Publish Nabla Path Tracer runtime compare report

Author: AnastaZIuk

README.md

@@ -0,0 +1,86 @@
+# Nabla Path Tracer runtime compare from Nsight Graphics
+
+This directory contains one paired `Nsight Graphics` `GPU Trace` probe for Nabla Path Tracer.
+
+Protocol:
+- `1` run per variant
+- same capture point: `frame 1000`
+- same effective render path:
+  - geometry: `sphere`
+  - effective method: `solid angle`
+- runtime numbers below come directly from `Nsight Graphics` exports:
+  - `FRAME.xls`
+  - `GPUTRACE_FRAME.xls`
+- measurement machine:
+  - [`measurement_machine.md`](measurement_machine.md)
+
+## Variant matrix
+
+| Case | Checkout source | Nabla | DXC | SPIRV-Headers | SPIRV-Tools | Mode |
+| --- | --- | --- | --- | --- | --- | --- |
+| `master_source_off` | `master_runcheck` local worktree | [`e11b118dd2e80393b5b7eb309c6abb25f51a818c`](https://github.com/Devsh-Graphics-Programming/Nabla/commit/e11b118dd2e80393b5b7eb309c6abb25f51a818c) | [`d76c7890b19ce0b344ee0ce116dbc1c92220ccea`](https://github.com/Devsh-Graphics-Programming/DirectXShaderCompiler/commit/d76c7890b19ce0b344ee0ce116dbc1c92220ccea) | [`057230db28c7f7d1d571c9e61732da44815f2891`](https://github.com/Devsh-Graphics-Programming/SPIRV-Headers/commit/057230db28c7f7d1d571c9e61732da44815f2891) | [`91ac969ed599bfd0697a5b88cfae550318a04392`](https://github.com/Devsh-Graphics-Programming/SPIRV-Tools/commit/91ac969ed599bfd0697a5b88cfae550318a04392) | local `Release`, `SOURCE`, runtime `builtins OFF` |
+| `devshfixes_upstream` | `unroll_dxc_df_upstream_check` local worktree | [`c13c33662c3733b54d9014988a5ac602ab0c3245`](https://github.com/Devsh-Graphics-Programming/Nabla/commit/c13c33662c3733b54d9014988a5ac602ab0c3245) | [`74d6fbbad7388813c65ae269b20f15b4e971df9c`](https://github.com/Devsh-Graphics-Programming/DirectXShaderCompiler/commit/74d6fbbad7388813c65ae269b20f15b4e971df9c) | [`10b37414a3c9269b9bd8861cc759bd7fdf09760d`](https://github.com/Devsh-Graphics-Programming/SPIRV-Headers/commit/10b37414a3c9269b9bd8861cc759bd7fdf09760d) | [`2c75d08e3b31a673726ce6be80ab528250247064`](https://github.com/Devsh-Graphics-Programming/SPIRV-Tools/commit/2c75d08e3b31a673726ce6be80ab528250247064) | local `Release`, `SOURCE`, runtime `builtins OFF` |
+| `unroll_artifact` | CI install artifact from [`run 23599197849`](https://github.com/Devsh-Graphics-Programming/Nabla/actions/runs/23599197849) | [`262a8b72f295ec95d3cf83170f1768a43972c9ab`](https://github.com/Devsh-Graphics-Programming/Nabla/commit/262a8b72f295ec95d3cf83170f1768a43972c9ab) | [`07f06e9d48807ef8e7cabc41ae6acdeb26c68c09`](https://github.com/Devsh-Graphics-Programming/DirectXShaderCompiler/commit/07f06e9d48807ef8e7cabc41ae6acdeb26c68c09) | [`c141151dd53cbd5b1ced0665ad95ae3e91e8f916`](https://github.com/Devsh-Graphics-Programming/SPIRV-Headers/commit/c141151dd53cbd5b1ced0665ad95ae3e91e8f916) | [`2a730e127a32ac8b0713f5e1490d7b9be9d1cc9a`](https://github.com/Devsh-Graphics-Programming/SPIRV-Tools/commit/2a730e127a32ac8b0713f5e1490d7b9be9d1cc9a) | CI `Release install` artifact |
