/* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
* linkage_arm.s for PCSX *
- * Copyright (C) 2009-2010 Ari64 *
+ * Copyright (C) 2009-2011 Ari64 *
* Copyright (C) 2010-2011 GraÅžvydas "notaz" Ignotas *
* *
* This program is free software; you can redistribute it and/or modify *
* 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA. *
* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * */
-.equiv HAVE_ARMV7, 1
+/* .equiv HAVE_ARMV7, 1 */
-.if HAVE_ARMV7
- .cpu cortex-a8
- .fpu vfp
-.else
- .cpu arm9tdmi
- .fpu softvfp
-.endif
.global rdram
rdram = 0x80000000
.global dynarec_local
.global psxRegs
.global nd_pcsx_io
.global psxH_ptr
+ .global inv_code_start
+ .global inv_code_end
.bss
.align 4
psxH_ptr = nd_pcsx_io_end
.type psxH_ptr, %object
.size psxH_ptr, 4
-align0 = psxH_ptr + 4 /* just for alignment */
+inv_code_start = psxH_ptr + 4
+ .type inv_code_start, %object
+ .size inv_code_start, 4
+inv_code_end = inv_code_start + 4
+ .type inv_code_end, %object
+ .size inv_code_end, 4
+align0 = inv_code_end + 4 /* just for alignment */
.type align0, %object
- .size align0, 4
-branch_target = align0 + 4
+ .size align0, 12
+branch_target = align0 + 12
.type branch_target, %object
.size branch_target, 4
mini_ht = branch_target + 4
.size memory_map, 4194304
dynarec_local_end = memory_map + 4194304
- .text
- .align 2
- .global dyna_linker
- .type dyna_linker, %function
-dyna_linker:
+.macro load_var_adr reg var
+.if HAVE_ARMV7
+ movw \reg, #:lower16:\var
+ movt \reg, #:upper16:\var
+.else
+ ldr \reg, =\var
+.endif
+.endm
+
+.macro dyna_linker_main
/* r0 = virtual target address */
/* r1 = instruction to patch */
- mov r12, r0
- mov r6, #4096
- mov r2, #0x80000
ldr r3, .jiptr
+ /* get_page */
+ lsr r2, r0, #12
+ mov r6, #4096
+ bic r2, r2, #0xe0000
sub r6, r6, #1
+ cmp r2, #0x1000
ldr r7, [r1]
- eor r2, r2, r12, lsr #12
- and r6, r6, r12, lsr #12
+ biclt r2, #0x0e00
+ and r6, r6, r2
cmp r2, #2048
add r12, r7, #2
orrcs r2, r6, #2048
ldr r5, [r3, r2, lsl #2]
lsl r12, r12, #8
+ add r6, r1, r12, asr #6
+ mov r8, #0
/* jump_in lookup */
-.A1:
+1:
movs r4, r5
- beq .A3
+ beq 2f
ldr r3, [r5]
ldr r5, [r4, #12]
teq r3, r0
- bne .A1
+ bne 1b
ldr r3, [r4, #4]
ldr r4, [r4, #8]
tst r3, r3
- bne .A1
-.A2:
- mov r5, r1
- add r1, r1, r12, asr #6
- teq r1, r4
+ bne 1b
+ teq r4, r6
moveq pc, r4 /* Stale i-cache */
+ mov r8, r4
+ b 1b /* jump_in may have dupes, continue search */
+2:
+ tst r8, r8
+ beq 3f /* r0 not in jump_in */
+
+ mov r5, r1
+ mov r1, r6
bl add_link
- sub r2, r4, r5
+ sub r2, r8, r5
and r1, r7, #0xff000000
lsl r2, r2, #6
sub r1, r1, #2
add r1, r1, r2, lsr #8
str r1, [r5]
- mov pc, r4
-.A3:
+ mov pc, r8
+3:
/* hash_table lookup */
cmp r2, #2048
ldr r3, .jdptr
teq r7, r0
ldreq pc, [r6, #12]
/* jump_dirty lookup */
-.A6:
+6:
movs r4, r5
- beq .A8
+ beq 8f
ldr r3, [r5]
ldr r5, [r4, #12]
teq r3, r0
- bne .A6
-.A7:
+ bne 6b
+7:
ldr r1, [r4, #8]
/* hash_table insert */
ldr r2, [r6]
str r2, [r6, #8]
str r3, [r6, #12]
mov pc, r1
-.A8:
+8:
+.endm
+
+ .text
+ .align 2
+ .global dyna_linker
+ .type dyna_linker, %function
