mirror of
https://github.com/partic2/libffi
synced 2026-11-10 10:07:19 +00:00
arm: Rewrite ffi_closure
Move the push of the argument registers into ffi_closure_SYSV, reducing the size of the trampoline.
This commit is contained in:
Richard Henderson
+105
-116
@@ -189,139 +189,128 @@ ARM_FUNC_END(ffi_call_SYSV)
|
||||
|
||||
ARM_FUNC_START(ffi_closure_SYSV, 1)
|
||||
UNWIND .fnstart
|
||||
UNWIND .pad #16
|
||||
add ip, sp, #16
|
||||
stmfd sp!, {ip, lr}
|
||||
UNWIND .save {r0, lr}
|
||||
add r2, sp, #8
|
||||
UNWIND .pad #16
|
||||
sub sp, sp, #16
|
||||
str sp, [sp, #8]
|
||||
add r1, sp, #8
|
||||
bl CNAME(ffi_closure_inner)
|
||||
cmp r0, #FFI_TYPE_INT
|
||||
beq .Lretint
|
||||
cfi_startproc
|
||||
stmdb sp!, {r0-r3} @ save argument regs
|
||||
cfi_adjust_cfa_offset(16)
|
||||
mov r0, ip @ load closure
|
||||
add ip, sp, #16 @ compute entry sp
|
||||
sub sp, sp, #32 @ allocate rvalue space
|
||||
stmdb sp!, {sp,lr}
|
||||
|
||||
cmp r0, #FFI_TYPE_FLOAT
|
||||
#if defined(__SOFTFP__) || defined(__ARM_EABI__)
|
||||
beq .Lretint
|
||||
#else
|
||||
beq .Lretfloat
|
||||
#endif
|
||||
/* Remember that EABI unwind info only applies at call sites.
|
||||
We need do nothing except note the save of the stack pointer
|
||||
and the link registers. */
|
||||
UNWIND .save {sp,lr}
|
||||
cfi_adjust_cfa_offset(8)
|
||||
cfi_rel_offset(lr, 4)
|
||||
|
||||
cmp r0, #FFI_TYPE_DOUBLE
|
||||
#if defined(__SOFTFP__) || defined(__ARM_EABI__)
|
||||
beq .Lretlonglong
|
||||
#else
|
||||
beq .Lretdouble
|
||||
#endif
|
||||
add r1, sp, #8 @ load respp
|
||||
add r2, sp, #8+32 @ load args
|
||||
mov r3, #0 @ load vfp_args
|
||||
|
||||
cmp r0, #FFI_TYPE_LONGDOUBLE
|
||||
#if defined(__SOFTFP__) || defined(__ARM_EABI__)
|
||||
beq .Lretlonglong
|
||||
#else
|
||||
beq .Lretlongdouble
|
||||
#endif
|
||||
bl CNAME(ffi_closure_inner_SYSV)
|
||||
|
||||
cmp r0, #FFI_TYPE_SINT64
|
||||
beq .Lretlonglong
|
||||
.Lclosure_epilogue:
|
||||
add sp, sp, #16
|
||||
ldmfd sp, {sp, pc}
|
||||
.Lretint:
|
||||
ldr r0, [sp]
|
||||
b .Lclosure_epilogue
|
||||
.Lretlonglong:
|
||||
ldr r0, [sp]
|
||||
ldr r1, [sp, #4]
|
||||
b .Lclosure_epilogue
|
||||
|
||||
#if !defined(__SOFTFP__) && !defined(__ARM_EABI__)
|
||||
.Lretfloat:
|
||||
ldfs f0, [sp]
|
||||
b .Lclosure_epilogue
|
||||
.Lretdouble:
|
||||
ldfd f0, [sp]
|
||||
b .Lclosure_epilogue
|
||||
.Lretlongdouble:
|
||||
ldfd f0, [sp]
|
||||
b .Lclosure_epilogue
|
||||
#endif
|
||||
UNWIND .fnend
|
||||
@ Load values returned in registers.