+
+## Main Nsight result
+
+| Variant | GPU frame ms | Dispatch count | Compute active | SM throughput | PCIe write GB/s |
+| --- | ---: | ---: | ---: | ---: | ---: |
+| `master_source_off` | `21.4304` | `2` | `83.2501%` | `35.5388%` | `2.62710` |
+| `devshfixes_upstream` | `19.6157` | `2` | `82.9923%` | `38.2916%` | `2.64694` |
+| `unroll_artifact` | `21.5935` | `2` | `83.9945%` | `34.3346%` | `2.62212` |
+
+### Runtime deltas
+
+| Comparison | Delta ms | Delta % |
+| --- | ---: | ---: |
+| `devshfixes_upstream` vs `master_source_off` | `-1.8147` | `-8.47%` |
+| `unroll_artifact` vs `master_source_off` | `+0.1631` | `+0.76%` |
+| `unroll_artifact` vs `devshfixes_upstream` | `+1.9778` | `+10.08%` |
+
+## Main conclusion
+
+The measured `latest upstream refresh` baseline is faster than `master_source_off` in this probe. At the same time `unroll_artifact` is effectively at parity with `master_source_off` here at only `+0.76%`, while the remaining gap appears only against `devshfixes_upstream`.
+
+Taken together, the measured runtime cost points at the `unroll` side of the experiment, not at the generic `DXC/SPIRV-Tools upstream refresh`. That tradeoff is also aligned with the intent of the experiment: reduce shader build time aggressively while accepting a small runtime cost.
+
+In practice this is also a strong argument for a development-oriented DXC optimization profile, for example an `-O1`-style mode. For the Nabla Path Tracer builds behind this comparison the shader-build wall time is about `10x` worse without that profile, while the measured runtime delta stays at `+0.76%` against `master_source_off` in this probe and at `+10.08%` against `devshfixes_upstream`; on other machines a simpler `average FPS` check across multiple methods, modes, and shapes placed the same runtime cost in the `5-8%` range. That is exactly the profile proposed in the paired PRs: for development use it delivers a major build-time win while keeping runtime impact effectively negligible in practice.
+
+## Deeper Nsight signals from the same exports
+
+Frame-level exports also show:
+- `dispatch_count = 2` and `gr__ctas_launched_queue_sync.sum = 14401` in all three variants
+- `unroll_artifact` has lower `SM throughput` than `devshfixes_upstream`
+- `unroll_artifact` also shows higher total executed instructions and much higher `L1/LSU/shared` pressure than `devshfixes_upstream`
+
+This points at a `compute-side codegen / execution-mix` difference with higher `L1/LSU/shared` pressure on the `unroll` side.
