iverilog/tgt-vvp/eval_vec4.c

1428 lines
38 KiB
C

/*
* Copyright (c) 2013-2020 Stephen Williams (steve@icarus.com)
*
* This source code is free software; you can redistribute it
* and/or modify it in source code form under the terms of the GNU
* General Public License as published by the Free Software
* Foundation; either version 2 of the License, or (at your option)
* any later version.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
*/
/*
* This file includes functions for evaluating VECTOR expressions.
*/
# include "vvp_priv.h"
# include <string.h>
# include <stdlib.h>
# include <math.h>
# include <assert.h>
# include <stdbool.h>
void resize_vec4_wid(ivl_expr_t expr, unsigned wid)
{
if (ivl_expr_width(expr) == wid)
return;
if (ivl_expr_signed(expr))
fprintf(vvp_out, " %%pad/s %u;\n", wid);
else
fprintf(vvp_out, " %%pad/u %u;\n", wid);
}
/*
* Test if the draw_immediate_vec4 instruction can be used.
*/
int test_immediate_vec4_ok(ivl_expr_t re)
{
const char*bits;
unsigned idx;
if (ivl_expr_type(re) != IVL_EX_NUMBER)
return 0;
if (ivl_expr_width(re) <= 32)
return 1;
bits = ivl_expr_bits(re);
for (idx = 32 ; idx < ivl_expr_width(re) ; idx += 1) {
if (bits[idx] != '0')
return 0;
}
return 1;
}
static void make_immediate_vec4_words(ivl_expr_t re, unsigned long*val0p, unsigned long*valxp, unsigned*widp)
{
unsigned long val0 = 0;
unsigned long valx = 0;
unsigned wid = ivl_expr_width(re);
const char*bits = ivl_expr_bits(re);
unsigned idx;
for (idx = 0 ; idx < wid ; idx += 1) {
assert( ((val0|valx)&0x80000000UL) == 0UL );
val0 <<= 1;
valx <<= 1;
switch (bits[wid-idx-1]) {
case '0':
break;
case '1':
val0 |= 1;
break;
case 'x':
val0 |= 1;
valx |= 1;
break;
case 'z':
val0 |= 0;
valx |= 1;
break;
default:
assert(0);
break;
}
}
*val0p = val0;
*valxp = valx;
*widp = wid;
}
void draw_immediate_vec4(ivl_expr_t re, const char*opcode)
{
unsigned long val0, valx;
unsigned wid;
make_immediate_vec4_words(re, &val0, &valx, &wid);
fprintf(vvp_out, " %s %lu, %lu, %u;\n", opcode, val0, valx, wid);
}
static void draw_binary_vec4_arith(ivl_expr_t expr)
{
ivl_expr_t le = ivl_expr_oper1(expr);
ivl_expr_t re = ivl_expr_oper2(expr);
unsigned lwid = ivl_expr_width(le);
unsigned rwid = ivl_expr_width(re);
unsigned ewid = ivl_expr_width(expr);
int is_power_op = ivl_expr_opcode(expr) == 'p' ? 1 : 0;
/* The power operation differs from the other arithmetic operations
in that we only use the signed version of the operation if the
right hand operand (the exponent) is signed. */
int signed_flag = (ivl_expr_signed(le) || is_power_op) && ivl_expr_signed(re) ? 1 : 0;
const char*signed_string = signed_flag? "/s" : "";
/* All the arithmetic operations handled here (except for the power
operation) require that the operands (and the result) be the same
width. We further assume that the core has not given us an operand
wider then the expression width. So pad operands as needed. */
draw_eval_vec4(le);
if (lwid != ewid) {
fprintf(vvp_out, " %%pad/%c %u;\n", ivl_expr_signed(le)? 's' : 'u', ewid);
}
/* Special case: If the re expression can be collected into an
immediate operand, and the instruction supports it, then
generate an immediate instruction instead of the generic
version. */
if (rwid==ewid && test_immediate_vec4_ok(re)) {
switch (ivl_expr_opcode(expr)) {
case '+':
draw_immediate_vec4(re, "%addi");
return;
case '-':
draw_immediate_vec4(re, "%subi");
return;
case '*':
draw_immediate_vec4(re, "%muli");
return;
default:
break;
}
}
draw_eval_vec4(re);
if ((rwid != ewid) && !is_power_op) {
fprintf(vvp_out, " %%pad/%c %u;\n", ivl_expr_signed(re)? 's' : 'u', ewid);
}
switch (ivl_expr_opcode(expr)) {
case '+':
fprintf(vvp_out, " %%add;\n");
break;
case '-':
fprintf(vvp_out, " %%sub;\n");
break;
case '*':
fprintf(vvp_out, " %%mul;\n");
break;
case '/':
fprintf(vvp_out, " %%div%s;\n", signed_string);
break;
case '%':
fprintf(vvp_out, " %%mod%s;\n", signed_string);
break;
case 'p':
fprintf(vvp_out, " %%pow%s;\n", signed_string);
break;
default:
assert(0);
break;
}
}
static void draw_binary_vec4_bitwise(ivl_expr_t expr)
{
draw_eval_vec4(ivl_expr_oper1(expr));
draw_eval_vec4(ivl_expr_oper2(expr));
switch (ivl_expr_opcode(expr)) {
case '&':
fprintf(vvp_out, " %%and;\n");
break;
case '|':
fprintf(vvp_out, " %%or;\n");
