summaryrefslogtreecommitdiff
path: root/src/arch/x86
diff options
context:
space:
mode:
Diffstat (limited to 'src/arch/x86')
-rw-r--r--src/arch/x86/emulenv.cc3
-rw-r--r--src/arch/x86/insts/microregop.cc2
-rw-r--r--src/arch/x86/isa/decoder/one_byte_opcodes.isa10
-rw-r--r--src/arch/x86/isa/insts/general_purpose/arithmetic/multiply_and_divide.py132
-rw-r--r--src/arch/x86/isa/insts/general_purpose/flags/load_and_store.py20
-rw-r--r--src/arch/x86/isa/insts/general_purpose/rotate_and_shift/rotate.py72
-rw-r--r--src/arch/x86/isa/microasm.isa4
-rw-r--r--src/arch/x86/isa/microops/regop.isa108
-rw-r--r--src/arch/x86/predecoder.cc3
-rw-r--r--src/arch/x86/types.hh9
-rw-r--r--src/arch/x86/utility.hh3
11 files changed, 189 insertions, 177 deletions
diff --git a/src/arch/x86/emulenv.cc b/src/arch/x86/emulenv.cc
index 8e4600a14..7e4a9fc55 100644
--- a/src/arch/x86/emulenv.cc
+++ b/src/arch/x86/emulenv.cc
@@ -118,8 +118,7 @@ void EmulEnv::doModRM(const ExtMachInst & machInst)
//Figure out what segment to use. This won't be entirely accurate since
//the presence of a displacement is supposed to make the instruction
//default to the data segment.
- if ((base != INTREG_RBP && base != INTREG_RSP) ||
- 0/*Has an immediate offset*/) {
+ if ((base != INTREG_RBP && base != INTREG_RSP) || machInst.dispSize) {
seg = SEGMENT_REG_DS;
//Handle any segment override that might have been in the instruction
int segFromInst = machInst.legacy.seg;
diff --git a/src/arch/x86/insts/microregop.cc b/src/arch/x86/insts/microregop.cc
index 2edd3ba87..5982dff7a 100644
--- a/src/arch/x86/insts/microregop.cc
+++ b/src/arch/x86/insts/microregop.cc
@@ -78,7 +78,7 @@ namespace X86ISA
if(subtract)
flags ^= (flagMask & (ECFBit | CFBit));
}
- if(flagMask & PFBit && findParity(dataSize*8, _dest))
+ if(flagMask & PFBit && !findParity(8, _dest))
flags |= PFBit;
if(flagMask & AFBit)
{
diff --git a/src/arch/x86/isa/decoder/one_byte_opcodes.isa b/src/arch/x86/isa/decoder/one_byte_opcodes.isa
index 84d18441d..f365ed4b0 100644
--- a/src/arch/x86/isa/decoder/one_byte_opcodes.isa
+++ b/src/arch/x86/isa/decoder/one_byte_opcodes.isa
@@ -330,14 +330,8 @@
//The 64 bit versions of both of these should be illegal only
//if CPUID says it isn't supported. For now, we'll just assume
//that it's supported.
- 0x6: decode MODE_SUBMODE {
- 0x0: SAHF_64();
- default: SAHF();
- }
- 0x7: decode MODE_SUBMODE {
- 0x0: LAHF_64();
- default: LAHF();
- }
+ 0x6: SAHF();
+ 0x7: LAHF();
}
0x14: decode OPCODE_OP_BOTTOM3 {
0x0: MOV(rAb, Ob);
diff --git a/src/arch/x86/isa/insts/general_purpose/arithmetic/multiply_and_divide.py b/src/arch/x86/isa/insts/general_purpose/arithmetic/multiply_and_divide.py
index 19d1c7789..dbc803350 100644
--- a/src/arch/x86/isa/insts/general_purpose/arithmetic/multiply_and_divide.py
+++ b/src/arch/x86/isa/insts/general_purpose/arithmetic/multiply_and_divide.py
@@ -61,29 +61,26 @@ microcode = '''
def macroop MUL_B_R
{
- mul1u rax, reg
+ mul1u rax, reg, flags=(OF,CF)
mulel rax
- # Really ah
- muleh rsi, flags=(OF,CF)
+ muleh ah
};
def macroop MUL_B_M
{
ld t1, seg, sib, disp
- mul1u rax, t1
+ mul1u rax, t1, flags=(OF,CF)
mulel rax
- # Really ah
- muleh rsi, flags=(OF,CF)
+ muleh ah
};
def macroop MUL_B_P
{
rdip t7
ld t1, seg, riprel, disp
- mul1u rax, t1
+ mul1u rax, t1, flags=(OF,CF)
mulel rax
- # Really ah
- muleh rsi, flags=(OF,CF)
+ muleh ah
};
#
@@ -92,26 +89,26 @@ def macroop MUL_B_P
def macroop MUL_R
{
- mul1u rax, reg
+ mul1u rax, reg, flags=(OF,CF)
mulel rax
- muleh rdx, flags=(OF,CF)
+ muleh rdx
};
def macroop MUL_M
{
ld t1, seg, sib, disp
- mul1u rax, t1
+ mul1u rax, t1, flags=(OF,CF)
mulel rax
