--- /dev/null
+# mips test sanity, expected to pass.
+# mach: mips64 sb1
+# as: -mabi=eabi
+# ld: -N -Ttext=0x80010000
+# output: *\\npass\\n
+
+ .include "testutils.inc"
+
+ .macro check_ps psval, upperval, lowerval
+ .set push
+ .set noreorder
+ cvt.s.pu $f0, \psval # upper
+ cvt.s.pl $f2, \psval # lower
+ li.s $f4, \upperval
+ li.s $f6, \lowerval
+ c.eq.s $fcc0, $f0, $f4
+ bc1f $fcc0, _fail
+ c.eq.s $fcc0, $f2, $f6
+ bc1f $fcc0, _fail
+ nop
+ .set pop
+ .endm
+
+ setup
+
+ .set noreorder
+
+ .ent DIAG
+DIAG:
+
+ # make sure that Status.FR and .CU1 are set.
+ mfc0 $2, $12
+ or $2, $2, (1 << 26) | (1 << 29)
+ mtc0 $2, $12
+
+
+ writemsg "ldc1"
+
+ .data
+1: .dword 0xc1a8000042200000 # -21.0, 40.0
+ .text
+ la $2, 1b
+ ldc1 $f8, 0($2)
+ check_ps $f8, -21.0, 40.0
+
+
+ writemsg "cvt.ps.s"
+
+ li.s $f10, 1.0
+ li.s $f12, 3.0
+ cvt.ps.s $f8, $f10, $f12 # upper, lower
+ check_ps $f8, 1.0, 3.0
+
+
+ writemsg "cvt.ps.s, sdc1, copy, ldc1"
+
+ .data
+1: .dword 0
+ .dword 0
+ .text
+ la $2, 1b
+ li.s $f12, -4.0
+ li.s $f14, 32.0
+ cvt.ps.s $f10, $f12, $f14 # upper, lower
+ sdc1 $f10, 8($2)
+ lw $3, 8($2)
+ lw $4, 12($2)
+ sw $3, 0($2)
+ sw $4, 4($2)
+ ldc1 $f8, 0($2)
+ check_ps $f8, -4.0, 32.0
+
+
+ # Load some constants for later use
+
+ li.s $f10, 4.0
+ li.s $f12, 16.0
+ cvt.ps.s $f20, $f10, $f12 # $f20: u=4.0, l=16.0
+
+ li.s $f10, -1.0
+ li.s $f12, 2.0
+ cvt.ps.s $f22, $f10, $f12 # $f22: u=-1.0, l=2.0
+
+ li.s $f10, 17.0
+ li.s $f12, -8.0
+ cvt.ps.s $f24, $f10, $f12 # $f24: u=17.0, l=-8.0
+
+
+ writemsg "pll.ps"
+
+ pll.ps $f8, $f20, $f22
+ check_ps $f8, 16.0, 2.0
+
+
+ writemsg "plu.ps"
+
+ plu.ps $f8, $f20, $f22
+ check_ps $f8, 16.0, -1.0
+
+
+ writemsg "pul.ps"
+
+ pul.ps $f8, $f20, $f22
+ check_ps $f8, 4.0, 2.0
+
+
+ writemsg "puu.ps"
+
+ puu.ps $f8, $f20, $f22
+ check_ps $f8, 4.0, -1.0
+
+
+ writemsg "abs.ps"
+
+ abs.ps $f8, $f22
+ check_ps $f8, 1.0, 2.0
+
+
+ writemsg "mov.ps"
+
+ mov.ps $f8, $f22
+ check_ps $f8, -1.0, 2.0
+
+
+ writemsg "neg.ps"
+
+ neg.ps $f8, $f22
+ check_ps $f8, 1.0, -2.0
+
+
+ writemsg "add.ps"
+
+ add.ps $f8, $f20, $f22
+ check_ps $f8, 3.0, 18.0
+
+
+ writemsg "mul.ps"
+
+ mul.ps $f8, $f20, $f22
