1 /// -*- mode: asm; asm-comment-char: ?/ -*-
3 /// Common definitions for asesembler source files
5 /// (c) 2015 Straylight/Edgeware
8 ///----- Licensing notice ---------------------------------------------------
10 /// This file is part of Catacomb.
12 /// Catacomb is free software; you can redistribute it and/or modify
13 /// it under the terms of the GNU Library General Public License as
14 /// published by the Free Software Foundation; either version 2 of the
15 /// License, or (at your option) any later version.
17 /// Catacomb is distributed in the hope that it will be useful,
18 /// but WITHOUT ANY WARRANTY; without even the implied warranty of
19 /// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
20 /// GNU Library General Public License for more details.
22 /// You should have received a copy of the GNU Library General Public
23 /// License along with Catacomb; if not, write to the Free
24 /// Software Foundation, Inc., 59 Temple Place - Suite 330, Boston,
25 /// MA 02111-1307, USA.
27 #ifndef CATACOMB_ASM_COMMON_H
28 #define CATACOMB_ASM_COMMON_H
30 ///--------------------------------------------------------------------------
31 /// General definitions.
33 // Preprocessor hacks.
34 #define STRINGY(x) _STRINGY(x, y)
35 #define _STRINGY(x) #x
36 #define GLUE(x, y) _GLUE(x, y)
37 #define _GLUE(x, y) x##y
40 // Some useful variables.
43 // Literal pools done the hard way.
44 #define _LIT .text .L$_subsec + 1
45 #define _ENDLIT .text .L$_subsec
46 #define _LTORG .L$_subsec = .L$_subsec + 2; .text .L$_subsec
51 # define _SECTTY(ty) %ty
53 # define _SECTTY(ty) @ty
58 #define TEXT .text .L$_subsec
60 # define RODATA .section .rdata, "dr"
62 # define RODATA .section .rodata, "a", _SECTTY(progbits)
68 // Announcing an internal function.
69 #define INTFUNC(name) \
71 .macro ENDFUNC; _ENDFUNC(name); .endm; \
72 .L$_prologue_p = 0; .L$_frameptr_p = 0; \
77 // Announcing an external function.
82 // Marking the end of a function.
83 #define _ENDFUNC(name) \
84 .if ~ .L$_prologue_p; .error "Missing `endprologue'"; .endif; \
85 .if .L$_frameptr_p; .purgem dropfp; .endif; \
91 // Make a helper function, if necessary.
93 .ifndef .L$_auxfn_def.name; \
95 .macro _ENDAUXFN; _ENDAUXFN_TAIL(name); .endm; \
98 #define _ENDAUXFN_TAIL(name) \
101 .L$_auxfn_def.name = 1
102 #define ENDAUXFN _ENDAUXFN; .endif
104 ///--------------------------------------------------------------------------
105 /// ELF-specific hacking.
109 // Special arrangements for position-independent code.
110 #if __PIC__ || __PIE__
114 // Additional symbol metadata.
115 #define TYPE_FUNC(name) .type name, STT_FUNC
116 #define TYPE_OBJ(name) .type name, STT_OBJECT
117 #define SIZE_OBJ(name) .size name, . - name
121 ///--------------------------------------------------------------------------
122 /// Windows-specific hacking.
126 // Function names need decorating on 32-bit i386.
128 # define F(name) _##name
133 ///--------------------------------------------------------------------------
134 /// x86- and amd64-specific hacking.
136 /// It's (slightly) easier to deal with both of these in one go.
138 #if CPUFAM_X86 || CPUFAM_AMD64
148 // Set the function hooks.
149 #define FUNC_PREHOOK(_) .balign 16
151 // On Windows, arrange to install stack-unwinding data.
152 #if CPUFAM_AMD64 && ABI_WIN
153 # define FUNC_POSTHOOK(name) .seh_proc name
154 # define ENDFUNC_HOOK(_) .seh_endproc
155 // Procedures are expected to invoke `.seh_setframe' if necessary, and
156 // `.seh_pushreg' and friends, and `.seh_endprologue'.
160 # define FUNC_POSTHOOK(_) .cfi_startproc
161 # define ENDFUNC_HOOK(_) .cfi_endproc
164 // Don't use the wretched AT&T syntax. It's festooned with pointless
165 // punctuation, and all of the data movement is backwards. Ugh!
166 .intel_syntax noprefix
168 // Call external subroutine at ADDR, possibly via PLT.
177 // Do I need to arrange a spare GOT register?
178 #if WANT_PIC && CPUFAM_X86
181 #define GOTREG ebx // Not needed in AMD64 so don't care.
183 // Maybe load GOT address into GOT.