- muleh rdx, flags=(OF,CF)
+ muleh rdx
};
def macroop MUL_P
{
rdip t7
ld t1, seg, riprel, disp
- mul1u rax, t1
+ mul1u rax, t1, flags=(OF,CF)
mulel rax
- muleh rdx, flags=(OF,CF)
+ muleh rdx
};
#
@@ -120,29 +117,26 @@ def macroop MUL_P
def macroop IMUL_B_R
{
- mul1s rax, reg
+ mul1s rax, reg, flags=(OF,CF)
mulel rax
- # Really ah
- muleh rsi, flags=(OF,CF)
+ muleh ah
};
def macroop IMUL_B_M
{
ld t1, seg, sib, disp
- mul1s rax, t1
+ mul1s rax, t1, flags=(OF,CF)
mulel rax
- # Really ah
- muleh rsi, flags=(OF,CF)
+ muleh ah
};
def macroop IMUL_B_P
{
rdip t7
ld t1, seg, riprel, disp
- mul1s rax, t1
+ mul1s rax, t1, flags=(OF,CF)
mulel rax
- # Really ah
- muleh rsi, flags=(OF,CF)
+ muleh ah
};
#
@@ -151,50 +145,50 @@ def macroop IMUL_B_P
def macroop IMUL_R
{
- mul1s rax, reg
+ mul1s rax, reg, flags=(OF,CF)
mulel rax
- muleh rdx, flags=(OF,CF)
+ muleh rdx
};
def macroop IMUL_M
{
ld t1, seg, sib, disp
- mul1s rax, t1
+ mul1s rax, t1, flags=(OF,CF)
mulel rax
- muleh rdx, flags=(OF,CF)
+ muleh rdx
};
def macroop IMUL_P
{
rdip t7
ld t1, seg, riprel, disp
- mul1s rax, t1
+ mul1s rax, t1, flags=(OF,CF)
mulel rax
- muleh rdx, flags=(OF,CF)
+ muleh rdx
};
def macroop IMUL_R_R
{
- mul1s reg, regm
+ mul1s reg, regm, flags=(OF,CF)
mulel reg
- muleh t0, flags=(CF,OF)
+ muleh t0
};
def macroop IMUL_R_M
{
ld t1, seg, sib, disp
- mul1s reg, t1
+ mul1s reg, t1, flags=(CF,OF)
mulel reg
- muleh t0, flags=(CF,OF)
+ muleh t0
};
def macroop IMUL_R_P
{
rdip t7
ld t1, seg, riprel, disp
- mul1s reg, t1
+ mul1s reg, t1, flags=(CF,OF)
mulel reg
- muleh t0, flags=(CF,OF)
+ muleh t0
};
#
@@ -204,18 +198,18 @@ def macroop IMUL_R_P
def macroop IMUL_R_R_I
{
limm t1, imm
- mul1s regm, t1
+ mul1s regm, t1, flags=(OF,CF)
mulel reg
- muleh t0, flags=(OF,CF)
+ muleh t0
};
def macroop IMUL_R_M_I
{
limm t1, imm
ld t2, seg, sib, disp
- mul1s t2, t1
+ mul1s t2, t1, flags=(OF,CF)
mulel reg
- muleh t0, flags=(OF,CF)
+ muleh t0
};
def macroop IMUL_R_P_I
@@ -223,9 +217,9 @@ def macroop IMUL_R_P_I
rdip t7
limm t1, imm
ld t2, seg, riprel
- mul1s t2, t1
+ mul1s t2, t1, flags=(OF,CF)
mulel reg
- muleh t0, flags=(OF,CF)
+ muleh t0
};
#
@@ -235,7 +229,7 @@ def macroop IMUL_R_P_I
def macroop DIV_B_R
{
# Do the initial part of the division
- div1 rsi, reg, dataSize=1
+ div1 ah, reg, dataSize=1
#These are split out so we can initialize the number of bits in the
#second register
@@ -250,7 +244,7 @@ divLoopTop:
#Unload the answer
divq rax, dataSize=1
- divr rsi, dataSize=1
+ divr ah, dataSize=1
};
def macroop DIV_B_M
@@ -258,7 +252,7 @@ def macroop DIV_B_M
ld t2, seg, sib, disp
# Do the initial part of the division
- div1 rsi, t2, dataSize=1
+ div1 ah, t2, dataSize=1
#These are split out so we can initialize the number of bits in the
#second register
@@ -273,7 +267,7 @@ divLoopTop:
#Unload the answer
divq rax, dataSize=1
- divr rsi, dataSize=1
+ divr ah, dataSize=1
};
def macroop DIV_B_P
@@ -282,7 +276,7 @@ def macroop DIV_B_P
ld t2, seg, riprel, disp
# Do the initial part of the division
- div1 rsi, t2, dataSize=1
+ div1 ah, t2, dataSize=1
#These are split out so we can initialize the number of bits in the
#second register
@@ -297,7 +291,7 @@ divLoopTop:
#Unload the answer
divq rax, dataSize=1
- divr rsi, dataSize=1
+ divr ah, dataSize=1
};
#
@@ -390,7 +384,7 @@ def macroop IDIV_B_R
# Negate dividend
sub t1, t0, rax, flags=(ECF,), dataSize=1
ruflag t4, 3
- sub t2, t0, rsi, dataSize=1
+ sub t2, t0, ah, dataSize=1
sub t2, t2, t4
#Find the sign of the divisor
@@ -404,11 +398,11 @@ def macroop IDIV_B_R
#Find the sign of the dividend
#FIXME!!! This depends on shifts setting the carry flag correctly.
