[llvm] [PatternMatch] Use `m_SpecificCmp` matchers. NFC. (PR #100878)

[Yingwei Zheng via llvm-commits]

dtcxzyw wrote:

Can you please check what kind of results you get if you run the compile-time test against a no-op change to see what the noise level looks like?

Tested with 803eaf29267c6aae9162d1a83a4a2ae508b440d3

Test1 vs Test2
```
Top 5 improvements:
  openjdk/ad_x86_gen.ll 19142207641 18261250223 -4.60%
  lightgbm/metadata.cpp.ll 5027690725 4825166371 -4.03%
  lightgbm/metric.cpp.ll 7180151122 6963045147 -3.02%
  lightgbm/gradient_discretizer.cpp.ll 1382530605 1340875607 -3.01%
  gromacs/atomdata.cpp.ll 2513395847 2437906279 -3.00%
Top 5 regressions:
  lightgbm/gbdt.cpp.ll 5331964084 5518023064 +3.49%
  faiss/IndexIVFFlat.cpp.ll 1736529706 1796315649 +3.44%
  faiss/lattice_Zn.cpp.ll 2131208372 2186387362 +2.59%
  lightgbm/linear_tree_learner.cpp.ll 9769789458 10022534163 +2.59%
  gromacs/constr.cpp.ll 2078220850 2125769072 +2.29%
Overall: -0.00069448%
```
Test1 vs Test3
```
Top 5 improvements:
  faiss/IndexIVFFastScan.cpp.ll 7434390364 7077625815 -4.80%
  lightgbm/metadata.cpp.ll 5027690725 4827333025 -3.99%
  lightgbm/sample_strategy.cpp.ll 1047493568 1008880792 -3.69%
  lightgbm/gradient_discretizer.cpp.ll 1382530605 1337101867 -3.29%
  faiss/IndexHNSW.cpp.ll 2671482829 2589963057 -3.05%
Top 5 regressions:
  openjdk/ad_x86_gen.ll 19142207641 23882411724 +24.76%
  faiss/IndexIVF.cpp.ll 2380453173 2463393700 +3.48%
  faiss/AdditiveQuantizer.cpp.ll 1411960257 1451771018 +2.82%
  lightgbm/gbdt.cpp.ll 5331964084 5475595898 +2.69%
  faiss/MetaIndexes.cpp.ll 1164233270 1194513258 +2.60%
Overall: 0.00045434%
```
Test2 vs Test3
```
Top 5 improvements:
  faiss/IndexIVFFastScan.cpp.ll 7413814621 7077625815 -4.53%
  lightgbm/linear_tree_learner.cpp.ll 10022534163 9677397451 -3.44%
  lightgbm/voting_parallel_tree_learner.cpp.ll 48690920539 47030359898 -3.41%
  faiss/IndexIVFFlat.cpp.ll 1796315649 1740205247 -3.12%
  lightgbm/data_parallel_tree_learner.cpp.ll 3577363719 3505678627 -2.00%
Top 5 regressions:
  openjdk/ad_x86_gen.ll 18261250223 23882411724 +30.78%
  faiss/IndexIVF.cpp.ll 2330588660 2463393700 +5.70%
  faiss/IndexIVFAdditiveQuantizerFastScan.cpp.ll 825778559 848334894 +2.73%
  gromacs/pullutil.cpp.ll 1468623225 1506443771 +2.58%
  gromacs/freeenergydispatch.cpp.ll 726020265 740090114 +1.94%
Overall: 0.00114883%
```


https://github.com/llvm/llvm-project/pull/100878