[BACKEND] Update LLVM version to 2fe947b4

[makslevental]

The only change here is due to deprecation of applyPatternsAndFoldGreedily. Note, possibly this regression is still an issue but I'm just catching up to API changes so that if/when the regression is resolved we're not overwhelmed.

Note, also the commit reflects where llvm-head is currently (so tars should be available...).

[peterbell10]

I'm seeing memory access errors in our internal workloads. For OAI people the log is here.

I don't have the bandwidth to investigate further just now.

[makslevental]

I have "bisected" this on A100:

git bisect start
# bad: [f5a35a31bfe6cbc16bec0c130f2bb3632dbf1fbf] [LV] Add test cases with incorrect IV live-outs.
git bisect bad f5a35a31bfe6cbc16bec0c130f2bb3632dbf1fbf
# good: [86b69c31642e98f8357df62c09d118ad1da4e16a] [SPIR-V] Fix SPIR-V extension SPV_INTEL_function_pointers: introduce CodeSectionINTEL (#117250)
git bisect good 86b69c31642e98f8357df62c09d118ad1da4e16a
# bad: [9359625ba99dfbce8d8c27373ade544df16bee34] [OpenACC] 'create' clause sema for data/enter data constructs
git bisect bad 9359625ba99dfbce8d8c27373ade544df16bee34
# bad: [4675db5f394ebadd8a56bb17e2344da5c7779a68] [DAGCombiner] Add support for scalarising extracts of a vector setcc (#117566)
git bisect bad 4675db5f394ebadd8a56bb17e2344da5c7779a68
# bad: [4675db5f394ebadd8a56bb17e2344da5c7779a68] [DAGCombiner] Add support for scalarising extracts of a vector setcc (#117566)
git bisect bad 4675db5f394ebadd8a56bb17e2344da5c7779a68
# bad: [e3fdc3aa81c7e8196d804fbde1a3f76796dbd5dd] [RISCV] Allow hoisting VXRM writes out of loops speculatively (#110044)
git bisect bad e3fdc3aa81c7e8196d804fbde1a3f76796dbd5dd
# bad: [bb88fd171a6be892cec36969860a9034b48b8656] [DirectX] Calculate resource binding offsets using the lower bound (#117303)
git bisect bad bb88fd171a6be892cec36969860a9034b48b8656
# bad: [ae01e3a7c66d68544c85dd6e2c5498d3c5896e29] [nfc][sancov] Remove unnecessary default argument (#117463)
git bisect bad ae01e3a7c66d68544c85dd6e2c5498d3c5896e29
# bad: [8f50321256ecc1e7d433ca00baff47bd8a8bc81e] [libc++] Add a .clang-tidy file to libcxx/src
git bisect bad 8f50321256ecc1e7d433ca00baff47bd8a8bc81e
# good: [689c53219280151b6421d633d993ed886827849b] [CUDA] pass -fno-threadsafe-statics to GPU sub-compilations. (#117074)
git bisect good 689c53219280151b6421d633d993ed886827849b
# good: [8d3435f8a111a39dc333c0ffeafd5ffe953f1f02] AMDGPU: Add v_smfmac_f32_32x32x64_bf8_fp8 for gfx950 (#117257)
git bisect good 8d3435f8a111a39dc333c0ffeafd5ffe953f1f02
# bad: [4ab5e90c9c52fcf9b4ed297c887b31219e41f7d8] [flang][Driver] Add correct libraries to driver
git bisect bad 4ab5e90c9c52fcf9b4ed297c887b31219e41f7d8
# good: [2fe947b47798de1ad20553be4e162e332428ad91] [Clang] Add Doug Wyatt and myself as maintainers for function effect analysis (#117324)
git bisect good 2fe947b47798de1ad20553be4e162e332428ad91
# bad: [7523086a050d679370dfd86a0166d5f7168ffa09] [SLP]Use getExtendedReduction cost and fix reduction cost calculations
git bisect bad 7523086a050d679370dfd86a0166d5f7168ffa09
# bad: [b8703369daf777706196ff914c0376c27adde3cf] [SLP] Match poison as instruction with the same opcode
git bisect bad b8703369daf777706196ff914c0376c27adde3cf
# first bad commit: [b8703369daf777706196ff914c0376c27adde3cf] [SLP] Match poison as instruction with the same opcode

The failing commit is b870336 and the last working commit is 2fe947b4. It's not clear to me the relationship but will investigate.

@peterbell10 I will send a PR to llvm-head so you/we can verify this matches the internal fail.

cc @antiagainst @jataylo @ThomasRaoux

[peterbell10]

Looks like that was it, passing internal CI now.

[antiagainst]

Awesome! Thanks @makslevental for the bisecting! I'll leave @peterbell10 to approve and merge.