+ check_ps $f8, -4.0, 32.0
+
+
+ writemsg "sub.ps"
+
+ sub.ps $f8, $f20, $f22
+ check_ps $f8, 5.0, 14.0
+
+
+ writemsg "madd.ps"
+
+ madd.ps $f8, $f24, $f20, $f22
+ check_ps $f8, 13.0, 24.0
+
+
+ writemsg "msub.ps"
+
+ msub.ps $f8, $f24, $f20, $f22
+ check_ps $f8, -21.0, 40.0
+
+
+ writemsg "nmadd.ps"
+
+ nmadd.ps $f8, $f24, $f20, $f22
+ check_ps $f8, -13.0, -24.0
+
+
+ writemsg "nmsub.ps"
+
+ nmsub.ps $f8, $f24, $f20, $f22
+ check_ps $f8, 21.0, -40.0
+
+
+ writemsg "movn.ps (n)"
+
+ li $2, 0
+ mov.ps $f8, $f20
+ movn.ps $f8, $f22, $2 # doesn't move
+ check_ps $f8, 4.0, 16.0
+
+
+ writemsg "movn.ps (y)"
+
+ li $2, 1
+ mov.ps $f8, $f20
+ movn.ps $f8, $f22, $2 # does move
+ check_ps $f8, -1.0, 2.0
+
+
+ writemsg "movz.ps (y)"
+
+ li $2, 0
+ mov.ps $f8, $f20
+ movz.ps $f8, $f22, $2 # does move
+ check_ps $f8, -1.0, 2.0
+
+
+ writemsg "movz.ps (n)"
+
+ li $2, 1
+ mov.ps $f8, $f20
+ movz.ps $f8, $f22, $2 # doesn't move
+ check_ps $f8, 4.0, 16.0
+
+
+ writemsg "movf.ps (y,y)"
+
+ cfc1 $2, $31
+ or $2, $2, (1 << 23) | (1 << 25)
+ xor $2, $2, (1 << 23) | (1 << 25)
+ ctc1 $2, $31 # clear fcc0, clear fcc1
+ mov.ps $f8, $f20
+ movf.ps $f8, $f22, $fcc0 # moves both halves
+ check_ps $f8, -1.0, 2.0
+
+
+ writemsg "movf.ps (y,n)"
+
+ cfc1 $2, $31
+ or $2, $2, (1 << 23) | (1 << 25)
+ xor $2, $2, (0 << 23) | (1 << 25)
+ ctc1 $2, $31 # set fcc0, clear fcc1
+ mov.ps $f8, $f20
+ movf.ps $f8, $f22, $fcc0 # moves upper half only
+ check_ps $f8, -1.0, 16.0
+
+
+ writemsg "movf.ps (n,y)"
+
+ cfc1 $2, $31
+ or $2, $2, (1 << 23) | (1 << 25)
+ xor $2, $2, (1 << 23) | (0 << 25)
+ ctc1 $2, $31 # clear fcc0, set fcc1
+ mov.ps $f8, $f20
+ movf.ps $f8, $f22, $fcc0 # moves lower half only
+ check_ps $f8, 4.0, 2.0
+
+
+ writemsg "movf.ps (n,n)"
+
+ cfc1 $2, $31
+ or $2, $2, (1 << 23) | (1 << 25)
+ xor $2, $2, (0 << 23) | (0 << 25)
+ ctc1 $2, $31 # set fcc0, set fcc1
+ mov.ps $f8, $f20
+ movf.ps $f8, $f22, $fcc0 # doesn't move either half
+ check_ps $f8, 4.0, 16.0
+
+
+ writemsg "movt.ps (n,n)"
+
+ cfc1 $2, $31
+ or $2, $2, (1 << 23) | (1 << 25)
+ xor $2, $2, (1 << 23) | (1 << 25)
+ ctc1 $2, $31 # clear fcc0, clear fcc1
+ mov.ps $f8, $f20
