Dramatic Compilation slow-down on higher Optimizaitons

[llvmbot]

Bugzilla Link	13651
Version	unspecified
OS	All
Attachments	File that exhibits problem, Smaller Example File
Reporter	LLVM Bugzilla Contributor
CC	@d0k,@nico,@stoklund

Extended Description

The attached file contains a few function definitions followed by a main method with a single line of code duplicated 2500 times.

At optimization levels O0 and O1, the code compiles relatively quickly (~5s). However, at optimization levels O2 and O3, the compilation time jumps to over a minute.

[llvmbot]

gcc version of this bug: http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54337

clang is slow on the gcc test file, but gcc is fast on the test file attached to this bug.

[llvmbot]

-ftime-report results for -O0 and -O3:
At -O0:
===-------------------------------------------------------------------------===
Instruction Selection and Scheduling
===-------------------------------------------------------------------------===
Total Execution Time: 0.5012 seconds (0.5051 wall clock)

---User Time--- --System Time-- --User+System-- ---Wall Time--- --- Name ---
0.0936 ( 20.3%) 0.0156 ( 40.2%) 0.1093 ( 21.8%) 0.1091 ( 21.6%) Instruction Creation
0.1024 ( 22.1%) 0.0027 ( 6.9%) 0.1050 ( 21.0%) 0.1049 ( 20.8%) Instruction Scheduling
0.0999 ( 21.6%) 0.0021 ( 5.3%) 0.1020 ( 20.3%) 0.1019 ( 20.2%) Instruction Selection
0.0381 ( 8.2%) 0.0082 ( 21.2%) 0.0463 ( 9.2%) 0.0495 ( 9.8%) DAG Combining 1
0.0348 ( 7.5%) 0.0022 ( 5.7%) 0.0370 ( 7.4%) 0.0369 ( 7.3%) Vector Legalization
0.0335 ( 7.2%) 0.0021 ( 5.3%) 0.0356 ( 7.1%) 0.0368 ( 7.3%) DAG Legalization
0.0261 ( 5.6%) 0.0020 ( 5.2%) 0.0281 ( 5.6%) 0.0281 ( 5.6%) Type Legalization
0.0197 ( 4.3%) 0.0020 ( 5.2%) 0.0218 ( 4.3%) 0.0217 ( 4.3%) DAG Combining 2
0.0142 ( 3.1%) 0.0020 ( 5.1%) 0.0162 ( 3.2%) 0.0162 ( 3.2%) Instruction Scheduling Cleanup
0.4623 (100.0%) 0.0389 (100.0%) 0.5012 (100.0%) 0.5051 (100.0%) Total

===-------------------------------------------------------------------------===
DWARF Emission
===-------------------------------------------------------------------------===
Total Execution Time: 0.0319 seconds (0.0319 wall clock)

---User Time--- --System Time-- --User+System-- ---Wall Time--- --- Name ---
0.0266 ( 97.8%) 0.0047 ( 99.6%) 0.0313 ( 98.0%) 0.0313 ( 98.0%) DWARF Exception Writer
0.0006 ( 2.2%) 0.0000 ( 0.4%) 0.0006 ( 2.0%) 0.0006 ( 2.0%) DWARF Debug Writer
0.0272 (100.0%) 0.0047 (100.0%) 0.0319 (100.0%) 0.0319 (100.0%) Total

===-------------------------------------------------------------------------===
... Pass execution timing report ...
===-------------------------------------------------------------------------===
Total Execution Time: 1.1464 seconds (1.1513 wall clock)

