aniani/nasm
0
0
Fork 1
mirror of https://github.com/netwide-assembler/nasm.git synced 2025-09-22 10:43:39 -04:00
Code Activity

Revert "Improve process_ea and introduce -OL"

Browse Source 
This reverts commit ac732cb6a5.

Resolved Conflicts:

	doc/nasmdoc.src

Signed-off-by: H. Peter Anvin <hpa@zytor.com>
This commit is contained in:
H. Peter Anvin
2010-07-25 12:43:30 -07:00
parent b453c866ab
commit ab5bd05d82
7 changed files with 72 additions and 143 deletions
Download Patch File Download Diff File Expand all files Collapse all files

99
assemble.c
View File

@@ -751,59 +751,36 @@ int64_t insn_size(int32_t segment, int64_t offset, int bits, uint32_t cp,
} }
} }
static bool possible_sbyte(operand *o, int min_optimizing) static bool possible_sbyte(operand *o)
{ {
return o->wrt == NO_SEG && o->segment == NO_SEG && return o->wrt == NO_SEG && o->segment == NO_SEG &&
!(o->opflags & OPFLAG_UNKNOWN) && !(o->opflags & OPFLAG_UNKNOWN) &&
optimizing >= min_optimizing && !(o->type & STRICT); optimizing >= 0 && !(o->type & STRICT);
} }
/* check that opn[op] is a signed byte of size 16 or 32 */ /* check that opn[op] is a signed byte of size 16 or 32 */
static bool is_sbyte16(operand *o, int min_optimizing) static bool is_sbyte16(operand *o)
{ {
int16_t v; int16_t v;
if (!possible_sbyte(o, min_optimizing)) if (!possible_sbyte(o))
return false; return false;
v = o->offset; v = o->offset;
return v >= -128 && v <= 127; return v >= -128 && v <= 127;
} }
static bool is_sbyte32(operand *o, int min_optimizing) static bool is_sbyte32(operand *o)
{ {
int32_t v; int32_t v;
if (!possible_sbyte(o, min_optimizing)) if (!possible_sbyte(o))
return false; return false;
v = o->offset; v = o->offset;
return v >= -128 && v <= 127; return v >= -128 && v <= 127;
} }
/* Check if o is zero of size 16 or 32 */
static bool is_zero16(operand *o, int min_optimizing)
{
int16_t v;
if (!possible_sbyte(o, min_optimizing))
return false;
v = o->offset;
return v == 0;
}
static bool is_zero32(operand *o, int min_optimizing)
{
int32_t v;
if (!possible_sbyte(o, min_optimizing))
return false;
v = o->offset;
return v == 0;
}
/* Common construct */ /* Common construct */
#define case4(x) case (x): case (x)+1: case (x)+2: case (x)+3 #define case4(x) case (x): case (x)+1: case (x)+2: case (x)+3
@@ -905,7 +882,7 @@ static int64_t calcsize(int32_t segment, int64_t offset, int bits,
break; break;
case4(0140): case4(0140):
length += is_sbyte16(opx, 0) ? 1 : 2; length += is_sbyte16(opx) ? 1 : 2;
break; break;
case4(0144): case4(0144):
@@ -914,7 +891,7 @@ static int64_t calcsize(int32_t segment, int64_t offset, int bits,
break; break;
case4(0150): case4(0150):
length += is_sbyte32(opx, 0) ? 1 : 4; length += is_sbyte32(opx) ? 1 : 4;
break; break;
case4(0154): case4(0154):
@@ -945,7 +922,7 @@ static int64_t calcsize(int32_t segment, int64_t offset, int bits,
break; break;
case4(0250): case4(0250):
length += is_sbyte32(opx, 0) ? 1 : 4; length += is_sbyte32(opx) ? 1 : 4;
break; break;
case4(0254): case4(0254):
@@ -1441,7 +1418,7 @@ static void gencode(int32_t segment, int64_t offset, int bits,
case4(0140): case4(0140):
data = opx->offset; data = opx->offset;
warn_overflow_opd(opx, 2); warn_overflow_opd(opx, 2);
if (is_sbyte16(opx, 0)) { if (is_sbyte16(opx)) {
bytes[0] = data; bytes[0] = data;
out(offset, segment, bytes, OUT_RAWDATA, 1, NO_SEG, out(offset, segment, bytes, OUT_RAWDATA, 1, NO_SEG,
NO_SEG); NO_SEG);
@@ -1456,7 +1433,7 @@ static void gencode(int32_t segment, int64_t offset, int bits,
case4(0144): case4(0144):
EMIT_REX(); EMIT_REX();
bytes[0] = *codes++; bytes[0] = *codes++;
if (is_sbyte16(opx, 0)) if (is_sbyte16(opx))
bytes[0] |= 2; /* s-bit */ bytes[0] |= 2; /* s-bit */
out(offset, segment, bytes, OUT_RAWDATA, 1, NO_SEG, NO_SEG); out(offset, segment, bytes, OUT_RAWDATA, 1, NO_SEG, NO_SEG);
offset++; offset++;
@@ -1465,7 +1442,7 @@ static void gencode(int32_t segment, int64_t offset, int bits,
case4(0150): case4(0150):
data = opx->offset; data = opx->offset;
warn_overflow_opd(opx, 4); warn_overflow_opd(opx, 4);
if (is_sbyte32(opx, 0)) { if (is_sbyte32(opx)) {
bytes[0] = data; bytes[0] = data;
out(offset, segment, bytes, OUT_RAWDATA, 1, NO_SEG, out(offset, segment, bytes, OUT_RAWDATA, 1, NO_SEG,
NO_SEG); NO_SEG);
@@ -1480,7 +1457,7 @@ static void gencode(int32_t segment, int64_t offset, int bits,
case4(0154): case4(0154):
EMIT_REX(); EMIT_REX();
bytes[0] = *codes++; bytes[0] = *codes++;
if (is_sbyte32(opx, 0)) if (is_sbyte32(opx))
bytes[0] |= 2; /* s-bit */ bytes[0] |= 2; /* s-bit */
out(offset, segment, bytes, OUT_RAWDATA, 1, NO_SEG, NO_SEG); out(offset, segment, bytes, OUT_RAWDATA, 1, NO_SEG, NO_SEG);