- slli t0, rsi, 1, flags=(ECF,), dataSize=1
+ slli t0, ah, 1, flags=(ECF,), dataSize=1
# Put the dividend's absolute value into t1 and t2
mov t1, t1, rax, flags=(nCECF,), dataSize=1
- mov t2, t2, rsi, flags=(nCECF,), dataSize=1
+ mov t2, t2, ah, flags=(nCECF,), dataSize=1
# Do the initial part of the division
div1 t2, t3, dataSize=1
@@ -435,10 +429,10 @@ divLoopTop:
# Negate the remainder
sub t4, t0, t6, dataSize=1
- # If the dividend was negitive, put the negated remainder in rsi.
- mov rsi, rsi, t4, (CECF,), dataSize=1
- # Otherwise put the regular remainder in rsi.
- mov rsi, rsi, t6, (nCECF,), dataSize=1
+ # If the dividend was negitive, put the negated remainder in ah.
+ mov ah, ah, t4, (CECF,), dataSize=1
+ # Otherwise put the regular remainder in ah.
+ mov ah, ah, t6, (nCECF,), dataSize=1
# Negate the quotient.
sub t4, t0, t5, dataSize=1
@@ -461,7 +455,7 @@ def macroop IDIV_B_M
# Negate dividend
sub t1, t0, rax, flags=(ECF,), dataSize=1
ruflag t4, 3
- sub t2, t0, rsi, dataSize=1
+ sub t2, t0, ah, dataSize=1
sub t2, t2, t4
ld t3, seg, sib, disp
@@ -477,11 +471,11 @@ def macroop IDIV_B_M
#Find the sign of the dividend
#FIXME!!! This depends on shifts setting the carry flag correctly.
- slli t0, rsi, 1, flags=(ECF,), dataSize=1
+ slli t0, ah, 1, flags=(ECF,), dataSize=1
# Put the dividend's absolute value into t1 and t2
mov t1, t1, rax, flags=(nCECF,), dataSize=1
- mov t2, t2, rsi, flags=(nCECF,), dataSize=1
+ mov t2, t2, ah, flags=(nCECF,), dataSize=1
# Do the initial part of the division
div1 t2, t3, dataSize=1
@@ -508,10 +502,10 @@ divLoopTop:
# Negate the remainder
sub t4, t0, t6, dataSize=1
- # If the dividend was negitive, put the negated remainder in rsi.
- mov rsi, rsi, t4, (CECF,), dataSize=1
- # Otherwise put the regular remainder in rsi.
- mov rsi, rsi, t6, (nCECF,), dataSize=1
+ # If the dividend was negitive, put the negated remainder in ah.
+ mov ah, ah, t4, (CECF,), dataSize=1
+ # Otherwise put the regular remainder in ah.
+ mov ah, ah, t6, (nCECF,), dataSize=1
# Negate the quotient.
sub t4, t0, t5, dataSize=1
@@ -534,7 +528,7 @@ def macroop IDIV_B_P
# Negate dividend
sub t1, t0, rax, flags=(ECF,), dataSize=1
ruflag t4, 3
- sub t2, t0, rsi, dataSize=1
+ sub t2, t0, ah, dataSize=1
sub t2, t2, t4
rdip t7
@@ -551,11 +545,11 @@ def macroop IDIV_B_P
#Find the sign of the dividend
#FIXME!!! This depends on shifts setting the carry flag correctly.
- slli t0, rsi, 1, flags=(ECF,), dataSize=1
+ slli t0, ah, 1, flags=(ECF,), dataSize=1
# Put the dividend's absolute value into t1 and t2
mov t1, t1, rax, flags=(nCECF,), dataSize=1
- mov t2, t2, rsi, flags=(nCECF,), dataSize=1
+ mov t2, t2, ah, flags=(nCECF,), dataSize=1
# Do the initial part of the division
div1 t2, t3, dataSize=1
@@ -582,10 +576,10 @@ divLoopTop:
# Negate the remainder
sub t4, t0, t6, dataSize=1
- # If the dividend was negitive, put the negated remainder in rsi.
- mov rsi, rsi, t4, (CECF,), dataSize=1
- # Otherwise put the regular remainder in rsi.
- mov rsi, rsi, t6, (nCECF,), dataSize=1
+ # If the dividend was negitive, put the negated remainder in ah.
+ mov ah, ah, t4, (CECF,), dataSize=1
+ # Otherwise put the regular remainder in ah.