---User Time--- --System Time-- --User+System-- ---Wall Time--- --- Name ---
0.7807 ( 73.5%) 0.0690 ( 81.7%) 0.8498 ( 74.1%) 0.8547 ( 74.2%) X86 DAG->DAG Instruction Selection
0.0816 ( 7.7%) 0.0075 ( 8.9%) 0.0891 ( 7.8%) 0.0891 ( 7.7%) X86 AT&T-Style Assembly Printer
0.0404 ( 3.8%) 0.0011 ( 1.3%) 0.0415 ( 3.6%) 0.0414 ( 3.6%) Machine Function Analysis
0.0380 ( 3.6%) 0.0014 ( 1.6%) 0.0394 ( 3.4%) 0.0394 ( 3.4%) Fast Register Allocator
0.0201 ( 1.9%) 0.0004 ( 0.5%) 0.0205 ( 1.8%) 0.0205 ( 1.8%) Prologue/Epilogue Insertion & Frame Finalization
0.0156 ( 1.5%) 0.0027 ( 3.2%) 0.0183 ( 1.6%) 0.0183 ( 1.6%) Dominator Tree Construction
0.0152 ( 1.4%) 0.0010 ( 1.2%) 0.0162 ( 1.4%) 0.0162 ( 1.4%) Dominator Tree Construction
0.0154 ( 1.4%) 0.0007 ( 0.8%) 0.0161 ( 1.4%) 0.0161 ( 1.4%) Dominator Tree Construction
0.0143 ( 1.3%) 0.0000 ( 0.0%) 0.0144 ( 1.3%) 0.0144 ( 1.2%) Two-Address instruction pass
0.0073 ( 0.7%) 0.0000 ( 0.0%) 0.0073 ( 0.6%) 0.0073 ( 0.6%) Module Verifier
0.0072 ( 0.7%) 0.0000 ( 0.0%) 0.0072 ( 0.6%) 0.0072 ( 0.6%) Module Verifier
0.0072 ( 0.7%) 0.0000 ( 0.0%) 0.0072 ( 0.6%) 0.0072 ( 0.6%) Module Verifier
0.0040 ( 0.4%) 0.0000 ( 0.0%) 0.0040 ( 0.3%) 0.0040 ( 0.3%) Post-RA pseudo instruction expansion pass
0.0038 ( 0.4%) 0.0001 ( 0.1%) 0.0039 ( 0.3%) 0.0039 ( 0.3%) Basic CallGraph Construction
0.0025 ( 0.2%) 0.0001 ( 0.1%) 0.0026 ( 0.2%) 0.0026 ( 0.2%) Remove unreachable blocks from the CFG
0.0021 ( 0.2%) 0.0000 ( 0.0%) 0.0022 ( 0.2%) 0.0022 ( 0.2%) Expand ISel Pseudo-instructions
0.0019 ( 0.2%) 0.0000 ( 0.0%) 0.0019 ( 0.2%) 0.0019 ( 0.2%) Inliner for always_inline functions
0.0016 ( 0.1%) 0.0002 ( 0.2%) 0.0018 ( 0.2%) 0.0018 ( 0.2%) Bundle Machine CFG Edges
0.0009 ( 0.1%) 0.0000 ( 0.0%) 0.0009 ( 0.1%) 0.0009 ( 0.1%) Preliminary module verification
0.0005 ( 0.0%) 0.0000 ( 0.0%) 0.0005 ( 0.0%) 0.0005 ( 0.0%) Insert stack protectors
0.0004 ( 0.0%) 0.0000 ( 0.0%) 0.0004 ( 0.0%) 0.0004 ( 0.0%) Eliminate PHI nodes for register allocation
0.0004 ( 0.0%) 0.0000 ( 0.0%) 0.0004 ( 0.0%) 0.0004 ( 0.0%) Preliminary module verification
0.0003 ( 0.0%) 0.0000 ( 0.0%) 0.0003 ( 0.0%) 0.0003 ( 0.0%) Exception handling preparation
0.0003 ( 0.0%) 0.0000 ( 0.0%) 0.0003 ( 0.0%) 0.0003 ( 0.0%) Preliminary module verification
0.0001 ( 0.0%) 0.0000 ( 0.0%) 0.0001 ( 0.0%) 0.0001 ( 0.0%) X86 Maximal Stack Alignment Check
0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0001 ( 0.0%) 0.0001 ( 0.0%) Basic Alias Analysis (stateless AA impl)
0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0000 ( 0.0%) No Alias Analysis (always returns 'may' alias)
0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0000 ( 0.0%) Local Stack Slot Allocation
0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0000 ( 0.0%) Lower Garbage Collection Instructions
0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0000 ( 0.0%) Delete Garbage Collector Information
0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0000 ( 0.0%) X86 FP Stackifier
0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0000 ( 0.0%) Analyze Machine Code For Garbage Collection
0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0000 ( 0.0%) Create Garbage Collector Module Metadata
0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0000 ( 0.0%) Target Library Information
0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0000 ( 0.0%) Target Pass Configuration
0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0000 ( 0.0%) Machine Module Information
1.0619 (100.0%) 0.0844 (100.0%) 1.1464 (100.0%) 1.1513 (100.0%) Total