offset++; offset++;
@@ -1544,7 +1521,7 @@ static void gencode(int32_t segment, int64_t offset, int bits,
errfunc(ERR_WARNING | ERR_PASS2 | ERR_WARN_NOV, errfunc(ERR_WARNING | ERR_PASS2 | ERR_WARN_NOV,
"signed dword immediate exceeds bounds"); "signed dword immediate exceeds bounds");
} }
if (is_sbyte32(opx, 0)) { if (is_sbyte32(opx)) {
bytes[0] = data; bytes[0] = data;
out(offset, segment, bytes, OUT_RAWDATA, 1, NO_SEG, out(offset, segment, bytes, OUT_RAWDATA, 1, NO_SEG,
NO_SEG); NO_SEG);
@@ -2226,9 +2203,7 @@ static enum match_result matches(const struct itemplate *itemp,
static ea *process_ea(operand * input, ea * output, int bits, static ea *process_ea(operand * input, ea * output, int bits,
int addrbits, int rfield, opflags_t rflags) int addrbits, int rfield, opflags_t rflags)
{ {
bool byte_offs = !!(input->eaflags & EAF_BYTEOFFS); bool forw_ref = !!(input->opflags & OPFLAG_UNKNOWN);
bool word_offs = !!(input->eaflags & EAF_WORDOFFS);
bool no_offs = !!(input->eaflags & EAF_NO_OFFS);
output->rip = false; output->rip = false;
@@ -2288,6 +2263,7 @@ static ea *process_ea(operand * input, ea * output, int bits,
} }
} else { /* it's an indirection */ } else { /* it's an indirection */
int i = input->indexreg, b = input->basereg, s = input->scale; int i = input->indexreg, b = input->basereg, s = input->scale;
int32_t seg = input->segment;
int hb = input->hintbase, ht = input->hinttype; int hb = input->hintbase, ht = input->hinttype;
int t, it, bt; /* register numbers */ int t, it, bt; /* register numbers */
opflags_t x, ix, bx; /* register flags */ opflags_t x, ix, bx; /* register flags */
@@ -2315,7 +2291,7 @@ static ea *process_ea(operand * input, ea * output, int bits,
if ((ix|bx) & (BITS32|BITS64)) { if ((ix|bx) & (BITS32|BITS64)) {
/* it must be a 32/64-bit memory reference. Firstly we have /* it must be a 32/64-bit memory reference. Firstly we have
* to check that all registers involved are type E/Rxx. */ * to check that all registers involved are type E/Rxx. */
int32_t sok = BITS32|BITS64; int32_t sok = BITS32|BITS64, o = input->offset;
if (it != -1) { if (it != -1) {
if (!(REG64 & ~ix) || !(REG32 & ~ix)) if (!(REG64 & ~ix) || !(REG32 & ~ix))
@@ -2385,13 +2361,15 @@ static ea *process_ea(operand * input, ea * output, int bits,
mod = 0; mod = 0;
} else { } else {
rm = (bt & 7); rm = (bt & 7);
if (rm != REG_NUM_EBP && if (rm != REG_NUM_EBP && o == 0 &&
(no_offs || is_zero32(input, -1)) && seg == NO_SEG && !forw_ref &&
!(byte_offs || word_offs)) !(input->eaflags &
(EAF_BYTEOFFS | EAF_WORDOFFS)))
mod = 0; mod = 0;
else if (byte_offs || else if (input->eaflags & EAF_BYTEOFFS ||
(! word_offs && is_sbyte32(input, -1)) || (o >= -128 && o <= 127 && seg == NO_SEG
(rm == REG_NUM_EBP && no_offs)) && !forw_ref
&& !(input->eaflags & EAF_WORDOFFS)))
mod = 1; mod = 1;
else else
mod = 2; mod = 2;
@@ -2431,13 +2409,15 @@ static ea *process_ea(operand * input, ea * output, int bits,
mod = 0; mod = 0;
} else { } else {
base = (bt & 7); base = (bt & 7);
if (base != REG_NUM_EBP && if (base != REG_NUM_EBP && o == 0 &&
(no_offs || is_zero32(input, -1)) && seg == NO_SEG && !forw_ref &&
!(byte_offs || word_offs)) !(input->eaflags &
(EAF_BYTEOFFS | EAF_WORDOFFS)))
mod = 0; mod = 0;
else if (byte_offs || else if (input->eaflags & EAF_BYTEOFFS ||
(! word_offs && is_sbyte32(input, -1)) || (o >= -128 && o <= 127 && seg == NO_SEG
(base == REG_NUM_EBP && no_offs)) && !forw_ref
&& !(input->eaflags & EAF_WORDOFFS)))
mod = 1; mod = 1;
else else
mod = 2; mod = 2;
@@ -2450,6 +2430,7 @@ static ea *process_ea(operand * input, ea * output, int bits,
} }
} else { /* it's 16-bit */ } else { /* it's 16-bit */
int mod, rm; int mod, rm;
int16_t o = input->offset;
/* check for 64-bit long mode */ /* check for 64-bit long mode */
if (addrbits == 64) if (addrbits == 64)
@@ -2519,13 +2500,13 @@ static ea *process_ea(operand * input, ea * output, int bits,
if (rm == -1) /* can't happen, in theory */ if (rm == -1) /* can't happen, in theory */
return NULL; /* so panic if it does */ return NULL; /* so panic if it does */
if (rm != 6 && if (o == 0 && seg == NO_SEG && !forw_ref && rm != 6 &&
(no_offs || is_zero16(input, -1)) && !(input->eaflags & (EAF_BYTEOFFS | EAF_WORDOFFS)))
!(byte_offs || word_offs))
mod = 0; mod = 0;
else if (byte_offs || else if (input->eaflags & EAF_BYTEOFFS ||
(! word_offs && is_sbyte16(input, -1)) || (o >= -128 && o <= 127 && seg == NO_SEG
(rm == 6 && no_offs)) && !forw_ref
&& !(input->eaflags & EAF_WORDOFFS)))
mod = 1; mod = 1;
else else
mod = 2; mod = 2;

