sha256": "dde076b83db484998513f57f4c8668a77d9504ffd55427da1c7d685b6170593a", "trace_sha256": "d20204150920d4575348e6a556ddd79af83d5413a05e8054d105c2d721903991", "full_tensor_sha256": "96e46136de03fef076f36f5738fa726ee3268b8398388fd7b1df799ac1da0a7e", "full_tensor_file": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_same_gpu_v1_r1/fp32_ieee_8192_full.pt" }, { "repeat": 1, "mode": "fp32_ieee", "n": 12288, "median_ms": 73.72660827636719, "tflops": 50.33259810126788, "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, "sample_tolerance": true, "policy_matches": true }, "passed": true, "manifest": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_same_gpu_v1_r1/manifest.json", "sample_sha256": "115241b05cd87f4ca3f52554ffb4fb65496254f03872e1d5dcf4f0ce17ac12bf", "trace_sha256": "95283d63c4cbf91d6006fef2ec3b8703141120ce7eaa88b18983c62d3c7950cc", "full_tensor_sha256": "0ea2d93d684eec26c335e982cac6203cddc962ae88aafba6cc168166eae0c580", "full_tensor_file": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_same_gpu_v1_r1/fp32_ieee_12288_full.pt" }, { "repeat": 1, "mode": "fp32_tf32_allowed", "n": 4096, "median_ms": 0.3458079993724823, "tflops": 397.44295597962594, "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, "sample_tolerance": true, "policy_matches": true }, "passed": true, "manifest": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_same_gpu_v1_r1/manifest.json", "sample_sha256": "73404de058399df07678ec601175d3522e035357e611166665e22da9f51bd074", "trace_sha256": "ad3c425c47684442e9a6eb0c39ad9d26e7dc4c326013844c0833e63eace5c320", "full_tensor_sha256": "47bf60c3f95c189fd9b05b89fa86ad660f0e06e6b777219311e1d47c77f19f51", "full_tensor_file": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_same_gpu_v1_r1/fp32_tf32_allowed_4096_full.pt" }, { "repeat": 1, "mode": "fp32_tf32_allowed", "n": 8192, "median_ms": 3.058527946472168, "tflops": 359.4904630654828, "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, "sample_tolerance": true, "policy_matches": true }, "passed": true, "manifest": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_same_gpu_v1_r1/manifest.json", "sample_sha256": "6854e7a3023501d559c7df3df704ab9f8f2dc4b7060d4aef535e9900242a3bd0", "trace_sha256": "9405ec5ef26b19dd136ed0a01b1baec17df0c33f5b1615792f40c099962c4d23", "full_tensor_sha256": "d5014779ef629267d03a9b2796cbde33ce3f875133ce47d2f081222a80c4e4ce", "full_tensor_file": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_same_gpu_v1_r1/fp32_tf32_allowed_8192_full.pt" }, { "repeat": 1, "mode": "fp32_tf32_allowed", "n": 12288, "median_ms": 10.384111881256104, "tflops": 357.3586057409775, "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, "sample_tolerance": true, "policy_matches": true }, "passed": true, "manifest": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_same_gpu_v1_r1/manifest.json", "sample_sha256": "e88bb5e9531399c6aa9317ea83d886cd8d4d09e361049c00e5cac9fea7713da3", "trace_sha256": "15b41004b4b65775144eb6d3adfb09c3b3d405a570b6b9eb26bfc05ed24848b1", "full_tensor_sha256": "d2bff3f5249b9bd4da2ba7fa4769d2d6c440c0da4ffd1f124bebfcd0e70490c3", "full_tensor_file": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_same_gpu_v1_r1/fp32_tf32_allowed_12288_full.pt" }, { "repeat": 2, "mode": "bf16", "n": 4096, "median_ms": 0.17417599260807037, "tflops": 789.0809256432037, "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, "sample_tolerance": true, "policy_matches": true }, "passed": true, "manifest": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_same_gpu_v1_r2/manifest.json", "sample_sha256": "41879a27c3bbc5af65d2045530858ba3ce440fa57ef39eb5460da5a993c2c8ac", "trace_sha256": "06c6d9fc1238cea1aad8315fe368e6c456d7a609267369e26b2742aca992137a", "full_tensor_sha256": "504c586c2e66ef7e60ba02090192e4106ab5003333dde3303d24639b2614f193", "full_tensor_file": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_same_gpu_v1_r2/bf16_4096_full.pt" }, { "repeat": 2, "mode": "bf16", "n": 8192, "median_ms": 1.5998560190200806, "tflops": 687.2566122853082, "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, "sample_tolerance": true, "policy_matches": true }, "passed": true, "manifest": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_same_gpu_v1_r2/manifest.json", "sample_sha256": "c8ec82ddc420f3b09f36c67359d348a736834d7f8e1a34dc1a79ea9b037f14e5", "trace_sha256": "a1a4e5b8fea263d3b4ed8392d9537dc9286f174aa845e08a099593347c6d5176", "full_tensor_sha256": "703679569fc56438886833c0346a15f8a936d55771592b3346ea22ef53f492ad", "full_tensor_file": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_same_gpu_v1_r2/bf16_8192_full.pt" }, { "repeat": 2, "mode": "bf16", "n": 12288, "median_ms": 5.297232151031494, "tflops": 700.5265463061518, "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, "sample_tolerance": true, "policy_matches": true }, "passed": true, "manifest": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_same_gpu_v1_r2/manifest.json", "sample_sha256": "327de61023cd329fc3688eb8477b8f6e3295e279e9126aaf172f6099caeba72b", "trace_sha256": "10ce11c141cd6af75e7570c17e0f946cd3bce1d2527f5cccec44b402b6955aff", "full_tensor_sha256": "5e3f51868bec24da111e733f28f3b8602ba9cbea12a714b0383f2d9848bccd1b", "full_tensor_file": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_same_gpu_v1_r2/bf16_12288_full.pt" }, { "repeat": 2, "mode": "fp32_ieee", "n": 4096, "median_ms": 2.6845760345458984, "tflops": 51.19577605677616, "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, "sample_tolerance": true, "policy_matches": true }, "passed": true, "manifest": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_same_gpu_v1_r2/manifest.json", "sample_sha256": "c5d4cd32738feb4dfe1e6fc068460cab1c361bce38965fcb581ef98f88c091be", "trace_sha256": "f52a16dee96ddf7d58b11f8e49b03c9413fb891464f079feff878764d89f60fb", "full_tensor_sha256": "50b1a19971a7be37e9f9933391a08ecb9f9c51e10aeb4bc6ca35e3755e12598a", "full_tensor_file": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_same_gpu_v1_r2/fp32_ieee_4096_full.pt" }, { "repeat": 2, "mode": "fp32_ieee", "n": 8192, "median_ms": 21.453824043273926, "tflops": 51.250146619931485, "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, "sample_tolerance": true, "policy_matches": true }, "passed": true, "manifest": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_same_gpu_v1_r2/manifest.json", "sample_sha256": "dde076b83db484998513f57f4c8668a77d9504ffd55427da1c7d685b6170593a", "trace_sha256": "8a3a425288212bb3cca2ddfc2727b336edfe5704521afff5e4c36b546f822cb1", "full_tensor_sha256": "96e46136de03fef076f36f5738fa726ee3268b8398388fd7b1df799ac1da0a7e", "full_tensor_file": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_same_gpu_v1_r2/fp32_ieee_8192_full.pt" }, { "repeat": 2, "mode": "fp32_ieee", "n": 12288, "median_ms": 73.78628921508789, "tflops": 50.29188733054218, "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, "sample_tolerance": true, "policy_matches": true }, "passed": true, "manifest": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_same_gpu_v1_r2/manifest.json", "sample_sha256": "115241b05cd87f4ca3f52554ffb4fb65496254f03872e1d5dcf4f0ce17ac12bf", "trace_sha256": "ceb8acf8d9ba951e064bb7047db54049d5fb1908398440c6aafea28fc7653f40", "full_tensor_sha256": "0ea2d93d684eec26c335e982cac6203cddc962ae88aafba6cc168166eae0c580", "full_tensor_file": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_same_gpu_v1_r2/fp32_ieee_12288_full.pt" }, { "repeat": 2, "mode": "fp32_tf32_allowed", "n": 4096, "median_ms": 0.3458240032196045, "tflops": 397.4245633398784, "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, "sample_tolerance": true, "policy_matches": true }, "passed": true, "manifest": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_same_gpu_v1_r2/manifest.json", "sample_sha256": "73404de058399df07678ec601175d3522e035357e611166665e22da9f51bd074", "trace_sha256": "86b95afd95b36d366ff0dbfbff37ed250391b9f5f1a1c8884a73b1853ce2179a", "full_tensor_sha256": "47bf60c3f95c189fd9b05b89fa86ad660f0e06e6b777219311e1d47c77f19f51", "full_tensor_file": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_same_gpu_v1_r2/fp32_tf32_allowed_4096_full.pt" }, { "repeat": 2, "mode": "fp32_tf32_allowed", "n": 8192, "median_ms": 3.0707839727401733, "tflops": 358.0556748818984, "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, "sample_tolerance": true, "policy_matches": true }, "passed": true, "manifest": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_same_gpu_v1_r2/manifest.json", "sample_sha256": "6854e7a3023501d559c7df3df704ab9f8f2dc4b7060d4aef535e9900242a3bd0", "trace_sha256": "d091b0d3461bedcaecb830b16fcac0a65ce6cf218075443c52292cef2ea98710", "full_tensor_sha256": "d5014779ef629267d03a9b2796cbde33ce3f875133ce47d2f081222a80c4e4ce", "full_tensor_file": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_same_gpu_v1_r2/fp32_tf32_allowed_8192_full.pt" }, { "repeat": 2, "mode": "fp32_tf32_allowed", "n": 12288, "median_ms": 10.48030424118042, "tflops": 354.07862771415483, "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, "sample_tolerance": true, "policy_matches": true }, "passed": true, "manifest": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_same_gpu_v1_r2/manifest.json", "sample_sha256": "e88bb5e9531399c6aa9317ea83d886cd8d4d09e361049c00e5cac9fea7713da3", "trace_sha256": "73bf8be992164210e8a6ec407bb2c1458ecbcea4dbfd28ea3b21e3b5e4f8ffc9", "full_tensor_sha256": "d2bff3f5249b9bd4da2ba7fa4769d2d6c440c0da4ffd1f124bebfcd0e70490c3", "full_tensor_file": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/raw/local_gemm_same_gpu_v1_r2/fp32_tf32_allowed_12288_full.pt" } ] }