===-------------------------------------------------------------------------===
Miscellaneous Ungrouped Timers
===-------------------------------------------------------------------------===

---User Time--- --System Time-- --User+System-- ---Wall Time--- --- Name ---
1.5405 ( 57.0%) 0.1169 ( 55.9%) 1.6575 ( 56.9%) 1.7984 ( 58.0%) Clang front-end timer
1.0853 ( 40.1%) 0.0860 ( 41.1%) 1.1713 ( 40.2%) 1.1927 ( 38.5%) Code Generation Time
0.0776 ( 2.9%) 0.0063 ( 3.0%) 0.0839 ( 2.9%) 0.1071 ( 3.5%) LLVM IR Generation Time
2.7034 (100.0%) 0.2093 (100.0%) 2.9127 (100.0%) 3.0983 (100.0%) Total

At -O3:
===-------------------------------------------------------------------------===
Register Allocation
===-------------------------------------------------------------------------===
Total Execution Time: 0.0059 seconds (0.0059 wall clock)

---User Time--- --System Time-- --User+System-- ---Wall Time--- --- Name ---
0.0057 ( 99.9%) 0.0002 ( 97.7%) 0.0059 ( 99.8%) 0.0059 ( 99.8%) Seed Live Regs
0.0000 ( 0.1%) 0.0000 ( 2.3%) 0.0000 ( 0.2%) 0.0000 ( 0.2%) Evict
0.0057 (100.0%) 0.0002 (100.0%) 0.0059 (100.0%) 0.0059 (100.0%) Total

===-------------------------------------------------------------------------===
Instruction Selection and Scheduling
===-------------------------------------------------------------------------===
Total Execution Time: 1.4591 seconds (1.4608 wall clock)

---User Time--- --System Time-- --User+System-- ---Wall Time--- --- Name ---
0.2975 ( 21.6%) 0.0069 ( 8.6%) 0.3043 ( 20.9%) 0.3038 ( 20.8%) Instruction Scheduling
0.2495 ( 18.1%) 0.0059 ( 7.4%) 0.2554 ( 17.5%) 0.2548 ( 17.4%) Instruction Selection
0.1795 ( 13.0%) 0.0196 ( 24.5%) 0.1991 ( 13.6%) 0.1985 ( 13.6%) Instruction Creation
0.1305 ( 9.5%) 0.0164 ( 20.5%) 0.1469 ( 10.1%) 0.1509 ( 10.3%) DAG Combining 1
0.1389 ( 10.1%) 0.0055 ( 6.9%) 0.1444 ( 9.9%) 0.1446 ( 9.9%) Type Legalization
0.1344 ( 9.7%) 0.0056 ( 7.0%) 0.1400 ( 9.6%) 0.1395 ( 9.6%) DAG Legalization
0.0816 ( 5.9%) 0.0035 ( 4.4%) 0.0851 ( 5.8%) 0.0848 ( 5.8%) DAG Combining after legalize types
0.0695 ( 5.0%) 0.0054 ( 6.7%) 0.0749 ( 5.1%) 0.0749 ( 5.1%) Vector Legalization
0.0568 ( 4.1%) 0.0056 ( 7.1%) 0.0624 ( 4.3%) 0.0623 ( 4.3%) DAG Combining 2
0.0411 ( 3.0%) 0.0055 ( 6.9%) 0.0465 ( 3.2%) 0.0466 ( 3.2%) Instruction Scheduling Cleanup
1.3793 (100.0%) 0.0799 (100.0%) 1.4591 (100.0%) 1.4608 (100.0%) Total

===-------------------------------------------------------------------------===
DWARF Emission
===-------------------------------------------------------------------------===
Total Execution Time: 0.0252 seconds (0.0252 wall clock)

---User Time--- --System Time-- --User+System-- ---Wall Time--- --- Name ---
0.0185 ( 75.6%) 0.0006 ( 96.8%) 0.0191 ( 76.1%) 0.0192 ( 76.1%) DWARF Exception Writer
0.0060 ( 24.4%) 0.0000 ( 3.2%) 0.0060 ( 23.9%) 0.0060 ( 23.9%) DWARF Debug Writer
0.0245 (100.0%) 0.0006 (100.0%) 0.0252 (100.0%) 0.0252 (100.0%) Total

