PPC/S390: [compiler] Delay allocation of code-embedded heap numbers.

PPC/S390: [compiler] Delay allocation of code-embedded heap numbers.

Port 659e8f7b5cd318dd486f6853c4da47a9a8587aa6

Original Commit Message:

    Instead of allocating and embedding certain heap numbers into the code
    during code assembly, emit dummies but record the allocation requests.
    Later then, in Assembler::GetCode, allocate the heap numbers and patch
    the code by replacing the dummies with the actual objects. The
    RelocInfos for the embedded objects are already recorded correctly when
    emitting the dummies.

R=neis@chromium.org, joransiu@ca.ibm.com, jyan@ca.ibm.com, bjaideep@ca.ibm.com, michael_dawson@ca.ibm.com
BUG=v8:6048
LOG=N

[neis, 2017-06-01 06:59:49]

Hi, this is missing the changes to src/compiler/ppc/code-generator-ppc.cc.
(Also, is this CL supposed to include the S390 changes?)

[Sampson, 2017-06-01 21:07:37]

On 2017/06/01 06:59:49, neis wrote:
> Hi, this is missing the changes to src/compiler/ppc/code-generator-ppc.cc.
> (Also, is this CL supposed to include the S390 changes?)

For the changes in code-generator-ppc.cc and code-generator-s390.cc, why are we
setting is_heap_number to true through EmbeddedNumber(double) and expect it to
be false in immediate()?

[neis, 2017-06-02 07:19:28]

On 2017/06/01 21:07:37, Sampson wrote:
> On 2017/06/01 06:59:49, neis wrote:
> > Hi, this is missing the changes to src/compiler/ppc/code-generator-ppc.cc.
> > (Also, is this CL supposed to include the S390 changes?)
> 
> For the changes in code-generator-ppc.cc and code-generator-s390.cc, why are
we
> setting is_heap_number to true through EmbeddedNumber(double) and expect it to
> be false in immediate()?

In the case where the operand is a heap number, it must not be accessed using
the immediate() getter. The CL is missing the changes to assembler-*.cc that
handle this case, analogous to the changes to move_32_bit_immediate in
https://codereview.chromium.org/2900683002/diff/180001/src/arm/assembler-arm.cc.
From a quick look at PPC it looks like this should happen inside the "  if
(use_constant_pool_for_mov(dst, src, canOptimize)) {                            
                                                                                
        " branch in mov().

Making immediate() fail for heap numbers rather than return the dummy (0) is
just a precaution.