break;
case '^':
fprintf(vvp_out, " %%xor;\n");
break;
case 'A': /* ~& */
fprintf(vvp_out, " %%nand;\n");
break;
case 'O': /* ~| */
fprintf(vvp_out, " %%nor;\n");
break;
case 'X': /* ~^ */
fprintf(vvp_out, " %%xnor;\n");
break;
default:
assert(0);
break;
}
}
static void draw_binary_vec4_compare_real(ivl_expr_t expr)
{
draw_eval_real(ivl_expr_oper1(expr));
draw_eval_real(ivl_expr_oper2(expr));
switch (ivl_expr_opcode(expr)) {
case 'e': /* == */
fprintf(vvp_out, " %%cmp/wr;\n");
fprintf(vvp_out, " %%flag_get/vec4 4;\n");
break;
case 'n': /* != */
fprintf(vvp_out, " %%cmp/wr;\n");
fprintf(vvp_out, " %%flag_inv 4;\n");
fprintf(vvp_out, " %%flag_get/vec4 4;\n");
break;
default:
assert(0);
}
}
static void draw_binary_vec4_compare_string(ivl_expr_t expr)
{
draw_eval_string(ivl_expr_oper1(expr));
draw_eval_string(ivl_expr_oper2(expr));
switch (ivl_expr_opcode(expr)) {
case 'e': /* == */
fprintf(vvp_out, " %%cmp/str;\n");
fprintf(vvp_out, " %%flag_get/vec4 4;\n");
break;
case 'n': /* != */
fprintf(vvp_out, " %%cmp/str;\n");
fprintf(vvp_out, " %%flag_inv 4;\n");
fprintf(vvp_out, " %%flag_get/vec4 4;\n");
break;
default:
assert(0);
}
}
static void draw_binary_vec4_compare_class(ivl_expr_t expr)
{
ivl_expr_t le = ivl_expr_oper1(expr);
ivl_expr_t re = ivl_expr_oper2(expr);
if (ivl_expr_type(le) == IVL_EX_NULL) {
ivl_expr_t tmp = le;
le = re;
re = tmp;
}
/* Special case: If both operands are null, then the
expression has a constant value. */
if (ivl_expr_type(le)==IVL_EX_NULL && ivl_expr_type(re)==IVL_EX_NULL) {
switch (ivl_expr_opcode(expr)) {
case 'e': /* == */
fprintf(vvp_out, " %%pushi/vec4 1, 0, 1;\n");
break;
case 'n': /* != */
fprintf(vvp_out, " %%pushi/vec4 0, 0, 1;\n");
break;
default:
assert(0);
break;
}
return;
}
/* A signal/variable is compared to null. Implement this with
the %test_nul statement, which peeks at the variable
contents directly. */
if (ivl_expr_type(re)==IVL_EX_NULL && ivl_expr_type(le)==IVL_EX_SIGNAL) {
ivl_signal_t sig= ivl_expr_signal(le);
if (ivl_signal_dimensions(sig) == 0) {
fprintf(vvp_out, " %%test_nul v%p_0;\n", sig);
} else {
ivl_expr_t word_ex = ivl_expr_oper1(le);
int word_ix = allocate_word();
draw_eval_expr_into_integer(word_ex, word_ix);
fprintf(vvp_out, " %%test_nul/a v%p, %d;\n", sig, word_ix);
clr_word(word_ix);
}
if (ivl_expr_opcode(expr) == 'n')
fprintf(vvp_out, " %%flag_inv 4;\n");
fprintf(vvp_out, " %%flag_get/vec4 4;\n");
return;
}
if (ivl_expr_type(re)==IVL_EX_NULL && ivl_expr_type(le)==IVL_EX_PROPERTY) {
ivl_signal_t sig = ivl_expr_signal(le);
unsigned pidx = ivl_expr_property_idx(le);
ivl_expr_t idx_expr = ivl_expr_oper1(le);
int idx = 0;
/* If the property has an array index, then evaluate it
into an index register. */
if ( idx_expr ) {
idx = allocate_word();
draw_eval_expr_into_integer(idx_expr, idx);
}
fprintf(vvp_out, " %%load/obj v%p_0;\n", sig);
fprintf(vvp_out, " %%test_nul/prop %u, %d;\n", pidx, idx);
fprintf(vvp_out, " %%pop/obj 1, 0;\n");
if (ivl_expr_opcode(expr) == 'n')
fprintf(vvp_out, " %%flag_inv 4;\n");
fprintf(vvp_out, " %%flag_get/vec4 4;\n");
if (idx != 0) clr_word(idx);
return;
}
fprintf(stderr, "SORRY: Compare class handles not implemented\n");
fprintf(vvp_out, " ; XXXX compare class handles. re-type=%d, le-type=%d\n",
ivl_expr_type(re), ivl_expr_type(le));
vvp_errors += 1;
}
static void draw_binary_vec4_compare(ivl_expr_t expr)
{
ivl_expr_t le = ivl_expr_oper1(expr);
ivl_expr_t re = ivl_expr_oper2(expr);
if ((ivl_expr_value(le) == IVL_VT_REAL)
|| (ivl_expr_value(re) == IVL_VT_REAL)) {
draw_binary_vec4_compare_real(expr);
return;
}
if ((ivl_expr_value(le)==IVL_VT_STRING)
&& (ivl_expr_value(re)==IVL_VT_STRING)) {
draw_binary_vec4_compare_string(expr);
return;
}
if ((ivl_expr_value(le)==IVL_VT_STRING)
&& (ivl_expr_type(re)==IVL_EX_STRING)) {
draw_binary_vec4_compare_string(expr);
return;
}
if ((ivl_expr_type(le)==IVL_EX_STRING)
&& (ivl_expr_value(re)==IVL_VT_STRING)) {
draw_binary_vec4_compare_string(expr);
return;
}
if ((ivl_expr_value(le)==IVL_VT_CLASS)
&& (ivl_expr_value(re)==IVL_VT_CLASS)) {
draw_binary_vec4_compare_class(expr);
return;
}
unsigned use_wid = ivl_expr_width(le);
if (ivl_expr_width(re) > use_wid)
use_wid = ivl_expr_width(re);
draw_eval_vec4(le);
resize_vec4_wid(le, use_wid);