===-------------------------------------------------------------------------===
... Pass execution timing report ...
===-------------------------------------------------------------------------===
Total Execution Time: 44.8214 seconds (44.8240 wall clock)

---User Time--- --System Time-- --User+System-- ---Wall Time--- --- Name ---
36.6471 ( 82.3%) 0.0243 ( 7.9%) 36.6714 ( 81.8%) 36.6719 ( 81.8%) Eliminate PHI nodes for register allocation
2.1293 ( 4.8%) 0.1606 ( 52.5%) 2.2899 ( 5.1%) 2.2902 ( 5.1%) X86 DAG->DAG Instruction Selection
1.2961 ( 2.9%) 0.0005 ( 0.2%) 1.2966 ( 2.9%) 1.2966 ( 2.9%) Simple Register Coalescing
0.5791 ( 1.3%) 0.0001 ( 0.0%) 0.5792 ( 1.3%) 0.5792 ( 1.3%) Machine code sinking
0.4765 ( 1.1%) 0.0004 ( 0.1%) 0.4769 ( 1.1%) 0.4770 ( 1.1%) Simplify the CFG
0.4065 ( 0.9%) 0.0146 ( 4.8%) 0.4211 ( 0.9%) 0.4211 ( 0.9%) Function Integration/Inlining
0.2742 ( 0.6%) 0.0020 ( 0.6%) 0.2761 ( 0.6%) 0.2765 ( 0.6%) Greedy Register Allocator
0.2141 ( 0.5%) 0.0048 ( 1.6%) 0.2189 ( 0.5%) 0.2189 ( 0.5%) Combine redundant instructions
0.1784 ( 0.4%) 0.0253 ( 8.3%) 0.2037 ( 0.5%) 0.2037 ( 0.5%) Global Value Numbering
0.1390 ( 0.3%) 0.0006 ( 0.2%) 0.1396 ( 0.3%) 0.1396 ( 0.3%) Branch Probability Basic Block Placement
0.1273 ( 0.3%) 0.0024 ( 0.8%) 0.1297 ( 0.3%) 0.1297 ( 0.3%) Live Variable Analysis
0.1078 ( 0.2%) 0.0015 ( 0.5%) 0.1093 ( 0.2%) 0.1094 ( 0.2%) Combine redundant instructions
0.1076 ( 0.2%) 0.0011 ( 0.4%) 0.1087 ( 0.2%) 0.1087 ( 0.2%) Combine redundant instructions
0.1068 ( 0.2%) 0.0014 ( 0.5%) 0.1082 ( 0.2%) 0.1082 ( 0.2%) Combine redundant instructions
0.0956 ( 0.2%) 0.0009 ( 0.3%) 0.0965 ( 0.2%) 0.0965 ( 0.2%) Machine Common Subexpression Elimination
0.0949 ( 0.2%) 0.0003 ( 0.1%) 0.0952 ( 0.2%) 0.0952 ( 0.2%) Control Flow Optimizer
0.0887 ( 0.2%) 0.0013 ( 0.4%) 0.0900 ( 0.2%) 0.0900 ( 0.2%) X86 AT&T-Style Assembly Printer
0.0670 ( 0.2%) 0.0092 ( 3.0%) 0.0762 ( 0.2%) 0.0762 ( 0.2%) Machine Function Analysis
0.0619 ( 0.1%) 0.0032 ( 1.0%) 0.0651 ( 0.1%) 0.0651 ( 0.1%) Jump Threading
0.0621 ( 0.1%) 0.0007 ( 0.2%) 0.0628 ( 0.1%) 0.0628 ( 0.1%) Early CSE
0.0590 ( 0.1%) 0.0026 ( 0.9%) 0.0616 ( 0.1%) 0.0616 ( 0.1%) Sparse Conditional Constant Propagation
0.0557 ( 0.1%) 0.0023 ( 0.7%) 0.0579 ( 0.1%) 0.0579 ( 0.1%) Value Propagation
0.0527 ( 0.1%) 0.0041 ( 1.3%) 0.0568 ( 0.1%) 0.0567 ( 0.1%) Live Interval Analysis
0.0561 ( 0.1%) 0.0005 ( 0.2%) 0.0566 ( 0.1%) 0.0566 ( 0.1%) Value Propagation
0.0453 ( 0.1%) 0.0000 ( 0.0%) 0.0453 ( 0.1%) 0.0453 ( 0.1%) Optimize for code generation
0.0417 ( 0.1%) 0.0008 ( 0.2%) 0.0425 ( 0.1%) 0.0425 ( 0.1%) Jump Threading
0.0382 ( 0.1%) 0.0000 ( 0.0%) 0.0382 ( 0.1%) 0.0382 ( 0.1%) Module Verifier
0.0381 ( 0.1%) 0.0000 ( 0.0%) 0.0381 ( 0.1%) 0.0381 ( 0.1%) Module Verifier
0.0366 ( 0.1%) 0.0006 ( 0.2%) 0.0372 ( 0.1%) 0.0372 ( 0.1%) Reassociate expressions
0.0354 ( 0.1%) 0.0000 ( 0.0%) 0.0354 ( 0.1%) 0.0354 ( 0.1%) Virtual Register Rewriter
0.0330 ( 0.1%) 0.0009 ( 0.3%) 0.0339 ( 0.1%) 0.0339 ( 0.1%) Two-Address instruction pass