+
+## Directory map
+
+### Runtime stats
+- [`master_source_off/stats.json`](master_source_off/stats.json)
+- [`devshfixes_upstream/stats.json`](devshfixes_upstream/stats.json)
+- [`unroll_artifact/stats.json`](unroll_artifact/stats.json)
+
+### Machine spec
+- [`measurement_machine.md`](measurement_machine.md)
+
+### Executable locations
+- `master_source_off`: [`runnable/master_source_off_minimal/31_hlslpathtracer.exe`](runnable/master_source_off_minimal/31_hlslpathtracer.exe)
+- `devshfixes_upstream`: [`runnable/devshfixes_upstream_minimal/31_hlslpathtracer.exe`](runnable/devshfixes_upstream_minimal/31_hlslpathtracer.exe)
+- `unroll_artifact`: [`runnable/unroll_artifact_minimal/31_hlslpathtracer.exe`](runnable/unroll_artifact_minimal/31_hlslpathtracer.exe)
+
+### Capture files
+- [`master_source_off/run01/master_source_off_frame1000_run01.ngfx-capture`](master_source_off/run01/master_source_off_frame1000_run01.ngfx-capture)
+- [`devshfixes_upstream/run01/devshfixes_upstream_frame1000_run01.ngfx-capture`](devshfixes_upstream/run01/devshfixes_upstream_frame1000_run01.ngfx-capture)
+- [`unroll_artifact/run01/unroll_artifact_frame1000_run01.ngfx-capture`](unroll_artifact/run01/unroll_artifact_frame1000_run01.ngfx-capture)
+
+### Raw Nsight exports
+- [`master_source_off/run01/gpu-trace/BASE/FRAME.xls`](master_source_off/run01/gpu-trace/BASE/FRAME.xls)
+- [`master_source_off/run01/gpu-trace/BASE/GPUTRACE_FRAME.xls`](master_source_off/run01/gpu-trace/BASE/GPUTRACE_FRAME.xls)
+- [`devshfixes_upstream/run01/gpu-trace/BASE/GPUTRACE_FRAME.xls`](devshfixes_upstream/run01/gpu-trace/BASE/GPUTRACE_FRAME.xls)
+- [`unroll_artifact/run01/gpu-trace/BASE/GPUTRACE_FRAME.xls`](unroll_artifact/run01/gpu-trace/BASE/GPUTRACE_FRAME.xls)
+
+### Startup logs
+- [`startup_devshfixes_upstream/stats.json`](startup_devshfixes_upstream/stats.json)
+- [`startup_unroll_artifact/stats.json`](startup_unroll_artifact/stats.json)

devshfixes_upstream/run01/capture_stdout.txt

@@ -0,0 +1 @@
+event_text	time_ms

devshfixes_upstream/run01/devshfixes_upstream_frame1000_run01.log

@@ -0,0 +1 @@
+GPU frame time	19.6157

devshfixes_upstream/run01/devshfixes_upstream_frame1000_run01.ngfx-capture

@@ -0,0 +1,153 @@
+FE_B.TriageSCG.gr__cycles_active.avg.pct_of_peak_sustained_elapsed	93.1858
+HOST.TriageSCG.gpu__engine_cycles_active_any_syncce.avg.pct_of_peak_sustained_elapsed	0.475694
+FE_A.TriageSCG.gpu__scheduler_engine_asyncce0_cycles_active.avg.pct_of_peak_sustained_elapsed	0
+FE_A.TriageSCG.gpu__scheduler_engine_asyncce1_cycles_active.avg.pct_of_peak_sustained_elapsed	0.0764008
+HOST.TriageSCG.gpu__scheduler_engine_asyncce2_cycles_active.avg.pct_of_peak_sustained_elapsed	100
+GPUTrace.sm__throughput.avg.pct_of_peak_sustained_elapsed	38.2916