draw_eval_vec4(re);
resize_vec4_wid(re, use_wid);
switch (ivl_expr_opcode(expr)) {
case 'e': /* == */
fprintf(vvp_out, " %%cmp/e;\n");
fprintf(vvp_out, " %%flag_get/vec4 4;\n");
break;
case 'n': /* != */
fprintf(vvp_out, " %%cmp/ne;\n");
fprintf(vvp_out, " %%flag_get/vec4 4;\n");
break;
case 'E': /* === */
fprintf(vvp_out, " %%cmp/e;\n");
fprintf(vvp_out, " %%flag_get/vec4 6;\n");
break;
case 'N': /* !== */
fprintf(vvp_out, " %%cmp/ne;\n");
fprintf(vvp_out, " %%flag_get/vec4 6;\n");
break;
case 'w': /* ==? */
fprintf(vvp_out, " %%cmp/we;\n");
fprintf(vvp_out, " %%flag_get/vec4 4;\n");
break;
case 'W': /* !=? */
fprintf(vvp_out, " %%cmp/wne;\n");
fprintf(vvp_out, " %%flag_get/vec4 4;\n");
break;
default:
assert(0);
}
}
/*
* Handle the logical implication:
* <le> -> <re>
* which is the same as the expression
* !<le> || <re>
*
*/
static void draw_binary_vec4_limpl(ivl_expr_t expr)
{
ivl_expr_t le = ivl_expr_oper1(expr);
ivl_expr_t re = ivl_expr_oper2(expr);
/* The arguments should have alreacy been reduced. */
assert(ivl_expr_width(le) == 1);
assert(ivl_expr_width(re) == 1);
draw_eval_vec4(le);
fprintf(vvp_out, " %%inv;\n");
draw_eval_vec4(re);
fprintf(vvp_out, " %%or;\n");
}
static void draw_binary_vec4_lequiv(ivl_expr_t expr)
{
ivl_expr_t le = ivl_expr_oper1(expr);
ivl_expr_t re = ivl_expr_oper2(expr);
/* The arguments should have already been reduced. */
assert(ivl_expr_width(le) == 1);
draw_eval_vec4(le);
assert(ivl_expr_width(re) == 1);
draw_eval_vec4(re);
fprintf(vvp_out, " %%xnor;\n");
assert(ivl_expr_width(expr) == 1);
}
static void draw_binary_vec4_logical(ivl_expr_t expr, char op)
{
const char *opcode;
const char *jmp_type;
switch (op) {
case 'a':
opcode = "and";
jmp_type = "0";
break;
case 'o':
opcode = "or";
jmp_type = "1";
break;
default:
assert(0);
break;
}
unsigned label_out = local_count++;
ivl_expr_t le = ivl_expr_oper1(expr);
ivl_expr_t re = ivl_expr_oper2(expr);
/* Evaluate the left expression as a conditon and skip the right expression
* if the left is false. */
int flag = draw_eval_condition(le);
fprintf(vvp_out, " %%flag_get/vec4 %d;\n", flag);
fprintf(vvp_out, " %%jmp/%s T_%u.%u, %d;\n", jmp_type, thread_count,
label_out, flag);
clr_flag(flag);
/* Now push the right expression. Reduce to a single bit if necessary. */
draw_eval_vec4(re);
if (ivl_expr_width(re) > 1)
fprintf(vvp_out, " %%or/r;\n");
fprintf(vvp_out, " %%%s;\n", opcode);
fprintf(vvp_out, "T_%u.%u;\n", thread_count, label_out);
if (ivl_expr_width(expr) > 1)
fprintf(vvp_out, " %%pad/u %u;\n", ivl_expr_width(expr));
}
static void draw_binary_vec4_le_real(ivl_expr_t expr)
{
ivl_expr_t le = ivl_expr_oper1(expr);
ivl_expr_t re = ivl_expr_oper2(expr);
switch (ivl_expr_opcode(expr)) {
case '<':
draw_eval_real(le);
draw_eval_real(re);
fprintf(vvp_out, " %%cmp/wr;\n");
fprintf(vvp_out, " %%flag_get/vec4 5;\n");
break;
case 'L': /* <= */
draw_eval_real(le);
draw_eval_real(re);
fprintf(vvp_out, " %%cmp/wr;\n");
fprintf(vvp_out, " %%flag_get/vec4 4;\n");
fprintf(vvp_out, " %%flag_get/vec4 5;\n");
fprintf(vvp_out, " %%or;\n");
break;
case '>':
draw_eval_real(re);
draw_eval_real(le);
fprintf(vvp_out, " %%cmp/wr;\n");
fprintf(vvp_out, " %%flag_get/vec4 5;\n");
break;
case 'G': /* >= */
draw_eval_real(re);
draw_eval_real(le);
fprintf(vvp_out, " %%cmp/wr;\n");
fprintf(vvp_out, " %%flag_get/vec4 4;\n");
fprintf(vvp_out, " %%flag_get/vec4 5;\n");
fprintf(vvp_out, " %%or;\n");
break;
default:
assert(0);
break;
}
}
static void draw_binary_vec4_le_string(ivl_expr_t expr)
{
ivl_expr_t le = ivl_expr_oper1(expr);
ivl_expr_t re = ivl_expr_oper2(expr);
switch (ivl_expr_opcode(expr)) {
case '<':
draw_eval_string(le);
draw_eval_string(re);
fprintf(vvp_out, " %%cmp/str;\n");
fprintf(vvp_out, " %%flag_get/vec4 5;\n");
break;
case 'L': /* <= */
draw_eval_string(le);
draw_eval_string(re);
fprintf(vvp_out, " %%cmp/str;\n");
fprintf(vvp_out, " %%flag_get/vec4 4;\n");
fprintf(vvp_out, " %%flag_get/vec4 5;\n");
fprintf(vvp_out, " %%or;\n");
break;
case '>':
draw_eval_string(re);
draw_eval_string(le);
fprintf(vvp_out, " %%cmp/str;\n");
fprintf(vvp_out, " %%flag_get/vec4 5;\n");
break;
case 'G': /* >= */
draw_eval_string(re);
draw_eval_string(le);
fprintf(vvp_out, " %%cmp/str;\n");
fprintf(vvp_out, " %%flag_get/vec4 4;\n");
