[misc] Add auto-profiling to IR passes

[xumingkuan]

Related issue = #926

Usage: ti.core.print_profile_info()

Example:
taichi_elements/demo_2d.py

>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
[Profiler thread 5444]
      1.674  s taichi::lang::TaichiLLVMContext::clone_runtime_module [1 x   1.674  s]
          1.581  s 94.43%  taichi::lang::compile_runtime_bitcode [1 x   1.581  s]
          0.033  s  1.95%  taichi::lang::module_from_bitcode_file [1 x  32.674 ms]
          0.060  s  3.60%  clone module          [1 x  60.314 ms]
    605.888 ms taichi::lang::StructCompilerLLVM::run [1 x 605.888 ms]
          0.217 ms  0.04%  taichi::lang::StructCompilerLLVM::generate_types [33 x   6.576 us]
          0.458 ms  0.08%  taichi::lang::StructCompilerLLVM::generate_child_accessors [1 x 458.000 us]
             71.000 us 15.50%  taichi::lang::StructCompilerLLVM::generate_refine_coordinates [1 x  71.000 us]
            384.000 us 83.84%  taichi::lang::StructCompilerLLVM::generate_child_accessors [7 x  54.857 us]
                192.000 us 50.00%  taichi::lang::StructCompilerLLVM::generate_refine_coordinates [7 x  27.429 us]
                139.000 us 36.20%  taichi::lang::StructCompilerLLVM::generate_child_accessors [25 x   5.560 us]
                 53.000 us 13.80%  [unaccounted]
              3.000 us  0.66%  [unaccounted]
         13.805 ms  2.28%  taichi::lang::TaichiLLVMContext::clone_struct_module [1 x  13.805 ms]
          2.006 ms  0.33%  taichi::lang::TaichiLLVMContext::eliminate_unused_functions [1 x   2.006 ms]
        414.628 ms 68.43%  taichi::lang::JITSessionCPU::global_optimize_module_cpu [1 x 414.628 ms]
             10.094 ms  2.43%  llvm_function_pass    [1 x  10.094 ms]
            398.991 ms 96.23%  llvm_module_pass      [1 x 398.991 ms]
              5.543 ms  1.34%  [unaccounted]
        174.774 ms 28.85%  [unaccounted]
      4.163  s taichi::lang::Program::compile [19 x 219.123 ms]
          1.498  s 35.98%  taichi::lang::irpass::compile_to_offloads [19 x  78.841 ms]
              0.101  s  6.72%  taichi::lang::irpass::lower [19 x   5.300 ms]
              0.022  s  1.45%  taichi::lang::irpass::typecheck [76 x 285.053 us]
              0.028  s  1.89%  taichi::lang::irpass::analysis::verify [304 x  93.352 us]
              0.000  s  0.00%  taichi::lang::irpass::loop_vectorize [19 x   3.421 us]
              0.000  s  0.00%  taichi::lang::irpass::vector_split [19 x   2.421 us]
              0.312  s 20.84%  taichi::lang::irpass::simplify [19 x  16.434 ms]
                  0.807 ms  0.26%  taichi::lang::irpass::typecheck [84 x   9.607 us]
                311.438 ms 99.74%  [unaccounted]
              0.342  s 22.80%  taichi::lang::irpass::variable_optimization [38 x   8.989 ms]
              0.000  s  0.03%  taichi::lang::irpass::flag_access [57 x   7.807 us]
              0.650  s 43.41%  taichi::lang::irpass::full_simplify [38 x  17.113 ms]
                  2.777 ms  0.43%  taichi::lang::irpass::extract_constant [79 x  35.152 us]
                  1.253 ms  0.19%  taichi::lang::irpass::binary_op_simplify [79 x  15.861 us]
                 61.359 ms  9.44%  taichi::lang::irpass::constant_fold [79 x 776.696 us]
                     56.805 ms 92.58%  taichi::lang::Program::compile [2 x  28.403 ms]
                          0.295 ms  0.52%  taichi::lang::irpass::compile_to_offloads [2 x 147.500 us]
                             11.000 us  3.73%  taichi::lang::irpass::lower [2 x   5.500 us]
                             10.000 us  3.39%  taichi::lang::irpass::typecheck [2 x   5.000 us]
                             21.000 us  7.12%  taichi::lang::irpass::analysis::verify [4 x   5.250 us]
                            245.000 us 83.05%  taichi::lang::irpass::offload [2 x 122.500 us]
                                 13.000 us  5.31%  taichi::lang::irpass::typecheck [4 x   3.250 us]
                                232.000 us 94.69%  [unaccounted]
                              8.000 us  2.71%  [unaccounted]
                         56.487 ms 99.44%  taichi::lang::KernelCodeGen::compile [2 x  28.244 ms]
                             56.483 ms 99.99%  taichi::lang::CodeGenCPU::codegen [2 x  28.242 ms]
                                 26.030 ms 46.08%  taichi::lang::TaichiLLVMContext::clone_struct_module [2 x  13.015 ms]