+rtcore__cycles_executed.avg.pct_of_peak_sustained_elapsed	0
+LTS.TriageSCG.lts__throughput.avg.pct_of_peak_sustained_elapsed	5.72954
+FBSP.TriageSCG.dramc__throughput.avg.pct_of_peak_sustained_elapsed	3.40994
+PCI.TriageSCG.pcie__throughput.avg.pct_of_peak_sustained_elapsed	16.8333
+GPUTrace.WorldPipe_throughput.avg.pct_of_peak_sustained_elapsed	0.00115412
+GPUTrace.ScreenPipe_throughput.avg.pct_of_peak_sustained_elapsed	0.132693
+fe__throughput.avg.pct_of_peak_sustained_elapsed	0.013713
+PCI.TriageSCG.pcie__read_bytes.avg.pct_of_peak_sustained_elapsed	0.579002
+PCI.TriageSCG.pcie__read_bytes.sum	1.78591e+06
+PCI.TriageSCG.pcie__read_bytes.sum.per_second	0.0910447
+PCI.TriageSCG.pcie__write_bytes.avg.pct_of_peak_sustained_elapsed	16.8333
+PCI.TriageSCG.pcie__write_bytes.sum	5.19217e+07
+PCI.TriageSCG.pcie__write_bytes.sum.per_second	2.64694
+pcie__rx_requests_aperture_bar0_op_read.sum.per_second	0.0604108
+pcie__rx_requests_aperture_bar0_op_read.sum	1185
+pcie__rx_requests_aperture_bar0_op_write.sum.per_second	0.0291603
+pcie__rx_requests_aperture_bar0_op_write.sum	572
+pcie__rx_requests_aperture_bar1_op_read.sum.per_second	0.00209016
+pcie__rx_requests_aperture_bar1_op_read.sum	41
+pcie__rx_requests_aperture_bar1_op_write.sum.per_second	0.042415
+pcie__rx_requests_aperture_bar1_op_write.sum	832
+pcie__rx_requests_aperture_bar2_op_read.sum.per_second	0.0196271
+pcie__rx_requests_aperture_bar2_op_read.sum	385
+pcie__rx_requests_aperture_bar2_op_write.sum.per_second	0.0279368
+pcie__rx_requests_aperture_bar2_op_write.sum	548
+FBSP.TriageSCG.dramc__read_throughput.avg.pct_of_peak_sustained_elapsed	3.12
+FBSP.TriageSCG.dramc__write_throughput.avg.pct_of_peak_sustained_elapsed	0.289944
+LTS.TriageSCG.lts__t_sector_throughput_srcunit_tex.avg.pct_of_peak_sustained_elapsed	2.40237
+LTS.TriageSCG.lts__t_sector_throughput_srcunit_gcc.avg.pct_of_peak_sustained_elapsed	1.62815
+LTS.TriageSCG.lts__t_sector_throughput_srcunit_pe.avg.pct_of_peak_sustained_elapsed	0.00356426
+ROP.TriageSCG.lts__t_sector_throughput_srcunit_crop.avg.pct_of_peak_sustained_elapsed	0.0296463
+ROP.TriageSCG.lts__t_sector_throughput_srcunit_zrop.avg.pct_of_peak_sustained_elapsed	0
+TriageSCG.lts__t_sector_throughput_srcunit_raster.avg.pct_of_peak_sustained_elapsed	0.330087
+LTS.TriageSCG.lts__t_sector_throughput_srcnode_fbp.avg.pct_of_peak_sustained_elapsed	0.00158123
+LTS.TriageSCG.lts__t_sector_throughput_srcnode_hub.avg.pct_of_peak_sustained_elapsed	1.69742
+LTS.TriageSCG.lts__average_t_sector_hit_rate_realtime.pct	81.2163
+SM_C.TriageSCG.smsp__inst_executed_pipe_fma.avg.pct_of_peak_sustained_elapsed	10.5253
+SM_C.TriageSCG.smsp__inst_executed_pipe_fma.sum	7.29385e+08
+SM_C.TriageSCG.smsp__inst_executed_pipe_fma.avg.per_cycle_elapsed	0.421011