+ movt.ps $f8, $f22, $fcc0 # doesn't move either half
+ check_ps $f8, 4.0, 16.0
+
+
+ writemsg "movt.ps (n,y)"
+
+ cfc1 $2, $31
+ or $2, $2, (1 << 23) | (1 << 25)
+ xor $2, $2, (0 << 23) | (1 << 25)
+ ctc1 $2, $31 # set fcc0, clear fcc1
+ mov.ps $f8, $f20
+ movt.ps $f8, $f22, $fcc0 # moves lower half only
+ check_ps $f8, 4.0, 2.0
+
+
+ writemsg "movt.ps (y,n)"
+
+ cfc1 $2, $31
+ or $2, $2, (1 << 23) | (1 << 25)
+ xor $2, $2, (1 << 23) | (0 << 25)
+ ctc1 $2, $31 # clear fcc0, set fcc1
+ mov.ps $f8, $f20
+ movt.ps $f8, $f22, $fcc0 # moves upper half only
+ check_ps $f8, -1.0, 16.0
+
+
+ writemsg "movt.ps (y,y)"
+
+ cfc1 $2, $31
+ or $2, $2, (1 << 23) | (1 << 25)
+ xor $2, $2, (0 << 23) | (0 << 25)
+ ctc1 $2, $31 # set fcc0, set fcc1
+ mov.ps $f8, $f20
+ movt.ps $f8, $f22, $fcc0 # moves both halves
+ check_ps $f8, -1.0, 2.0
+
+
+ writemsg "alnv.ps (aligned)"
+
+ .data
+1: .dword 0xc1a8000042200000 # -21.0, 40.0
+ .dword 0xc228000041a00000 # -42.0, 20.0
+ .text
+ la $2, 1b
+ li $3, 0
+ addu $4, $3, 8
+ luxc1 $f10, $3($2)
+ luxc1 $f12, $4($2)
+ alnv.ps $f8, $f10, $f12, $3
+ check_ps $f8, -21.0, 40.0
+
+
+ writemsg "alnv.ps (unaligned)"
+
+ .data
+1: .dword 0xc1a8000042200000 # -21.0, 40.0
+ .dword 0xc228000041a00000 # -42.0, 20.0
+ .hword 0x0001
+ .text
+ la $2, 1b
+ li $3, 4
+ addu $4, $3, 8
+ luxc1 $f10, $3($2)
+ luxc1 $f12, $4($2)
+ alnv.ps $f8, $f10, $f12, $3
+
+ lb $5, 16($2)
+ bnez $5, 2f # little endian
+ nop
+
+ # big endian
+ check_ps $f8, 40.0, -42.0
+ b 3f
+ nop
+2:
+ # little endian
+ check_ps $f8, 20.0, -21.0
+3:
+
+
+ # We test c.cond.ps only lightly, just to make sure it modifies
+ # two bits and compares the halves separately. Perhaps it should
+ # be tested more thoroughly.
+
+ writemsg "c.f.ps"
+
+ cfc1 $2, $31
+ or $2, $2, (1 << 23) | (0x7f << 25)
+ ctc1 $2, $31 # set all fcc bits
+ c.f.ps $fcc0, $f8, $f8 # -> f, f
+ bc1t $fcc0, _fail
+ nop
+ bc1t $fcc1, _fail
+ nop
+
+
+ writemsg "c.olt.ps"
+
+ cfc1 $2, $31
+ or $2, $2, (1 << 23) | (0x7f << 25)
+ xor $2, $2, (1 << 23) | (0x7f << 25)
+ ctc1 $2, $31 # clear all fcc bits
+ c.lt.ps $fcc0, $f22, $f24 # -> f, t
+ bc1t $fcc0, _fail
+ nop
+ bc1f $fcc1, _fail
+ nop
+
+
+ pass
+
+ .end DIAG