Huge code bloat with wasi-sdk-23

Hi,

I've just updated my benchmark binaries to compile with 23, instead of 22. I'm seeing code size changes that I don't understand, maybe changes in how the libraries are built? Has anyone noticed similar changes? Below is my git pull log which caught my attention!

cheers,
Sam

 benchmarks/adobe/wasm/functionobjects.wasm         | Bin 508408 -> 646733 bytes
 benchmarks/adobe/wasm/loop_unroll.wasm             | Bin 477734 -> 631955 bytes
 benchmarks/adobe/wasm/stepanov_vector.wasm         | Bin 122282 -> 274559 bytes
 benchmarks/aha/wasm/aha.wasm                       | Bin 33029 -> 152764 bytes
 benchmarks/alac/wasm/encode.wasm                   | Bin 163918 -> 374160 bytes
 benchmarks/bullet/wasm/1000-convex.wasm            | Bin 889848 -> 1080895 bytes
 benchmarks/bullet/wasm/1000-stack.wasm             | Bin 889847 -> 1080894 bytes
 benchmarks/bullet/wasm/136-ragdolls.wasm           | Bin 889849 -> 1080896 bytes
 benchmarks/bullet/wasm/3000-fall.wasm              | Bin 889846 -> 1080893 bytes
 benchmarks/bullet/wasm/convex-trimesh.wasm         | Bin 889851 -> 1080898 bytes
 benchmarks/bullet/wasm/prim-trimesh.wasm           | Bin 889849 -> 1080896 bytes
 benchmarks/bullet/wasm/raytests.wasm               | Bin 889845 -> 1080892 bytes
 benchmarks/jetstream2/wasm/float-mm.wasm           | Bin 21266 -> 102339 bytes
 benchmarks/jetstream2/wasm/gcc-loops.wasm          | Bin 2415704 -> 2777604 bytes
 benchmarks/jetstream2/wasm/hashset.wasm            | Bin 2716050 -> 3078733 bytes
 benchmarks/jetstream2/wasm/quicksort.wasm          | Bin 20233 -> 104079 bytes
 benchmarks/jetstream2/wasm/richards.wasm           | Bin 29787 -> 150670 bytes
 benchmarks/llama2/wasm/run.wasm                    | Bin 91974 -> 364223 bytes
 benchmarks/lzma/wasm/lzma.wasm                     | Bin 79858 -> 202837 bytes  
 benchmarks/meshoptimizer/wasm/codecbench.wasm      | Bin 93145 -> 219666 bytes   
 benchmarks/minisat/wasm/minisat.wasm               | Bin 120254 -> 361591 bytes
 benchmarks/oggenc/wasm/oggenc.wasm                 | Bin 1198354 -> 1491969 bytes
 benchmarks/olden/wasm/bisort.wasm                  | Bin 30221 -> 150810 bytes
 benchmarks/olden/wasm/em3d.wasm                    | Bin 32955 -> 157059 bytes
 benchmarks/olden/wasm/health.wasm                  | Bin 31865 -> 154573 bytes
 benchmarks/olden/wasm/mst.wasm                     | Bin 31427 -> 152016 bytes   
 benchmarks/olden/wasm/perimeter.wasm               | Bin 30454 -> 148273 bytes
 benchmarks/olden/wasm/power.wasm                   | Bin 32960 -> 152747 bytes   
 benchmarks/olden/wasm/treeadd.wasm                 | Bin 29255 -> 149844 bytes  
 benchmarks/olden/wasm/tsp.wasm                     | Bin 36749 -> 162879 bytes 
 benchmarks/olden/wasm/voronoi.wasm                 | Bin 46459 -> 175376 bytes 
 benchmarks/pairlocalalign/wasm/pairlocalalign.wasm | Bin 297406 -> 579510 bytes
 benchmarks/peanut-gb/wasm/peanut-gb.wasm           | Bin 146626 -> 245030 bytes
 benchmarks/pocket-nn/wasm/pocketnn.wasm            | Bin 2435239 -> 2797863 bytes
 benchmarks/polybench/wasm/2mm.wasm                 | Bin 33802 -> 154904 bytes   
 benchmarks/polybench/wasm/3mm.wasm                 | Bin 34214 -> 155316 bytes 
 benchmarks/polybench/wasm/adi.wasm                 | Bin 33920 -> 155022 bytes 
 benchmarks/polybench/wasm/atax.wasm                | Bin 33041 -> 154143 bytes   
 benchmarks/polybench/wasm/bicg.wasm                | Bin 32687 -> 153789 bytes
 benchmarks/polybench/wasm/cholesky.wasm            | Bin 33011 -> 154113 bytes
 benchmarks/polybench/wasm/correlation.wasm         | Bin 34013 -> 155115 bytes