+ mov ah, ah, t6, (nCECF,), dataSize=1
# Negate the quotient.
sub t4, t0, t5, dataSize=1
diff --git a/src/arch/x86/isa/insts/general_purpose/flags/load_and_store.py b/src/arch/x86/isa/insts/general_purpose/flags/load_and_store.py
index 0915bf819..01908ca7b 100644
--- a/src/arch/x86/isa/insts/general_purpose/flags/load_and_store.py
+++ b/src/arch/x86/isa/insts/general_purpose/flags/load_and_store.py
@@ -55,26 +55,10 @@
microcode = '''
def macroop SAHF {
- # This will fold to ah since this never executes in 64 bit mode.
- ruflags rsp, dataSize=1
-};
-
-# This is allows the instruction to write to ah in 64 bit mode.
-def macroop SAHF_64 {
- ruflags t1
- slli t1, t1, 8
- mov t1, t1, rax, dataSize=1
- mov rax, rax, t1, dataSize=2
+ ruflags ah, dataSize=1
};
def macroop LAHF {
- # This will fold to ah since this never executes in 64 bit mode.
- wruflags rsp, t0, dataSize=1
-};
-
-# This is allows the instruction to read from ah in 64 bit mode.
-def macroop LAHF_64 {
- srli t1, rax, 8, dataSize=2
- wruflags t1, t0, dataSize=1
+ wruflags ah, t0, dataSize=1
};
'''
diff --git a/src/arch/x86/isa/insts/general_purpose/rotate_and_shift/rotate.py b/src/arch/x86/isa/insts/general_purpose/rotate_and_shift/rotate.py
index b5ae9560e..3be954768 100644
--- a/src/arch/x86/isa/insts/general_purpose/rotate_and_shift/rotate.py
+++ b/src/arch/x86/isa/insts/general_purpose/rotate_and_shift/rotate.py
@@ -56,13 +56,13 @@
microcode = '''
def macroop ROL_R_I
{
- roli reg, reg, imm
+ roli reg, reg, imm, flags=(OF,CF)
};
def macroop ROL_M_I
{
ldst t1, seg, sib, disp
- roli t1, t1, imm
+ roli t1, t1, imm, flags=(OF,CF)
st t1, seg, sib, disp
};
@@ -70,19 +70,19 @@ def macroop ROL_P_I
{
rdip t7
ldst t1, seg, riprel, disp
- roli t1, t1, imm
+ roli t1, t1, imm, flags=(OF,CF)
st t1, seg, riprel, disp
};
def macroop ROL_1_R
{
- roli reg, reg, 1
+ roli reg, reg, 1, flags=(OF,CF)
};
def macroop ROL_1_M
{
ldst t1, seg, sib, disp
- roli t1, t1, 1
+ roli t1, t1, 1, flags=(OF,CF)
st t1, seg, sib, disp
};
@@ -90,19 +90,19 @@ def macroop ROL_1_P
{
rdip t7
ldst t1, seg, riprel, disp
- roli t1, t1, 1
+ roli t1, t1, 1, flags=(OF,CF)
st t1, seg, riprel, disp
};
def macroop ROL_R_R
{
- rol reg, reg, regm
+ rol reg, reg, regm, flags=(OF,CF)
};
def macroop ROL_M_R
{
ldst t1, seg, sib, disp
- rol t1, t1, reg
+ rol t1, t1, reg, flags=(OF,CF)
st t1, seg, sib, disp
};
@@ -110,19 +110,19 @@ def macroop ROL_P_R
{
rdip t7
ldst t1, seg, riprel, disp
- rol t1, t1, reg
+ rol t1, t1, reg, flags=(OF,CF)
st t1, seg, riprel, disp
};
def macroop ROR_R_I
{
- rori reg, reg, imm
+ rori reg, reg, imm, flags=(OF,CF)
};
def macroop ROR_M_I
{
ldst t1, seg, sib, disp
- rori t1, t1, imm
+ rori t1, t1, imm, flags=(OF,CF)
st t1, seg, sib, disp
};
@@ -130,19 +130,19 @@ def macroop ROR_P_I
{
rdip t7
ldst t1, seg, riprel, disp
- rori t1, t1, imm
+ rori t1, t1, imm, flags=(OF,CF)
st t1, seg, riprel, disp
};
def macroop ROR_1_R
{
- rori reg, reg, 1
+ rori reg, reg, 1, flags=(OF,CF)
};
def macroop ROR_1_M
{
ldst t1, seg, sib, disp
- rori t1, t1, 1
+ rori t1, t1, 1, flags=(OF,CF)
st t1, seg, sib, disp
};
@@ -150,19 +150,19 @@ def macroop ROR_1_P
{
rdip t7
ldst t1, seg, riprel, disp
- rori t1, t1, 1
+ rori t1, t1, 1, flags=(OF,CF)
st t1, seg, riprel, disp
};