fprintf(vvp_out, " %%flag_get/vec4 5;\n");
fprintf(vvp_out, " %%or;\n");
break;
default:
assert(0);
break;
}
}
static void draw_binary_vec4_le(ivl_expr_t expr)
{
ivl_expr_t le = ivl_expr_oper1(expr);
ivl_expr_t re = ivl_expr_oper2(expr);
ivl_expr_t tmp;
if ((ivl_expr_value(le) == IVL_VT_REAL)
|| (ivl_expr_value(re) == IVL_VT_REAL)) {
draw_binary_vec4_le_real(expr);
return;
}
char use_opcode = ivl_expr_opcode(expr);
char s_flag = (ivl_expr_signed(le) && ivl_expr_signed(re)) ? 's' : 'u';
/* If this is a > or >=, then convert it to < or <= by
swapping the operands. Adjust the opcode to match. */
switch (use_opcode) {
case 'G':
tmp = le;
le = re;
re = tmp;
use_opcode = 'L';
break;
case '>':
tmp = le;
le = re;
re = tmp;
use_opcode = '<';
break;
}
if ((ivl_expr_value(le)==IVL_VT_STRING)
&& (ivl_expr_value(re)==IVL_VT_STRING)) {
draw_binary_vec4_le_string(expr);
return;
}
if ((ivl_expr_value(le)==IVL_VT_STRING)
&& (ivl_expr_type(re)==IVL_EX_STRING)) {
draw_binary_vec4_le_string(expr);
return;
}
if ((ivl_expr_type(le)==IVL_EX_STRING)
&& (ivl_expr_value(re)==IVL_VT_STRING)) {
draw_binary_vec4_le_string(expr);
return;
}
/* NOTE: I think I would rather the elaborator handle the
operand widths. When that happens, take this code out. */
unsigned use_wid = ivl_expr_width(le);
if (ivl_expr_width(re) > use_wid)
use_wid = ivl_expr_width(re);
draw_eval_vec4(le);
resize_vec4_wid(le, use_wid);
if (ivl_expr_width(re)==use_wid && test_immediate_vec4_ok(re)) {
/* Special case: If the right operand can be handled as
an immediate operand, then use that instead. */
char opcode[8];
snprintf(opcode, sizeof opcode, "%%cmpi/%c", s_flag);
draw_immediate_vec4(re, opcode);
} else {
draw_eval_vec4(re);
resize_vec4_wid(re, use_wid);
fprintf(vvp_out, " %%cmp/%c;\n", s_flag);
}
switch (use_opcode) {
case 'L':
fprintf(vvp_out, " %%flag_get/vec4 4;\n");
fprintf(vvp_out, " %%flag_get/vec4 5;\n");
fprintf(vvp_out, " %%or;\n");
break;
case '<':
fprintf(vvp_out, " %%flag_get/vec4 5;\n");
break;
default:
assert(0);
break;
}
}
static void draw_binary_vec4_lrs(ivl_expr_t expr)
{
ivl_expr_t le = ivl_expr_oper1(expr);
ivl_expr_t re = ivl_expr_oper2(expr);
// Push the left expression onto the stack.
draw_eval_vec4(le);
// Calculate the shift amount into an index register.
int use_index_reg = allocate_word();
assert(use_index_reg >= 0);
draw_eval_expr_into_integer(re, use_index_reg);
// Emit the actual shift instruction. This will pop the top of
// the stack and replace it with the result of the shift.
switch (ivl_expr_opcode(expr)) {
case 'l': /* << */
fprintf(vvp_out, " %%shiftl %d;\n", use_index_reg);
break;
case 'r': /* >> */
fprintf(vvp_out, " %%shiftr %d;\n", use_index_reg);
break;
case 'R': /* >>> */
if (ivl_expr_signed(le))
fprintf(vvp_out, " %%shiftr/s %d;\n", use_index_reg);
else
fprintf(vvp_out, " %%shiftr %d;\n", use_index_reg);
break;
default:
assert(0);
break;
}
clr_word(use_index_reg);
}
static void draw_binary_vec4(ivl_expr_t expr)
{
switch (ivl_expr_opcode(expr)) {
case 'a': /* Logical && */
case 'o': /* || (logical or) */
draw_binary_vec4_logical(expr, ivl_expr_opcode(expr));
break;
case '+':
case '-':
case '*':
case '/':
case '%':
case 'p': /* ** (power) */
draw_binary_vec4_arith(expr);
break;
case '&':
case '|':
case '^':
case 'A': /* NAND (~&) */
case 'O': /* NOR (~|) */
case 'X': /* exclusive nor (~^) */
draw_binary_vec4_bitwise(expr);
break;
case 'e': /* == */
case 'E': /* === */
case 'n': /* != */
case 'N': /* !== */
case 'w': /* ==? */
case 'W': /* !=? */
draw_binary_vec4_compare(expr);
break;
case 'G': /* >= */
case 'L': /* <= */
case '>':
case '<':
draw_binary_vec4_le(expr);
break;
case 'l': /* << */
case 'r': /* >> */
case 'R': /* >>> */
draw_binary_vec4_lrs(expr);
break;
case 'q': /* -> (logical implication) */
draw_binary_vec4_limpl(expr);
break;
case 'Q': /* <-> (logical equivalence) */
draw_binary_vec4_lequiv(expr);
break;
default:
fprintf(stderr, "vvp.tgt error: unsupported binary (%c)\n",
ivl_expr_opcode(expr));
assert(0);
}
}
/*
* This handles two special cases:
* 1) Making a large IVL_EX_NUMBER as an immediate value. In this
* case, start with a %pushi/vec4 to get the stack started, then
* continue with %concati/vec4 instructions to build that number
* up.
