quakeforge/libs/gamecode/test/test-float.c
Bill Currie 8050c7bd77 [gamecode] Rearrange several instructions
ANY/ALL/NONE have been temporarily removed until I implement the HOPS
(horizontal operations) sub-instructions, which will all both 32-bit and
64-bit operands and several other operations (eg, horizontal add).

All the fancy addressing modes for the conditional branch instructions
have been permanently removed: I decided the gain was too little for the
cost (24 instructions vs 6). JUMP and CALL retain their addressing
modes, though.

Other instructions have been shuffled around a little to fill most of
the holes in the upper block of 256 instructions: just a single small
7-instruction hole.

Rearrangements in the actual engine are mostly just to keep the code
organized. The only real changes were the various IF statements and
dealing with the resulting changes in their addressing.
2022-01-16 14:22:04 +09:00

333 lines
11 KiB
C

#include "head.c"
#include "QF/mathlib.h"
#define sq(x) ((float)(x)*(float)(x))
static pr_vec4_t float_binop_init[] = {
{ 5, -5, 5, -5},
{ 3, 3, -3, -3},
{ 0, 0, 0, 0},
{ 0, 0, 0, 0},
{ 0, 0, 0, 0},
{ 0, 0, 0, 0},
{ 0, 0, 0, 0},
{ 0, 0, 0, 0},
};
static pr_vec4_t float_binop_expect[] = {
{ 5, -5, 5, -5},
{ 3, 3, -3, -3},
{ 15, -15, -15, 15},
{ 1.666666627, -1.666666627, -1.666666627, 1.666666627},
{ 2, -2, 2, -2},
{ 2, 1, -1, -2},
{ 8, -2, 2, -8},
{ 2, -8, 8, -2},
};
static dstatement_t float_binop_1_statements[] = {
{ OP(0, 0, 0, OP_LEA_A), 4, 0, 32 }, // init index
//loop:
{ OP(0, 0, 0, OP_LEA_C), 32, -1, 32 }, // dec index
{ OP(0, 0, 0, OP_IFAE), 2, 0, 32 },
{ OP(0, 0, 0, OP_BREAK), 0, 0, 0 },
{ OP(0, 0, 0, OP_WITH), 4, 32, 1 },
{ OP(1, 1, 1, OP_MUL_F_1), 0, 4, 8 },
{ OP(1, 1, 1, OP_DIV_F_1), 0, 4, 12 },
{ OP(1, 1, 1, OP_REM_F_1), 0, 4, 16 },
{ OP(1, 1, 1, OP_MOD_F_1), 0, 4, 20 },
{ OP(1, 1, 1, OP_ADD_F_1), 0, 4, 24 },
{ OP(1, 1, 1, OP_SUB_F_1), 0, 4, 28 },
{ OP(1, 1, 1, OP_JUMP_A), -10, 0, 0 },
};
static dstatement_t float_binop_2_statements[] = {
{ OP(0, 0, 0, OP_LEA_A), 4, 0, 32 }, // index
//loop:
{ OP(0, 0, 0, OP_LEA_C), 32, -2, 32 }, // dec index
{ OP(0, 0, 0, OP_IFAE), 2, 0, 32 },
{ OP(0, 0, 0, OP_BREAK), 0, 0, 0 },
{ OP(0, 0, 0, OP_WITH), 4, 32, 1 },
{ OP(1, 1, 1, OP_MUL_F_2), 0, 4, 8 },
{ OP(1, 1, 1, OP_DIV_F_2), 0, 4, 12 },
{ OP(1, 1, 1, OP_REM_F_2), 0, 4, 16 },
{ OP(1, 1, 1, OP_MOD_F_2), 0, 4, 20 },
{ OP(1, 1, 1, OP_ADD_F_2), 0, 4, 24 },
{ OP(1, 1, 1, OP_SUB_F_2), 0, 4, 28 },
{ OP(1, 1, 1, OP_JUMP_A), -10, 0, 0 },
};
static dstatement_t float_binop_3a_statements[] = {
{ OP(1, 1, 1, OP_MUL_F_3), 0, 4, 8 },
{ OP(1, 1, 1, OP_MUL_F_1), 3, 7, 11 },
{ OP(1, 1, 1, OP_DIV_F_3), 0, 4, 12 },
{ OP(1, 1, 1, OP_DIV_F_1), 3, 7, 15 },
{ OP(1, 1, 1, OP_REM_F_3), 0, 4, 16 },
{ OP(1, 1, 1, OP_REM_F_1), 3, 7, 19 },
{ OP(1, 1, 1, OP_MOD_F_3), 0, 4, 20 },
{ OP(1, 1, 1, OP_MOD_F_1), 3, 7, 23 },