 benchmarks/polybench/wasm/covariance.wasm          | Bin 33598 -> 154700 bytes
 benchmarks/polybench/wasm/doitgen.wasm             | Bin 33348 -> 154450 bytes
 benchmarks/polybench/wasm/durbin.wasm              | Bin 33498 -> 154600 bytes
 benchmarks/polybench/wasm/dynprog.wasm             | Bin 33953 -> 155055 bytes
 benchmarks/polybench/wasm/fdtd-2d.wasm             | Bin 34449 -> 155551 bytes
 benchmarks/polybench/wasm/floyd-warshall.wasm      | Bin 32575 -> 153677 bytes
 benchmarks/polybench/wasm/gemm.wasm                | Bin 33111 -> 154213 bytes
 benchmarks/polybench/wasm/gemver.wasm              | Bin 34009 -> 155111 bytes
 benchmarks/polybench/wasm/gesummv.wasm             | Bin 32724 -> 153826 bytes
 benchmarks/polybench/wasm/gramschmidt.wasm         | Bin 33973 -> 155075 bytes
 benchmarks/polybench/wasm/jacobi-1d-imper.wasm     | Bin 32894 -> 153996 bytes
 benchmarks/polybench/wasm/jacobi-2d-imper.wasm     | Bin 33088 -> 154190 bytes
 benchmarks/polybench/wasm/lu.wasm                  | Bin 32943 -> 154045 bytes
 benchmarks/polybench/wasm/ludcmp.wasm              | Bin 33955 -> 155057 bytes
 benchmarks/polybench/wasm/mvt.wasm                 | Bin 33158 -> 154260 bytes
 benchmarks/polybench/wasm/reg_detect.wasm          | Bin 37088 -> 158190 bytes
 benchmarks/polybench/wasm/seidel-2d.wasm           | Bin 32531 -> 153633 bytes
 benchmarks/polybench/wasm/symm.wasm                | Bin 33087 -> 154189 bytes
 benchmarks/polybench/wasm/syr2k.wasm               | Bin 33416 -> 154518 bytes
 benchmarks/polybench/wasm/syrk.wasm                | Bin 33047 -> 154149 bytes
 benchmarks/polybench/wasm/trisolv.wasm             | Bin 32509 -> 153611 bytes
 benchmarks/polybench/wasm/trmm.wasm                | Bin 32845 -> 153947 bytes
 benchmarks/quickjs/wasm/quickjs.wasm               | Bin 928646 -> 1272433 bytes
 benchmarks/raytracing/wasm/ray.wasm                | Bin 2399738 -> 2762424 bytes
 benchmarks/raytracing/wasm/smallpt.wasm            | Bin 96937 -> 317912 bytes
 benchmarks/raytracing/wasm/sphereflake.wasm        | Bin 2399958 -> 2762644 bytes
 benchmarks/shootout/wasm/binarytrees.wasm          | Bin 32025 -> 154471 bytes
 benchmarks/shootout/wasm/fannkuchredux.wasm        | Bin 30707 -> 150624 bytes
 benchmarks/shootout/wasm/fasta.wasm                | Bin 49609 -> 203531 bytes
 benchmarks/shootout/wasm/meteor.wasm               | Bin 2403848 -> 2766607 bytes
 benchmarks/shootout/wasm/nbody.wasm                | Bin 32955 -> 150771 bytes
 benchmarks/shootout/wasm/spectralnorm.wasm         | Bin 2396611 -> 2760153 bytes
 benchmarks/spec2017/wasm/508.namd_r.wasm           | Bin 926211 -> 1207218 bytes
 benchmarks/spec2017/wasm/525.x264_r.wasm           | Bin 668771 -> 984164 bytes
 benchmarks/spec2017/wasm/531.deepsjeng_r.wasm      | Bin 141565 -> 392130 bytes
 benchmarks/spec2017/wasm/544.nab_r.wasm            | Bin 188731 -> 490742 bytes
 benchmarks/spiff/wasm/spiff.wasm                   | Bin 69690 -> 287388 bytes
 benchmarks/sqlite3/wasm/sqlite3.wasm               | Bin 1395264 -> 1704213 bytes
 benchmarks/tramp3d-v4/wasm/tramp3d-v4.wasm         | Bin 4768150 -> 5157573 bytes
 benchmarks/tsvc/wasm/tsvc.wasm                     | Bin 231640 -> 364655 bytes
 benchmarks/wasm3/wasm/wasm3.wasm                   | Bin 179305 -> 434474 bytes
 benchmarks/zlib-bench/wasm/zlib-bench.wasm         | Bin 2856892 -> 3218373 bytes

Start by comparing the listed benchmarks//wasm/.wasm outputs built with wasi-sdk-22 and wasi-sdk-23, then inspect the SDK or library build changes that could explain the size increases. Done means identifying the cause of the regression and documenting or validating a fix against representative benchmark binaries.