*
* 2) Concatenating a large IVL_EX_NUMBER to the current top of the
* stack. In this case, start with %concati/vec4 and continue
* generating %concati/vec4 instructions to finish up the large number.
*/
static void draw_concat_number_vec4(ivl_expr_t expr, int as_concati)
{
unsigned long val0 = 0;
unsigned long valx = 0;
unsigned wid = ivl_expr_width(expr);
const char*bits = ivl_expr_bits(expr);
unsigned idx;
int accum = 0;
int count_pushi = as_concati? 1 : 0;
/* Scan the literal bits, MSB first. */
for (idx = 0 ; idx < wid ; idx += 1) {
val0 <<= 1;
valx <<= 1;
switch (bits[wid-idx-1]) {
case '0':
break;
case '1':
val0 |= 1;
break;
case 'x':
val0 |= 1;
valx |= 1;
break;
case 'z':
val0 |= 0;
valx |= 1;
break;
default:
assert(0);
break;
}
accum += 1;
/* Collect as many bits as can be written by a single
%pushi/vec4 instruction. This may be more than 32 if
the higher bits are zero, but if the currently
accumulated value fills what a %pushi/vec4 can do,
then write it out, generate a %concat/vec4, and set
up to handle more bits. */
if ( (val0|valx) & 0x80000000UL ) {
if (count_pushi) {
fprintf(vvp_out, " %%concati/vec4 %lu, %lu, %d;\n",
val0, valx, accum);
} else {
fprintf(vvp_out, " %%pushi/vec4 %lu, %lu, %d;\n",
val0, valx, accum);
}
accum = 0;
val0 = 0;
valx = 0;
count_pushi += 1;
}
}
if (accum) {
if (count_pushi) {
fprintf(vvp_out, " %%concati/vec4 %lu, %lu, %d;\n",
val0, valx, accum);
} else {
fprintf(vvp_out, " %%pushi/vec4 %lu, %lu, %d;\n",
val0, valx, accum);
}
}
}
static void draw_concat_vec4(ivl_expr_t expr)
{
/* Repeat the concatenation this many times to make a
super-concatenation. */
unsigned repeat = ivl_expr_repeat(expr);
/* This is the number of expressions that go into the
concatenation. */
unsigned num_sube = ivl_expr_parms(expr);
unsigned sub_idx = 0;
ivl_expr_t sube;
assert(num_sube > 0);
/* Start with the most-significant bits. */
sube = ivl_expr_parm(expr, sub_idx);
draw_eval_vec4(sube);
/* Evaluate, but skip any zero replication expressions at the
* head of this concatenation. */
while ((ivl_expr_type(sube) == IVL_EX_CONCAT) &&
(ivl_expr_repeat(sube) == 0)) {
fprintf(vvp_out, " %%pop/vec4 1; skip zero replication\n");
sub_idx += 1;
sube = ivl_expr_parm(expr, sub_idx);
draw_eval_vec4(sube);
}
for ( sub_idx += 1 ; sub_idx < num_sube ; sub_idx += 1) {
/* Concatenate progressively lower parts. */
sube = ivl_expr_parm(expr, sub_idx);
/* Special case: The next expression is a NUMBER that
can be concatenated using %concati/vec4
instructions. */
if (ivl_expr_type(sube) == IVL_EX_NUMBER) {
draw_concat_number_vec4(sube, 1);
continue;
}
draw_eval_vec4(sube);
/* Evaluate, but skip any zero replication expressions in the
* rest of this concatenation. */
if ((ivl_expr_type(sube) == IVL_EX_CONCAT) &&
(ivl_expr_repeat(sube) == 0)) {
fprintf(vvp_out, " %%pop/vec4 1; skip zero replication\n");
continue;
}
fprintf(vvp_out, " %%concat/vec4; draw_concat_vec4\n");
}
if (repeat > 1) {
fprintf(vvp_out, " %%replicate %u;\n", repeat);
}
}
/*
* Push a number into the vec4 stack using %pushi/vec4
* instructions. The %pushi/vec4 instruction can only handle up to 32
* non-zero bits, so if there are more than that, then generate
* multiple %pushi/vec4 statements, and use %concat/vec4 statements to
* concatenate the vectors into the desired result.