0.0329 ( 0.1%) 0.0008 ( 0.3%) 0.0337 ( 0.1%) 0.0337 ( 0.1%) Prologue/Epilogue Insertion & Frame Finalization
0.0297 ( 0.1%) 0.0008 ( 0.3%) 0.0305 ( 0.1%) 0.0305 ( 0.1%) Aggressive Dead Code Elimination
0.0290 ( 0.1%) 0.0011 ( 0.4%) 0.0301 ( 0.1%) 0.0301 ( 0.1%) Dominator Tree Construction
0.0293 ( 0.1%) 0.0007 ( 0.2%) 0.0300 ( 0.1%) 0.0300 ( 0.1%) Dominator Tree Construction
0.0288 ( 0.1%) 0.0010 ( 0.3%) 0.0299 ( 0.1%) 0.0299 ( 0.1%) Dominator Tree Construction
0.0239 ( 0.1%) 0.0054 ( 1.8%) 0.0294 ( 0.1%) 0.0294 ( 0.1%) Slot index numbering
0.0274 ( 0.1%) 0.0003 ( 0.1%) 0.0278 ( 0.1%) 0.0292 ( 0.1%) Combine redundant instructions
0.0267 ( 0.1%) 0.0023 ( 0.7%) 0.0289 ( 0.1%) 0.0289 ( 0.1%) MachineDominator Tree Construction
0.0268 ( 0.1%) 0.0014 ( 0.5%) 0.0282 ( 0.1%) 0.0282 ( 0.1%) Dominator Tree Construction
0.0171 ( 0.0%) 0.0106 ( 3.5%) 0.0277 ( 0.1%) 0.0277 ( 0.1%) Early CSE
0.0262 ( 0.1%) 0.0008 ( 0.3%) 0.0270 ( 0.1%) 0.0270 ( 0.1%) Dominator Tree Construction
0.0238 ( 0.1%) 0.0000 ( 0.0%) 0.0238 ( 0.1%) 0.0238 ( 0.1%) Calculate spill weights
0.0217 ( 0.0%) 0.0001 ( 0.0%) 0.0218 ( 0.0%) 0.0218 ( 0.0%) Dead Store Elimination
0.0207 ( 0.0%) 0.0006 ( 0.2%) 0.0213 ( 0.0%) 0.0213 ( 0.0%) MachineDominator Tree Construction
0.0210 ( 0.0%) 0.0000 ( 0.0%) 0.0210 ( 0.0%) 0.0210 ( 0.0%) Machine Copy Propagation Pass
0.0200 ( 0.0%) 0.0000 ( 0.0%) 0.0200 ( 0.0%) 0.0200 ( 0.0%) Simplify the CFG
0.0197 ( 0.0%) 0.0001 ( 0.0%) 0.0197 ( 0.0%) 0.0197 ( 0.0%) Simplify the CFG
0.0166 ( 0.0%) 0.0027 ( 0.9%) 0.0193 ( 0.0%) 0.0193 ( 0.0%) Dominator Tree Construction
0.0182 ( 0.0%) 0.0008 ( 0.3%) 0.0191 ( 0.0%) 0.0191 ( 0.0%) Execution dependency fix
0.0178 ( 0.0%) 0.0000 ( 0.0%) 0.0178 ( 0.0%) 0.0178 ( 0.0%) Tail Duplication
0.0170 ( 0.0%) 0.0006 ( 0.2%) 0.0176 ( 0.0%) 0.0176 ( 0.0%) Lazy Value Information Analysis
0.0173 ( 0.0%) 0.0003 ( 0.1%) 0.0176 ( 0.0%) 0.0176 ( 0.0%) Lazy Value Information Analysis
0.0160 ( 0.0%) 0.0000 ( 0.0%) 0.0160 ( 0.0%) 0.0160 ( 0.0%) Remove dead machine instructions
0.0137 ( 0.0%) 0.0003 ( 0.1%) 0.0140 ( 0.0%) 0.0140 ( 0.0%) Simplify the CFG
0.0117 ( 0.0%) 0.0006 ( 0.2%) 0.0124 ( 0.0%) 0.0124 ( 0.0%) Natural Loop Information
0.0115 ( 0.0%) 0.0006 ( 0.2%) 0.0121 ( 0.0%) 0.0121 ( 0.0%) Natural Loop Information
0.0115 ( 0.0%) 0.0003 ( 0.1%) 0.0118 ( 0.0%) 0.0118 ( 0.0%) Interprocedural Sparse Conditional Constant Propagation
0.0110 ( 0.0%) 0.0007 ( 0.2%) 0.0118 ( 0.0%) 0.0118 ( 0.0%) Machine Block Frequency Analysis
0.0107 ( 0.0%) 0.0006 ( 0.2%) 0.0113 ( 0.0%) 0.0113 ( 0.0%) X86 FP Stackifier
0.0107 ( 0.0%) 0.0006 ( 0.2%) 0.0113 ( 0.0%) 0.0113 ( 0.0%) Branch Probability Analysis
0.0106 ( 0.0%) 0.0007 ( 0.2%) 0.0112 ( 0.0%) 0.0112 ( 0.0%) Natural Loop Information
0.0108 ( 0.0%) 0.0004 ( 0.1%) 0.0112 ( 0.0%) 0.0112 ( 0.0%) Scalar Replacement of Aggregates (DT)
0.0097 ( 0.0%) 0.0000 ( 0.0%) 0.0097 ( 0.0%) 0.0097 ( 0.0%) MemCpy Optimization
0.0090 ( 0.0%) 0.0000 ( 0.0%) 0.0090 ( 0.0%) 0.0090 ( 0.0%) Dead Global Elimination