{ OP(1, 1, 1, OP_ADD_F_3), 0, 4, 24 },
{ OP(1, 1, 1, OP_ADD_F_1), 3, 7, 27 },
{ OP(1, 1, 1, OP_SUB_F_3), 0, 4, 28 },
{ OP(1, 1, 1, OP_SUB_F_1), 3, 7, 31 },
};
static dstatement_t float_binop_3b_statements[] = {
{ OP(1, 1, 1, OP_MUL_F_1), 0, 4, 8 },
{ OP(1, 1, 1, OP_MUL_F_3), 1, 5, 9 },
{ OP(1, 1, 1, OP_DIV_F_1), 0, 4, 12 },
{ OP(1, 1, 1, OP_DIV_F_3), 1, 5, 13 },
{ OP(1, 1, 1, OP_REM_F_1), 0, 4, 16 },
{ OP(1, 1, 1, OP_REM_F_3), 1, 5, 17 },
{ OP(1, 1, 1, OP_MOD_F_1), 0, 4, 20 },
{ OP(1, 1, 1, OP_MOD_F_3), 1, 5, 21 },
{ OP(1, 1, 1, OP_ADD_F_1), 0, 4, 24 },
{ OP(1, 1, 1, OP_ADD_F_3), 1, 5, 25 },
{ OP(1, 1, 1, OP_SUB_F_1), 0, 4, 28 },
{ OP(1, 1, 1, OP_SUB_F_3), 1, 5, 29 },
};
static dstatement_t float_binop_4_statements[] = {
{ OP(1, 1, 1, OP_MUL_F_4), 0, 4, 8 },
{ OP(1, 1, 1, OP_DIV_F_4), 0, 4, 12 },
{ OP(1, 1, 1, OP_REM_F_4), 0, 4, 16 },
{ OP(1, 1, 1, OP_MOD_F_4), 0, 4, 20 },
{ OP(1, 1, 1, OP_ADD_F_4), 0, 4, 24 },
{ OP(1, 1, 1, OP_SUB_F_4), 0, 4, 28 },
};
static pr_vec4_t float_cossin_init[] = {
{ 1, 2, 3, 4 }, // 0: output
{ M_PI/6, 0, 0, 0 }, // 4: x
{ 1, 2, 0, 0 }, // 8: f
{ 1, 1, 0, 25 }, // 12: f inc and f0 max
{ 0, 0, 0, 0 }, // 16: x2 -> [xx, xx]
// { } // 20: xn
};
static pr_vec4_t float_cossin_expect[] = {
{ 0.866025388, 0.5, 0, 0 }, // 0: output
{ M_PI/6, 0, 0, 0 }, // 4: x
{ 25, 26, 0, 0 }, // 8: f
{ 1, 1, 0, 25 }, // 12: f inc and f0 max
{ -sq(M_PI/6), -sq(M_PI/6), 0, 0 }, // 16: x2 -> [xx, xx]
};
static dstatement_t float_cossin_statements[] = {
{ OP(0, 0, 0, OP_STORE_A_1), 21, 0, 4 }, // init xn -> [?, x]
{ OP(0, 0, 0, OP_STORE_A_1), 20, 0, 8 }, // init xn -> [1, x]
{ OP(0, 0, 0, OP_SWIZZLE_F), 4, 0xc000, 16 },// init x2 -> [x, x, 0, 0]
{ OP(0, 0, 0, OP_MUL_F_2), 16, 16, 16 }, // x2 -> [x*x, x*x, 0, 0]
{ OP(0, 0, 0, OP_SWIZZLE_F), 16, 0xc3e4, 16 },// init x2 -> -x2
{ OP(0, 0, 0, OP_SUB_F_4), 0, 0, 0 }, // init acc (output) to 0
// loop:
{ OP(0, 0, 0, OP_ADD_F_2), 0, 20, 0 }, // acc += xn
{ OP(0, 0, 0, OP_MUL_F_2), 20, 16, 20 }, // xn *= x2
{ OP(0, 0, 0, OP_DIV_F_2), 20, 8, 20 }, // xn /= f
{ OP(0, 0, 0, OP_ADD_F_2), 8, 12, 8 }, // f += inc
{ OP(0, 0, 0, OP_DIV_F_2), 20, 8, 20 }, // xn /= f
{ OP(0, 0, 0, OP_ADD_F_2), 8, 12, 8 }, // f += inc
{ OP(0, 0, 0, OP_LT_F_1), 8, 15, 23 }, // f0 < fmax
{ OP(0, 0, 0, OP_IFNZ), -7, 0, 23 }, // f0 < fmax
};
static pr_vec4_t float_cmpop_init[] = {
{ 5, -5, 5, -5},
{ 5, 5, -5, -5},
{ 0, 0, 0, 0},
{ 0, 0, 0, 0},
{ 0, 0, 0, 0},
{ 0, 0, 0, 0},
{ 0, 0, 0, 0},
{ 0, 0, 0, 0},
};
// 5.0 as 32-bit int
#define F 0x40a00000
#define mF 0xc0a00000
static pr_ivec4_t float_cmpop_expect[] = {
{ F, mF, F, mF},
{ F, F, mF, mF},
{ -1, 0, 0, -1},
{ 0, -1, 0, 0},
{ 0, 0, -1, 0},
{ 0, -1, -1, 0},
{ -1, 0, -1, -1},
{ -1, -1, 0, -1},
};
static dstatement_t float_cmpop_1_statements[] = {
{ OP(0, 0, 0, OP_LEA_A), 4, 0, 32 }, // init index
//loop:
{ OP(0, 0, 0, OP_LEA_C), 32, -1, 32 }, // dec index
{ OP(0, 0, 0, OP_IFAE), 2, 0, 32 },
{ OP(0, 0, 0, OP_BREAK), 0, 0, 0 },
{ OP(0, 0, 0, OP_WITH), 4, 32, 1 },
{ OP(1, 1, 1, OP_EQ_F_1), 0, 4, 8 },
{ OP(1, 1, 1, OP_LT_F_1), 0, 4, 12 },
{ OP(1, 1, 1, OP_GT_F_1), 0, 4, 16 },
{ OP(1, 1, 1, OP_NE_F_1), 0, 4, 20 },
{ OP(1, 1, 1, OP_GE_F_1), 0, 4, 24 },
{ OP(1, 1, 1, OP_LE_F_1), 0, 4, 28 },
{ OP(1, 1, 1, OP_JUMP_A), -10, 0, 0 },
};
static dstatement_t float_cmpop_2_statements[] = {
{ OP(0, 0, 0, OP_LEA_A), 4, 0, 32 }, // index
//loop:
{ OP(0, 0, 0, OP_LEA_C), 32, -2, 32 }, // dec index
{ OP(0, 0, 0, OP_IFAE), 2, 0, 32 },
{ OP(0, 0, 0, OP_BREAK), 0, 0, 0 },
{ OP(0, 0, 0, OP_WITH), 4, 32, 1 },
{ OP(1, 1, 1, OP_EQ_F_2), 0, 4, 8 },
{ OP(1, 1, 1, OP_LT_F_2), 0, 4, 12 },
{ OP(1, 1, 1, OP_GT_F_2), 0, 4, 16 },
{ OP(1, 1, 1, OP_NE_F_2), 0, 4, 20 },
{ OP(1, 1, 1, OP_GE_F_2), 0, 4, 24 },
{ OP(1, 1, 1, OP_LE_F_2), 0, 4, 28 },
{ OP(1, 1, 1, OP_JUMP_A), -10, 0, 0 },
};
static dstatement_t float_cmpop_3a_statements[] = {
{ OP(1, 1, 1, OP_EQ_F_3), 0, 4, 8 },
{ OP(1, 1, 1, OP_EQ_F_1), 3, 7, 11 },
{ OP(1, 1, 1, OP_LT_F_3), 0, 4, 12 },
{ OP(1, 1, 1, OP_LT_F_1), 3, 7, 15 },
{ OP(1, 1, 1, OP_GT_F_3), 0, 4, 16 },
{ OP(1, 1, 1, OP_GT_F_1), 3, 7, 19 },
{ OP(1, 1, 1, OP_NE_F_3), 0, 4, 20 },
{ OP(1, 1, 1, OP_NE_F_1), 3, 7, 23 },
{ OP(1, 1, 1, OP_GE_F_3), 0, 4, 24 },
{ OP(1, 1, 1, OP_GE_F_1), 3, 7, 27 },
{ OP(1, 1, 1, OP_LE_F_3), 0, 4, 28 },
{ OP(1, 1, 1, OP_LE_F_1), 3, 7, 31 },
};
static dstatement_t float_cmpop_3b_statements[] = {
{ OP(1, 1, 1, OP_EQ_F_1), 0, 4, 8 },
{ OP(1, 1, 1, OP_EQ_F_3), 1, 5, 9 },
{ OP(1, 1, 1, OP_LT_F_1), 0, 4, 12 },
{ OP(1, 1, 1, OP_LT_F_3), 1, 5, 13 },
{ OP(1, 1, 1, OP_GT_F_1), 0, 4, 16 },
{ OP(1, 1, 1, OP_GT_F_3), 1, 5, 17 },
{ OP(1, 1, 1, OP_NE_F_1), 0, 4, 20 },
{ OP(1, 1, 1, OP_NE_F_3), 1, 5, 21 },
{ OP(1, 1, 1, OP_GE_F_1), 0, 4, 24 },
{ OP(1, 1, 1, OP_GE_F_3), 1, 5, 25 },
{ OP(1, 1, 1, OP_LE_F_1), 0, 4, 28 },
{ OP(1, 1, 1, OP_LE_F_3), 1, 5, 29 },
};
static dstatement_t float_cmpop_4_statements[] = {
{ OP(1, 1, 1, OP_EQ_F_4), 0, 4, 8 },
{ OP(1, 1, 1, OP_LT_F_4), 0, 4, 12 },
{ OP(1, 1, 1, OP_GT_F_4), 0, 4, 16 },
{ OP(1, 1, 1, OP_NE_F_4), 0, 4, 20 },
{ OP(1, 1, 1, OP_GE_F_4), 0, 4, 24 },
{ OP(1, 1, 1, OP_LE_F_4), 0, 4, 28 },
};
test_t tests[] = {
{
.desc = "float binop 1",
.extra_globals = 4 * 1,
.num_globals = num_globals(float_binop_init,float_binop_expect),
.num_statements = num_statements (float_binop_1_statements),
.statements = float_binop_1_statements,
.init_globals = (pr_int_t *) float_binop_init,
.expect_globals = (pr_int_t *) float_binop_expect,
},
{
.desc = "float binop 2",
.extra_globals = 4 * 1,
.num_globals = num_globals(float_binop_init,float_binop_expect),
.num_statements = num_statements (float_binop_2_statements),
.statements = float_binop_2_statements,
.init_globals = (pr_int_t *) float_binop_init,
.expect_globals = (pr_int_t *) float_binop_expect,
},
{
.desc = "float binop 3a",
.extra_globals = 4 * 1,
.num_globals = num_globals(float_binop_init,float_binop_expect),
.num_statements = num_statements (float_binop_3a_statements),
.statements = float_binop_3a_statements,
.init_globals = (pr_int_t *) float_binop_init,
.expect_globals = (pr_int_t *) float_binop_expect,
},
{
.desc = "float binop 3b",
.extra_globals = 4 * 1,
.num_globals = num_globals(float_binop_init,float_binop_expect),
.num_statements = num_statements (float_binop_3b_statements),
.statements = float_binop_3b_statements,
.init_globals = (pr_int_t *) float_binop_init,
.expect_globals = (pr_int_t *) float_binop_expect,
},
{
.desc = "float binop 4",
.extra_globals = 4 * 1,
.num_globals = num_globals(float_binop_init,float_binop_expect),
.num_statements = num_statements (float_binop_4_statements),
.statements = float_binop_4_statements,
.init_globals = (pr_int_t *) float_binop_init,
.expect_globals = (pr_int_t *) float_binop_expect,
},
{
.desc = "float cos sin",
.extra_globals = 4 * 1,
.num_globals = num_globals (float_cossin_init, float_cossin_expect),
.num_statements = num_statements (float_cossin_statements),
.statements = float_cossin_statements,
.init_globals = (pr_int_t *) float_cossin_init,
.expect_globals = (pr_int_t *) float_cossin_expect,
},
{
.desc = "float cmpop 1",
.extra_globals = 4 * 1,
.num_globals = num_globals(float_cmpop_init,float_cmpop_expect),
.num_statements = num_statements (float_cmpop_1_statements),
.statements = float_cmpop_1_statements,
.init_globals = (pr_int_t *) float_cmpop_init,
.expect_globals = (pr_int_t *) float_cmpop_expect,
},
{
.desc = "float cmpop 2",
.extra_globals = 4 * 1,
.num_globals = num_globals(float_cmpop_init,float_cmpop_expect),
.num_statements = num_statements (float_cmpop_2_statements),
.statements = float_cmpop_2_statements,
.init_globals = (pr_int_t *) float_cmpop_init,
.expect_globals = (pr_int_t *) float_cmpop_expect,
},
{
.desc = "float cmpop 3a",
.extra_globals = 4 * 1,
.num_globals = num_globals(float_cmpop_init,float_cmpop_expect),
.num_statements = num_statements (float_cmpop_3a_statements),
.statements = float_cmpop_3a_statements,
.init_globals = (pr_int_t *) float_cmpop_init,
.expect_globals = (pr_int_t *) float_cmpop_expect,
},
{
.desc = "float cmpop 3b",
.extra_globals = 4 * 1,
.num_globals = num_globals(float_cmpop_init,float_cmpop_expect),
.num_statements = num_statements (float_cmpop_3b_statements),
.statements = float_cmpop_3b_statements,
.init_globals = (pr_int_t *) float_cmpop_init,
.expect_globals = (pr_int_t *) float_cmpop_expect,
},
{
.desc = "float cmpop 4",
.extra_globals = 4 * 1,
.num_globals = num_globals(float_cmpop_init,float_cmpop_expect),
.num_statements = num_statements (float_cmpop_4_statements),
.statements = float_cmpop_4_statements,
.init_globals = (pr_int_t *) float_cmpop_init,
.expect_globals = (pr_int_t *) float_cmpop_expect,
},
};
#include "main.c"