*/
static void draw_number_vec4(ivl_expr_t expr)
{
draw_concat_number_vec4(expr, 0);
}
static void draw_property_vec4(ivl_expr_t expr)
{
ivl_signal_t sig = ivl_expr_signal(expr);
unsigned pidx = ivl_expr_property_idx(expr);
fprintf(vvp_out, " %%load/obj v%p_0;\n", sig);
fprintf(vvp_out, " %%prop/v %u;\n", pidx);
fprintf(vvp_out, " %%pop/obj 1, 0;\n");
}
static void draw_select_vec4(ivl_expr_t expr)
{
// This is the sub-expression to part-select.
ivl_expr_t subexpr = ivl_expr_oper1(expr);
// This is the base of the part select
ivl_expr_t base = ivl_expr_oper2(expr);
// This is the part select width
unsigned wid = ivl_expr_width(expr);
// Is the select base expression signed or unsigned?
char sign_suff = ivl_expr_signed(base)? 's' : 'u';
// Special Case: If the sub-expression is a STRING, then this
// is a select from that string.
if (ivl_expr_value(subexpr)==IVL_VT_STRING) {
assert(base);
assert(wid==8);
draw_eval_string(subexpr);
int base_idx = allocate_word();
draw_eval_expr_into_integer(base, base_idx);
fprintf(vvp_out, " %%substr/vec4 %d, %u;\n", base_idx, wid);
fprintf(vvp_out, " %%pop/str 1;\n");
clr_word(base_idx);
return;
}
if (ivl_expr_value(subexpr)==IVL_VT_DARRAY) {
ivl_signal_t sig = ivl_expr_signal(subexpr);
assert(sig);
assert( (ivl_signal_data_type(sig)==IVL_VT_DARRAY)
|| (ivl_signal_data_type(sig)==IVL_VT_QUEUE) );
assert(base);
draw_eval_expr_into_integer(base, 3);
fprintf(vvp_out, " %%load/dar/vec4 v%p_0;\n", sig);
if (ivl_expr_value(expr) == IVL_VT_BOOL)
fprintf(vvp_out, " %%cast2;\n");
return;
}
if (test_immediate_vec4_ok(base)) {
unsigned long val0, valx;
unsigned base_wid;
make_immediate_vec4_words(base, &val0, &valx, &base_wid);
assert(valx == 0);
draw_eval_vec4(subexpr);
fprintf(vvp_out, " %%parti/%c %u, %lu, %u;\n",
sign_suff, wid, val0, base_wid);
} else {
draw_eval_vec4(subexpr);
draw_eval_vec4(base);
fprintf(vvp_out, " %%part/%c %u;\n", sign_suff, wid);
}
}
static void draw_select_pad_vec4(ivl_expr_t expr)
{
// This is the sub-expression to pad/truncate
ivl_expr_t subexpr = ivl_expr_oper1(expr);
// This is the target width of the expression
unsigned wid = ivl_expr_width(expr);
// Push the sub-expression onto the stack.
draw_eval_vec4(subexpr);
// Special case: The expression is already the correct width,
// so there is nothing to be done.
if (wid == ivl_expr_width(subexpr))
return;
if (ivl_expr_signed(expr))
fprintf(vvp_out, " %%pad/s %u;\n", wid);
else
fprintf(vvp_out, " %%pad/u %u;\n", wid);
}
/*
* This function handles the special case of a call to the internal
* functions $ivl_queue_method$pop_back et al. The first (and only)
* argument is the signal that represents a dynamic queue. Generate a
* %qpop instruction to pop a value and push it to the vec4 stack.
*/
static void draw_darray_pop(ivl_expr_t expr)
{
const char*fb;
if (strcmp(ivl_expr_name(expr), "$ivl_queue_method$pop_back")==0)
fb = "b";
else
fb = "f";
ivl_expr_t arg = ivl_expr_parm(expr, 0);
assert(ivl_expr_type(arg) == IVL_EX_SIGNAL);
fprintf(vvp_out, " %%qpop/%s/v v%p_0, %u;\n", fb, ivl_expr_signal(arg),
ivl_expr_width(expr));
}
static void draw_sfunc_vec4(ivl_expr_t expr)
{
unsigned parm_count = ivl_expr_parms(expr);
/* Special case: If there are no arguments to print, then the
%vpi_call statement is easy to draw. */
if (parm_count == 0) {
assert(ivl_expr_value(expr)==IVL_VT_LOGIC
|| ivl_expr_value(expr)==IVL_VT_BOOL);
fprintf(vvp_out, " %%vpi_func %u %u \"%s\" %u {0 0 0};\n",
ivl_file_table_index(ivl_expr_file(expr)),
ivl_expr_lineno(expr), ivl_expr_name(expr),
ivl_expr_width(expr));
return;
}
if (strcmp(ivl_expr_name(expr), "$ivl_queue_method$pop_back")==0) {
draw_darray_pop(expr);
return;
}
if (strcmp(ivl_expr_name(expr),"$ivl_queue_method$pop_front")==0) {
draw_darray_pop(expr);
return;
}
draw_vpi_func_call(expr);
}
static void draw_signal_vec4(ivl_expr_t expr)
{
ivl_signal_t sig = ivl_expr_signal(expr);
/* Special Case: If the signal is the return value of the function,
then use a different opcode to get the value. */
if (signal_is_return_value(sig)) {
assert(ivl_signal_dimensions(sig) == 0);
