146 lines
4.3 KiB
ArmAsm
146 lines
4.3 KiB
ArmAsm
/* Copyright 2017 The Chromium OS Authors. All rights reserved.
|
|
* Use of this source code is governed by a BSD-style license that can be
|
|
* found in the LICENSE file.
|
|
*/
|
|
|
|
#include "config_chip.h"
|
|
|
|
/*
|
|
* DLMB register = 0x80189:
|
|
* Disable all interrupts and switching CPU's
|
|
* ALU (Arithmetic Logic Unit) to floating point operation mode.
|
|
* (IEEE standard 754 floating point)
|
|
*
|
|
* DLMB register = 0x80009:
|
|
* Restore interrupts and ALU.
|
|
*/
|
|
.text
|
|
.align 2
|
|
.global __addsf3
|
|
.type __addsf3, @function
|
|
__addsf3:
|
|
sethi $r2, 0x80 /* r2 = 0x80000 */
|
|
addi $r3, $r2, 0x189 /* r3 = 0x80189 */
|
|
addi45 $r2, 0x9 /* r2 = 0x80009 */
|
|
mtsr $r3, $dlmb /* dlmb = 0x80189 */
|
|
dsb
|
|
/* Floating-point addition single-precision */
|
|
add45 $r0, $r1
|
|
mtsr $r2, $dlmb /* dlmb = 0x80009 */
|
|
dsb
|
|
ret5 $lp
|
|
.size __addsf3, .-__addsf3
|
|
|
|
.text
|
|
.align 2
|
|
.global __subsf3
|
|
.type __subsf3, @function
|
|
__subsf3:
|
|
sethi $r2, 0x80 /* r2 = 0x80000 */
|
|
addi $r3, $r2, 0x189 /* r3 = 0x80189 */
|
|
addi45 $r2, 0x9 /* r2 = 0x80009 */
|
|
mtsr $r3, $dlmb /* dlmb = 0x80189 */
|
|
dsb
|
|
/* Floating-point subtraction single-precision */
|
|
sub45 $r0, $r1
|
|
mtsr $r2, $dlmb /* dlmb = 0x80009 */
|
|
dsb
|
|
ret5 $lp
|
|
.size __subsf3, .-__subsf3
|
|
|
|
.text
|
|
.align 2
|
|
.global __mulsf3
|
|
.type __mulsf3, @function
|
|
__mulsf3:
|
|
#ifdef IT83XX_FPU_MUL_BY_DIV
|
|
#define SIGN $r2
|
|
#define EXPOA $r3
|
|
#define MANTA $r4
|
|
#define VALUA $r5
|
|
#define EXPOB $r6
|
|
#define MANTB $r7
|
|
#define VALUB $r8
|
|
#define SPROD $r15
|
|
/* save r6-r8 */
|
|
smw.adm $r6, [$sp], $r8, #0x0
|
|
xor SPROD, $r1, $r0 /* sign(A xor B) */
|
|
move SIGN, #0x80000000
|
|
and SPROD, SPROD, SIGN /* store sign bit */
|
|
slli VALUA, $r0, 1 /* A<<1, (exponent and mantissa) */
|
|
slli VALUB, $r1, 1 /* B<<1, (exponent and mantissa) */
|
|
srli EXPOA, VALUA, 24 /* exponent(A) */
|
|
srli EXPOB, VALUB, 24 /* exponent(B) */
|
|
slli MANTA, VALUA, 7 /* A<<8, mantissa(A) with exponent's LSB */
|
|
slli MANTB, VALUB, 7 /* A<<8, mantissa(B) with exponent's LSB */
|
|
beqz VALUA, .LFzeroA /* exponent(A) and mantissa (A) are zero */
|
|
beqc EXPOA, 0xff, .LFinfnanA /* A is inf or NaN */
|
|
beqz VALUB, .LFzeroB /* exponent(B) and mantissa (B) are zero */
|
|
beqc EXPOB, 0xff, .LFinfnanB /* B is inf or NaN */
|
|
/* A*B = A/(1/B) */
|
|
sethi $r2, 0x80 /* r2 = 0x80000 */
|
|
addi $r3, $r2, 0x189 /* r3 = 0x80189 */
|
|
addi45 $r2, 0x9 /* r2 = 0x80009 */
|
|
mtsr $r3, $dlmb /* dlmb = 0x80189 */
|
|
dsb
|
|
sethi $r5, #0x3f800 /* r5 = 1.0f */
|
|
divsr $r1,$r1,$r5,$r1 /* r1 = 1.0f / r1 */
|
|
divsr $r0,$r0,$r0,$r1 /* r0 = r0 / r1 */
|
|
mtsr $r2, $dlmb /* dlmb = 0x80009 */
|
|
dsb
|
|
.LFret:
|
|
/* restore r6-r8 */
|
|
lmw.bim $r6, [$sp], $r8, #0x0
|
|
ret5 $lp
|
|
|
|
.LFzeroA: /* A is zero */
|
|
beqc EXPOB, 0xff, .LFnan/*zero * inf = zero * NaN = NaN */
|
|
.LFzero:
|
|
move $r0, SPROD /* return 0.0f or -0.0f */
|
|
b .LFret
|
|
.LFinfnanA: /* exponent(A) is 0xff */
|
|
bne MANTA, SIGN, .LFnan/* A is NaN: NaN * B = NaN */
|
|
beqz VALUB, .LFnan /* A is inf and B is zero: inf * zero = NaN */
|
|
bnec EXPOB, 0xff, .LFinf/* B is finite: inf * B = inf */
|
|
.LFinfnanB: /* exponent(B) is 0xff */
|
|
bne MANTB, SIGN, .LFnan/* B is NaN: A * NaN = NaN */
|
|
.LFinf:
|
|
move $r0, #0x7f800000
|
|
or $r0, $r0, SPROD /* return inf or -inf */
|
|
b .LFret
|
|
.LFzeroB: /* B is zero and A is finit */
|
|
b .LFzero /* B is zero */
|
|
.LFnan:
|
|
move $r0, #0xffc00000 /* return NaN */
|
|
b .LFret
|
|
#else /* !IT83XX_FPU_MUL_BY_DIV */
|
|
sethi $r2, 0x80 /* r2 = 0x80000 */
|
|
addi $r3, $r2, 0x189 /* r3 = 0x80189 */
|
|
addi45 $r2, 0x9 /* r2 = 0x80009 */
|
|
mtsr $r3, $dlmb /* dlmb = 0x80189 */
|
|
dsb
|
|
/* Floating-point multiplication single-precision */
|
|
mul33 $r0, $r1
|
|
mtsr $r2, $dlmb /* dlmb = 0x80009 */
|
|
dsb
|
|
ret5 $lp
|
|
#endif /* IT83XX_FPU_MUL_BY_DIV */
|
|
.size __mulsf3, .-__mulsf3
|
|
|
|
.text
|
|
.align 2
|
|
.global __divsf3
|
|
.type __divsf3, @function
|
|
__divsf3:
|
|
sethi $r2, 0x80 /* r2 = 0x80000 */
|
|
addi $r3, $r2, 0x189 /* r3 = 0x80189 */
|
|
addi45 $r2, 0x9 /* r2 = 0x80009 */
|
|
mtsr $r3, $dlmb /* dlmb = 0x80189 */
|
|
dsb
|
|
/* Floating-point division single-precision */
|
|
divsr $r0,$r0,$r0,$r1
|
|
mtsr $r2, $dlmb /* dlmb = 0x80009 */
|
|
dsb
|
|
ret5 $lp
|
|
.size __divsf3, .-__divsf3
|