Split instcombine test file into one file per pattern (#519).

[sishtiaq]

I've split Test/Passes/instcombine.mlir into 16 files in Test/Passes/InstCombine.
What I don't know, so would appreciate feedback on:

1. Is this the right granularity to split at? It's basically file-per-pattern, but one person's pattern is another person's multi-pattern:)
2. The names of the files. The other tests in Test/Passes aren't super descriptive, so I've kept the new names high-level.

[github-actions]

VeIR Benchmarks

Details

Benchmark suite	Current: 4d589ab	Previous: 758a751	Ratio
add-fold-worklist/create	2209000 ns (± 64817)	1832000 ns (± 92681)	1.21
add-fold-worklist/rewrite	3772000 ns (± 46208)	3432000 ns (± 108244)	1.10
add-fold-worklist-local/create	2251000 ns (± 109363)	1980000 ns (± 89350)	1.14
add-fold-worklist-local/rewrite	3216000 ns (± 30372)	2864000 ns (± 24201)	1.12
add-zero-worklist/create	2347000 ns (± 97712)	1825000 ns (± 20379)	1.29
add-zero-worklist/rewrite	2604000 ns (± 55969)	2166000 ns (± 24366)	1.20
add-zero-reuse-worklist/create	1844000 ns (± 92673)	1558000 ns (± 32059)	1.18
add-zero-reuse-worklist/rewrite	2084500 ns (± 49605)	1768000 ns (± 19705)	1.18
mul-two-worklist/create	2248000 ns (± 32538)	1832000 ns (± 73493)	1.23
mul-two-worklist/rewrite	5443000 ns (± 63057)	4780000 ns (± 35704)	1.14
add-fold-forwards/create	2300000 ns (± 40856)	1993000 ns (± 82291)	1.15
add-fold-forwards/rewrite	3110000 ns (± 27727)	2711500 ns (± 136664)	1.15
add-zero-forwards/create	2301000 ns (± 22832)	1922000 ns (± 92695)	1.20
add-zero-forwards/rewrite	2095000 ns (± 32129)	1793000 ns (± 68165)	1.17
add-zero-reuse-forwards/create	1947500 ns (± 93571)	1558500 ns (± 78046)	1.25
add-zero-reuse-forwards/rewrite	1673500 ns (± 30651)	1388000 ns (± 64317)	1.21
mul-two-forwards/create	2345000 ns (± 54429)	1824000 ns (± 71272)	1.29
mul-two-forwards/rewrite	3796000 ns (± 75108)	3254000 ns (± 90075)	1.17
add-zero-reuse-first/create	1962000 ns (± 82770)	1563500 ns (± 122105)	1.25
add-zero-reuse-first/rewrite	8000 ns (± 405)	10000 ns (± 2383)	0.80
add-zero-lots-of-reuse-first/create	1939000 ns (± 92326)	1549000 ns (± 25383)	1.25
add-zero-lots-of-reuse-first/rewrite	888000 ns (± 26529)	788000 ns (± 19741)	1.13

This comment was automatically generated by workflow using github-action-benchmark.

[regehr]

this LGTM! although I don't really have any particular strong feelings about files containing fewer vs. more test cases

[luisacicolini]

Thank you, lgtm!
I don't have particularly strong feelings, although looking at llvm's test organization I wonder if we might want one test per operation and some "special" files for particular rewrites (e.g., de morgan)? Not sure.

[sishtiaq]

Added test.test in mult_two.mlir to use the mult_two variable.

[math-fehr]

Nice, thanks!