fprintf(vvp_out, " %%retload/vec4 0; Load %s (draw_signal_vec4)\n",
ivl_signal_basename(sig));
return;
}
/* Handle the simple case, a signal expression that is a
simple vector, no array dimensions. */
if (ivl_signal_dimensions(sig) == 0) {
fprintf(vvp_out, " %%load/vec4 v%p_0;\n", sig);
return;
}
/* calculate the array index... */
int addr_index = allocate_word();
draw_eval_expr_into_integer(ivl_expr_oper1(expr), addr_index);
fprintf(vvp_out, " %%load/vec4a v%p, %d;\n", sig, addr_index);
clr_word(addr_index);
}
static void draw_string_vec4(ivl_expr_t expr)
{
unsigned wid = ivl_expr_width(expr);
char*fp = process_octal_codes(ivl_expr_string(expr), wid);
char*p = fp;
unsigned long tmp = 0;
unsigned tmp_wid = 0;
int push_flag = 0;
for (unsigned idx = 0 ; idx < wid ; idx += 8) {
tmp <<= 8;
tmp |= (unsigned long)*p;
p += 1;
tmp_wid += 8;
if (tmp_wid == 32) {
fprintf(vvp_out, " %%pushi/vec4 %lu, 0, 32; draw_string_vec4\n", tmp);
tmp = 0;
tmp_wid = 0;
if (push_flag == 0)
push_flag += 1;
else
fprintf(vvp_out, " %%concat/vec4; draw_string_vec4\n");
}
}
if (tmp_wid > 0) {
fprintf(vvp_out, " %%pushi/vec4 %lu, 0, %u; draw_string_vec4\n", tmp, tmp_wid);
if (push_flag != 0)
fprintf(vvp_out, " %%concat/vec4; draw_string_vec4\n");
}
free(fp);
}
static void draw_ternary_vec4(ivl_expr_t expr)
{
ivl_expr_t cond = ivl_expr_oper1(expr);
ivl_expr_t true_ex = ivl_expr_oper2(expr);
ivl_expr_t false_ex = ivl_expr_oper3(expr);
unsigned lab_true = local_count++;
unsigned lab_out = local_count++;
int use_flag = draw_eval_condition(cond);
/* The condition flag is used after possibly other statements,
so we need to put it into a non-common place. Allocate a
safe flag bit and move the condition to the flag position. */
if (use_flag < 8) {
int tmp_flag = allocate_flag();
assert(tmp_flag >= 8);
fprintf(vvp_out, " %%flag_mov %d, %d;\n", tmp_flag, use_flag);
use_flag = tmp_flag;
}
fprintf(vvp_out, " %%jmp/0 T_%u.%u, %d;\n", thread_count, lab_true, use_flag);
/* If the condition is true or xz (not false), we need the true
expression. If the condition is true, then we ONLY need the
true expression. */
draw_eval_vec4(true_ex);
fprintf(vvp_out, " %%jmp/1 T_%u.%u, %d;\n", thread_count, lab_out, use_flag);
fprintf(vvp_out, "T_%u.%u ; End of true expr.\n", thread_count, lab_true);
/* If the condition is false or xz (not true), we need the false
expression. If the condition is false, then we ONLY need
the false expression. */
draw_eval_vec4(false_ex);
fprintf(vvp_out, " %%jmp/0 T_%u.%u, %d;\n", thread_count, lab_out, use_flag);
fprintf(vvp_out, " ; End of false expr.\n");
/* Here, the condition is not true or false, it is xz. Both
the true and false expressions have been pushed onto the
stack, we just need to blend the bits. */
fprintf(vvp_out, " %%blend;\n");
fprintf(vvp_out, "T_%u.%u;\n", thread_count, lab_out);
clr_flag(use_flag);
}
static void draw_unary_inc_dec(ivl_expr_t sub, bool incr, bool pre)
{
ivl_signal_t sig = 0;
unsigned wid = 0;
switch (ivl_expr_type(sub)) {
case IVL_EX_SELECT: {
ivl_expr_t e1 = ivl_expr_oper1(sub);
sig = ivl_expr_signal(e1);
wid = ivl_expr_width(e1);
break;
}
case IVL_EX_SIGNAL:
sig = ivl_expr_signal(sub);
wid = ivl_expr_width(sub);
break;
default:
assert(0);
break;
}
draw_eval_vec4(sub);
const char*cmd = incr? "%add" : "%sub";
if (pre) {
/* prefix means we add the result first, and store the
result, as well as leaving a copy on the stack. */
fprintf(vvp_out, " %si 1, 0, %u;\n", cmd, wid);
fprintf(vvp_out, " %%dup/vec4;\n");
fprintf(vvp_out, " %%store/vec4 v%p_0, 0, %u;\n", sig, wid);
} else {
/* The post-fix decrement returns the non-decremented
version, so there is a slight re-arrange. */
fprintf(vvp_out, " %%dup/vec4;\n");
fprintf(vvp_out, " %si 1, 0, %u;\n", cmd, wid);
fprintf(vvp_out, " %%store/vec4 v%p_0, 0, %u;\n", sig, wid);
}
}
static void draw_unary_vec4(ivl_expr_t expr)
{
ivl_expr_t sub = ivl_expr_oper1(expr);
if (debug_draw) {
fprintf(vvp_out, " ; %s:%u:draw_unary_vec4: opcode=%c\n",
ivl_expr_file(expr), ivl_expr_lineno(expr),
ivl_expr_opcode(expr));
}
switch (ivl_expr_opcode(expr)) {
case '&':
draw_eval_vec4(sub);
fprintf(vvp_out, " %%and/r;\n");
break;
