[PATCH v3 03/11] [x86] Add OpSize16 bit, for instructions which need 0x66 prefix in 16-bit mode
David Woodhouse
dwmw2 at infradead.org
Fri Dec 20 06:58:00 PST 2013
From: David Woodhouse <David.Woodhouse at intel.com>
The 0x66 prefix toggles between 16-bit and 32-bit addressing mode.
So in 32-bit mode it is used to switch to 16-bit addressing mode for the
following instruction, while in 16-bit mode it's the other way round — it's
used to switch to 32-bit mode instead.
Thus, emit the 0x66 prefix byte for OpSize only in 32-bit (and 64-bit) mode,
and introduce a new OpSize16 bit which is used in 16-bit mode instead.
This is just the basic infrastructure for that change; a subsequent patch
will add the new OpSize16 bit to the 32-bit instructions that need it.
---
lib/Target/X86/MCTargetDesc/X86BaseInfo.h | 8 +-
lib/Target/X86/MCTargetDesc/X86MCCodeEmitter.cpp | 3 +-
lib/Target/X86/X86InstrFormats.td | 59 ++++----
test/MC/X86/x86-16.s | 170 +++++++++++++++++++++++
4 files changed, 207 insertions(+), 33 deletions(-)
diff --git a/lib/Target/X86/MCTargetDesc/X86BaseInfo.h b/lib/Target/X86/MCTargetDesc/X86BaseInfo.h
index 1ef9814..dcb9a8a 100644
--- a/lib/Target/X86/MCTargetDesc/X86BaseInfo.h
+++ b/lib/Target/X86/MCTargetDesc/X86BaseInfo.h
@@ -299,13 +299,15 @@ namespace X86II {
// OpSize - Set if this instruction requires an operand size prefix (0x66),
// which most often indicates that the instruction operates on 16 bit data
- // instead of 32 bit data.
+ // instead of 32 bit data. OpSize16 in 16 bit mode indicates that the
+ // instruction operates on 32 bit data instead of 16 bit data.
OpSize = 1 << 6,
+ OpSize16 = 1 << 7,
// AsSize - Set if this instruction requires an operand size prefix (0x67),
// which most often indicates that the instruction address 16 bit address
// instead of 32 bit address (or 32 bit address in 64 bit mode).
- AdSize = 1 << 7,
+ AdSize = 1 << 8,
//===------------------------------------------------------------------===//
// Op0Mask - There are several prefix bytes that are used to form two byte
@@ -313,7 +315,7 @@ namespace X86II {
// used to obtain the setting of this field. If no bits in this field is
// set, there is no prefix byte for obtaining a multibyte opcode.
//
- Op0Shift = 8,
+ Op0Shift = 9,
Op0Mask = 0x1F << Op0Shift,
// TB - TwoByte - Set if this instruction has a two byte opcode, which
diff --git a/lib/Target/X86/MCTargetDesc/X86MCCodeEmitter.cpp b/lib/Target/X86/MCTargetDesc/X86MCCodeEmitter.cpp
index 876138e..06a4882 100644
--- a/lib/Target/X86/MCTargetDesc/X86MCCodeEmitter.cpp
+++ b/lib/Target/X86/MCTargetDesc/X86MCCodeEmitter.cpp
@@ -1178,8 +1178,7 @@ void X86MCCodeEmitter::EmitOpcodePrefix(uint64_t TSFlags, unsigned &CurByte,
EmitByte(0x67, CurByte, OS);
// Emit the operand size opcode prefix as needed.
- // FIXME for is16BitMode().
- if (TSFlags & X86II::OpSize)
+ if (TSFlags & (is16BitMode() ? X86II::OpSize16 : X86II::OpSize))
EmitByte(0x66, CurByte, OS);
bool Need0FPrefix = false;
diff --git a/lib/Target/X86/X86InstrFormats.td b/lib/Target/X86/X86InstrFormats.td
index 5022259..dc6273d 100644
--- a/lib/Target/X86/X86InstrFormats.td
+++ b/lib/Target/X86/X86InstrFormats.td
@@ -113,6 +113,7 @@ def CD8VT8 : CD8VForm<7>; // v := 8
// Prefix byte classes which are used to indicate to the ad-hoc machine code
// emitter that various prefix bytes are required.
class OpSize { bit hasOpSizePrefix = 1; }
+class OpSize16 { bit hasOpSize16Prefix = 1; }
class AdSize { bit hasAdSizePrefix = 1; }
class REX_W { bit hasREX_WPrefix = 1; }
class LOCK { bit hasLockPrefix = 1; }
@@ -187,8 +188,9 @@ class X86Inst<bits<8> opcod, Format f, ImmType i, dag outs, dag ins,
//
// Attributes specific to X86 instructions...