def macroop ROR_R_R
{
- ror reg, reg, regm
+ ror reg, reg, regm, flags=(OF,CF)
};
def macroop ROR_M_R
{
ldst t1, seg, sib, disp
- ror t1, t1, reg
+ ror t1, t1, reg, flags=(OF,CF)
st t1, seg, sib, disp
};
@@ -170,19 +170,19 @@ def macroop ROR_P_R
{
rdip t7
ldst t1, seg, riprel, disp
- ror t1, t1, reg
+ ror t1, t1, reg, flags=(OF,CF)
st t1, seg, riprel, disp
};
def macroop RCL_R_I
{
- rcli reg, reg, imm
+ rcli reg, reg, imm, flags=(OF,CF)
};
def macroop RCL_M_I
{
ldst t1, seg, sib, disp
- rcli t1, t1, imm
+ rcli t1, t1, imm, flags=(OF,CF)
st t1, seg, sib, disp
};
@@ -190,19 +190,19 @@ def macroop RCL_P_I
{
rdip t7
ldst t1, seg, riprel, disp
- rcli t1, t1, imm
+ rcli t1, t1, imm, flags=(OF,CF)
st t1, seg, riprel, disp
};
def macroop RCL_1_R
{
- rcli reg, reg, 1
+ rcli reg, reg, 1, flags=(OF,CF)
};
def macroop RCL_1_M
{
ldst t1, seg, sib, disp
- rcli t1, t1, 1
+ rcli t1, t1, 1, flags=(OF,CF)
st t1, seg, sib, disp
};
@@ -210,19 +210,19 @@ def macroop RCL_1_P
{
rdip t7
ldst t1, seg, riprel, disp
- rcli t1, t1, 1
+ rcli t1, t1, 1, flags=(OF,CF)
st t1, seg, riprel, disp
};
def macroop RCL_R_R
{
- rcl reg, reg, regm
+ rcl reg, reg, regm, flags=(OF,CF)
};
def macroop RCL_M_R
{
ldst t1, seg, sib, disp
- rcl t1, t1, reg
+ rcl t1, t1, reg, flags=(OF,CF)
st t1, seg, sib, disp
};
@@ -230,19 +230,19 @@ def macroop RCL_P_R
{
rdip t7
ldst t1, seg, riprel, disp
- rcl t1, t1, reg
+ rcl t1, t1, reg, flags=(OF,CF)
st t1, seg, riprel, disp
};
def macroop RCR_R_I
{
- rcri reg, reg, imm
+ rcri reg, reg, imm, flags=(OF,CF)
};
def macroop RCR_M_I
{
ldst t1, seg, sib, disp
- rcri t1, t1, imm
+ rcri t1, t1, imm, flags=(OF,CF)
st t1, seg, sib, disp
};
@@ -250,19 +250,19 @@ def macroop RCR_P_I
{
rdip t7
ldst t1, seg, riprel, disp
- rcri t1, t1, imm
+ rcri t1, t1, imm, flags=(OF,CF)
st t1, seg, riprel, disp
};
def macroop RCR_1_R
{
- rcri reg, reg, 1
+ rcri reg, reg, 1, flags=(OF,CF)
};
def macroop RCR_1_M
{
ldst t1, seg, sib, disp
- rcri t1, t1, 1
+ rcri t1, t1, 1, flags=(OF,CF)
st t1, seg, sib, disp
};
@@ -270,19 +270,19 @@ def macroop RCR_1_P
{
rdip t7
ldst t1, seg, riprel, disp
- rcri t1, t1, 1
+ rcri t1, t1, 1, flags=(OF,CF)
st t1, seg, riprel, disp
};
def macroop RCR_R_R
{
- rcr reg, reg, regm
+ rcr reg, reg, regm, flags=(OF,CF)
};
def macroop RCR_M_R
{
ldst t1, seg, sib, disp
- rcr t1, t1, reg
+ rcr t1, t1, reg, flags=(OF,CF)
st t1, seg, sib, disp
};
@@ -290,7 +290,7 @@ def macroop RCR_P_R
{
rdip t7
ldst t1, seg, riprel, disp
- rcr t1, t1, reg
+ rcr t1, t1, reg, flags=(OF,CF)
st t1, seg, riprel, disp
};
'''
diff --git a/src/arch/x86/isa/microasm.isa b/src/arch/x86/isa/microasm.isa
index 0cc72bf7b..c6f5e9cdd 100644
--- a/src/arch/x86/isa/microasm.isa
+++ b/src/arch/x86/isa/microasm.isa
@@ -151,6 +151,10 @@ let {{
assembler.symbols["r%s" % reg] = \
regIdx("INTREG_R%s" % reg.upper())
+ for reg in ('ah', 'bh', 'ch', 'dh'):
+ assembler.symbols[reg] = \
+ regIdx("INTREG_FOLDED(INTREG_%s, IntFoldBit)" % reg.upper())
+
for reg in range(16):
assembler.symbols["cr%d" % reg] = regIdx("MISCREG_CR%d" % reg)
diff --git a/src/arch/x86/isa/microops/regop.isa b/src/arch/x86/isa/microops/regop.isa
index 698216139..dc6819886 100644
--- a/src/arch/x86/isa/microops/regop.isa
+++ b/src/arch/x86/isa/microops/regop.isa
@@ -525,18 +525,25 @@ let {{
