Commit 7410e440 authored by Damien's avatar Damien
Browse files

Add basic complex number support.

parent fecee2b0
......@@ -218,7 +218,7 @@ static void push_result_token(parser_t *parser, const py_lexer_t *lex) {
int_val = base * int_val + str[i] - 'a' + 10;
} else if (base == 16 && 'F' <= str[i] && str[i] <= 'F') {
int_val = base * int_val + str[i] - 'A' + 10;
} else if (str[i] == '.' || str[i] == 'e' || str[i] == 'E') {
} else if (str[i] == '.' || str[i] == 'e' || str[i] == 'E' || str[i] == 'j' || str[i] == 'J') {
dec = true;
break;
} else {
......
......@@ -14,6 +14,11 @@
#include "runtime.h"
#include "bc.h"
#if MICROPY_ENABLE_FLOAT
// for sqrt
#include <math.h>
#endif
#if 0 // print debugging info
#define DEBUG_PRINT (1)
#define WRITE_NATIVE (1)
......@@ -32,7 +37,7 @@ typedef machine_int_t py_small_int_t;
#define TO_SMALL_INT(o) ((py_obj_t)(((o) << 1) | 1))
#if MICROPY_ENABLE_FLOAT
typedef machine_float_t float_t;
typedef machine_float_t py_float_t;
#endif
typedef enum {
......@@ -40,6 +45,7 @@ typedef enum {
O_STR,
#if MICROPY_ENABLE_FLOAT
O_FLOAT,
O_COMPLEX,
#endif
O_EXCEPTION_0,
O_EXCEPTION_N,
......@@ -91,7 +97,11 @@ struct _py_obj_base_t {
const char *id;
qstr u_str;
#if MICROPY_ENABLE_FLOAT
float_t u_flt;
py_float_t u_float; // for O_FLOAT
struct { // for O_COMPLEX
py_float_t real;
py_float_t imag;
} u_complex;
#endif
struct { // for O_EXCEPTION_0
qstr id;
......@@ -176,6 +186,7 @@ static qstr q_IndexError;
static qstr q_KeyError;
static qstr q_NameError;
static qstr q_TypeError;
static qstr q_SyntaxError;
py_obj_t py_const_none;
py_obj_t py_const_false;
......@@ -350,10 +361,18 @@ py_obj_t py_obj_new_str(qstr qstr) {
}
#if MICROPY_ENABLE_FLOAT
py_obj_t py_obj_new_float(float_t val) {
py_obj_t py_obj_new_float(py_float_t val) {
py_obj_base_t *o = m_new(py_obj_base_t, 1);
o->kind = O_FLOAT;
o->u_flt = val;
o->u_float = val;
return (py_obj_t)o;
}
py_obj_t py_obj_new_complex(py_float_t real, py_float_t imag) {
py_obj_base_t *o = m_new(py_obj_base_t, 1);
o->kind = O_COMPLEX;
o->u_complex.real = real;
o->u_complex.imag = imag;
return (py_obj_t)o;
}
#endif
......@@ -533,6 +552,30 @@ py_obj_t py_builtin_len(py_obj_t o_in) {
return TO_SMALL_INT(len);
}
py_obj_t py_builtin_abs(py_obj_t o_in) {
if (IS_SMALL_INT(o_in)) {
py_small_int_t val = FROM_SMALL_INT(o_in);
if (val < 0) {
val = -val;
}
return TO_SMALL_INT(val);
} else if (IS_O(o_in, O_FLOAT)) {
py_obj_base_t *o = o_in;
// TODO check for NaN etc
if (o->u_float < 0) {
return py_obj_new_float(-o->u_float);
} else {
return o_in;
}
} else if (IS_O(o_in, O_COMPLEX)) {
py_obj_base_t *o = o_in;
return py_obj_new_float(sqrt(o->u_complex.real*o->u_complex.real + o->u_complex.imag*o->u_complex.imag));
} else {
assert(0);
return py_const_none;
}
}
py_obj_t py_builtin___build_class__(py_obj_t o_class_fun, py_obj_t o_class_name) {
// we differ from CPython: we set the new __locals__ object here