//
- bit hasOpSizePrefix = 0; // Does this inst have a 0x66 prefix?
- bit hasAdSizePrefix = 0; // Does this inst have a 0x67 prefix?
+ bit hasOpSizePrefix = 0; // Does this inst have a 0x66 prefix?
+ bit hasOpSize16Prefix = 0; // Does this inst have a 0x66 prefix in 16-bit mode?
+ bit hasAdSizePrefix = 0; // Does this inst have a 0x67 prefix?
bits<5> Prefix = 0; // Which prefix byte does this inst have?
bit hasREX_WPrefix = 0; // Does this inst require the REX.W prefix?
@@ -219,32 +221,33 @@ class X86Inst<bits<8> opcod, Format f, ImmType i, dag outs, dag ins,
// TSFlags layout should be kept in sync with X86InstrInfo.h.
let TSFlags{5-0} = FormBits;
let TSFlags{6} = hasOpSizePrefix;
- let TSFlags{7} = hasAdSizePrefix;
- let TSFlags{12-8} = Prefix;
- let TSFlags{13} = hasREX_WPrefix;
- let TSFlags{16-14} = ImmT.Value;
- let TSFlags{19-17} = FPForm.Value;
- let TSFlags{20} = hasLockPrefix;
- let TSFlags{22-21} = SegOvrBits;
- let TSFlags{24-23} = ExeDomain.Value;
- let TSFlags{32-25} = Opcode;
- let TSFlags{33} = hasVEXPrefix;
- let TSFlags{34} = hasVEX_WPrefix;
- let TSFlags{35} = hasVEX_4VPrefix;
- let TSFlags{36} = hasVEX_4VOp3Prefix;
- let TSFlags{37} = hasVEX_i8ImmReg;
- let TSFlags{38} = hasVEX_L;
- let TSFlags{39} = ignoresVEX_L;
- let TSFlags{40} = hasEVEXPrefix;
- let TSFlags{41} = hasEVEX_K;
- let TSFlags{42} = hasEVEX_Z;
- let TSFlags{43} = hasEVEX_L2;
- let TSFlags{44} = hasEVEX_B;
- let TSFlags{46-45} = EVEX_CD8E;
- let TSFlags{49-47} = EVEX_CD8V;
- let TSFlags{50} = has3DNow0F0FOpcode;
- let TSFlags{51} = hasMemOp4Prefix;
- let TSFlags{52} = hasXOP_Prefix;
+ let TSFlags{7} = hasOpSize16Prefix;
+ let TSFlags{8} = hasAdSizePrefix;
+ let TSFlags{13-9} = Prefix;
+ let TSFlags{14} = hasREX_WPrefix;
+ let TSFlags{17-15} = ImmT.Value;
+ let TSFlags{20-18} = FPForm.Value;
+ let TSFlags{21} = hasLockPrefix;
+ let TSFlags{23-22} = SegOvrBits;
+ let TSFlags{25-24} = ExeDomain.Value;
+ let TSFlags{33-26} = Opcode;
+ let TSFlags{34} = hasVEXPrefix;
+ let TSFlags{35} = hasVEX_WPrefix;
+ let TSFlags{36} = hasVEX_4VPrefix;
+ let TSFlags{37} = hasVEX_4VOp3Prefix;
+ let TSFlags{38} = hasVEX_i8ImmReg;
+ let TSFlags{39} = hasVEX_L;
+ let TSFlags{40} = ignoresVEX_L;
+ let TSFlags{41} = hasEVEXPrefix;
+ let TSFlags{42} = hasEVEX_K;
+ let TSFlags{43} = hasEVEX_Z;
+ let TSFlags{44} = hasEVEX_L2;
+ let TSFlags{45} = hasEVEX_B;
+ let TSFlags{47-46} = EVEX_CD8E;
+ let TSFlags{50-48} = EVEX_CD8V;
+ let TSFlags{51} = has3DNow0F0FOpcode;
+ let TSFlags{52} = hasMemOp4Prefix;
+ let TSFlags{53} = hasXOP_Prefix;
}
