Handle UINT64_MAX edge case in switch lowering.

Handle UINT64_MAX edge case in switch lowering.

BUG=
R=jvoung@chromium.org, jvoung

Committed: https://gerrit.chromium.org/gerrit/gitweb?p=native_client/pnacl-subzero.git;a=commit;h=016c56d966eb913007f781b5b90c676322180679

[ascull, 2015-07-23 19:32:31]

The 64-bit lowering is still done the old way (even though it is
analysed and identified as a jump table opportunity) so you
won't get the performance benefit for those.

[jvoung (off chromium), 2015-07-23 21:26:34]

LGTM thanks

https://codereview.chromium.org/1247833003/diff/1/tests_lit/llvm2ice_tests/ad...
File tests_lit/llvm2ice_tests/adv-switch-opt.ll (right):

https://codereview.chromium.org/1247833003/diff/1/tests_lit/llvm2ice_tests/ad...
tests_lit/llvm2ice_tests/adv-switch-opt.ll:200: i64 18446744073709551615, label
%sw.bb3
I think technically you could say "-1" also, if that seems more "realistic", but
this is probably okay too

https://codereview.chromium.org/1247833003/diff/1/tests_lit/llvm2ice_tests/ad...
tests_lit/llvm2ice_tests/adv-switch-opt.ll:218: }
Maybe move the "TODO" to the bottom here, add the

; CHECK-LABEL: testJumpTable64
; TODO(ascull): this should generate a jump table. For now, just make
; sure it doesn't crash the compiler.

??

[ascull, 2015-07-23 21:31:16]

Committed patchset #2 (id:20001) manually as
016c56d966eb913007f781b5b90c676322180679 (presubmit successful).