7
doc/changes.src
View File

@@ -28,7 +28,7 @@ since 2007.
\b Macros parameters range expansion introduced. See \k{mlmacrange}. \b Macros parameters range expansion introduced. See \k{mlmacrange}.
\b Backward compatibility on expanding of local single line macros restored. \b Backward compatibility on expanging of local sigle macros restored.
\b 8 bit relocations for \c{elf} and \c{bin} output formats are introduced. \b 8 bit relocations for \c{elf} and \c{bin} output formats are introduced.
@@ -59,10 +59,7 @@ since 2007.
\b Tighten EA checks. We warn a user if there overflow in EA addressing. \b Tighten EA checks. We warn a user if there overflow in EA addressing.
\b Make \c{-Ox} the default optimization level. For the legacy \b Make \c{-Ox} the default optimization level. For the legacy
behavior, specify \c{-OL} explicitly. See \k{opt-O}. behavior, specify \c{-O0} explicitly. See \k{opt-O}.
\b Don't optimize displacements in \c{-O0}. For the legacy
behavior, specify \c{-OL}. See \k{opt-O}.
\b Environment variables read with \c{%!} or tested with \c{%ifenv} \b Environment variables read with \c{%!} or tested with \c{%ifenv}
can now contain non-identifier characters if surrounded by quotes. can now contain non-identifier characters if surrounded by quotes.