184 .macro ldgot got
=GOTREG
185 #if WANT_PIC && CPUFAM_X86
191 add \got
, offset _GLOBAL_OFFSET_TABLE_
195 // Load address of external symbol ADDR into REG, maybe using GOT.
196 .macro leaext reg
, addr
, got
=GOTREG
199 mov
\reg
, [\got
+ \addr@GOT
]
202 mov
\reg
, \addr@GOTPCREL
[rip
]
206 mov
\reg
, offset
\addr
214 // Address expression (possibly using a base register, and a displacement)
215 // referring to ADDR, which is within our module, maybe using GOT.
216 #define INTADDR(...) INTADDR__0(__VA_ARGS__, GOTREG, dummy)
217 #define INTADDR__0(addr, got, ...) INTADDR__1(addr, got)
219 # define INTADDR__1(addr, got) addr + rip
221 # define INTADDR__1(addr, got) got + addr@GOTOFF
223 # define INTADDR__1(addr, got) addr
226 // Permutations for SIMD instructions. SHUF(A, B, C, D) is an immediate,
227 // suitable for use in `pshufd' or `shufpd', which copies element A
228 // (0 <= A < 4) of the source to element 0 of the destination, element B to
229 // element 1, element C to element 2, and element D to element 3.
230 #define SHUF(a, b, c, d) ((a) + 4*(b) + 16*(c) + 64*(d))
232 // Map register names to their individual pieces.
234 // Apply decoration decor to (internal) register name reg of type ty.
236 // See `R_...' for internal register names. Decorations are as follows.
238 // b low byte (e.g., `al', `r8b')
239 // h high byte (e.g., `ah')
240 // w word (e.g., `ax', `r8w')
241 // d doubleword (e.g., `eax', `r8d')
242 // q quadword (e.g., `rax', `r8')
243 // r whole register (doubleword on x86, quadword on amd64)
245 // And types are as follows.
247 // abcd the four traditional registers `a', `b', `c', `d'
248 // xp the four pointer registers `si', `di', `bp', `sp'
249 // ip the instruction pointer `ip'
250 // rn the AMD64 numbered registers `r8'--`r15'
251 #define _DECOR(ty, decor, reg) _DECOR_##ty##_##decor(reg)
253 // Internal macros: _DECOR_ty_decor(reg) applies decoration decor to
254 // (internal) register name reg of type ty.
256 #define _DECOR_abcd_b(reg) reg##l
257 #define _DECOR_abcd_h(reg) reg##h
258 #define _DECOR_abcd_w(reg) reg##x
259 #define _DECOR_abcd_d(reg) e##reg##x
261 # define _DECOR_abcd_q(reg) r##reg##x
264 #define _DECOR_xp_w(reg) reg
265 #define _DECOR_xp_d(reg) e##reg
267 # define _DECOR_xp_b(reg) reg##l
268 # define _DECOR_xp_q(reg) r##reg
271 #define _DECOR_ip_w(reg) reg
272 #define _DECOR_ip_d(reg) e##reg
274 # define _DECOR_ip_q(reg) r##reg
278 # define _DECOR_rn_b(reg) reg##b
279 # define _DECOR_rn_w(reg) reg##w
280 # define _DECOR_rn_d(reg) reg##d
281 # define _DECOR_rn_q(reg) reg
282 # define _DECOR_rn_r(reg) reg
285 #define _DECOR_mem_b(addr) byte ptr addr
286 #define _DECOR_mem_w(addr) word ptr addr
287 #define _DECOR_mem_d(addr) dword ptr addr
289 # define _DECOR_mem_q(addr) qword ptr addr
292 #define _DECOR_imm_b(imm) byte imm
293 #define _DECOR_imm_w(imm) word imm
294 #define _DECOR_imm_d(imm) dword imm
296 # define _DECOR_imm_q(imm) qword imm
300 # define _DECOR_abcd_r(reg) e##reg##x
301 # define _DECOR_xp_r(reg) e##reg
302 # define _DECOR_ip_r(reg) e##reg
303 # define _DECOR_mem_r(addr) dword ptr addr
304 # define _DECOR_imm_r(imm) dword imm
307 # define _DECOR_abcd_r(reg) r##reg##x
308 # define _DECOR_xp_r(reg) r##reg
309 # define _DECOR_ip_r(reg) r##reg
310 # define _DECOR_mem_r(addr) qword ptr addr
311 # define _DECOR_imm_r(imm) qword imm
314 // R_r(decor) applies decoration decor to register r, which is an internal
315 // register name. The internal register names are: `ip', `a', `b', `c', `d',
316 // `si', `di', `bp', `sp', `r8'--`r15'.
