x86-64: Inline small array copy size for int and long arrays

(1)
Inline small array copy size without using `rep movs`
for int array and long array
  - If AVX-512 is supported and copy size <= 128 bytes,
    the inlined copy sequence is used.
  - If AVX is supported, the threshold for inlined sequence
    is 64 bytes instead.
  - The enhancement for 32 bit and 64 bit arrays will be reused
    in the downstream OpenJ9 project for reference arrays.

(2)
Refactor char array enhancement implementation to support
both `rep movsw` and `rep movsd` if the copysize is greater
than the supported enhancement threshold.

(3)
Add the following options to disable the enhancement
  - `disableArrayCopyIntArrayInlineSmallSizeWithoutREPMOVS`:
    disable the enhancement for int array

  - `disableArrayCopyLongArrayInlineSmallSizeWithoutREPMOVS`:
    disable the enhancement for long array

  - `disableArrayCopyReferenceArrayInlineSmallSizeWithoutREPMOVS`:
    disable the enhancement for reference array, which will be
    implemented in downstream project OpenJ9.

Signed-off-by: Annabelle Huo <Annabelle.Huo@ibm.com>

compiler/control/OMROptions.cpp

@@ -253,7 +253,10 @@ TR::OptionTable OMR::Options::_jitOptions[] = {
    {"disableAOTWarmRunThroughputImprovement", "O\tdisable change iprofiler entry choosing heuristic to improve aot warm run throughput",                      SET_OPTION_BIT(TR_DisableAOTWarmRunThroughputImprovement), "F"},
    {"disableArrayCopyByteArrayInlineSmallSizeWithoutREPMOVS", "O\tdisable array copy optimizations enhancement for 8 bit primitive array", SET_OPTION_BIT(TR_Disable8BitPrimitiveArrayCopyInlineSmallSizeWithoutREPMOVS), "F"},
    {"disableArrayCopyCharArrayInlineSmallSizeWithoutREPMOVS", "O\tdisable array copy optimizations enhancement for 16 bit primitive array", SET_OPTION_BIT(TR_Disable16BitPrimitiveArrayCopyInlineSmallSizeWithoutREPMOVS), "F"},
+   {"disableArrayCopyIntArrayInlineSmallSizeWithoutREPMOVS", "O\tdisable array copy optimizations enhancement for 32 bit primitive array", SET_OPTION_BIT(TR_Disable32BitPrimitiveArrayCopyInlineSmallSizeWithoutREPMOVS), "F"},
+   {"disableArrayCopyLongArrayInlineSmallSizeWithoutREPMOVS", "O\tdisable array copy optimizations enhancement for 64 bit primitive array", SET_OPTION_BIT(TR_Disable64BitPrimitiveArrayCopyInlineSmallSizeWithoutREPMOVS), "F"},
    {"disableArrayCopyOpts",               "O\tdisable array copy optimizations",                SET_OPTION_BIT(TR_DisableArrayCopyOpts), "F"},
+   {"disableArrayCopyReferenceArrayInlineSmallSizeWithoutREPMOVS", "O\tdisable array copy optimizations enhancement for reference array", SET_OPTION_BIT(TR_DisableReferenceArrayCopyInlineSmallSizeWithoutREPMOVS), "F"},
    {"disableArraySetOpts",                "O\tdisable array set optimizations",                 SET_OPTION_BIT(TR_DisableArraySetOpts), "F"},
    {"disableArraySetStoreElimination",     "O\tdisable arrayset store elimination",                SET_OPTION_BIT(TR_DisableArraysetStoreElimination), "F"},
    {"disableArrayStoreCheckOpts",          "O\tdisable array store check optimizations",SET_OPTION_BIT(TR_DisableArrayStoreCheckOpts), "F"},

compiler/control/OMROptions.hpp

@@ -344,20 +344,20 @@ enum TR_CompilationOptions
    TR_DisablePeephole                     = 0x00200000 + 8,
    TR_NoOptServer                         = 0x00400000 + 8,
    TR_DisableDLTrecompilationPrevention   = 0x00800000 + 8,
-   // Available                           = 0x01000000 + 8,
+   TR_Disable32BitPrimitiveArrayCopyInlineSmallSizeWithoutREPMOVS = 0x01000000 + 8,
    TR_DisablePeekAOTResolutions           = 0x02000000 + 8,
    TR_UseFlattenedArrayElementRuntimeHelpers = 0x04000000 + 8,
    TR_UseFlattenedFieldRuntimeHelpers     = 0x08000000 + 8,
    TR_DisableLiveRangeSplitter            = 0x10000000 + 8,
    TR_DisableHalfSlotSpills               = 0x20000000 + 8,
    TR_DisableMHInlineWithoutPeeking       = 0x40000000 + 8,
-   // Available                           = 0x80000000 + 8,
+   TR_Disable64BitPrimitiveArrayCopyInlineSmallSizeWithoutREPMOVS = 0x80000000 + 8,
 
 
    // Option word 9
    //
    TR_SplitWarmAndColdBlocks              = 0x00000020 + 9,
-   // Available                           = 0x00000040 + 9,
+   TR_DisableReferenceArrayCopyInlineSmallSizeWithoutREPMOVS = 0x00000040 + 9,
    TR_DisableTLHPrefetch                  = 0x00000080 + 9,
    TR_DisableJProfilerThread              = 0x00000100 + 9,
    TR_DisableIProfilerThread              = 0x00000200 + 9,