2010-03-17 03:21:18 +01:00
|
|
|
#------------------------------------------------------------------------------
|
|
|
|
#
|
|
|
|
# Replacement for Math64.c that is coded to use older GCC intrinsics.
|
|
|
|
# Doing this reduces the number of intrinsics that are required when
|
|
|
|
# you port to a new version of gcc.
|
|
|
|
#
|
|
|
|
# Need to split this into multple files to size optimize the image.
|
|
|
|
#
|
2010-04-23 18:00:47 +02:00
|
|
|
# Copyright (c) 2009 - 2010, Apple Inc. All rights reserved.<BR>
|
|
|
|
# This program and the accompanying materials
|
2010-03-17 03:21:18 +01:00
|
|
|
# are licensed and made available under the terms and conditions of the BSD License
|
|
|
|
# which accompanies this distribution. The full text of the license may be found at
|
|
|
|
# http://opensource.org/licenses/bsd-license.php
|
|
|
|
#
|
|
|
|
# THE PROGRAM IS DISTRIBUTED UNDER THE BSD LICENSE ON AN "AS IS" BASIS,
|
|
|
|
# WITHOUT WARRANTIES OR REPRESENTATIONS OF ANY KIND, EITHER EXPRESS OR IMPLIED.
|
|
|
|
#
|
|
|
|
#------------------------------------------------------------------------------
|
|
|
|
|
|
|
|
.text
|
|
|
|
.align 2
|
|
|
|
.globl ASM_PFX(InternalMathLShiftU64)
|
|
|
|
ASM_PFX(InternalMathLShiftU64):
|
|
|
|
stmfd sp!, {r4, r5, r6}
|
|
|
|
mov r6, r1
|
|
|
|
rsb ip, r2, #32
|
|
|
|
mov r4, r6, asl r2
|
|
|
|
subs r1, r2, #32
|
|
|
|
orr r4, r4, r0, lsr ip
|
|
|
|
mov r3, r0, asl r2
|
|
|
|
movpl r4, r0, asl r1
|
|
|
|
mov r5, r0
|
|
|
|
mov r0, r3
|
|
|
|
mov r1, r4
|
|
|
|
ldmfd sp!, {r4, r5, r6}
|
|
|
|
bx lr
|
|
|
|
|
|
|
|
.align 2
|
|
|
|
.globl ASM_PFX(InternalMathRShiftU64)
|
|
|
|
ASM_PFX(InternalMathRShiftU64):
|
|
|
|
stmfd sp!, {r4, r5, r6}
|
|
|
|
mov r5, r0
|
|
|
|
rsb ip, r2, #32
|
|
|
|
mov r3, r5, lsr r2
|
|
|
|
subs r0, r2, #32
|
|
|
|
orr r3, r3, r1, asl ip
|
|
|
|
mov r4, r1, lsr r2
|
|
|
|
movpl r3, r1, lsr r0
|
|
|
|
mov r6, r1
|
|
|
|
mov r0, r3
|
|
|
|
mov r1, r4
|
|
|
|
ldmfd sp!, {r4, r5, r6}
|
|
|
|
bx lr
|
|
|
|
|
|
|
|
.align 2
|
|
|
|
.globl ASM_PFX(InternalMathARShiftU64)
|
|
|
|
ASM_PFX(InternalMathARShiftU64):
|
|
|
|
stmfd sp!, {r4, r5, r6}
|
|
|
|
mov r5, r0
|
|
|
|
rsb ip, r2, #32
|
|
|
|
mov r3, r5, lsr r2
|
|
|
|
subs r0, r2, #32
|
|
|
|
orr r3, r3, r1, asl ip
|
|
|
|
mov r4, r1, asr r2
|
|
|
|
movpl r3, r1, asr r0
|
|
|
|
mov r6, r1
|
|
|
|
mov r0, r3
|
|
|
|
mov r1, r4
|
|
|
|
ldmfd sp!, {r4, r5, r6}
|
|
|
|
bx lr
|
|
|
|
|
|
|
|
.align 2
|
|
|