27
doc/nasmdoc.src
View File

@@ -539,7 +539,7 @@ an intervening space. For example:
\c nasm -f bin driver.asm -odriver.sys \c nasm -f bin driver.asm -odriver.sys
Note that this is a small o, and is different from a capital O , which Note that this is a small o, and is different from a capital O , which
is used to specify the optimization level. See \k{opt-O}. is used to specify the number of optimisation passes required. See \k{opt-O}.
\S{opt-f} The \i\c{-f} Option: Specifying the \i{Output File Format} \S{opt-f} The \i\c{-f} Option: Specifying the \i{Output File Format}
@@ -840,29 +840,28 @@ argument, instructs NASM to replace its powerful \i{preprocessor}
with a \i{stub preprocessor} which does nothing. with a \i{stub preprocessor} which does nothing.
\S{opt-O} The \i\c{-O} Option: Specifying \i{Code Size Optimization} \S{opt-O} The \i\c{-O} Option: Specifying \i{Multipass Optimization}
NASM defaults to not optimizing operands which can fit into a signed byte.
This means that if you want the shortest possible object code,
you have to enable optimization.
Using the \c{-O} option, you can tell NASM to carry out different Using the \c{-O} option, you can tell NASM to carry out different
levels of optimization. The syntax is: levels of optimization. The syntax is:
\b \c{-O0}: No optimization. All operands take their long forms, \b \c{-O0}: No optimization. All operands take their long forms,
if a short form is not specified, except conditional jumps. if a short form is not specified, except conditional jumps.
\b \c{-OL}: Legacy optimization. Like \c{-O0}, but displacements
which will fit in a signed byte and do not contain references
to labels later in the file are optimized, unless the long form
is specified.
This is intended to match NASM 0.98 behavior. This is intended to match NASM 0.98 behavior.
\b \c{-O1}: Minimal optimization. Like \c{-O0}, but immediate operands \b \c{-O1}: Minimal optimization. As above, but immediate operands
which will fit in a signed byte are optimized, which will fit in a signed byte are optimized,
unless the long form is specified. Conditional jumps default unless the long form is specified. Conditional jumps default
to the long form unless otherwise specified. to the long form unless otherwise specified.
\b \c{-Ox} (where \c{x} is the actual letter \c{x}): Full optimization. \b \c{-Ox} (where \c{x} is the actual letter \c{x}): Multipass optimization.
Minimize branch offsets, displacements and immediates, Minimize branch offsets and signed immediate bytes,
overriding size specification unless the \c{strict} keyword overriding size specification unless the \c{strict} keyword
has been used (see \k{strict}). For compatibility with earlier has been used (see \k{strict}). For compatability with earlier
releases, the letter \c{x} may also be any number greater than releases, the letter \c{x} may also be any number greater than
one. This number has no effect on the actual number of passes. one. This number has no effect on the actual number of passes.
@@ -1844,8 +1843,8 @@ invent one using the macro processor.
\H{strict} \i\c{STRICT}: Inhibiting Optimization \H{strict} \i\c{STRICT}: Inhibiting Optimization
When assembling with the optimizer set to full (see \k{opt-O}), When assembling with the optimizer set to level 2 or higher (see
NASM will use size specifiers (\c{BYTE}, \c{WORD}, \k{opt-O}), NASM will use size specifiers (\c{BYTE}, \c{WORD},
\c{DWORD}, \c{QWORD}, \c{TWORD}, \c{OWORD} or \c{YWORD}), but will \c{DWORD}, \c{QWORD}, \c{TWORD}, \c{OWORD} or \c{YWORD}), but will
give them the smallest possible size. The keyword \c{STRICT} can be give them the smallest possible size. The keyword \c{STRICT} can be
used to inhibit optimization and force a particular operand to be used to inhibit optimization and force a particular operand to be
@@ -4953,7 +4952,7 @@ be specified, even if it is the same as the internal name. The
available attributes are: available attributes are:
\b \c{resident} indicates that the exported name is to be kept \b \c{resident} indicates that the exported name is to be kept
resident by the system loader. This is an optimization for resident by the system loader. This is an optimisation for
frequently used symbols imported by name. frequently used symbols imported by name.
\b \c{nodata} indicates that the exported symbol is a function which \b \c{nodata} indicates that the exported symbol is a function which