+SM_C.TriageSCG.smsp__inst_executed_pipe_fma.avg.peak_sustained	4
+SM_C.TriageSCG.smsp__inst_executed_pipe_fmaheavy.avg.pct_of_peak_sustained_elapsed	6.83763
+SM_C.TriageSCG.smsp__inst_executed_pipe_fmaheavy.sum	2.36918e+08
+SM_C.TriageSCG.smsp__inst_executed_pipe_fmaheavy.avg.per_cycle_elapsed	0.136753
+SM_C.TriageSCG.smsp__inst_executed_pipe_fmaheavy.avg.peak_sustained	2
+sm__pipe_tensor_cycles_active_realtime.avg.pct_of_peak_sustained_elapsed	0
+sm__pipe_tensor_cycles_active_realtime.sum	0
+smsp__thread_inst_executed_per_inst_executed.ratio	28.292
+smsp__thread_inst_executed_per_inst_executed.pct	88.4125
+LTS.TriageSCG.lts__average_t_sector_hit_rate_srcunit_tex_realtime.pct	99.6297
+FE_B.TriageSCG.gr__ctas_launched_queue_sync.sum.per_second	0.734156
+FE_B.TriageSCG.gr__ctas_launched_queue_sync.sum	14401
+FE_B.TriageSCG.gr__ctas_launched_queue_async.sum.per_second	0
+FE_B.TriageSCG.gr__ctas_launched_queue_async.sum	0
+FE_A.TriageSCG.gr__compute_cycles_active_queue_sync.avg.pct_of_peak_sustained_elapsed	82.9923
+FE_A.TriageSCG.gr__compute_cycles_active_queue_async.avg.pct_of_peak_sustained_elapsed	0.0318724
+FE_A.TriageSCG.gr__dispatch_count.sum	2
+GPC_A.TriageSCG.raster__throughput.avg.pct_of_peak_sustained_elapsed	0.0301417
+GPC_A.TriageSCG.prop__throughput.avg.pct_of_peak_sustained_elapsed	0.131119
+ROP.TriageSCG.crop__throughput.avg.pct_of_peak_sustained_elapsed	0.132693
+ROP.TriageSCG.zrop__throughput.avg.pct_of_peak_sustained_elapsed	0
+crop__read_throughput.avg.pct_of_peak_sustained_elapsed	0.00358251
+crop__write_throughput.avg.pct_of_peak_sustained_elapsed	0.055826
+zrop__read_throughput.avg.pct_of_peak_sustained_elapsed	0
+zrop__write_throughput.avg.pct_of_peak_sustained_elapsed	0
+GPC_A.TriageSCG.prop__prop2zrop_pixels_stage_latez_realtime.avg.pct_of_peak_sustained_elapsed	0
+GPC_A.TriageSCG.raster__zcull_input_samples_realtime.avg.per_cycle_elapsed	0.0250038
+GPC_A.TriageSCG.raster__zcull_input_samples_realtime.avg.pct_of_peak_sustained_elapsed	0.0097671
+GPC_A.TriageSCG.raster__zcull_input_samples_realtime.sum	3.76678e+06
+GPC_A.TriageSCG.raster__zcull_input_samples_op_accepted_realtime.avg.per_cycle_elapsed	0.0250038
+GPC_A.TriageSCG.raster__zcull_input_samples_op_accepted_realtime.avg.pct_of_peak_sustained_elapsed	0.0097671
+GPC_A.TriageSCG.raster__zcull_input_samples_op_accepted_realtime.sum	3.76678e+06
+GPC_A.TriageSCG.prop__input_pixels_type_3d_realtime.avg.per_cycle_elapsed	0.0141987
+GPC_A.TriageSCG.prop__input_pixels_type_3d_realtime.avg.pct_of_peak_sustained_elapsed	0.0221854
+GPC_A.TriageSCG.prop__input_pixels_type_3d_realtime.sum	2.13901e+06
+GPC_A.TriageSCG.prop__prop2zrop_pixels_realtime.avg.per_cycle_elapsed	0
+GPC_A.TriageSCG.prop__prop2zrop_pixels_realtime.avg.pct_of_peak_sustained_elapsed	0