uint64_t hiResult;
uint64_t psrc1_h = psrc1 / shifter;
uint64_t psrc1_l = psrc1 & mask(halfSize);
- uint64_t psrc2_h = op2 / shifter;
+ uint64_t psrc2_h = (op2 / shifter) & mask(halfSize);
uint64_t psrc2_l = op2 & mask(halfSize);
hiResult = ((psrc1_l * psrc2_h + psrc1_h * psrc2_l +
((psrc1_l * psrc2_l) / shifter)) /shifter) +
psrc1_h * psrc2_h;
- if (spsrc1 < 0)
+ if (bits(psrc1, dataSize * 8 - 1))
hiResult -= op2;
- int64_t bigSop2 = sop2;
- if (bigSop2 < 0)
+ if (bits(op2, dataSize * 8 - 1))
hiResult -= psrc1;
ProdHi = hiResult;
'''
+ flag_code = '''
+ if ((-ProdHi & mask(dataSize * 8)) !=
+ bits(ProdLow, dataSize * 8 - 1)) {
+ ccFlagBits = ccFlagBits | (ext & (CFBit | OFBit | ECFBit));
+ } else {
+ ccFlagBits = ccFlagBits & ~(ext & (CFBit | OFBit | ECFBit));
+ }
+ '''
class Mul1u(WrRegOp):
code = '''
@@ -545,12 +552,19 @@ let {{
uint64_t shifter = (1ULL << halfSize);
uint64_t psrc1_h = psrc1 / shifter;
uint64_t psrc1_l = psrc1 & mask(halfSize);
- uint64_t psrc2_h = op2 / shifter;
+ uint64_t psrc2_h = (op2 / shifter) & mask(halfSize);
uint64_t psrc2_l = op2 & mask(halfSize);
ProdHi = ((psrc1_l * psrc2_h + psrc1_h * psrc2_l +
((psrc1_l * psrc2_l) / shifter)) / shifter) +
psrc1_h * psrc2_h;
'''
+ flag_code = '''
+ if (ProdHi) {
+ ccFlagBits = ccFlagBits | (ext & (CFBit | OFBit | ECFBit));
+ } else {
+ ccFlagBits = ccFlagBits & ~(ext & (CFBit | OFBit | ECFBit));
+ }
+ '''
class Mulel(RdRegOp):
code = 'DestReg = merge(SrcReg1, ProdLow, dataSize);'
@@ -562,12 +576,6 @@ let {{
super(RdRegOp, self).__init__(dest, src1, \
"InstRegIndex(NUM_INTREGS)", flags, dataSize)
code = 'DestReg = merge(SrcReg1, ProdHi, dataSize);'
- flag_code = '''
- if (ProdHi)
- ccFlagBits = ccFlagBits | (ext & (CFBit | OFBit | ECFBit));
- else
- ccFlagBits = ccFlagBits & ~(ext & (CFBit | OFBit | ECFBit));
- '''
# One or two bit divide
class Div1(WrRegOp):
@@ -631,7 +639,7 @@ let {{
class Mov(CondRegOp):
code = 'DestReg = merge(SrcReg1, op2, dataSize)'
- else_code = 'DestReg=DestReg;'
+ else_code = 'DestReg = merge(DestReg, DestReg, dataSize);'
# Shift instructions
@@ -648,8 +656,10 @@ let {{
ccFlagBits = ccFlagBits & ~(ext & (CFBit | ECFBit | OFBit));
int CFBits = 0;
//Figure out if we -would- set the CF bits if requested.
- if (bits(SrcReg1, dataSize * 8 - shiftAmt))
+ if (shiftAmt <= dataSize * 8 &&
+ bits(SrcReg1, dataSize * 8 - shiftAmt)) {
CFBits = 1;
+ }
//If some combination of the CF bits need to be set, set them.
if ((ext & (CFBit | ECFBit)) && CFBits)
ccFlagBits = ccFlagBits | (ext & (CFBit | ECFBit));
@@ -678,8 +688,11 @@ let {{
//worry about setting them.
ccFlagBits = ccFlagBits & ~(ext & (CFBit | ECFBit | OFBit));
//If some combination of the CF bits need to be set, set them.
- if ((ext & (CFBit | ECFBit)) && bits(SrcReg1, shiftAmt - 1))
+ if ((ext & (CFBit | ECFBit)) &&
+ shiftAmt <= dataSize * 8 &&
+ bits(SrcReg1, shiftAmt - 1)) {
ccFlagBits = ccFlagBits | (ext & (CFBit | ECFBit));
+ }
//Figure out what the OF bit should be.
if ((ext & OFBit) && bits(SrcReg1, dataSize * 8 - 1))
ccFlagBits = ccFlagBits | OFBit;
@@ -695,7 +708,7 @@ let {{
// Because what happens to the bits shift -in- on a right shift
// is not defined in the C/C++ standard, we have to sign extend
// them manually to be sure.