+dyna_linker:
+ /* r0 = virtual target address */
+ /* r1 = instruction to patch */
+ dyna_linker_main
+
mov r4, r0
mov r5, r1
bl new_recompile_block
dyna_linker_ds:
/* r0 = virtual target address */
/* r1 = instruction to patch */
- mov r12, r0
- mov r6, #4096
- mov r2, #0x80000
- ldr r3, .jiptr
- sub r6, r6, #1
- ldr r7, [r1]
- eor r2, r2, r12, lsr #12
- and r6, r6, r12, lsr #12
- cmp r2, #2048
- add r12, r7, #2
- orrcs r2, r6, #2048
- ldr r5, [r3, r2, lsl #2]
- lsl r12, r12, #8
- /* jump_in lookup */
-.B1:
- movs r4, r5
- beq .B3
- ldr r3, [r5]
- ldr r5, [r4, #12]
- teq r3, r0
- bne .B1
- ldr r3, [r4, #4]
- ldr r4, [r4, #8]
- tst r3, r3
- bne .B1
-.B2:
- mov r5, r1
- add r1, r1, r12, asr #6
- teq r1, r4
- moveq pc, r4 /* Stale i-cache */
- bl add_link
- sub r2, r4, r5
- and r1, r7, #0xff000000
- lsl r2, r2, #6
- sub r1, r1, #2
- add r1, r1, r2, lsr #8
- str r1, [r5]
- mov pc, r4
-.B3:
- /* hash_table lookup */
- cmp r2, #2048
- ldr r3, .jdptr
- eor r4, r0, r0, lsl #16
- lslcc r2, r0, #9
- ldr r6, .htptr
- lsr r4, r4, #12
- lsrcc r2, r2, #21
- bic r4, r4, #15
- ldr r5, [r3, r2, lsl #2]
- ldr r7, [r6, r4]!
- teq r7, r0
- ldreq pc, [r6, #4]
- ldr r7, [r6, #8]
- teq r7, r0
- ldreq pc, [r6, #12]
- /* jump_dirty lookup */
-.B6:
- movs r4, r5
- beq .B8
- ldr r3, [r5]
- ldr r5, [r4, #12]
- teq r3, r0
- bne .B6
-.B7:
- ldr r1, [r4, #8]
- /* hash_table insert */
- ldr r2, [r6]
- ldr r3, [r6, #4]
- str r0, [r6]
- str r1, [r6, #4]
- str r2, [r6, #8]
- str r3, [r6, #12]
- mov pc, r1
-.B8:
+ dyna_linker_main
+
mov r4, r0
bic r0, r0, #7
mov r5, r1
.global do_interrupt
.type do_interrupt, %function
do_interrupt:
- /* FIXME: cycles already calculated, not needed? */
ldr r0, [fp, #pcaddr-dynarec_local]
bl get_addr_ht
- ldr r1, [fp, #next_interupt-dynarec_local]
- ldr r10, [fp, #cycle-dynarec_local]
- str r1, [fp, #last_count-dynarec_local]
- sub r10, r10, r1
add r10, r10, #2
mov pc, r0
.size do_interrupt, .-do_interrupt
+
.align 2
.global fp_exception
.type fp_exception, %function
ldr r2, [fp, #last_count-dynarec_local]
str r0, [fp, #pcaddr-dynarec_local]
add r2, r2, r10
- str r2, [fp, #cycle-dynarec_local] /* PCSX cycle counter */
adr lr, pcsx_return
+ str r2, [fp, #cycle-dynarec_local] /* PCSX cycle counter */
bx r1
.size jump_hlecall, .-jump_hlecall
+ .align 2
+ .global jump_intcall
+ .type jump_intcall, %function
+jump_intcall:
+ ldr r2, [fp, #last_count-dynarec_local]
+ str r0, [fp, #pcaddr-dynarec_local]
+ add r2, r2, r10
+ adr lr, pcsx_return
+ str r2, [fp, #cycle-dynarec_local] /* PCSX cycle counter */
+ b execI
+ .size jump_hlecall, .-jump_hlecall
+
new_dyna_leave:
.align 2
.global new_dyna_leave
.size indirect_jump, .-indirect_jump
.size indirect_jump_indexed, .-indirect_jump_indexed
+ .align 2
+ .global invalidate_addr_r0
+ .type invalidate_addr_r0, %function
+invalidate_addr_r0:
+ stmia fp, {r0, r1, r2, r3, r12, lr}
+ b invalidate_addr_call
+ .size invalidate_addr_r0, .-invalidate_addr_r0
+ .align 2
+ .global invalidate_addr_r1
+ .type invalidate_addr_r1, %function
+invalidate_addr_r1:
+ stmia fp, {r0, r1, r2, r3, r12, lr}
+ mov r0, r1
+ b invalidate_addr_call
+ .size invalidate_addr_r1, .-invalidate_addr_r1
+ .align 2
+ .global invalidate_addr_r2
+ .type invalidate_addr_r2, %function
+invalidate_addr_r2:
+ stmia fp, {r0, r1, r2, r3, r12, lr}
+ mov r0, r2
+ b invalidate_addr_call
+ .size invalidate_addr_r2, .-invalidate_addr_r2
+ .align 2
+ .global invalidate_addr_r3
+ .type invalidate_addr_r3, %function
+invalidate_addr_r3:
+ stmia fp, {r0, r1, r2, r3, r12, lr}
+ mov r0, r3
+ b invalidate_addr_call
+ .size invalidate_addr_r3, .-invalidate_addr_r3
+ .align 2
+ .global invalidate_addr_r4
+ .type invalidate_addr_r4, %function
+invalidate_addr_r4:
+ stmia fp, {r0, r1, r2, r3, r12, lr}
+ mov r0, r4
+ b invalidate_addr_call
+ .size invalidate_addr_r4, .-invalidate_addr_r4
+ .align 2
+ .global invalidate_addr_r5
+ .type invalidate_addr_r5, %function
+invalidate_addr_r5:
+ stmia fp, {r0, r1, r2, r3, r12, lr}
+ mov r0, r5
+ b invalidate_addr_call
+ .size invalidate_addr_r5, .-invalidate_addr_r5
+ .align 2
+ .global invalidate_addr_r6
+ .type invalidate_addr_r6, %function
+invalidate_addr_r6:
+ stmia fp, {r0, r1, r2, r3, r12, lr}
+ mov r0, r6
+ b invalidate_addr_call
+ .size invalidate_addr_r6, .-invalidate_addr_r6
+ .align 2
+ .global invalidate_addr_r7
+ .type invalidate_addr_r7, %function
+invalidate_addr_r7:
+ stmia fp, {r0, r1, r2, r3, r12, lr}
+ mov r0, r7
+ b invalidate_addr_call
+ .size invalidate_addr_r7, .-invalidate_addr_r7
+ .align 2
+ .global invalidate_addr_r8
+ .type invalidate_addr_r8, %function
+invalidate_addr_r8:
+ stmia fp, {r0, r1, r2, r3, r12, lr}
+ mov r0, r8
+ b invalidate_addr_call
+ .size invalidate_addr_r8, .-invalidate_addr_r8
+ .align 2
+ .global invalidate_addr_r9
+ .type invalidate_addr_r9, %function
+invalidate_addr_r9:
+ stmia fp, {r0, r1, r2, r3, r12, lr}
+ mov r0, r9
+ b invalidate_addr_call
+ .size invalidate_addr_r9, .-invalidate_addr_r9
+ .align 2
+ .global invalidate_addr_r10
+ .type invalidate_addr_r10, %function
+invalidate_addr_r10:
+ stmia fp, {r0, r1, r2, r3, r12, lr}
+ mov r0, r10
+ b invalidate_addr_call
+ .size invalidate_addr_r10, .-invalidate_addr_r10
+ .align 2
+ .global invalidate_addr_r12
+ .type invalidate_addr_r12, %function
+invalidate_addr_r12:
+ stmia fp, {r0, r1, r2, r3, r12, lr}
+ mov r0, r12
+ .size invalidate_addr_r12, .-invalidate_addr_r12
+ .align 2
+ .global invalidate_addr_call
+ .type invalidate_addr_call, %function
+invalidate_addr_call:
+ ldr r12, [fp, #inv_code_start-dynarec_local]
+ ldr lr, [fp, #inv_code_end-dynarec_local]
+ cmp r0, r12
+ cmpcs lr, r0
+ blcc invalidate_addr
+ ldmia fp, {r0, r1, r2, r3, r12, pc}
+ .size invalidate_addr_call, .-invalidate_addr_call
+
.align 2
.global new_dyna_start
.type new_dyna_start, %function
new_dyna_start:
/* ip is stored to conform EABI alignment */
stmfd sp!, {r4, r5, r6, r7, r8, r9, sl, fp, ip, lr}
-.if HAVE_ARMV7
- movw fp, #:lower16:dynarec_local
- movt fp, #:upper16:dynarec_local
-.else
- ldr fp, .dlptr
-.endif
+ load_var_adr fp, dynarec_local
ldr r0, [fp, #pcaddr-dynarec_local]
bl get_addr_ht
ldr r1, [fp, #next_interupt-dynarec_local]