                                  0.004 ms  0.01%  taichi::lang::CodeGenLLVMCPU::CodeGenLLVMCPU [2 x   2.000 us]
                                  0.165 ms  0.29%  taichi::lang::CodeGenLLVM::emit_to_module [2 x  82.500 us]
                                 30.249 ms 53.55%  taichi::lang::CodeGenLLVM::compile_module_to_executable [2 x  15.125 ms]
                                      3.346 ms 11.06%  taichi::lang::TaichiLLVMContext::eliminate_unused_functions [2 x   1.673 ms]
                                     21.673 ms 71.65%  taichi::lang::JITSessionCPU::global_optimize_module_cpu [2 x  10.837 ms]
                                          4.268 ms 19.69%  llvm_function_pass    [2 x   2.134 ms]
                                         11.527 ms 53.19%  llvm_module_pass      [2 x   5.764 ms]
                                          5.878 ms 27.12%  [unaccounted]
                                      5.230 ms 17.29%  [unaccounted]
                      4.554 ms  7.42%  [unaccounted]
                  2.298 ms  0.35%  taichi::lang::irpass::alg_simp [79 x  29.089 us]
                 29.233 ms  4.50%  taichi::lang::irpass::die [237 x 123.346 us]
                462.993 ms 71.20%  taichi::lang::irpass::whole_kernel_cse [79 x   5.861 ms]
                 89.967 ms 13.83%  taichi::lang::irpass::simplify [79 x   1.139 ms]
                     16.684 ms 18.54%  taichi::lang::irpass::typecheck [74 x 225.459 us]
                     73.283 ms 81.46%  [unaccounted]
              0.020  s  1.31%  taichi::lang::irpass::offload [19 x   1.030 ms]
                  5.742 ms 29.33%  taichi::lang::irpass::typecheck [38 x 151.105 us]
                 13.836 ms 70.67%  [unaccounted]
              0.005  s  0.34%  taichi::lang::irpass::die [19 x 269.053 us]
              0.007  s  0.46%  taichi::lang::irpass::demote_atomics [19 x 362.789 us]
                  3.390 ms 49.18%  taichi::lang::irpass::typecheck [19 x 178.421 us]
                  3.503 ms 50.82%  [unaccounted]
              0.000  s  0.03%  taichi::lang::irpass::cfg_optimization [19 x  23.421 us]
              0.011  s  0.70%  [unaccounted]
          2.665  s 64.01%  taichi::lang::KernelCodeGen::compile [19 x 140.262 ms]
              2.665  s 100.00%  taichi::lang::CodeGenCPU::codegen [19 x 140.260 ms]
                  0.291  s 10.90%  taichi::lang::TaichiLLVMContext::clone_struct_module [19 x  15.293 ms]
                  0.000  s  0.00%  taichi::lang::CodeGenLLVMCPU::CodeGenLLVMCPU [19 x   1.368 us]
                  0.007  s  0.28%  taichi::lang::CodeGenLLVM::emit_to_module [19 x 387.789 us]
                  2.367  s 88.80%  taichi::lang::CodeGenLLVM::compile_module_to_executable [19 x 124.553 ms]
                      0.037  s  1.58%  taichi::lang::TaichiLLVMContext::eliminate_unused_functions [19 x   1.970 ms]
                      1.929  s 81.51%  taichi::lang::JITSessionCPU::global_optimize_module_cpu [19 x 101.523 ms]
                          0.093  s  4.82%  llvm_function_pass    [19 x   4.896 ms]
                          1.777  s 92.13%  llvm_module_pass      [19 x  93.535 ms]
                          0.059  s  3.05%  [unaccounted]
                      0.400  s 16.91%  [unaccounted]
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>

[Click here for the format server]

---

[archibate]

Removing this cause it's already done in Simplified II?

[xumingkuan]

Yes. (Removing to avoid printing a line 0.000 s 0.00% taichi::lang::irpass::constant_fold under the tab taichi::lang::irpass::compile_to_offloads.)

[codecov]

Codecov Report

Merging #1255 into master will not change coverage.
The diff coverage is n/a.

@@           Coverage Diff           @@
##           master    #1255   +/-   ##
=======================================
  Coverage   84.78%   84.78%           
=======================================
  Files          18       18           
  Lines        3267     3267           
  Branches      613      613           
=======================================
  Hits         2770     2770           
  Misses        360      360           
  Partials      137      137           

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 1448ca9...3c509a8. Read the comment docs.