317 #define R_nil(decor) nil
318 #define R_ip(decor) _DECOR(ip, decor, ip)
319 #define R_a(decor) _DECOR(abcd, decor, a)
320 #define R_b(decor) _DECOR(abcd, decor, b)
321 #define R_c(decor) _DECOR(abcd, decor, c)
322 #define R_d(decor) _DECOR(abcd, decor, d)
323 #define R_si(decor) _DECOR(xp, decor, si)
324 #define R_di(decor) _DECOR(xp, decor, di)
325 #define R_bp(decor) _DECOR(xp, decor, bp)
326 #define R_sp(decor) _DECOR(xp, decor, sp)
328 # define R_r8(decor) _DECOR(rn, decor, r8)
329 # define R_r9(decor) _DECOR(rn, decor, r9)
330 # define R_r10(decor) _DECOR(rn, decor, r10)
331 # define R_r11(decor) _DECOR(rn, decor, r11)
332 # define R_r12(decor) _DECOR(rn, decor, r12)
333 # define R_r13(decor) _DECOR(rn, decor, r13)
334 # define R_r14(decor) _DECOR(rn, decor, r14)
335 # define R_r15(decor) _DECOR(rn, decor, r15)
338 // Refer to an in-memory datum of the type implied by decor residing at
339 // address addr (which should supply its own square-brackets).
340 #define MEM(decor, addr) _DECOR(mem, decor, addr)
342 // Refer to an immediate datum of the type implied by decor.
343 #define IMM(decor, imm) _DECOR(mem, decor, imm)
345 // Applies decoration decor to assembler-level register name reg.
346 #define _REGFORM(reg, decor) _GLUE(_REGFORM_, reg)(decor)
348 // Internal macros: _REGFORM_r(decor) applies decoration decor to an
349 // assembler-level register name, in place of any decoration that register
352 #define _REGFORM_nil(decor) R_nil(decor)
354 #define _REGFORM_ip(decor) R_ip(decor)
355 #define _REGFORM_eip(decor) R_ip(decor)
357 #define _REGFORM_a(decor) R_a(decor)
358 #define _REGFORM_al(decor) R_a(decor)
359 #define _REGFORM_ah(decor) R_a(decor)
360 #define _REGFORM_ax(decor) R_a(decor)
361 #define _REGFORM_eax(decor) R_a(decor)
363 #define _REGFORM_b(decor) R_b(decor)
364 #define _REGFORM_bl(decor) R_b(decor)
365 #define _REGFORM_bh(decor) R_b(decor)
366 #define _REGFORM_bx(decor) R_b(decor)
367 #define _REGFORM_ebx(decor) R_b(decor)
369 #define _REGFORM_c(decor) R_c(decor)
370 #define _REGFORM_cl(decor) R_c(decor)
371 #define _REGFORM_ch(decor) R_c(decor)
372 #define _REGFORM_cx(decor) R_c(decor)
373 #define _REGFORM_ecx(decor) R_c(decor)
375 #define _REGFORM_d(decor) R_d(decor)
376 #define _REGFORM_dl(decor) R_d(decor)
377 #define _REGFORM_dh(decor) R_d(decor)
378 #define _REGFORM_dx(decor) R_d(decor)
379 #define _REGFORM_edx(decor) R_d(decor)
381 #define _REGFORM_si(decor) R_si(decor)
382 #define _REGFORM_sil(decor) R_si(decor)
383 #define _REGFORM_esi(decor) R_si(decor)
385 #define _REGFORM_di(decor) R_di(decor)
386 #define _REGFORM_dil(decor) R_di(decor)
387 #define _REGFORM_edi(decor) R_di(decor)
389 #define _REGFORM_bp(decor) R_bp(decor)
390 #define _REGFORM_bpl(decor) R_bp(decor)
391 #define _REGFORM_ebp(decor) R_bp(decor)
393 #define _REGFORM_sp(decor) R_sp(decor)
394 #define _REGFORM_spl(decor) R_sp(decor)
395 #define _REGFORM_esp(decor) R_sp(decor)
399 # define _REGFORM_rip(decor) R_ip(decor)
400 # define _REGFORM_rsp(decor) R_sp(decor)
401 # define _REGFORM_rbp(decor) R_bp(decor)
402 # define _REGFORM_rdi(decor) R_di(decor)
403 # define _REGFORM_rsi(decor) R_si(decor)
404 # define _REGFORM_rdx(decor) R_d(decor)
405 # define _REGFORM_rcx(decor) R_c(decor)