py_map_t *old_locals = map_locals;
......@@ -572,6 +615,7 @@ void rt_init(void) {
q_KeyError = qstr_from_str_static("KeyError");
q_NameError = qstr_from_str_static("NameError");
q_TypeError = qstr_from_str_static("TypeError");
q_SyntaxError = qstr_from_str_static("SyntaxError");
py_const_none = py_obj_new_const("None");
py_const_false = py_obj_new_const("False");
......@@ -586,6 +630,7 @@ void rt_init(void) {
py_qstr_map_lookup(&map_builtins, qstr_from_str_static("__repl_print__"), true)->value = rt_make_function_1(py_builtin___repl_print__);
py_qstr_map_lookup(&map_builtins, q_print, true)->value = rt_make_function_1(py_builtin_print);
py_qstr_map_lookup(&map_builtins, q_len, true)->value = rt_make_function_1(py_builtin_len);
py_qstr_map_lookup(&map_builtins, qstr_from_str_static("abs"), true)->value = rt_make_function_1(py_builtin_abs);
py_qstr_map_lookup(&map_builtins, q___build_class__, true)->value = rt_make_function_2(py_builtin___build_class__);
py_qstr_map_lookup(&map_builtins, qstr_from_str_static("range"), true)->value = rt_make_function_1(py_builtin_range);
......@@ -789,7 +834,14 @@ void py_obj_print(py_obj_t o_in) {
break;
#if MICROPY_ENABLE_FLOAT
case O_FLOAT:
printf("%f", o->u_flt);
printf("%.8g", o->u_float);
break;
case O_COMPLEX:
if (o->u_complex.real == 0) {
printf("%.8gj", o->u_complex.imag);
} else {
printf("(%.8g+%.8gj)", o->u_complex.real, o->u_complex.imag);
}
break;
#endif
case O_EXCEPTION_0:
......@@ -907,13 +959,36 @@ machine_float_t py_obj_get_float(py_obj_t arg) {
} else if (IS_SMALL_INT(arg)) {
return FROM_SMALL_INT(arg);
} else if (IS_O(arg, O_FLOAT)) {
return ((py_obj_base_t*)arg)->u_flt;
return ((py_obj_base_t*)arg)->u_float;
} else {
assert(0);
return 0;
}
}
void py_obj_get_complex(py_obj_t arg, py_float_t *real, py_float_t *imag) {
if (arg == py_const_false) {
*real = 0;
*imag = 0;
} else if (arg == py_const_true) {
*real = 1;
*imag = 0;
} else if (IS_SMALL_INT(arg)) {
*real = FROM_SMALL_INT(arg);
*imag = 0;
} else if (IS_O(arg, O_FLOAT)) {
*real = ((py_obj_base_t*)arg)->u_float;
*imag = 0;
} else if (IS_O(arg, O_COMPLEX)) {
*real = ((py_obj_base_t*)arg)->u_complex.real;
*imag = ((py_obj_base_t*)arg)->u_complex.imag;
} else {
assert(0);
*real = 0;
*imag = 0;
}
}
qstr py_get_qstr(py_obj_t arg) {
if (IS_O(arg, O_STR)) {
return ((py_obj_base_t*)arg)->u_str;
......@@ -935,6 +1010,74 @@ py_obj_t *py_get_array_fixed_n(py_obj_t o_in, machine_int_t n) {
}
}
#define PARSE_DEC_IN_INTG (1)
#define PARSE_DEC_IN_FRAC (2)
#define PARSE_DEC_IN_EXP (3)
py_obj_t rt_load_const_dec(qstr qstr) {
#if MICROPY_ENABLE_FLOAT
DEBUG_OP_printf("load '%s'\n", qstr_str(qstr));
const char *s = qstr_str(qstr);
int in = PARSE_DEC_IN_INTG;
py_float_t dec_val = 0;
bool exp_neg = false;
int exp_val = 0;
int exp_extra = 0;
bool imag = false;
for (; *s; s++) {
int dig = *s;
if ('0' <= dig && dig <= '9') {
