[arm64][turbofan]: Handle any immediate shift.

[arm64][turbofan]: Handle any immediate shift.

With this patch, we can generate simple immediate-shift instructions for
immediates outside the range "0 <= imm < width". Several related
instruction selectors have also been updated accordingly.

Example of generated code:

    ---- Before ---         ---- After ----
    movz w0, #33            lsr w0, w1, #1
    lsr  w0, w1, w0

BUG=

Committed: https://crrev.com/36d771bbfa4af5efcc1c1dcf5b234445cb7ee722
Cr-Commit-Position: refs/heads/master@{#28943}

[jbramley, 2015-06-11 13:09:41]

jacob.bramley@arm.com changed reviewers:
+ bmeurer@chromium.org, ulan@chromium.org

[jbramley, 2015-06-11 13:09:42]

This is relevant to the discussion in issue 1179503003.

As I see it, we have two options:
- Handle arbitrary immediate shifts. (That's what this patch does.)
- Disallow out-of-range immediate shifts.

I prefer the method of just taking bits 0-4/5 because that's what ARM64 hardware
does for variable shifts. Since we currently use variable shift instructions for
out-of-range immediates, the actual behaviour of the generated code is unchanged
by this patch. It's also how JavaScript shifts are defined, and it makes sense
for the kWord32ShiftIsSafe flag.

That is, kWord32ShiftIsSafe allows us to provide out-of-range values for
variable shifts so why not for immediate shifts?

For the record, (at least) the following JavaScript will generate out-of-range
immediate shifts:

    function f() {
      var x = 0;
      x -= 34;
      return x >>> (33 % x);
    }
    print(f());

This kind of pattern is found in the numops-fuzz tests, and in one or two
instruction selector tests, but that's all.

What are your thoughts on the matter?

[Benedikt Meurer, 2015-06-11 13:23:20]

Yeah, I see your point. I guess the argumentation makes sense, so let's go for
it. LGTM

[jbramley, 2015-06-11 13:26:33]

The CQ bit was checked by jacob.bramley@arm.com

[commit-bot: I haz the power, 2015-06-11 13:26:42]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1179733004/1

[commit-bot: I haz the power, 2015-06-11 13:50:56]

Committed patchset #1 (id:1)

[commit-bot: I haz the power, 2015-06-11 13:51:13]

Patchset 1 (id:??) landed as
https://crrev.com/36d771bbfa4af5efcc1c1dcf5b234445cb7ee722
Cr-Commit-Position: refs/heads/master@{#28943}

[jbramley, 2015-06-11 14:46:15]

A revert of this CL (patchset #1 id:1) has been created in
https://codereview.chromium.org/1176393002/ by jacob.bramley@arm.com.

The reason for reverting is: Breaks
InstructionSelectorTest.Word64ShrWithWord64AndWithImmediate on debug builds (but
not optdebug builds). I'll investigate..