Optimization of TEST

[nidud]

Added some optimizations for TEST and OR.

This applies to mem16, mem32, and mem64 as the first operand and immediate unsigned value as the second operand. Typically a bit-flag test:

.if ( ff.attrib & _A_SUBDIR )

In these cases a byte values will be used if possible.

test     byte ptr [rsi+16],0x00000010
test     word ptr [rsi+16],0x00000010
test    dword ptr [rsi+16],0x00000010
test    qword ptr [rsi+16],0x00000010
test    qword ptr [rsi+16],0x00001000
test    qword ptr [rsi+16],0x00100000
test    qword ptr [rsi+16],0x10000000

Generated code:

test    byte ptr [rsi+10H], 10H                 ; 0000 _ F6. 46, 10, 10
; Note: Length-changing prefix causes delay on Intel processors
test    word ptr [rsi+10H], 10H                 ; 0004 _ 66: F7. 46, 10, 0010
test    dword ptr [rsi+10H], 10H                ; 000A _ F7. 46, 10, 00000010
test    qword ptr [rsi+10H], 10H                ; 0011 _ 48: F7. 46, 10, 00000010
test    qword ptr [rsi+10H], 1000H              ; 0019 _ 48: F7. 46, 10, 00001000
test    qword ptr [rsi+10H], 100000H            ; 0021 _ 48: F7. 46, 10, 00100000
test    qword ptr [rsi+10H], 10000000H          ; 0029 _ 48: F7. 46, 10, 10000000

Optimized code:

test    byte ptr [rsi+10H], 10H                 ; 0000 _ F6. 46, 10, 10
test    byte ptr [rsi+10H], 10H                 ; 0004 _ F6. 46, 10, 10
test    byte ptr [rsi+10H], 10H                 ; 0008 _ F6. 46, 10, 10
test    byte ptr [rsi+10H], 10H                 ; 000C _ F6. 46, 10, 10
test    byte ptr [rsi+11H], 10H                 ; 0010 _ F6. 46, 11, 10
test    byte ptr [rsi+12H], 10H                 ; 0014 _ F6. 46, 12, 10
test    byte ptr [rsi+13H], 10H                 ; 0018 _ F6. 46, 13, 10

Everything else (2..4 bytes) is 32-bit.

test     word ptr [rsi+16],0x1234
test    dword ptr [rsi+16],0x123456
test    qword ptr [rsi+16],0x12345678
...
test    dword ptr [rsi+10H], 1234H              ; 0000 _ F7. 46, 10, 00001234
test    dword ptr [rsi+10H], 123456H            ; 0007 _ F7. 46, 10, 00123456
test    dword ptr [rsi+10H], 12345678H          ; 000E _ F7. 46, 10, 12345678