+GPC_A.TriageSCG.prop__prop2zrop_pixels_realtime.sum	0
+GPC_A.TriageSCG.prop__prop2zrop_pixels_op_passed_realtime.avg.per_cycle_elapsed	0.00611755
+GPC_A.TriageSCG.prop__prop2zrop_pixels_op_passed_realtime.avg.pct_of_peak_sustained_elapsed	0.00477934
+GPC_A.TriageSCG.prop__prop2zrop_pixels_op_passed_realtime.sum	921600
+GPC_A.TriageSCG.prop__prop2crop_pixels_realtime.avg.per_cycle_elapsed	0.014171
+GPC_A.TriageSCG.prop__prop2crop_pixels_realtime.avg.pct_of_peak_sustained_elapsed	0.0885691
+GPC_A.TriageSCG.prop__prop2crop_pixels_realtime.sum	2.13485e+06
+HUB_B.TriageSCG.pda__throughput.avg.pct_of_peak_sustained_elapsed	0.000367926
+TPC.TriageSCG.vaf__throughput.avg.pct_of_peak_sustained_elapsed	0.000587835
+GPC_B.TriageSCG.pes__throughput.avg.pct_of_peak_sustained_elapsed	0.00115412
+HUB_B.TriageSCG.pda__input_prims_realtime.avg.pct_of_peak_sustained_elapsed	0.000367926
+HUB_B.TriageSCG.pda__input_prims_realtime.avg	1888
+HUB_B.TriageSCG.pda__input_prims_realtime.sum	1888
+FE_B.TriageSCG.fe__draw_count.sum	58
+FE_B.TriageSCG.fe__pixel_shader_barriers.sum	19
+FE_B.TriageSCG.fe__output_ops_cmd_go_idle_queue_sync.sum	61
+FE_A.TriageSCG.fe__output_ops_cmd_subchsw_queue_sync.sum	28
+FE_B.TriageSCG.fe__cycles_stalled_cmd_wfi_queue_sync.avg.pct_of_peak_sustained_elapsed	3.20003
+FE_B.TriageSCG.fe__cycles_stalled_cmd_wfi_queue_sync.sum	1.0263e+06
+FE_B.TriageSCG.fe__output_ops_cmd_go_idle_queue_async.sum	1
+FE_A.TriageSCG.fe__output_ops_cmd_subchsw_queue_async.sum	1
+FE_B.TriageSCG.fe__cycles_stalled_cmd_wfi_queue_async.avg.pct_of_peak_sustained_elapsed	0.0334251
+FE_B.TriageSCG.fe__cycles_stalled_cmd_wfi_queue_async.sum	10720
+gpc__cycles_elapsed.avg.per_second	1920
+gpc__cycles_elapsed.avg	3.76621e+07
+sys__cycles_elapsed.avg.per_second	1635
+sys__cycles_elapsed.avg	3.20717e+07
+lts__cycles_elapsed.avg.per_second	1710
+lts__cycles_elapsed.avg	3.35428e+07
+dramc__cycles_elapsed.avg.per_second	10491.5
+dramc__cycles_elapsed.avg	2.05798e+08
+SM_A.TriageSCG.l1tex__throughput.avg.pct_of_peak_sustained_elapsed	15.2134
+SM_A.TriageSCG.sm__inst_executed_realtime.avg.pct_of_peak_sustained_elapsed	36.1949
+SM_A.TriageSCG.sm__inst_executed_realtime.sum	2.50824e+09
+SM_A.TriageSCG.sm__inst_executed_realtime.avg.per_cycle_elapsed	1.44779
+SM_A.TriageSCG.sm__inst_executed_realtime.avg.peak_sustained	4
+SM_A.TriageSCG.sm__inst_executed_pipe_alu_realtime.avg.pct_of_peak_sustained_elapsed	38.2916
+SM_A.TriageSCG.sm__inst_executed_pipe_alu_realtime.sum	1.32677e+09
+SM_A.TriageSCG.sm__inst_executed_pipe_alu_realtime.avg.per_cycle_elapsed	0.765832
+SM_A.TriageSCG.sm__inst_executed_pipe_alu_realtime.avg.peak_sustained	2
+SM_A.TriageSCG.sm__inst_executed_pipe_xu_realtime.avg.pct_of_peak_sustained_elapsed	10.6193