406 # define _REGFORM_rbx(decor) R_b(decor)
407 # define _REGFORM_rax(decor) R_a(decor)
409 # define _REGFORM_r8(decor) R_r8(decor)
410 # define _REGFORM_r8b(decor) R_r8(decor)
411 # define _REGFORM_r8w(decor) R_r8(decor)
412 # define _REGFORM_r8d(decor) R_r8(decor)
414 # define _REGFORM_r9(decor) R_r9(decor)
415 # define _REGFORM_r9b(decor) R_r9(decor)
416 # define _REGFORM_r9w(decor) R_r9(decor)
417 # define _REGFORM_r9d(decor) R_r9(decor)
419 # define _REGFORM_r10(decor) R_r10(decor)
420 # define _REGFORM_r10b(decor) R_r10(decor)
421 # define _REGFORM_r10w(decor) R_r10(decor)
422 # define _REGFORM_r10d(decor) R_r10(decor)
424 # define _REGFORM_r11(decor) R_r11(decor)
425 # define _REGFORM_r11b(decor) R_r11(decor)
426 # define _REGFORM_r11w(decor) R_r11(decor)
427 # define _REGFORM_r11d(decor) R_r11(decor)
429 # define _REGFORM_r12(decor) R_r12(decor)
430 # define _REGFORM_r12b(decor) R_r12(decor)
431 # define _REGFORM_r12w(decor) R_r12(decor)
432 # define _REGFORM_r12d(decor) R_r12(decor)
434 # define _REGFORM_r13(decor) R_r13(decor)
435 # define _REGFORM_r13b(decor) R_r13(decor)
436 # define _REGFORM_r13w(decor) R_r13(decor)
437 # define _REGFORM_r13d(decor) R_r13(decor)
439 # define _REGFORM_r14(decor) R_r14(decor)
440 # define _REGFORM_r14b(decor) R_r14(decor)
441 # define _REGFORM_r14w(decor) R_r14(decor)
442 # define _REGFORM_r14d(decor) R_r14(decor)
444 # define _REGFORM_r15(decor) R_r15(decor)
445 # define _REGFORM_r15b(decor) R_r15(decor)
446 # define _REGFORM_r15w(decor) R_r15(decor)
447 # define _REGFORM_r15d(decor) R_r15(decor)
451 // Macros for converting register names.
452 #define BYTE(reg) _REGFORM(reg, b)
453 #define HIBYTE(reg) _REGFORM(reg, h)
454 #define WORD(reg) _REGFORM(reg, w)
455 #define DWORD(reg) _REGFORM(reg, d)
457 # define QWORD(reg) _REGFORM(reg, q)
459 #define WHOLE(reg) _REGFORM(reg, r)
461 // Macros for some common registers.
471 // Stack management and unwinding.
472 .macro setfp fp
=BP
, offset
=0
476 .cfi_def_cfa_register
\fp
478 #if ABI_WIN && CPUFAM_AMD64
482 lea
\fp
, [SP
+ \offset
]
484 .cfi_def_cfa_register
\fp
485 .cfi_adjust_cfa_offset
-\offset
487 #if ABI_WIN && CPUFAM_AMD64
488 .seh_setframe
\fp
, \offset
492 .macro dropfp
; _dropfp
\fp
, \offset
; .endm
495 .macro _dropfp fp
, offset
=0
499 .cfi_def_cfa_register SP
502 lea SP
, [\fp
- \offset
]
504 .cfi_def_cfa_register SP
505 .cfi_adjust_cfa_offset
+\offset
515 .cfi_adjust_cfa_offset
+\n
517 #if ABI_WIN && CPUFAM_AMD64
525 .cfi_adjust_cfa_offset
-\n
532 .cfi_adjust_cfa_offset
+WORDSZ
533 .cfi_rel_offset
\r, 0
535 #if ABI_WIN && CPUFAM_AMD64
543 .cfi_adjust_cfa_offset
-WORDSZ
548 .macro savexmm r
, offset
549 movdqa
[SP
+ \offset
], \r
550 #if ABI_WIN && CPUFAM_AMD64
551 .seh_savexmm
\r, \offset
555 .macro rstrxmm r
, offset
556 movdqa
\r, [SP
+ \offset
]
560 #if ABI_WIN && CPUFAM_AMD64
568 ///--------------------------------------------------------------------------
569 /// ARM-specific hacking.
573 // ARM/Thumb mode things. Use ARM by default.
574 #define ARM .arm; .L$_pcoff = 8
575 #define THUMB .thumb; .L$_pcoff = 4
578 // Set the function hooks.
579 #define FUNC_PREHOOK(_) .balign 4; .fnstart
580 #define ENDFUNC_HOOK(_) .fnend; .ltorg
582 // Call external subroutine at ADDR, possibly via PLT.