|
.globl ASM_PFX(InternalMathLRotU64)
|
|
|
|
ASM_PFX(InternalMathLRotU64):
|
|
|
|
stmfd sp!, {r4, r5, r6, r7, lr}
|
|
|
|
add r7, sp, #12
|
|
|
|
mov r6, r1
|
|
|
|
rsb ip, r2, #32
|
|
|
|
mov r4, r6, asl r2
|
|
|
|
rsb lr, r2, #64
|
|
|
|
subs r1, r2, #32
|
|
|
|
orr r4, r4, r0, lsr ip
|
|
|
|
mov r3, r0, asl r2
|
|
|
|
movpl r4, r0, asl r1
|
|
|
|
sub ip, r2, #32
|
|
|
|
mov r5, r0
|
|
|
|
mov r0, r0, lsr lr
|
|
|
|
rsbs r2, r2, #32
|
|
|
|
orr r0, r0, r6, asl ip
|
|
|
|
mov r1, r6, lsr lr
|
|
|
|
movpl r0, r6, lsr r2
|
|
|
|
orr r1, r1, r4
|
|
|
|
orr r0, r0, r3
|
|
|
|
ldmfd sp!, {r4, r5, r6, r7, pc}
|
|
|
|
|
|
|
|
|
|
|
|
.align 2
|
|
|
|
.globl ASM_PFX(InternalMathRRotU64)
|
|
|
|
ASM_PFX(InternalMathRRotU64):
|
|
|
|
stmfd sp!, {r4, r5, r6, r7, lr}
|
|
|
|
add r7, sp, #12
|
|
|
|
mov r5, r0
|
|
|
|
rsb ip, r2, #32
|
|
|
|
mov r3, r5, lsr r2
|
|
|
|
rsb lr, r2, #64
|
|
|
|
subs r0, r2, #32
|
|
|
|
orr r3, r3, r1, asl ip
|
|
|
|
mov r4, r1, lsr r2
|
|
|
|
movpl r3, r1, lsr r0
|
|
|
|
sub ip, r2, #32
|
|
|
|
mov r6, r1
|
|
|
|
mov r1, r1, asl lr
|
|
|
|
rsbs r2, r2, #32
|
|
|
|
orr r1, r1, r5, lsr ip
|
|
|
|
mov r0, r5, asl lr
|
|
|
|
movpl r1, r5, asl r2
|
|
|
|
orr r0, r0, r3
|
|
|
|
orr r1, r1, r4
|
|
|
|
ldmfd sp!, {r4, r5, r6, r7, pc}
|
|
|
|
|
|
|
|
.align 2
|
|
|
|
.globl ASM_PFX(InternalMathMultU64x32)
|
|
|
|
ASM_PFX(InternalMathMultU64x32):
|
|
|
|
stmfd sp!, {r7, lr}
|
|
|
|
add r7, sp, #0
|
|
|
|
mov r3, #0
|
|
|
|
mov ip, r0
|
|
|
|
mov lr, r1
|
|
|
|
umull r0, r1, ip, r2
|
|
|
|
mla r1, lr, r2, r1
|
|
|
|
mla r1, ip, r3, r1
|
|
|
|
ldmfd sp!, {r7, pc}
|
|
|
|
|
|
|
|
.align 2
|
|
|
|
.globl ASM_PFX(InternalMathMultU64x64)
|
|
|
|
ASM_PFX(InternalMathMultU64x64):
|
|
|
|
stmfd sp!, {r7, lr}
|
|
|
|
add r7, sp, #0
|
|
|
|
mov ip, r0
|
|
|
|
mov lr, r1
|
|
|
|
umull r0, r1, ip, r2
|
|
|
|
mla r1, lr, r2, r1
|
|
|
|
mla r1, ip, r3, r1
|
|
|
|
ldmfd sp!, {r7, pc}
|
|
|
|
|
|
|
|
.align 2
|
|
|
|
.globl ASM_PFX(InternalMathDivU64x32)
|
|
|
|
ASM_PFX(InternalMathDivU64x32):
|
|
|
|
stmfd sp!, {r7, lr}
|
|
|
|
add r7, sp, #0
|
|
|
|
mov r3, #0
|
|
|
|
bl ASM_PFX(__udivdi3)
|
|
|
|
ldmfd sp!, {r7, pc}
|
|
|
|
|
|
|
|
|
|
|
|
.align 2
|
|
|
|
.globl ASM_PFX(InternalMathModU64x32)
|
|
|
|
ASM_PFX(InternalMathModU64x32):
|
|
|
|
stmfd sp!, {r7, lr}
|
|
|
|
add r7, sp, #0
|
|
|
|
mov r3, #0
|
|
|
|
bl ASM_PFX(__umoddi3)
|
|
|
|
ldmfd sp!, {r7, pc}
|
|
|
|
|
|
|
|
|
|
|
|
.align 2
|
|
|
|