- uint64_t arithMask =
+ uint64_t arithMask = (shiftAmt == 0) ? 0 :
-bits(psrc1, dataSize * 8 - 1) << (dataSize * 8 - shiftAmt);
DestReg = merge(DestReg, (psrc1 >> shiftAmt) | arithMask, dataSize);
'''
@@ -706,8 +719,12 @@ let {{
//worry about setting them.
ccFlagBits = ccFlagBits & ~(ext & (CFBit | ECFBit | OFBit));
//If some combination of the CF bits need to be set, set them.
- if ((ext & (CFBit | ECFBit)) && bits(SrcReg1, shiftAmt - 1))
+ uint8_t effectiveShift =
+ (shiftAmt <= dataSize * 8) ? shiftAmt : (dataSize * 8);
+ if ((ext & (CFBit | ECFBit)) &&
+ bits(SrcReg1, effectiveShift - 1)) {
ccFlagBits = ccFlagBits | (ext & (CFBit | ECFBit));
+ }
//Use the regular mechanisms to calculate the other flags.
ccFlagBits = genFlags(ccFlagBits, ext & ~(CFBit | ECFBit | OFBit),
DestReg, psrc1, op2);
@@ -718,14 +735,15 @@ let {{
code = '''
uint8_t shiftAmt =
(op2 & ((dataSize == 8) ? mask(6) : mask(5)));
- if(shiftAmt)
+ uint8_t realShiftAmt = shiftAmt % (dataSize * 8);
+ if(realShiftAmt)
{
- uint64_t top = psrc1 << (dataSize * 8 - shiftAmt);
- uint64_t bottom = bits(psrc1, dataSize * 8, shiftAmt);
+ uint64_t top = psrc1 << (dataSize * 8 - realShiftAmt);
+ uint64_t bottom = bits(psrc1, dataSize * 8, realShiftAmt);
DestReg = merge(DestReg, top | bottom, dataSize);
}
else
- DestReg = DestReg;
+ DestReg = merge(DestReg, DestReg, dataSize);
'''
flag_code = '''
// If the shift amount is zero, no flags should be modified.
@@ -752,31 +770,37 @@ let {{
code = '''
uint8_t shiftAmt =
(op2 & ((dataSize == 8) ? mask(6) : mask(5)));
- if(shiftAmt)
+ uint8_t realShiftAmt = shiftAmt % (dataSize * 8 + 1);
+ if(realShiftAmt)
{
CCFlagBits flags = ccFlagBits;
- uint64_t top = flags.cf << (dataSize * 8 - shiftAmt);
- if(shiftAmt > 1)
- top |= psrc1 << (dataSize * 8 - shiftAmt - 1);
- uint64_t bottom = bits(psrc1, dataSize * 8, shiftAmt);
+ uint64_t top = flags.cf << (dataSize * 8 - realShiftAmt);
+ if (realShiftAmt > 1)
+ top |= psrc1 << (dataSize * 8 - realShiftAmt + 1);
+ uint64_t bottom = bits(psrc1, dataSize * 8 - 1, realShiftAmt);
DestReg = merge(DestReg, top | bottom, dataSize);
}
else
- DestReg = DestReg;
+ DestReg = merge(DestReg, DestReg, dataSize);
'''
flag_code = '''
// If the shift amount is zero, no flags should be modified.
if (shiftAmt) {
+ int origCFBit = (ccFlagBits & CFBit) ? 1 : 0;
//Zero out any flags we might modify. This way we only have to
//worry about setting them.
ccFlagBits = ccFlagBits & ~(ext & (CFBit | ECFBit | OFBit));
//Figure out what the OF bit should be.
- if ((ext & OFBit) && ((ccFlagBits & CFBit) ^
- bits(SrcReg1, dataSize * 8 - 1)))
+ if ((ext & OFBit) && (origCFBit ^
+ bits(SrcReg1, dataSize * 8 - 1))) {
ccFlagBits = ccFlagBits | OFBit;
+ }
//If some combination of the CF bits need to be set, set them.
- if ((ext & (CFBit | ECFBit)) && bits(SrcReg1, shiftAmt - 1))
+ if ((ext & (CFBit | ECFBit)) &&
+ (realShiftAmt == 0) ? origCFBit :
+ bits(SrcReg1, realShiftAmt - 1)) {
ccFlagBits = ccFlagBits | (ext & (CFBit | ECFBit));
+ }
//Use the regular mechanisms to calculate the other flags.
ccFlagBits = genFlags(ccFlagBits, ext & ~(CFBit | ECFBit | OFBit),
DestReg, psrc1, op2);
@@ -787,15 +811,16 @@ let {{
code = '''
uint8_t shiftAmt =
(op2 & ((dataSize == 8) ? mask(6) : mask(5)));
- if(shiftAmt)
+ uint8_t realShiftAmt = shiftAmt % (dataSize * 8);
+ if(realShiftAmt)
{
- uint64_t top = psrc1 << shiftAmt;
+ uint64_t top = psrc1 << realShiftAmt;
uint64_t bottom =
- bits(psrc1, dataSize * 8 - 1, dataSize * 8 - shiftAmt);
+ bits(psrc1, dataSize * 8 - 1, dataSize * 8 - realShiftAmt);
DestReg = merge(DestReg, top | bottom, dataSize);
}
else
- DestReg = DestReg;
+ DestReg = merge(DestReg, DestReg, dataSize);
'''
flag_code = '''
// If the shift amount is zero, no flags should be modified.