0.0086 ( 0.0%) 0.0001 ( 0.0%) 0.0086 ( 0.0%) 0.0086 ( 0.0%) Scalar Replacement of Aggregates (SSAUp)
0.0082 ( 0.0%) 0.0004 ( 0.1%) 0.0086 ( 0.0%) 0.0086 ( 0.0%) Dominator Tree Construction
0.0081 ( 0.0%) 0.0000 ( 0.0%) 0.0081 ( 0.0%) 0.0081 ( 0.0%) Module Verifier
0.0068 ( 0.0%) 0.0004 ( 0.1%) 0.0072 ( 0.0%) 0.0072 ( 0.0%) Post-RA pseudo instruction expansion pass
0.0067 ( 0.0%) 0.0005 ( 0.2%) 0.0072 ( 0.0%) 0.0072 ( 0.0%) Machine Natural Loop Construction
0.0071 ( 0.0%) 0.0000 ( 0.0%) 0.0071 ( 0.0%) 0.0071 ( 0.0%) Peephole Optimizations
0.0061 ( 0.0%) 0.0000 ( 0.0%) 0.0061 ( 0.0%) 0.0061 ( 0.0%) Simplify well-known library calls
0.0055 ( 0.0%) 0.0005 ( 0.2%) 0.0060 ( 0.0%) 0.0060 ( 0.0%) Machine Natural Loop Construction
0.0057 ( 0.0%) 0.0002 ( 0.1%) 0.0059 ( 0.0%) 0.0059 ( 0.0%) Remove unreachable machine basic blocks
0.0054 ( 0.0%) 0.0000 ( 0.0%) 0.0054 ( 0.0%) 0.0054 ( 0.0%) Remove unreachable blocks from the CFG
0.0049 ( 0.0%) 0.0000 ( 0.0%) 0.0049 ( 0.0%) 0.0049 ( 0.0%) Insert stack protectors
0.0048 ( 0.0%) 0.0000 ( 0.0%) 0.0049 ( 0.0%) 0.0049 ( 0.0%) Tail Call Elimination
0.0042 ( 0.0%) 0.0000 ( 0.0%) 0.0042 ( 0.0%) 0.0042 ( 0.0%) Tail Duplication
0.0040 ( 0.0%) 0.0000 ( 0.0%) 0.0040 ( 0.0%) 0.0040 ( 0.0%) Debug Variable Analysis
0.0040 ( 0.0%) 0.0001 ( 0.0%) 0.0040 ( 0.0%) 0.0040 ( 0.0%) Basic CallGraph Construction
0.0029 ( 0.0%) 0.0001 ( 0.0%) 0.0030 ( 0.0%) 0.0030 ( 0.0%) Bundle Machine CFG Edges
0.0029 ( 0.0%) 0.0000 ( 0.0%) 0.0030 ( 0.0%) 0.0029 ( 0.0%) Simplify the CFG
0.0029 ( 0.0%) 0.0000 ( 0.0%) 0.0029 ( 0.0%) 0.0029 ( 0.0%) Bundle Machine CFG Edges
0.0020 ( 0.0%) 0.0000 ( 0.0%) 0.0020 ( 0.0%) 0.0020 ( 0.0%) Expand ISel Pseudo-instructions
0.0019 ( 0.0%) 0.0000 ( 0.0%) 0.0019 ( 0.0%) 0.0019 ( 0.0%) Process Implicit Definitions
0.0018 ( 0.0%) 0.0000 ( 0.0%) 0.0019 ( 0.0%) 0.0019 ( 0.0%) Remove unused exception handling info
0.0016 ( 0.0%) 0.0000 ( 0.0%) 0.0016 ( 0.0%) 0.0016 ( 0.0%) Preliminary module verification
0.0013 ( 0.0%) 0.0000 ( 0.0%) 0.0013 ( 0.0%) 0.0013 ( 0.0%) Memory Dependence Analysis
0.0011 ( 0.0%) 0.0001 ( 0.0%) 0.0012 ( 0.0%) 0.0012 ( 0.0%) Deduce function attributes
0.0011 ( 0.0%) 0.0000 ( 0.0%) 0.0011 ( 0.0%) 0.0012 ( 0.0%) Preliminary module verification
0.0010 ( 0.0%) 0.0000 ( 0.0%) 0.0010 ( 0.0%) 0.0010 ( 0.0%) Optimize machine instruction PHIs
0.0009 ( 0.0%) 0.0000 ( 0.0%) 0.0009 ( 0.0%) 0.0009 ( 0.0%) Preliminary module verification
0.0008 ( 0.0%) 0.0000 ( 0.0%) 0.0008 ( 0.0%) 0.0008 ( 0.0%) Spill Code Placement Analysis
0.0006 ( 0.0%) 0.0000 ( 0.0%) 0.0007 ( 0.0%) 0.0007 ( 0.0%) Global Variable Optimizer
0.0005 ( 0.0%) 0.0000 ( 0.0%) 0.0005 ( 0.0%) 0.0005 ( 0.0%) Lower 'expect' Intrinsics
0.0005 ( 0.0%) 0.0000 ( 0.0%) 0.0005 ( 0.0%) 0.0005 ( 0.0%) Exception handling preparation
0.0001 ( 0.0%) 0.0003 ( 0.1%) 0.0005 ( 0.0%) 0.0005 ( 0.0%) Virtual Register Map
0.0001 ( 0.0%) 0.0002 ( 0.1%) 0.0003 ( 0.0%) 0.0004 ( 0.0%) No Alias Analysis (always returns 'may' alias)
0.0003 ( 0.0%) 0.0000 ( 0.0%) 0.0003 ( 0.0%) 0.0003 ( 0.0%) Live Register Matrix