.globl ASM_PFX(InternalMathDivRemU64x32)
|
|
|
|
ASM_PFX(InternalMathDivRemU64x32):
|
|
|
|
stmfd sp!, {r4, r5, r6, r7, lr}
|
|
|
|
add r7, sp, #12
|
|
|
|
stmfd sp!, {r10, r11}
|
|
|
|
subs r6, r3, #0
|
|
|
|
mov r10, r0
|
|
|
|
mov r11, r1
|
|
|
|
moveq r4, r2
|
|
|
|
moveq r5, #0
|
|
|
|
beq L22
|
|
|
|
mov r4, r2
|
|
|
|
mov r5, #0
|
|
|
|
mov r3, #0
|
|
|
|
bl ASM_PFX(__umoddi3)
|
|
|
|
str r0, [r6, #0]
|
|
|
|
L22:
|
|
|
|
mov r0, r10
|
|
|
|
mov r1, r11
|
|
|
|
mov r2, r4
|
|
|
|
mov r3, r5
|
|
|
|
bl ASM_PFX(__udivdi3)
|
|
|
|
ldmfd sp!, {r10, r11}
|
|
|
|
ldmfd sp!, {r4, r5, r6, r7, pc}
|
|
|
|
|
|
|
|
|
|
|
|
.align 2
|
|
|
|
.globl ASM_PFX(InternalMathDivRemU64x64)
|
|
|
|
ASM_PFX(InternalMathDivRemU64x64):
|
|
|
|
stmfd sp!, {r4, r5, r6, r7, lr}
|
|
|
|
add r7, sp, #12
|
|
|
|
stmfd sp!, {r10, r11}
|
|
|
|
ldr r6, [sp, #28]
|
|
|
|
mov r4, r0
|
|
|
|
cmp r6, #0
|
|
|
|
mov r5, r1
|
|
|
|
mov r10, r2
|
|
|
|
mov r11, r3
|
|
|
|
beq L26
|
|
|
|
bl ASM_PFX(__umoddi3)
|
|
|
|
stmia r6, {r0-r1}
|
|
|
|
L26:
|
|
|
|
mov r0, r4
|
|
|
|
mov r1, r5
|
|
|
|
mov r2, r10
|
|
|
|
mov r3, r11
|
|
|
|
bl ASM_PFX(__udivdi3)
|
|
|
|
ldmfd sp!, {r10, r11}
|
|
|
|
ldmfd sp!, {r4, r5, r6, r7, pc}
|
|
|
|
|
|
|
|
|
|
|
|
.align 2
|
|
|
|
.globl ASM_PFX(InternalMathDivRemS64x64)
|
|
|
|
ASM_PFX(InternalMathDivRemS64x64):
|
|
|
|
stmfd sp!, {r4, r5, r6, r7, lr}
|
|
|
|
add r7, sp, #12
|
|
|
|
stmfd sp!, {r10, r11}
|
|
|
|
ldr r6, [sp, #28]
|
|
|
|
mov r4, r0
|
|
|
|
cmp r6, #0
|
|
|
|
mov r5, r1
|
|
|
|
mov r10, r2
|
|
|
|
mov r11, r3
|
|
|
|
beq L30
|
|
|
|
bl ASM_PFX(__moddi3)
|
|
|
|
stmia r6, {r0-r1}
|
|
|
|
L30:
|
|
|
|
mov r0, r4
|
|
|
|
mov r1, r5
|
|
|
|
mov r2, r10
|
|
|
|
mov r3, r11
|
|
|
|
bl ASM_PFX(__divdi3)
|
|
|
|
ldmfd sp!, {r10, r11}
|
|
|
|
ldmfd sp!, {r4, r5, r6, r7, pc}
|
|
|
|
|
|
|
|
|
|
|
|
.align 2
|
|
|
|
.globl ASM_PFX(InternalMathSwapBytes64)
|
|
|
|
ASM_PFX(InternalMathSwapBytes64):
|
|
|
|
@ args = 0, pretend = 0, frame = 0
|
|
|
|
@ frame_needed = 1, uses_anonymous_args = 0
|
|
|
|
stmfd sp!, {r4, r5, r6, r7, lr}
|
|
|
|
add r7, sp, #12
|
|
|
|
mov r5, r1
|
|
|
|
bl ASM_PFX(SwapBytes32)
|
|
|
|
mov r6, r0
|
|
|
|
mov r0, r5
|
|
|
|
bl ASM_PFX(SwapBytes32)
|
|
|
|
mov r4, r6
|
|
|
|
mov r5, r4, asr #31
|
|
|
|
mov r2, #0
|
|
|
|
mov r1, r0, asr #31
|
|
|
|
orr r0, r0, r2
|
|
|
|
orr r1, r1, r4
|
|
|
|
ldmfd sp!, {r4, r5, r6, r7, pc}
|
|
|
|
|
|
|
|
|
|
|
|
ASM_FUNCTION_REMOVE_IF_UNREFERENCED
|