|
||||
add r2, sp, #8 @ load respp
|
||||
adr r3, CNAME(ffi_closure_ret)
|
||||
add pc, r3, r0, lsl #3
|
||||
cfi_endproc
|
||||
UNWIND .fnend
|
||||
ARM_FUNC_END(ffi_closure_SYSV)
|
||||
|
||||
|
||||
/* Below are VFP hard-float ABI call and closure implementations.
|
||||
Add VFP FPU directive here. This is only compiled into the library
|
||||
under EABI. */
|
||||
#ifdef __ARM_EABI__
|
||||
ARM_FUNC_START(ffi_closure_VFP, 1)
|
||||
UNWIND .fnstart
|
||||
fstmfdd sp!, {d0-d7}
|
||||
@ r0-r3, then d0-d7
|
||||
UNWIND .pad #80
|
||||
add ip, sp, #80
|
||||
stmfd sp!, {ip, lr}
|
||||
UNWIND .save {r0, lr}
|
||||
add r2, sp, #72
|
||||
add r3, sp, #8
|
||||
UNWIND .pad #72
|
||||
sub sp, sp, #72
|
||||
str sp, [sp, #64]
|
||||
add r1, sp, #64
|
||||
bl CNAME(ffi_closure_inner)
|
||||
cfi_startproc
|
||||
stmdb sp!, {r0-r3} @ save argument regs
|
||||
cfi_adjust_cfa_offset(16)
|
||||
sub sp, sp, #64+32 @ allocate vfp+rvalue space
|
||||
cfi_adjust_cfa_offset(64+32)
|
||||
stc p11, cr0, [sp], {16} @ vstm sp, {d0-d7}
|
||||
mov r0, ip @ load closure
|
||||
add ip, sp, #16+64+32 @ compute entry sp
|
||||
stmdb sp!, {ip,lr}
|
||||
|
||||
cmp r0, #FFI_TYPE_INT
|
||||
beq .Lretint_vfp
|
||||
/* See above. */
|
||||
UNWIND .save {sp,lr}
|
||||
cfi_adjust_cfa_offset(8)
|
||||
cfi_rel_offset(sp, 0)
|
||||
cfi_rel_offset(lr, 4)
|
||||
|
||||
cmp r0, #FFI_TYPE_FLOAT
|
||||
beq .Lretfloat_vfp
|
||||
add r1, sp, #8+64 @ load respp
|
||||
add r2, sp, #8+64+32 @ load args
|
||||
add r3, sp, #8 @ load vfp_args
|
||||
|
||||
cmp r0, #FFI_TYPE_DOUBLE
|
||||
cmpne r0, #FFI_TYPE_LONGDOUBLE
|
||||
beq .Lretdouble_vfp
|
||||
bl CNAME(ffi_closure_inner_VFP)
|
||||
|
||||
cmp r0, #FFI_TYPE_SINT64
|
||||
beq .Lretlonglong_vfp
|
||||
|
||||
cmp r0, #FFI_TYPE_STRUCT_VFP_FLOAT
|
||||
beq .Lretfloat_struct_vfp
|
||||
|
||||
cmp r0, #FFI_TYPE_STRUCT_VFP_DOUBLE
|
||||
beq .Lretdouble_struct_vfp
|
||||
|
||||
.Lclosure_epilogue_vfp:
|
||||
add sp, sp, #72
|
||||
ldmfd sp, {sp, pc}
|
||||
|
||||
.Lretfloat_vfp:
|
||||
flds s0, [sp]
|
||||
b .Lclosure_epilogue_vfp
|
||||
.Lretdouble_vfp:
|
||||
fldd d0, [sp]
|
||||
b .Lclosure_epilogue_vfp
|
||||
.Lretint_vfp:
|
||||
ldr r0, [sp]
|
||||
b .Lclosure_epilogue_vfp
|
||||
.Lretlonglong_vfp:
|
||||
ldmia sp, {r0, r1}
|
||||
b .Lclosure_epilogue_vfp
|
||||
.Lretfloat_struct_vfp:
|
||||
fldmiad sp, {d0-d1}
|
||||
b .Lclosure_epilogue_vfp
|
||||
.Lretdouble_struct_vfp:
|
||||
fldmiad sp, {d0-d3}
|
||||
b .Lclosure_epilogue_vfp
|
||||
UNWIND .fnend
|
||||
@ Load values returned in registers.