class PseudoI<dag oops, dag iops, list<dag> pattern>
diff --git a/test/MC/X86/x86-16.s b/test/MC/X86/x86-16.s
index 36f3b19..115132d 100644
--- a/test/MC/X86/x86-16.s
+++ b/test/MC/X86/x86-16.s
@@ -27,6 +27,9 @@
// CHECK: testb %bl, %cl # encoding: [0x84,0xcb]
testb %bl, %cl
+// CHECK: addw %ax, %ax # encoding: [0x01,0xc0]
+ addw %ax, %ax
+
into
// CHECK: into
// CHECK: encoding: [0xce]
@@ -40,6 +43,10 @@ int $255
// CHECK: int $255
// CHECK: encoding: [0xcd,0xff]
+// CHECK: cmovbw %bx, %bx
+cmovnae %bx,%bx
+
+
// CHECK: fmul %st(0)
// CHECK: encoding: [0xd8,0xc8]
fmul %st(0), %st
@@ -64,10 +71,119 @@ int $255
// CHECK: encoding: [0xd8,0xf0]
fdiv %st(0), %st
+// CHECK: movw %cs, %ax
+// CHECK: encoding: [0x8c,0xc8]
+ movw %cs, %ax
+
+// CHECK: movw %cs, (%eax)
+// CHECK: encoding: [0x67,0x8c,0x08]
+ movw %cs, (%eax)
+
+// CHECK: movw (%eax), %cs
+// CHECK: encoding: [0x67,0x8e,0x08]
+ movw (%eax), %cs
+
+// CHECK: movl %cr0, %eax
+// CHECK: encoding: [0x0f,0x20,0xc0]
+ movl %cr0,%eax
+
+// CHECK: movl %cr1, %eax
+// CHECK: encoding: [0x0f,0x20,0xc8]
+ movl %cr1,%eax
+
+// CHECK: movl %cr2, %eax
+// CHECK: encoding: [0x0f,0x20,0xd0]
+ movl %cr2,%eax
+
+// CHECK: movl %cr3, %eax
+// CHECK: encoding: [0x0f,0x20,0xd8]
+ movl %cr3,%eax
+
+// CHECK: movl %cr4, %eax
+// CHECK: encoding: [0x0f,0x20,0xe0]
+ movl %cr4,%eax
+
+// CHECK: movl %dr0, %eax
+// CHECK: encoding: [0x0f,0x21,0xc0]
+ movl %dr0,%eax
+
+// CHECK: movl %dr1, %eax
+// CHECK: encoding: [0x0f,0x21,0xc8]
+ movl %dr1,%eax
+
+// CHECK: movl %dr1, %eax
+// CHECK: encoding: [0x0f,0x21,0xc8]
+ movl %dr1,%eax
+
+// CHECK: movl %dr2, %eax
+// CHECK: encoding: [0x0f,0x21,0xd0]
+ movl %dr2,%eax
+
+// CHECK: movl %dr3, %eax
+// CHECK: encoding: [0x0f,0x21,0xd8]
+ movl %dr3,%eax
+
+// CHECK: movl %dr4, %eax
+// CHECK: encoding: [0x0f,0x21,0xe0]
+ movl %dr4,%eax
+
+// CHECK: movl %dr5, %eax
+// CHECK: encoding: [0x0f,0x21,0xe8]
+ movl %dr5,%eax
+
+// CHECK: movl %dr6, %eax
+// CHECK: encoding: [0x0f,0x21,0xf0]
+ movl %dr6,%eax
+
+// CHECK: movl %dr7, %eax
+// CHECK: encoding: [0x0f,0x21,0xf8]
+ movl %dr7,%eax
+
// CHECK: wait
// CHECK: encoding: [0x9b]
fwait
+sysret
+// CHECK: sysretl
+// CHECK: encoding: [0x0f,0x07]
+sysretl
+// CHECK: sysretl
+// CHECK: encoding: [0x0f,0x07]
+
+testl %ecx, -24(%ebp)
+// CHECK: testl -24(%ebp), %ecx
+testl -24(%ebp), %ecx
+// CHECK: testl -24(%ebp), %ecx
+
+
+pushw %cs
+// CHECK: pushw %cs
+// CHECK: encoding: [0x0e]
+pushw %ds
+// CHECK: pushw %ds
+// CHECK: encoding: [0x1e]
+pushw %ss
+// CHECK: pushw %ss