dig -= '0';
if (in == PARSE_DEC_IN_EXP) {
exp_val = 10 * exp_val + dig;
} else {
dec_val = 10 * dec_val + dig;
if (in == PARSE_DEC_IN_FRAC) {
exp_extra -= 1;
}
}
} else if (in == PARSE_DEC_IN_INTG && dig == '.') {
in = PARSE_DEC_IN_FRAC;
} else if (in != PARSE_DEC_IN_EXP && (dig == 'E' || dig == 'e')) {
in = PARSE_DEC_IN_EXP;
if (s[1] == '+') {
s++;
} else if (s[1] == '-') {
s++;
exp_neg = true;
}
} else if (dig == 'J' || dig == 'j') {
s++;
imag = true;
break;
} else {
// unknown character
break;
}
}
if (*s != 0) {
nlr_jump(py_obj_new_exception_2(q_SyntaxError, "invalid syntax for number", NULL, NULL));
}
if (exp_neg) {
exp_val = -exp_val;
}
exp_val += exp_extra;
for (; exp_val > 0; exp_val--) {
dec_val *= 10;
}
for (; exp_val < 0; exp_val++) {
dec_val *= 0.1;
}
if (imag) {
return py_obj_new_complex(0, dec_val);
} else {
return py_obj_new_float(dec_val);
}
#else
nlr_jump(py_obj_new_exception_2(q_SyntaxError, "decimal numbers not supported", NULL, NULL));
#endif
}
py_obj_t rt_load_const_str(qstr qstr) {
DEBUG_OP_printf("load '%s'\n", qstr_str(qstr));
return py_obj_new_str(qstr);
......@@ -990,6 +1133,32 @@ void rt_store_global(qstr qstr, py_obj_t obj) {
}
py_obj_t rt_unary_op(int op, py_obj_t arg) {
DEBUG_OP_printf("unary %d %p\n", op, arg);
if (IS_SMALL_INT(arg)) {
py_small_int_t val = FROM_SMALL_INT(arg);
switch (op) {
case RT_UNARY_OP_NOT: if (val != 0) { return py_const_true;} else { return py_const_false; }
case RT_UNARY_OP_POSITIVE: break;
case RT_UNARY_OP_NEGATIVE: val = -val; break;
case RT_UNARY_OP_INVERT: val = ~val; break;
default: assert(0); val = 0;
}
if (fit_small_int(val)) {
return TO_SMALL_INT(val);
}
#if MICROPY_ENABLE_FLOAT
} else if (IS_O(arg, O_FLOAT)) {
py_float_t val = py_obj_get_float(arg);
switch (op) {
case RT_UNARY_OP_NOT: if (val != 0) { return py_const_true;} else { return py_const_false; }
case RT_UNARY_OP_POSITIVE: break;
case RT_UNARY_OP_NEGATIVE: val = -val; break;
case RT_UNARY_OP_INVERT: nlr_jump(py_obj_new_exception_2(q_TypeError, "bad operand type for unary ~: 'float'", NULL, NULL));
default: assert(0); val = 0;
}
return py_obj_new_float(val);
#endif
}
assert(0);
return py_const_none;
}
......@@ -1033,63 +1202,95 @@ py_obj_t rt_binary_op(int op, py_obj_t lhs, py_obj_t rhs) {
} else if (IS_SMALL_INT(lhs) && IS_SMALL_INT(rhs)) {
py_small_int_t lhs_val = FROM_SMALL_INT(lhs);
py_small_int_t rhs_val = FROM_SMALL_INT(rhs);
py_small_int_t val;
switch (op) {
case RT_BINARY_OP_OR:
case RT_BINARY_OP_INPLACE_OR: val = lhs_val | rhs_val; break;
case RT_BINARY_OP_INPLACE_OR: lhs_val |= rhs_val; break;
case RT_BINARY_OP_XOR:
case RT_BINARY_OP_INPLACE_XOR: val = lhs_val ^ rhs_val; break;
case RT_BINARY_OP_INPLACE_XOR: lhs_val ^= rhs_val; break;
case RT_BINARY_OP_AND:
case RT_BINARY_OP_INPLACE_AND: val = lhs_val & rhs_val; break;
case RT_BINARY_OP_INPLACE_AND: lhs_val &= rhs_val; break;