0.0002 ( 0.0%) 0.0000 ( 0.0%) 0.0002 ( 0.0%) 0.0002 ( 0.0%) Dead Argument Elimination
0.0002 ( 0.0%) 0.0000 ( 0.0%) 0.0002 ( 0.0%) 0.0002 ( 0.0%) X86 Maximal Stack Alignment Check
0.0001 ( 0.0%) 0.0001 ( 0.0%) 0.0001 ( 0.0%) 0.0001 ( 0.0%) Memory Dependence Analysis
0.0001 ( 0.0%) 0.0001 ( 0.0%) 0.0001 ( 0.0%) 0.0001 ( 0.0%) Memory Dependence Analysis
0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0001 ( 0.0%) 0.0001 ( 0.0%) Basic Alias Analysis (stateless AA impl)
0.0001 ( 0.0%) 0.0000 ( 0.0%) 0.0001 ( 0.0%) 0.0001 ( 0.0%) Scalar Evolution Analysis
0.0000 ( 0.0%) 0.0001 ( 0.0%) 0.0001 ( 0.0%) 0.0001 ( 0.0%) Basic Alias Analysis (stateless AA impl)
0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0001 ( 0.0%) 0.0001 ( 0.0%) Promote 'by reference' arguments to scalars
0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0001 ( 0.0%) 0.0001 ( 0.0%) No Alias Analysis (always returns 'may' alias)
0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0000 ( 0.0%) No Alias Analysis (always returns 'may' alias)
0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0000 ( 0.0%) Target Library Information
0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0000 ( 0.0%) Machine Loop Invariant Code Motion
0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0000 ( 0.0%) Post RA top-down list latency scheduler
0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0000 ( 0.0%) Scalar Evolution Analysis
0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0000 ( 0.0%) Live Stack Slot Analysis
0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0000 ( 0.0%) Strip Unused Function Prototypes
0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0000 ( 0.0%) Machine Loop Invariant Code Motion
0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0000 ( 0.0%) Stack Slot Coloring
0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0000 ( 0.0%) Local Stack Slot Allocation
0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0000 ( 0.0%) Analyze Machine Code For Garbage Collection
0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0000 ( 0.0%) Delete Garbage Collector Information
0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0000 ( 0.0%) Lower Garbage Collection Instructions
0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0000 ( 0.0%) Live Stack Slot Analysis
0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0000 ( 0.0%) Merge Duplicate Global Constants
0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0000 ( 0.0%) Target Pass Configuration
0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0000 ( 0.0%) Create Garbage Collector Module Metadata
0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0000 ( 0.0%) Target Library Information
0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0000 ( 0.0%) Machine Module Information
0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0000 ( 0.0%) Machine Branch Probability Analysis
44.5154 (100.0%) 0.3060 (100.0%) 44.8214 (100.0%) 44.8240 (100.0%) Total