+// CHECK: encoding: [0x16]
+pushw %es
+// CHECK: pushw %es
+// CHECK: encoding: [0x06]
+pushw %fs
+// CHECK: pushw %fs
+// CHECK: encoding: [0x0f,0xa0]
+pushw %gs
+// CHECK: pushw %gs
+// CHECK: encoding: [0x0f,0xa8]
+
+pushfd
+// CHECK: pushfl
+popfd
+// CHECK: popfl
+pushfl
+// CHECK: pushfl
+popfl
+// CHECK: popfl
+
setc %bl
setnae %bl
@@ -103,9 +219,15 @@ ljmpl $0x7ace,$0x7ace
// CHECK: incb %al # encoding: [0xfe,0xc0]
incb %al
+// CHECK: incw %ax # encoding: [0x40]
+ incw %ax
+
// CHECK: decb %al # encoding: [0xfe,0xc8]
decb %al
+// CHECK: decw %ax # encoding: [0x48]
+ decw %ax
+
// CHECK: pshufw $14, %mm4, %mm0 # encoding: [0x0f,0x70,0xc4,0x0e]
pshufw $14, %mm4, %mm0
@@ -152,6 +274,10 @@ pshufw $90, %mm4, %mm0
// CHECK: encoding: [0x2f]
das
+// CHECK: bound 2(%eax), %bx
+// CHECK: encoding: [0x67,0x62,0x58,0x02]
+ bound 2(%eax),%bx
+
// CHECK: arpl %bx, %bx
// CHECK: encoding: [0x63,0xdb]
arpl %bx,%bx
@@ -219,11 +345,23 @@ pshufw $90, %mm4, %mm0
outsb %ds:(%si), %dx
outsb (%si), %dx
+// CHECK: outsw # encoding: [0x6f]
+// CHECK: outsw
+// CHECK: outsw
+ outsw
+ outsw %ds:(%si), %dx
+ outsw (%si), %dx
+
// CHECK: insb # encoding: [0x6c]
// CHECK: insb
insb
insb %dx, %es:(%di)
+// CHECK: insw # encoding: [0x6d]
+// CHECK: insw
+ insw
+ insw %dx, %es:(%di)
+
// CHECK: movsb # encoding: [0xa4]
// CHECK: movsb
// CHECK: movsb
@@ -231,6 +369,13 @@ pshufw $90, %mm4, %mm0
movsb %ds:(%si), %es:(%di)
movsb (%si), %es:(%di)
+// CHECK: movsw # encoding: [0xa5]
+// CHECK: movsw
+// CHECK: movsw
+ movsw
+ movsw %ds:(%si), %es:(%di)
+ movsw (%si), %es:(%di)
+
// CHECK: lodsb # encoding: [0xac]
// CHECK: lodsb
// CHECK: lodsb
@@ -242,6 +387,17 @@ pshufw $90, %mm4, %mm0
lods %ds:(%si), %al
lods (%si), %al
+// CHECK: lodsw # encoding: [0xad]
+// CHECK: lodsw
+// CHECK: lodsw
+// CHECK: lodsw
+// CHECK: lodsw
+ lodsw
+ lodsw %ds:(%si), %ax
+ lodsw (%si), %ax
+ lods %ds:(%si), %ax
+ lods (%si), %ax
+
// CHECK: stosb # encoding: [0xaa]
// CHECK: stosb
// CHECK: stosb
@@ -249,6 +405,17 @@ pshufw $90, %mm4, %mm0
stosb %al, %es:(%di)
stos %al, %es:(%di)
+// CHECK: stosw # encoding: [0xab]
+// CHECK: stosw
+// CHECK: stosw
+ stosw
+ stosw %ax, %es:(%di)
+ stos %ax, %es:(%di)
+
+// CHECK: strw
+// CHECK: encoding: [0x0f,0x00,0xc8]
+ str %ax
+
// CHECK: fsubp
// CHECK: encoding: [0xde,0xe1]
fsubp %st,%st(1)
@@ -257,3 +424,6 @@ fsubp %st,%st(1)
// CHECK: encoding: [0xde,0xe2]
fsubp %st, %st(2)
+// CHECK: xchgw %ax, %ax
+// CHECK: encoding: [0x90]
+xchgw %ax, %ax
--
1.8.3.1
More information about the llvm-commits
mailing list