583 .macro callext addr
, cond
=
591 // Do I need to arrange a spare GOT register?
597 // Maybe load GOT address into GOT.
598 .macro ldgot cond
=, got
=GOTREG
600 ldr\cond \got
, .L$_ldgot$\@
602 add\cond \got
, pc
, \got
606 .word _GLOBAL_OFFSET_TABLE_
- .L$_ldgot_pc$\@
- .L$_pcoff
611 // Load address of external symbol ADDR into REG, maybe using GOT.
612 .macro leaext reg
, addr
, cond
=, got
=GOTREG
614 ldr\cond
\reg
, .L$_leaext$\@
615 ldr\cond
\reg
, [\got
, \reg
]
622 ldr\cond
\reg
, =\addr
626 // Load address of external symbol ADDR into REG directly.
627 .macro leaextq reg
, addr
, cond
=
629 ldr\cond
\reg
, .L$_leaextq$\@
632 ldr\cond
\reg
, [pc
, \reg
]
635 ldr\cond
\reg
, [\reg
]
640 .word
\addr
(GOT_PREL
) + (. - .L$_leaextq_pc$\@
- .L$_pcoff
)
643 ldr\cond
\reg
, =\addr
648 // Set VZ (default q15) to zero.
652 .macro vshl128 vd
, vn
, nbit
, vz
=q15
653 // Set VD to VN shifted left by NBIT. Assume VZ (default q15) is
654 // all-bits-zero. NBIT must be a multiple of 8.
656 .error
"shift quantity must be whole number of bytes"
658 vext
.8 \vd
, \vz
, \vn
, #16 - (\nbit >> 3)
661 .macro vshr128 vd
, vn
, nbit
, vz
=q15
662 // Set VD to VN shifted right by NBIT. Assume VZ (default q15) is
663 // all-bits-zero. NBIT must be a multiple of 8.
665 .error
"shift quantity must be whole number of bytes"
667 vext
.8 \vd
, \vn
, \vz
, #\nbit >> 3
670 // Apply decoration decor to register name reg.
671 #define _REGFORM(reg, decor) _GLUE(_REGFORM_, reg)(decor)
673 // Internal macros: `_REGFORM_r(decor)' applies decoration decor to register
676 #define _REGFORM_nil(decor) nil
678 #define _REGFORM_s0(decor) _DECOR(s, decor, 0)
679 #define _REGFORM_s1(decor) _DECOR(s, decor, 1)
680 #define _REGFORM_s2(decor) _DECOR(s, decor, 2)
681 #define _REGFORM_s3(decor) _DECOR(s, decor, 3)
682 #define _REGFORM_s4(decor) _DECOR(s, decor, 4)
683 #define _REGFORM_s5(decor) _DECOR(s, decor, 5)
684 #define _REGFORM_s6(decor) _DECOR(s, decor, 6)
685 #define _REGFORM_s7(decor) _DECOR(s, decor, 7)
686 #define _REGFORM_s8(decor) _DECOR(s, decor, 8)
687 #define _REGFORM_s9(decor) _DECOR(s, decor, 9)
688 #define _REGFORM_s10(decor) _DECOR(s, decor, 10)
689 #define _REGFORM_s11(decor) _DECOR(s, decor, 11)
690 #define _REGFORM_s12(decor) _DECOR(s, decor, 12)
691 #define _REGFORM_s13(decor) _DECOR(s, decor, 13)
692 #define _REGFORM_s14(decor) _DECOR(s, decor, 14)
693 #define _REGFORM_s15(decor) _DECOR(s, decor, 15)
694 #define _REGFORM_s16(decor) _DECOR(s, decor, 16)
695 #define _REGFORM_s17(decor) _DECOR(s, decor, 17)
696 #define _REGFORM_s18(decor) _DECOR(s, decor, 18)
697 #define _REGFORM_s19(decor) _DECOR(s, decor, 19)
698 #define _REGFORM_s20(decor) _DECOR(s, decor, 20)
699 #define _REGFORM_s21(decor) _DECOR(s, decor, 21)
700 #define _REGFORM_s22(decor) _DECOR(s, decor, 22)
701 #define _REGFORM_s23(decor) _DECOR(s, decor, 23)
702 #define _REGFORM_s24(decor) _DECOR(s, decor, 24)
703 #define _REGFORM_s25(decor) _DECOR(s, decor, 25)
704 #define _REGFORM_s26(decor) _DECOR(s, decor, 26)
705 #define _REGFORM_s27(decor) _DECOR(s, decor, 27)