case RT_BINARY_OP_LSHIFT:
case RT_BINARY_OP_INPLACE_LSHIFT: val = lhs_val << rhs_val; break;
case RT_BINARY_OP_INPLACE_LSHIFT: lhs_val <<= rhs_val; break;
case RT_BINARY_OP_RSHIFT:
case RT_BINARY_OP_INPLACE_RSHIFT: val = lhs_val >> rhs_val; break;
case RT_BINARY_OP_INPLACE_RSHIFT: lhs_val >>= rhs_val; break;
case RT_BINARY_OP_ADD:
case RT_BINARY_OP_INPLACE_ADD: val = lhs_val + rhs_val; break;
case RT_BINARY_OP_INPLACE_ADD: lhs_val += rhs_val; break;
case RT_BINARY_OP_SUBTRACT:
case RT_BINARY_OP_INPLACE_SUBTRACT: val = lhs_val - rhs_val; break;
case RT_BINARY_OP_INPLACE_SUBTRACT: lhs_val -= rhs_val; break;
case RT_BINARY_OP_MULTIPLY:
case RT_BINARY_OP_INPLACE_MULTIPLY: val = lhs_val * rhs_val; break;
case RT_BINARY_OP_INPLACE_MULTIPLY: lhs_val *= rhs_val; break;
case RT_BINARY_OP_FLOOR_DIVIDE:
case RT_BINARY_OP_INPLACE_FLOOR_DIVIDE: val = lhs_val / rhs_val; break;
case RT_BINARY_OP_INPLACE_FLOOR_DIVIDE: lhs_val /= rhs_val; break;
#if MICROPY_ENABLE_FLOAT
case RT_BINARY_OP_TRUE_DIVIDE:
case RT_BINARY_OP_INPLACE_TRUE_DIVIDE: return py_obj_new_float((float_t)lhs_val / (float_t)rhs_val);
case RT_BINARY_OP_INPLACE_TRUE_DIVIDE: return py_obj_new_float((py_float_t)lhs_val / (py_float_t)rhs_val);
#endif
case RT_BINARY_OP_POWER:
case RT_BINARY_OP_INPLACE_POWER:
// TODO
if (rhs_val == 2) {
val = lhs_val * lhs_val;
lhs_val = lhs_val * lhs_val;
break;
}
default: printf("%d\n", op); assert(0); val = 0;
default: printf("%d\n", op); assert(0);
}
if (fit_small_int(val)) {
return TO_SMALL_INT(val);
if (fit_small_int(lhs_val)) {
return TO_SMALL_INT(lhs_val);
}
#if MICROPY_ENABLE_FLOAT
} else if (IS_O(lhs, O_FLOAT) || IS_O(rhs, O_FLOAT)) {
float_t lhs_val = py_obj_get_float(lhs);
float_t rhs_val = py_obj_get_float(rhs);
float_t val;
} else if (IS_O(lhs, O_COMPLEX) || IS_O(rhs, O_COMPLEX)) {
py_float_t lhs_real, lhs_imag, rhs_real, rhs_imag;
py_obj_get_complex(lhs, &lhs_real, &lhs_imag);
py_obj_get_complex(rhs, &rhs_real, &rhs_imag);
switch (op) {
case RT_BINARY_OP_ADD:
case RT_BINARY_OP_INPLACE_ADD: val = lhs_val + rhs_val; break;
case RT_BINARY_OP_INPLACE_ADD:
lhs_real += rhs_real;
lhs_imag += rhs_imag;
break;
case RT_BINARY_OP_SUBTRACT:
case RT_BINARY_OP_INPLACE_SUBTRACT: val = lhs_val - rhs_val; break;
case RT_BINARY_OP_INPLACE_SUBTRACT:
lhs_real -= rhs_real;
lhs_imag -= rhs_imag;
break;
case RT_BINARY_OP_MULTIPLY:
case RT_BINARY_OP_INPLACE_MULTIPLY: val = lhs_val * rhs_val; break;
case RT_BINARY_OP_INPLACE_MULTIPLY:
{
py_float_t real = lhs_real * rhs_real - lhs_imag * rhs_imag;
lhs_imag = lhs_real * rhs_imag + lhs_imag * rhs_real;
lhs_real = real;
break;
}
/* TODO floor(?) the value
case RT_BINARY_OP_FLOOR_DIVIDE:
case RT_BINARY_OP_INPLACE_FLOOR_DIVIDE: val = lhs_val / rhs_val; break;
*/
/* TODO
case RT_BINARY_OP_TRUE_DIVIDE:
case RT_BINARY_OP_INPLACE_TRUE_DIVIDE: val = lhs_val / rhs_val; break;
default: printf("%d\n", op); assert(0); val = 0;