case '|':
draw_eval_vec4(sub);
fprintf(vvp_out, " %%or/r;\n");
break;
case '^':
draw_eval_vec4(sub);
fprintf(vvp_out, " %%xor/r;\n");
break;
case '~':
draw_eval_vec4(sub);
fprintf(vvp_out, " %%inv;\n");
break;
case '!':
draw_eval_vec4(sub);
fprintf(vvp_out, " %%nor/r;\n");
break;
case '-':
draw_eval_vec4(sub);
fprintf(vvp_out, " %%inv;\n");
fprintf(vvp_out, " %%addi 1, 0, %u;\n", ivl_expr_width(sub));
break;
case 'A': /* nand (~&) */
draw_eval_vec4(sub);
fprintf(vvp_out, " %%nand/r;\n");
break;
case 'D': /* pre-decrement (--x) */
draw_unary_inc_dec(sub, false, true);
break;
case 'd': /* post_decrement (x--) */
draw_unary_inc_dec(sub, false, false);
break;
case 'I': /* pre-increment (++x) */
draw_unary_inc_dec(sub, true, true);
break;
case 'i': /* post-increment (x++) */
draw_unary_inc_dec(sub, true, false);
break;
case 'N': /* nor (~|) */
draw_eval_vec4(sub);
fprintf(vvp_out, " %%nor/r;\n");
break;
case 'X': /* xnor (~^) */
draw_eval_vec4(sub);
fprintf(vvp_out, " %%xnor/r;\n");
break;
case 'm': /* abs(m) */
draw_eval_vec4(sub);
if (! ivl_expr_signed(sub))
break;
/* Test if (m) < 0 */
fprintf(vvp_out, " %%dup/vec4;\n");
fprintf(vvp_out, " %%cmpi/s 0, 0, %u;\n", ivl_expr_width(sub));
fprintf(vvp_out, " %%jmp/0xz T_%u.%u, 5;\n", thread_count, local_count);
/* If so, calculate -(m) */
fprintf(vvp_out, " %%inv;\n");
fprintf(vvp_out, " %%addi 1, 0, %u;\n", ivl_expr_width(sub));
fprintf(vvp_out, "T_%u.%u ;\n", thread_count, local_count);
local_count += 1;
break;
case 'v': /* Cast expression to vec4 */
switch (ivl_expr_value(sub)) {
case IVL_VT_REAL:
draw_eval_real(sub);
fprintf(vvp_out, " %%cvt/vr %u;\n", ivl_expr_width(expr));
break;
case IVL_VT_STRING:
draw_eval_string(sub);
fprintf(vvp_out, " %%cast/vec4/str %u;\n", ivl_expr_width(expr));
break;
case IVL_VT_DARRAY:
draw_eval_object(sub);
fprintf(vvp_out, " %%cast/vec4/dar %u;\n", ivl_expr_width(expr));
break;
default:
assert(0);
break;
}
break;
case '2': /* Cast expression to bool */
switch (ivl_expr_value(sub)) {
case IVL_VT_LOGIC:
draw_eval_vec4(sub);
fprintf(vvp_out, " %%cast2;\n");
resize_vec4_wid(sub, ivl_expr_width(expr));
break;
case IVL_VT_BOOL:
draw_eval_vec4(sub);
resize_vec4_wid(sub, ivl_expr_width(expr));
break;
case IVL_VT_REAL:
draw_eval_real(sub);
fprintf(vvp_out, " %%cvt/vr %u;\n", ivl_expr_width(expr));
break;
case IVL_VT_STRING:
draw_eval_string(sub);
fprintf(vvp_out, " %%cast/vec4/str %u;\n", ivl_expr_width(expr));
break;
case IVL_VT_DARRAY:
draw_eval_object(sub);
fprintf(vvp_out, " %%cast/vec2/dar %u;\n", ivl_expr_width(expr));
break;
default:
assert(0);
break;
}
break;
default:
fprintf(stderr, "XXXX Unary operator %c not implemented\n", ivl_expr_opcode(expr));
break;
}
}
void draw_eval_vec4(ivl_expr_t expr)
{
if (debug_draw) {
fprintf(vvp_out, " ; %s:%u:draw_eval_vec4: expr_type=%d\n",
ivl_expr_file(expr), ivl_expr_lineno(expr),
ivl_expr_type(expr));
}
assert(ivl_expr_value(expr) == IVL_VT_BOOL ||
ivl_expr_value(expr) == IVL_VT_VECTOR);
switch (ivl_expr_type(expr)) {
case IVL_EX_BINARY:
draw_binary_vec4(expr);
return;
case IVL_EX_CONCAT:
draw_concat_vec4(expr);
return;
case IVL_EX_NUMBER:
draw_number_vec4(expr);
return;
case IVL_EX_SELECT:
if (ivl_expr_oper2(expr)==0)
draw_select_pad_vec4(expr);
else
draw_select_vec4(expr);
return;
case IVL_EX_SFUNC:
draw_sfunc_vec4(expr);
return;
case IVL_EX_SIGNAL:
draw_signal_vec4(expr);
return;
case IVL_EX_STRING:
draw_string_vec4(expr);
return;
case IVL_EX_TERNARY:
draw_ternary_vec4(expr);
return;
case IVL_EX_UFUNC:
draw_ufunc_vec4(expr);
return;
case IVL_EX_UNARY:
draw_unary_vec4(expr);
return;
case IVL_EX_PROPERTY:
draw_property_vec4(expr);
return;
case IVL_EX_NULL:
fprintf(stderr, "%s:%u: Error: 'null' used in an expression\n",
ivl_expr_file(expr), ivl_expr_lineno(expr));
fprintf(vvp_out, "; Error 'null' used in an expression\n");
vvp_errors += 1;
return;
default:
break;
}
fprintf(stderr, "%s:%u: Sorry: cannot evaluate VEC4 expression (%d)\n",
ivl_expr_file(expr), ivl_expr_lineno(expr),
ivl_expr_type(expr));
fprintf(vvp_out, "; Sorry: cannot evaluate VEC4 expression (%d)\n",
ivl_expr_type(expr));
vvp_errors += 1;
}