706 #define _REGFORM_s28(decor) _DECOR(s, decor, 28)
707 #define _REGFORM_s29(decor) _DECOR(s, decor, 29)
708 #define _REGFORM_s30(decor) _DECOR(s, decor, 30)
709 #define _REGFORM_s31(decor) _DECOR(s, decor, 31)
711 #define _REGFORM_d0(decor) _DECOR(d, decor, 0)
712 #define _REGFORM_d1(decor) _DECOR(d, decor, 1)
713 #define _REGFORM_d2(decor) _DECOR(d, decor, 2)
714 #define _REGFORM_d3(decor) _DECOR(d, decor, 3)
715 #define _REGFORM_d4(decor) _DECOR(d, decor, 4)
716 #define _REGFORM_d5(decor) _DECOR(d, decor, 5)
717 #define _REGFORM_d6(decor) _DECOR(d, decor, 6)
718 #define _REGFORM_d7(decor) _DECOR(d, decor, 7)
719 #define _REGFORM_d8(decor) _DECOR(d, decor, 8)
720 #define _REGFORM_d9(decor) _DECOR(d, decor, 9)
721 #define _REGFORM_d10(decor) _DECOR(d, decor, 10)
722 #define _REGFORM_d11(decor) _DECOR(d, decor, 11)
723 #define _REGFORM_d12(decor) _DECOR(d, decor, 12)
724 #define _REGFORM_d13(decor) _DECOR(d, decor, 13)
725 #define _REGFORM_d14(decor) _DECOR(d, decor, 14)
726 #define _REGFORM_d15(decor) _DECOR(d, decor, 15)
727 #define _REGFORM_d16(decor) _DECOR(d, decor, 16)
728 #define _REGFORM_d17(decor) _DECOR(d, decor, 17)
729 #define _REGFORM_d18(decor) _DECOR(d, decor, 18)
730 #define _REGFORM_d19(decor) _DECOR(d, decor, 19)
731 #define _REGFORM_d20(decor) _DECOR(d, decor, 20)
732 #define _REGFORM_d21(decor) _DECOR(d, decor, 21)
733 #define _REGFORM_d22(decor) _DECOR(d, decor, 22)
734 #define _REGFORM_d23(decor) _DECOR(d, decor, 23)
735 #define _REGFORM_d24(decor) _DECOR(d, decor, 24)
736 #define _REGFORM_d25(decor) _DECOR(d, decor, 25)
737 #define _REGFORM_d26(decor) _DECOR(d, decor, 26)
738 #define _REGFORM_d27(decor) _DECOR(d, decor, 27)
739 #define _REGFORM_d28(decor) _DECOR(d, decor, 28)
740 #define _REGFORM_d29(decor) _DECOR(d, decor, 29)
741 #define _REGFORM_d30(decor) _DECOR(d, decor, 30)
742 #define _REGFORM_d31(decor) _DECOR(d, decor, 31)
744 #define _REGFORM_q0(decor) _DECOR(q, decor, 0)
745 #define _REGFORM_q1(decor) _DECOR(q, decor, 1)
746 #define _REGFORM_q2(decor) _DECOR(q, decor, 2)
747 #define _REGFORM_q3(decor) _DECOR(q, decor, 3)
748 #define _REGFORM_q4(decor) _DECOR(q, decor, 4)
749 #define _REGFORM_q5(decor) _DECOR(q, decor, 5)
750 #define _REGFORM_q6(decor) _DECOR(q, decor, 6)
751 #define _REGFORM_q7(decor) _DECOR(q, decor, 7)
752 #define _REGFORM_q8(decor) _DECOR(q, decor, 8)
753 #define _REGFORM_q9(decor) _DECOR(q, decor, 9)
754 #define _REGFORM_q10(decor) _DECOR(q, decor, 10)
755 #define _REGFORM_q11(decor) _DECOR(q, decor, 11)
756 #define _REGFORM_q12(decor) _DECOR(q, decor, 12)
757 #define _REGFORM_q13(decor) _DECOR(q, decor, 13)
758 #define _REGFORM_q14(decor) _DECOR(q, decor, 14)
759 #define _REGFORM_q15(decor) _DECOR(q, decor, 15)
761 // `_LOPART(n)' and `_HIPART(n)' return the numbers of the register halves of
762 // register n, i.e., 2*n and 2*n + 1 respectively.
763 #define _LOPART(n) _GLUE(_LOPART_, n)
764 #define _HIPART(n) _GLUE(_HIPART_, n)
766 // Internal macros: `_LOPART_n' and `_HIPART_n' return the numbers of the
767 // register halves of register n, i.e., 2*n and 2*n + 1 respectively.