*/
default: printf("%d\n", op); assert(0);
}
return py_obj_new_float(val);
return py_obj_new_complex(lhs_real, lhs_imag);
} else if (IS_O(lhs, O_FLOAT) || IS_O(rhs, O_FLOAT)) {
py_float_t lhs_val = py_obj_get_float(lhs);
py_float_t rhs_val = py_obj_get_float(rhs);
switch (op) {
case RT_BINARY_OP_ADD:
case RT_BINARY_OP_INPLACE_ADD: lhs_val += rhs_val; break;
case RT_BINARY_OP_SUBTRACT:
case RT_BINARY_OP_INPLACE_SUBTRACT: lhs_val -= rhs_val; break;
case RT_BINARY_OP_MULTIPLY:
case RT_BINARY_OP_INPLACE_MULTIPLY: lhs_val *= rhs_val; break;
/* TODO floor(?) the value
case RT_BINARY_OP_FLOOR_DIVIDE:
case RT_BINARY_OP_INPLACE_FLOOR_DIVIDE: val = lhs_val / rhs_val; break;
*/
case RT_BINARY_OP_TRUE_DIVIDE:
case RT_BINARY_OP_INPLACE_TRUE_DIVIDE: lhs_val /= rhs_val; break;
default: printf("%d\n", op); assert(0);
}
return py_obj_new_float(lhs_val);
#endif
} else if (IS_O(lhs, O_STR) && IS_O(rhs, O_STR)) {
const char *lhs_str = qstr_str(((py_obj_base_t*)lhs)->u_str);
......@@ -1148,8 +1349,8 @@ py_obj_t rt_compare_op(int op, py_obj_t lhs, py_obj_t rhs) {
#if MICROPY_ENABLE_FLOAT
// deal with floats
if (IS_O(lhs, O_FLOAT) || IS_O(rhs, O_FLOAT)) {
float_t lhs_val = py_obj_get_float(lhs);
float_t rhs_val = py_obj_get_float(rhs);
py_float_t lhs_val = py_obj_get_float(lhs);
py_float_t rhs_val = py_obj_get_float(rhs);
int cmp;
switch (op) {
case RT_COMPARE_OP_LESS: cmp = lhs_val < rhs_val; break;
......@@ -1288,7 +1489,7 @@ machine_uint_t rt_convert_obj_for_inline_asm(py_obj_t obj) {
#if MICROPY_ENABLE_FLOAT
case O_FLOAT:
// convert float to int (could also pass in float registers)
return (machine_int_t)o->u_flt;
return (machine_int_t)o->u_float;
#endif
case O_TUPLE:
......
......@@ -102,6 +102,7 @@ machine_float_t py_obj_get_float(py_obj_t arg);
qstr py_get_qstr(py_obj_t arg);
py_obj_t *py_get_array_fixed_n(py_obj_t o, machine_int_t n);
py_obj_t py_obj_new_int(machine_int_t value);
py_obj_t rt_load_const_dec(qstr qstr);
py_obj_t rt_load_const_str(qstr qstr);
py_obj_t rt_load_name(qstr qstr);
py_obj_t rt_load_global(qstr qstr);
......
......@@ -83,6 +83,11 @@ bool py_execute_byte_code_2(const byte *code, const byte **ip_in_out, py_obj_t *
PUSH((py_obj_t)(snum << 1 | 1));
break;
case PYBC_LOAD_CONST_DEC:
DECODE_QSTR;
PUSH(rt_load_const_dec(qstr));
break;
case PYBC_LOAD_CONST_ID:
DECODE_QSTR;
PUSH(rt_load_const_str(qstr)); // TODO
......@@ -267,6 +272,11 @@ bool py_execute_byte_code_2(const byte *code, const byte **ip_in_out, py_obj_t *
//sp += 3; // pop 3 exception values
break;
case PYBC_UNARY_OP:
unum = *ip++;
*sp = rt_unary_op(unum, *sp);
break;
case PYBC_BINARY_OP:
unum = *ip++;
obj2 = POP();
......
......@@ -3,7 +3,7 @@ BUILD=build
CC = gcc
CFLAGS = -I. -I$(PYSRC) -Wall -ansi -std=gnu99 -Os #-DNDEBUG
LDFLAGS =
LDFLAGS = -lm
SRC_C = \
main.c \
......
Supports Markdown
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment