void matrix_multiply2()
{
int i, j;
v2sf m1, m2;
v2sf result, temp;
for (i = 0; i < 4; i++)
{
result = (v2sf) {0.0, 0.0};
for (j = 0; j < 4; j+=2)
{
/* Load two float values into m1 */
m1 = (v2sf) {a[j], a[j+1]};
m2 = (v2sf) {b[j][i], b[j+1][i]};
/* Multiply and add */
result += m1 * m2;
}
/* Reduction add at the end */
temp = __builtin_mips_addr_ps (result, result);
d[i] = __builtin_mips_cvt_s_pl (temp);
}
}
void matrix_multiply4()
{
v2sf m1, m2;
v2sf n1, n2, n3, n4, n5, n6, n7, n8;
v2sf temp1, temp2, temp3, temp4;
v2sf result1, result2;
/* Load a[0] a[1] values into m1
Load a[2] a[3] values into m2 */
m1 = (v2sf) {a[0], a[1]};
m2 = (v2sf) {a[2], a[3]};
/* Load b[0][0] b[1][0] values into n1
Load b[2][0] b[3][0] values into n2
Load b[0][1] b[1][1] values into n3
Load b[2][1] b[3][1] values into n4
Load b[0][2] b[1][2] values into n5
Load b[2][2] b[3][2] values into n6
Load b[0][3] b[1][3] values into n7
Load b[2][3] b[3][3] values into n8 */
n1 = (v2sf) {b[0][0], b[1][0]};
n2 = (v2sf) {b[2][0], b[3][0]};
n3 = (v2sf) {b[0][1], b[1][1]};
n4 = (v2sf) {b[2][1], b[3][1]};
n5 = (v2sf) {b[0][2], b[1][2]};
n6 = (v2sf) {b[2][2], b[3][2]};
n7 = (v2sf) {b[0][3], b[1][3]};
n8 = (v2sf) {b[2][3], b[3][3]};
temp1 = m1 * n1 + m2 * n2;
temp2 = m1 * n3 + m2 * n4;
temp3 = m1 * n5 + m2 * n6;
temp4 = m1 * n7 + m2 * n8;
result1 = __builtin_mips_addr_ps (temp1, temp2);
result2 = __builtin_mips_addr_ps (temp3, temp4);
f[0] = __builtin_mips_cvt_s_pu (result1);
f[1] = __builtin_mips_cvt_s_pl (result1);
f[2] = __builtin_mips_cvt_s_pu (result2);
f[3] = __builtin_mips_cvt_s_pl (result2);
}
void matrix_multiply3()
{
int i;
v2sf m1, m2, n1, n2;
v2sf result, temp;
m1 = (v2sf) {a[0], a[1]};
m2 = (v2sf) {a[2], a[3]};
for (i = 0; i < 4; i++)
{
n1 = (v2sf) {b[0][i], b[1][i]};
n2 = (v2sf) {b[2][i], b[3][i]};
/* Multiply and add */
result = m1 * n1 + m2 * n2;
/* Reduction add at the end */
temp = __builtin_mips_addr_ps (result, result);
e[i] = __builtin_mips_cvt_s_pl (temp);
}
}
NOMIPS16 int main ()
{
int little_endian;
v2sf a, b, c, d;
float e,f;
int i;
union { long long ll; int i[2]; } endianness_test;
endianness_test.ll = 1;
little_endian = endianness_test.i[0];
/* pll.ps */
a = (v2sf) {1, 2};
b = (v2sf) {3, 4};
c = __builtin_mips_pll_ps (a, b);
if (little_endian) // little endian
d = (v2sf) {3, 1};
else // big endian
d = (v2sf) {2, 4};
if (!__builtin_mips_upper_c_eq_ps (c, d) ||
!__builtin_mips_lower_c_eq_ps (c, d))
abort ();
/* pul.ps */
a = (v2sf) {1, 2};
b = (v2sf) {3, 4};
c = __builtin_mips_pul_ps (a, b);
if (little_endian) // little endian
d = (v2sf) {3, 2};
else // big endian
d = (v2sf) {1, 4};
if (!__builtin_mips_upper_c_eq_ps (c, d) ||
!__builtin_mips_lower_c_eq_ps (c, d))
abort ();
/* plu.ps */
a = (v2sf) {1, 2};
b = (v2sf) {3, 4};
c = __builtin_mips_plu_ps (a, b);
if (little_endian) // little endian
d = (v2sf) {4, 1};
else // big endian
d = (v2sf) {2, 3};
if (!__builtin_mips_upper_c_eq_ps (c, d) ||
!__builtin_mips_lower_c_eq_ps (c, d))
abort ();
/* puu.ps */
a = (v2sf) {1, 2};
b = (v2sf) {3, 4};
c = __builtin_mips_puu_ps (a, b);
if (little_endian) // little endian
d = (v2sf) {4, 2};
else // big endian
d = (v2sf) {1, 3};
if (!__builtin_mips_upper_c_eq_ps (c, d) ||
!__builtin_mips_lower_c_eq_ps (c, d))
abort ();
/* cvt.ps.s */
e = 3.4;
f = 4.5;
a = __builtin_mips_cvt_ps_s (e, f);
if (little_endian) // little endian
b = (v2sf) {4.5, 3.4};
else // big endian
b = (v2sf) {3.4, 4.5};
if (!__builtin_mips_upper_c_eq_ps (a, b) ||
!__builtin_mips_lower_c_eq_ps (a, b))
abort ();
/* cvt.s.pl */
a = (v2sf) {35.1, 120.2};
e = __builtin_mips_cvt_s_pl (a);
if (little_endian) // little endian
f = 35.1;
else // big endian
f = 120.2;
if (e != f)
abort ();
/* cvt.s.pu */
a = (v2sf) {30.0, 100.0};
e = __builtin_mips_cvt_s_pu (a);
if (little_endian) // little endian
f = 100.0;
else // big endian
f = 30.0;
if (e != f)
abort ();
/* abs.ps */
a = (v2sf) {-3.4, -5.8};
b = __builtin_mips_abs_ps (a);
c = (v2sf) {3.4, 5.8};
if (!__builtin_mips_upper_c_eq_ps (b, c) ||
!__builtin_mips_lower_c_eq_ps (b, c))
abort ();
/* alnv.ps with rs = 4*/
a = (v2sf) {1, 2};
b = (v2sf) {3, 4};
i = 4;
c = __builtin_mips_alnv_ps (a, b, i);
d = (v2sf) {2, 3};
if (!__builtin_mips_upper_c_eq_ps (c, d) ||
!__builtin_mips_lower_c_eq_ps (c, d))
abort ();
/* alnv.ps with rs = 0 */
a = (v2sf) {5, 6};
b = (v2sf) {7, 8};
i = 0;
c = __builtin_mips_alnv_ps (a, b, i);
d = (v2sf) {5, 6};
if (!__builtin_mips_upper_c_eq_ps (c, d) ||
!__builtin_mips_lower_c_eq_ps (c, d))
abort ();
printf ("Test Passes\n");
exit (0);
}