str r1, [fp, #last_count-dynarec_local]
sub r10, r10, r1
mov pc, r0
-.dlptr:
- .word dynarec_local
.size new_dyna_start, .-new_dyna_start
/* --------------------------------------- */
.global ari_write_ram_mirror8
.global ari_write_ram_mirror16
.global ari_write_ram_mirror32
+.global ari_write_ram_mirror_ro32
.global ari_read_bios8
.global ari_read_bios16
.global ari_read_bios32
ari_read_ram_mirror (3<<11), ldr
/* invalidation is already taken care of by the caller */
-.macro ari_write_ram bic_const var op
+.macro ari_write_ram bic_const var pf
ldr r0, [fp, #address-dynarec_local]
- ldr r1, [fp, #\var-dynarec_local]
+ ldr\pf r1, [fp, #\var-dynarec_local]
.if \bic_const
bic r0, r0, #\bic_const
.endif
- \op r1, [r0]
+ str\pf r1, [r0]
mov pc, lr
.endm
ari_write_ram8:
- ari_write_ram 0, byte, strb
+ ari_write_ram 0, byte, b
ari_write_ram16:
- ari_write_ram 1, hword, strh
+ ari_write_ram 1, hword, h
ari_write_ram32:
- ari_write_ram 3, word, str
+ ari_write_ram 3, word,
-.macro ari_write_ram_mirror mvn_const var op
+.macro ari_write_ram_mirror mvn_const var pf
ldr r0, [fp, #address-dynarec_local]
mvn r3, #\mvn_const
- ldr r1, [fp, #\var-dynarec_local]
+ ldr\pf r1, [fp, #\var-dynarec_local]
and r0, r3, lsr #11
ldr r2, [fp, #invc_ptr-dynarec_local]
orr r0, r0, #1<<31
ldrb r2, [r2, r0, lsr #12]
- \op r1, [r0]
+ str\pf r1, [r0]
tst r2, r2
movne pc, lr
- lsr r0, r0, #12
- b invalidate_block
+ ldr r1, [fp, #inv_code_start-dynarec_local]
+ ldr r2, [fp, #inv_code_end-dynarec_local]
+ cmp r0, r1
+ cmpcs r2, r0
+ movcs pc, lr
+ nop
+ b invalidate_addr
.endm
ari_write_ram_mirror8:
- ari_write_ram_mirror 0, byte, strb
+ ari_write_ram_mirror 0, byte, b
ari_write_ram_mirror16:
- ari_write_ram_mirror (1<<11), hword, strh
+ ari_write_ram_mirror (1<<11), hword, h
ari_write_ram_mirror32:
- ari_write_ram_mirror (3<<11), word, str
+ ari_write_ram_mirror (3<<11), word,
+
+ari_write_ram_mirror_ro32:
+ load_var_adr r0, pcsx_ram_is_ro
+ ldr r0, [r0]
+ tst r0, r0
+ movne pc, lr
+ nop
+ b ari_write_ram_mirror32
.macro ari_read_bios_mirror bic_const op
.endif
.endm
-.macro ari_write_io opvl opst var mem_tab tab_shift
+.macro ari_write_io pf var mem_tab tab_shift
ldr r0, [fp, #address-dynarec_local]
- \opvl r1, [fp, #\var-dynarec_local]
+ ldr\pf r1, [fp, #\var-dynarec_local]
.if \tab_shift == 0
bic r0, r0, #3
.endif
bxne r12
0:
ldr r3, [fp, #psxH_ptr-dynarec_local]
- \opst r1, [r2, r3]
+ str\pf r1, [r2, r3]
mov pc, lr
1:
-.if \tab_shift == 1 @ write16
cmp r2, #0x1c00
blo 0b
cmp r2, #0x1e00
+.if \tab_shift != 0
ldrlo pc, [fp, #spu_writef-dynarec_local]
- nop
+.else
+ @ write32 to SPU - very rare case (is this correct?)
+ bhs 0b
+ add r2, r0, #2
+ mov r3, r1, lsr #16
+ push {r2,r3,lr}
+ mov lr, pc
+ ldr pc, [fp, #spu_writef-dynarec_local]
+ pop {r0,r1,lr}
+ ldr pc, [fp, #spu_writef-dynarec_local]
.endif
+ nop
b 0b
.endm
mov pc, lr
ari_write_io16:
- ari_write_io ldrh, strh, hword, tab_write16, 1
+ ari_write_io h, hword, tab_write16, 1
ari_write_io32:
- ari_write_io ldr, str, word, tab_write32, 0
+ ari_write_io , word, tab_write32, 0
@ vim:filetype=armasm