@@ -822,30 +847,33 @@ let {{
code = '''
uint8_t shiftAmt =
(op2 & ((dataSize == 8) ? mask(6) : mask(5)));
- if(shiftAmt)
+ uint8_t realShiftAmt = shiftAmt % (dataSize * 8 + 1);
+ if(realShiftAmt)
{
CCFlagBits flags = ccFlagBits;
- uint64_t top = psrc1 << shiftAmt;
- uint64_t bottom = flags.cf << (shiftAmt - 1);
+ uint64_t top = psrc1 << realShiftAmt;
+ uint64_t bottom = flags.cf << (realShiftAmt - 1);
if(shiftAmt > 1)
bottom |=
bits(psrc1, dataSize * 8 - 1,
- dataSize * 8 - shiftAmt + 1);
+ dataSize * 8 - realShiftAmt + 1);
DestReg = merge(DestReg, top | bottom, dataSize);
}
else
- DestReg = DestReg;
+ DestReg = merge(DestReg, DestReg, dataSize);
'''
flag_code = '''
// If the shift amount is zero, no flags should be modified.
if (shiftAmt) {
+ int origCFBit = (ccFlagBits & CFBit) ? 1 : 0;
//Zero out any flags we might modify. This way we only have to
//worry about setting them.
ccFlagBits = ccFlagBits & ~(ext & (CFBit | ECFBit | OFBit));
int msb = bits(DestReg, dataSize * 8 - 1);
- int CFBits = bits(SrcReg1, dataSize * 8 - shiftAmt);
+ int CFBits = bits(SrcReg1, dataSize * 8 - realShiftAmt);
//If some combination of the CF bits need to be set, set them.
- if ((ext & (CFBit | ECFBit)) && CFBits)
+ if ((ext & (CFBit | ECFBit)) &&
+ (realShiftAmt == 0) ? origCFBit : CFBits)
ccFlagBits = ccFlagBits | (ext & (CFBit | ECFBit));
//Figure out what the OF bit should be.
if ((ext & OFBit) && (msb ^ CFBits))
diff --git a/src/arch/x86/predecoder.cc b/src/arch/x86/predecoder.cc
index 24276f06c..f537f92af 100644
--- a/src/arch/x86/predecoder.cc
+++ b/src/arch/x86/predecoder.cc
@@ -77,6 +77,7 @@ namespace X86ISA
immediateCollected = 0;
emi.immediate = 0;
emi.displacement = 0;
+ emi.dispSize = 0;
emi.modRM = 0;
emi.sib = 0;
@@ -383,6 +384,8 @@ namespace X86ISA
emiIsReady = true;
nextState = ResetState;
}
+
+ emi.dispSize = displacementSize;
}
else
nextState = DisplacementState;
diff --git a/src/arch/x86/types.hh b/src/arch/x86/types.hh
index bdf3a814e..956ec3216 100644
--- a/src/arch/x86/types.hh
+++ b/src/arch/x86/types.hh
@@ -175,6 +175,8 @@ namespace X86ISA
uint8_t addrSize;
//The effective stack size.
uint8_t stackSize;
+ //The size of the displacement
+ uint8_t dispSize;
//Mode information
OperatingMode mode;
@@ -187,12 +189,13 @@ namespace X86ISA
"op = {\n\t\tnum = %d,\n\t\top = %#x,\n\t\t"
"prefixA = %#x,\n\t\tprefixB = %#x\n\t},\n\t"
"modRM = %#x,\n\tsib = %#x,\n\t"
- "immediate = %#x,\n\tdisplacement = %#x\n}\n",
+ "immediate = %#x,\n\tdisplacement = %#x\n\t"
+ "dispSize = %d}\n",
(uint8_t)emi.legacy, (uint8_t)emi.rex,
emi.opcode.num, (uint8_t)emi.opcode.op,
emi.opcode.prefixA, emi.opcode.prefixB,
(uint8_t)emi.modRM, (uint8_t)emi.sib,
- emi.immediate, emi.displacement);
+ emi.immediate, emi.displacement, emi.dispSize);
return os;
}
@@ -227,6 +230,8 @@ namespace X86ISA
return false;
if(emi1.stackSize != emi2.stackSize)
return false;
+ if(emi1.dispSize != emi2.dispSize)
+ return false;
return true;
}
diff --git a/src/arch/x86/utility.hh b/src/arch/x86/utility.hh
index d305e2599..4388dd416 100644
--- a/src/arch/x86/utility.hh
+++ b/src/arch/x86/utility.hh
@@ -82,7 +82,8 @@ namespace __hash_namespace {
((uint64_t)emi.opcode.op)) ^
emi.immediate ^ emi.displacement ^
emi.mode ^
- emi.opSize ^ emi.addrSize ^ emi.stackSize;
+ emi.opSize ^ emi.addrSize ^
+ emi.stackSize ^ emi.dispSize;
};
};
}
generated by cgit v1.2.3 (git 2.25.1) at 2024-12-28 10:53:58 +0000