789 #define _LOPART_10 20
790 #define _HIPART_10 21
791 #define _LOPART_11 22
792 #define _HIPART_11 23
793 #define _LOPART_12 24
794 #define _HIPART_12 25
795 #define _LOPART_13 26
796 #define _HIPART_13 27
797 #define _LOPART_14 28
798 #define _HIPART_14 29
799 #define _LOPART_15 30
800 #define _HIPART_15 31
802 // Return the register number of the pair containing register n, i.e.,
804 #define _PAIR(n) _GLUE(_PAIR_, n)
806 // Internal macros: `_PAIR_n' returns the register number of the pair
807 // containing register n, i.e., floor(n/2).
841 // Apply decoration decor to register number n of type ty. Decorations are
844 // decor types meaning
845 // Q s, d the NEON qN register containing this one
846 // D s the NEON dN register containing this one
847 // D0 q the low 64-bit half of this one
848 // D1 q the high 64-bit half of this one
849 // S0 d, q the first 32-bit piece of this one
850 // S1 d, q the second 32-bit piece of this one
851 // S2 q the third 32-bit piece of this one
852 // S3 q the fourth 32-bit piece of this one
853 // Bn q the nth byte of this register, as a scalar
854 // Hn q the nth halfword of this register, as a scalar
855 // Wn q the nth word of this register, as a scalar
856 #define _DECOR(ty, decor, n) _DECOR_##ty##_##decor(n)
858 // Internal macros: `_DECOR_ty_decor(n)' applies decoration decor to register
859 // number n of type ty.
861 #define _DECOR_s_Q(n) GLUE(q, _PAIR(_PAIR(n)))
862 #define _DECOR_s_D(n) GLUE(d, _PAIR(n))
864 #define _DECOR_d_Q(n) GLUE(q, _PAIR(n))
865 #define _DECOR_d_S0(n) GLUE(s, _LOPART(n))
866 #define _DECOR_d_S1(n) GLUE(s, _LOPART(n))
868 #define _DECOR_q_D0(n) GLUE(d, _LOPART(n))
869 #define _DECOR_q_D1(n) GLUE(d, _HIPART(n))
870 #define _DECOR_q_S0(n) GLUE(s, _LOPART(_LOPART(n)))
871 #define _DECOR_q_S1(n) GLUE(s, _HIPART(_LOPART(n)))
872 #define _DECOR_q_S2(n) GLUE(s, _LOPART(_HIPART(n)))
873 #define _DECOR_q_S3(n) GLUE(s, _HIPART(_HIPART(n)))
874 #define _DECOR_q_W0(n) GLUE(d, _LOPART(n))[0]
875 #define _DECOR_q_W1(n) GLUE(d, _LOPART(n))[1]
876 #define _DECOR_q_W2(n) GLUE(d, _HIPART(n))[0]
877 #define _DECOR_q_W3(n) GLUE(d, _HIPART(n))[1]
878 #define _DECOR_q_H0(n) GLUE(d, _LOPART(n))[0]
879 #define _DECOR_q_H1(n) GLUE(d, _LOPART(n))[1]
880 #define _DECOR_q_H2(n) GLUE(d, _LOPART(n))[2]
881 #define _DECOR_q_H3(n) GLUE(d, _LOPART(n))[3]
882 #define _DECOR_q_H4(n) GLUE(d, _HIPART(n))[0]
883 #define _DECOR_q_H5(n) GLUE(d, _HIPART(n))[1]
884 #define _DECOR_q_H6(n) GLUE(d, _HIPART(n))[2]
885 #define _DECOR_q_H7(n) GLUE(d, _HIPART(n))[3]
886 #define _DECOR_q_B0(n) GLUE(d, _LOPART(n))[0]
887 #define _DECOR_q_B1(n) GLUE(d, _LOPART(n))[1]
888 #define _DECOR_q_B2(n) GLUE(d, _LOPART(n))[2]
889 #define _DECOR_q_B3(n) GLUE(d, _LOPART(n))[3]
890 #define _DECOR_q_B4(n) GLUE(d, _LOPART(n))[4]
891 #define _DECOR_q_B5(n) GLUE(d, _LOPART(n))[5]
892 #define _DECOR_q_B6(n) GLUE(d, _LOPART(n))[6]
893 #define _DECOR_q_B7(n) GLUE(d, _LOPART(n))[7]
894 #define _DECOR_q_B8(n) GLUE(d, _HIPART(n))[0]
895 #define _DECOR_q_B9(n) GLUE(d, _HIPART(n))[1]
896 #define _DECOR_q_B10(n) GLUE(d, _HIPART(n))[2]
897 #define _DECOR_q_B11(n) GLUE(d, _HIPART(n))[3]
898 #define _DECOR_q_B12(n) GLUE(d, _HIPART(n))[4]
899 #define _DECOR_q_B13(n) GLUE(d, _HIPART(n))[5]
900 #define _DECOR_q_B14(n) GLUE(d, _HIPART(n))[6]
901 #define _DECOR_q_B15(n) GLUE(d, _HIPART(n))[7]
903 // Macros for navigating the NEON register hierarchy.