+SM_A.TriageSCG.sm__inst_executed_pipe_xu_realtime.sum	9.19873e+07
+SM_A.TriageSCG.sm__inst_executed_pipe_xu_realtime.avg.per_cycle_elapsed	0.0530964
+SM_A.TriageSCG.sm__inst_executed_pipe_xu_realtime.avg.peak_sustained	0.5
+SM_B.TriageSCG.tpc__warps_active_shader_vtg_realtime.avg.per_cycle_elapsed	0.00014169
+SM_B.TriageSCG.tpc__warps_active_shader_vtg_realtime.avg.pct_of_peak_sustained_elapsed	0.000295187
+TPC.TriageSCG.tpc__warps_active_shader_ps_realtime.avg.per_cycle_elapsed	0.0245833
+TPC.TriageSCG.tpc__warps_active_shader_ps_realtime.avg.pct_of_peak_sustained_elapsed	0.0512152
+TPC.TriageSCG.tpc__warps_active_shader_cs_realtime.avg.per_cycle_elapsed	5.4674
+TPC.TriageSCG.tpc__warps_active_shader_cs_realtime.avg.pct_of_peak_sustained_elapsed	11.3904
+TriageSCG.tpc__warps_inactive_sm_active_realtime.avg.per_cycle_elapsed	31.0897
+TriageSCG.tpc__warps_inactive_sm_active_realtime.avg.pct_of_peak_sustained_elapsed	64.7702
+TPC.TriageSCG.tpc__sm_rf_registers_allocated_shader_3d_realtime.avg.pct_of_peak_sustained_elapsed	0.0423113
+TPC.TriageSCG.tpc__sm_rf_registers_allocated_shader_3d_realtime.avg.per_cycle_elapsed	27.7291
+TPC.TriageSCG.tpc__sm_rf_registers_allocated_shader_cs_realtime.avg.pct_of_peak_sustained_elapsed	68.3443
+TPC.TriageSCG.tpc__sm_rf_registers_allocated_shader_cs_realtime.avg.per_cycle_elapsed	44790.1
+TPC.TriageSCG.tpc__l1tex_mem_shared_data_tram_bytes_allocated_realtime.avg.per_cycle_elapsed	0.38392
+TPC.TriageSCG.tpc__l1tex_mem_shared_data_compute_bytes_allocated_queue_sync_realtime.avg.per_cycle_elapsed	41510.2
+TPC.TriageSCG.tpc__l1tex_mem_shared_data_compute_bytes_allocated_queue_async_realtime.avg.per_cycle_elapsed	0
+TPC.TriageSCG.tpc__l1tex_mem_shared_data_isbe_bytes_allocated_realtime.avg.per_cycle_elapsed	0.33932
+SM_B.TriageSCG.sm__ctas_active.avg.pct_of_peak_sustained_elapsed	11.852
+SM_B.TriageSCG.sm__ctas_active.avg.per_cycle_elapsed	2.84447
+SM_B.TriageSCG.l1tex__t_sector_hit_rate.pct	84.5574
+SM_A.TriageSCG.l1tex__data_pipe_lsu_wavefronts.avg.pct_of_peak_sustained_elapsed	15.2134
+SM_A.TriageSCG.l1tex__lsu_writeback_active.avg.pct_of_peak_sustained_elapsed	8.87206
+SM_A.TriageSCG.l1tex__data_pipe_tex_wavefronts.avg.pct_of_peak_sustained_elapsed	0.0448575
+SM_A.TriageSCG.l1tex__f_wavefronts_realtime.avg.pct_of_peak_sustained_elapsed	0.0905451
+SM_A.TriageSCG.l1tex__tex_writeback_active.avg.pct_of_peak_sustained_elapsed	0.0643766
+SM_A.TriageSCG.l1tex__data_pipe_lsu_wavefronts_mem_lg.avg.pct_of_peak_sustained_elapsed	2.49725
+SM_A.TriageSCG.l1tex__data_pipe_lsu_wavefronts_mem_shared_realtime.avg.pct_of_peak_sustained_elapsed	12.7029
+SM_A.TriageSCG.l1tex__data_pipe_lsu_wavefronts_mem_surface_realtime.avg.pct_of_peak_sustained_elapsed	0.0133034