23
nasm.c
View File

@@ -674,22 +674,14 @@ static bool process_arg(char *p, char *q)
case '5': case '6': case '7': case '8': case '9': case '5': case '6': case '7': case '8': case '9':
opt = strtoul(param, &param, 10); opt = strtoul(param, &param, 10);
if (opt == 0) /* -O0 -> optimizing == -1, 0.98 behaviour */
/* no optimization */ /* -O1 -> optimizing == 0, 0.98.09 behaviour */
optimizing = -2; if (opt < 2)
else if (opt == 1) optimizing = opt - 1;
/* 0.98.09 behaviour */
optimizing = 0;
else else
optimizing = opt; optimizing = opt;
break; break;
case 'L':
/* 0.98 behaviour */
param++;
optimizing = -1;
break;
case 'v': case 'v':
case '+': case '+':
param++; param++;
@@ -791,11 +783,10 @@ static bool process_arg(char *p, char *q)
" -F format select a debugging format\n\n" " -F format select a debugging format\n\n"
" -I<path> adds a pathname to the include file path\n"); " -I<path> adds a pathname to the include file path\n");
printf printf
(" -O<digit> optimize code size\n" (" -O<digit> optimize branch offsets\n"
" -O0: No optimization\n" " -O0: No optimization (default)\n"
" -OL: Legacy optimization\n"
" -O1: Minimal optimization\n" " -O1: Minimal optimization\n"
" -Ox: Full optimization (default)\n\n" " -Ox: Multipass optimization (recommended)\n\n"
" -P<file> pre-includes a file\n" " -P<file> pre-includes a file\n"
" -D<macro>[=<value>] pre-defines a macro\n" " -D<macro>[=<value>] pre-defines a macro\n"
" -U<macro> undefines a macro\n" " -U<macro> undefines a macro\n"

3
nasm.h
View File

@@ -484,8 +484,7 @@ enum ea_flags { /* special EA flags */
EAF_TIMESTWO = 4, /* really do EAX*2 not EAX+EAX */ EAF_TIMESTWO = 4, /* really do EAX*2 not EAX+EAX */
EAF_REL = 8, /* IP-relative addressing */ EAF_REL = 8, /* IP-relative addressing */
EAF_ABS = 16, /* non-IP-relative addressing */ EAF_ABS = 16, /* non-IP-relative addressing */
EAF_FSGS = 32, /* fs/gs segment override present */ EAF_FSGS = 32 /* fs/gs segment override present */
EAF_NO_OFFS = 64 /* no explicit offset in source */
}; };
enum eval_hint { /* values for `hinttype' */ enum eval_hint { /* values for `hinttype' */

1
parser.c
View File

@@ -804,7 +804,6 @@ is_expression:
} }
} }
} else { } else {
result->oprs[operand].eaflags |= EAF_NO_OFFS;
o = 0; o = 0;
result->oprs[operand].wrt = NO_SEG; result->oprs[operand].wrt = NO_SEG;
result->oprs[operand].segment = NO_SEG; result->oprs[operand].segment = NO_SEG;

37
test/zero_displacement.asm
View File

@@ -1,37 +0,0 @@
;Testname=O0; Arguments=-O0 -fbin -ozero_displacement.bin; Files=stdout stderr zero_displacement.bin
;Testname=OL; Arguments=-OL -fbin -ozero_displacement.bin; Files=stdout stderr zero_displacement.bin
;Testname=O1; Arguments=-O1 -fbin -ozero_displacement.bin; Files=stdout stderr zero_displacement.bin
;Testname=Ox; Arguments=-Ox -fbin -ozero_displacement.bin; Files=stdout stderr zero_displacement.bin
bits 16
mov ax, [bx]
mov ax, [bx+0]
mov ax, [bx+di]
mov ax, [bx+di+0]
mov ax, [bp]
mov ax, [bp+0]
bits 32
mov eax, [eax]
mov eax, [eax+0]
mov eax, [eax+ebx]
mov eax, [eax+ebx+0]
mov eax, [ebp]
mov eax, [ebp+0]
bits 64
mov eax, [rax]
mov eax, [rax+0]
mov eax, [rax+rbx]
mov eax, [rax+rbx+0]
mov eax, [rbp]
mov eax, [rbp+0]