|
||||
add r2, sp, #8+64 @ load respp
|
||||
adr r3, CNAME(ffi_closure_ret)
|
||||
add pc, r3, r0, lsl #3
|
||||
cfi_endproc
|
||||
UNWIND .fnend
|
||||
ARM_FUNC_END(ffi_closure_VFP)
|
||||
#endif
|
||||
|
||||
/* Load values returned in registers for both closure entry points.
|
||||
Note that we use LDM with SP in the register set. This is deprecated
|
||||
by ARM, but not yet unpredictable. */
|
||||
|
||||
ARM_FUNC_START(ffi_closure_ret, 0)
|
||||
cfi_startproc
|
||||
cfi_rel_offset(sp, 0)
|
||||
cfi_rel_offset(lr, 4)
|
||||
0:
|
||||
E ARM_TYPE_VFP_S
|
||||
ldc p10, cr0, [r2] @ vldr s0, [r2]
|
||||
ldm sp, {sp,pc}
|
||||
E ARM_TYPE_VFP_D
|
||||
ldc p11, cr0, [r2] @ vldr d0, [r2]
|
||||
ldm sp, {sp,pc}
|
||||
E ARM_TYPE_VFP_N
|
||||
ldc p11, cr0, [r2], {8} @ vldm r2, {d0-d3}
|
||||
ldm sp, {sp,pc}
|
||||
E ARM_TYPE_INT64
|
||||
ldr r1, [r2, #4]
|
||||
nop
|
||||
E ARM_TYPE_INT
|
||||
ldr r0, [r2]
|
||||
ldm sp, {sp,pc}
|
||||
E ARM_TYPE_VOID
|
||||
ldm sp, {sp,pc}
|
||||
nop
|
||||
E ARM_TYPE_STRUCT
|
||||
ldm sp, {sp,pc}
|
||||
cfi_endproc
|
||||
ARM_FUNC_END(ffi_closure_ret)
|
||||
|
||||
#if FFI_EXEC_TRAMPOLINE_TABLE
|
||||
|
||||
/* ??? The iOS support should be updated. The first insn used to
|
||||
be STMFD, but that's been moved into ffi_closure_SYSV. If the
|
||||
writable page is put after this one we can make use of the
|
||||
pc+8 feature of the architecture. We can also reduce the size
|
||||
of the thunk to 8 and pack more of these into the page.
|
||||
|
||||
In the meantime, simply replace the STMFD with a NOP so as to
|
||||
keep all the magic numbers the same within ffi.c. */
|
||||
|
||||
.align 12
|
||||
ARM_FUNC_START(ffi_closure_trampoline_table_page)
|
||||
.rept 4096 / 12
|
||||
nop
|
||||
ldr ip, [pc, #-4092]
|
||||
ldr pc, [pc, #-4092]
|
||||
.endr
|
||||
|
||||
#else
|
||||
|
||||
ARM_FUNC_START(ffi_arm_trampoline, 1)
|
||||
stmfd sp!, {r0-r3}
|
||||
ldr r0, [pc]
|
||||
ldr pc, [pc]
|
||||
0: adr ip, 0b
|
||||
ldr pc, 1f
|
||||
1: .long 0
|
||||
ARM_FUNC_END(ffi_arm_trampoline)
|
||||
|
||||
#endif /* FFI_EXEC_TRAMPOLINE_TABLE */
|
||||
|
||||
#if defined __ELF__ && defined __linux__
|
||||
.section .note.GNU-stack,"",%progbits
|
||||
#endif
|
||||
|
||||
Reference in New Issue
Block a user