===-------------------------------------------------------------------------===
Miscellaneous Ungrouped Timers
===-------------------------------------------------------------------------===

---User Time--- --System Time-- --User+System-- ---Wall Time--- --- Name ---
45.0637 ( 50.2%) 0.3487 ( 52.4%) 45.4124 ( 50.2%) 45.4247 ( 50.2%) Clang front-end timer
44.5841 ( 49.7%) 0.3102 ( 46.6%) 44.8943 ( 49.7%) 44.8969 ( 49.7%) Code Generation Time
0.0806 ( 0.1%) 0.0061 ( 0.9%) 0.0867 ( 0.1%) 0.0869 ( 0.1%) LLVM IR Generation Time
89.7284 (100.0%) 0.6650 (100.0%) 90.3934 (100.0%) 90.4086 (100.0%) Total

[nico]

+Jakob, since a lot of the time is spent in "Eliminate PHI nodes for register allocation".

[stoklund]

When PHI-elim takes a lot of time, it is usually because of critical edge splitting. Does llc -disable-phi-elim-edge-splitting go a lot faster?

[llvmbot]

When PHI-elim takes a lot of time, it is usually because of critical edge
splitting. Does llc -disable-phi-elim-edge-splitting go a lot faster?

Yeah, that speeds it up considerably. It then takes around 2 seconds.

[llvmbot]

When PHI-elim takes a lot of time, it is usually because of critical edge
splitting. Does llc -disable-phi-elim-edge-splitting go a lot faster?

Yeah, that speeds it up considerably. It then takes around 2 seconds.

Err, wait no sorry. Running llc on the byte code is that fast regardless of the inclusion of -disable-phi-elim-edge-splitting.

[llvmbot]

Did you compile LLVM with optimization on and assertions off?

[llvmbot]

Did you compile LLVM with optimization on and assertions off?

I had optimizations on. I tried again after rebuilding while disabling asserts, but there was no difference in performance.

[d0k]

Profiling shows that >80% of the time is spent updating LiveVariables when splitting critical edges.

[d0k]

r163478 brings this from 40s to 15s at -O2.