904 #define S0(reg) _REGFORM(reg, S0)
905 #define S1(reg) _REGFORM(reg, S1)
906 #define S2(reg) _REGFORM(reg, S2)
907 #define S3(reg) _REGFORM(reg, S3)
908 #define D(reg) _REGFORM(reg, D)
909 #define D0(reg) _REGFORM(reg, D0)
910 #define D1(reg) _REGFORM(reg, D1)
911 #define Q(reg) _REGFORM(reg, Q)
913 // Macros for indexing quadword registers.
914 #define QB(reg, i) _REGFORM(reg, B##i)
915 #define QH(reg, i) _REGFORM(reg, H##i)
916 #define QW(reg, i) _REGFORM(reg, W##i)
918 // Macros for converting vldm/vstm ranges.
919 #define QQ(qlo, qhi) D0(qlo)-D1(qhi)
921 // Stack management and unwinding.
922 .macro setfp fp
=r11
, offset
=0
927 add
\fp
, sp
, #\offset
928 .setfp
\fp
, sp
, #\offset
930 .macro dropfp
; _dropfp
\fp
, \offset
; .endm
934 .macro _dropfp fp
, offset
=0
938 sub sp
, \fp
, #\offset
954 .macro pushreg rr
:vararg
959 .macro popreg rr
:vararg
963 .macro pushvfp rr
:vararg
968 .macro popvfp rr
:vararg
975 // No need for prologue markers on ARM.
976 #define FUNC_POSTHOOK(_) .L$_prologue_p = -1
980 ///--------------------------------------------------------------------------
981 /// AArch64-specific hacking.
985 // Set the function hooks.
986 #define FUNC_PREHOOK(_) .balign 4
987 #define FUNC_POSTHOOK(_) .cfi_startproc; .L$_prologue_p = -1
988 #define ENDFUNC_HOOK(_) .cfi_endproc
990 // Call external subroutine at ADDR, possibly via PLT.
995 // Load address of external symbol ADDR into REG.
996 .macro leaext reg
, addr
998 adrp
\reg
, :got
:\addr
999 ldr
\reg
, [\reg
, #:got_lo12:\addr]
1002 add
\reg
, \reg
, #:lo12:\addr
1007 // Set VZ (default v31) to zero.
1011 .macro vshl128 vd
, vn
, nbit
, vz
=v31
1012 // Set VD to VN shifted left by NBIT. Assume VZ (default v31) is
1013 // all-bits-zero. NBIT must be a multiple of 8.
1015 .error
"shift quantity must be whole number of bytes"
1017 ext
\vd\
().16b
, \vz\
().16b
, \vn\
().16b
, #16 - (\nbit >> 3)
1020 .macro vshr128 vd
, vn
, nbit
, vz
=v31
1021 // Set VD to VN shifted right by NBIT. Assume VZ (default v31) is
1022 // all-bits-zero. NBIT must be a multiple of 8.
1024 .error
"shift quantity must be whole number of bytes"
1026 ext
\vd\
().16b
, \vn\
().16b
, \vz\
().16b
, #\nbit >> 3
1029 // Stack management and unwinding.
1030 .macro setfp fp
=x29
, offset
=0
1031 // If you're just going through the motions with a fixed-size stack frame,
1032 // then you want to say `add x29, sp, #OFFSET' directly, which will avoid
1033 // pointlessly restoring sp later.
1036 .cfi_def_cfa_register
\fp
1038 add
\fp
, sp
, #\offset
1039 .cfi_def_cfa_register
\fp
1040 .cfi_adjust_cfa_offset
-\offset
1042 .macro dropfp
; _dropfp
\fp
, \offset
; .endm
1046 .macro _dropfp fp
, offset
=0
1049 .cfi_def_cfa_register sp
1051 sub sp
, \fp
, #\offset
1052 .cfi_def_cfa_register sp
1053 .cfi_adjust_cfa_offset
+\offset
1061 .cfi_adjust_cfa_offset
+\n
1066 .cfi_adjust_cfa_offset
-\n
1069 .macro pushreg x
, y
=nil
1072 .cfi_adjust_cfa_offset
+16
1073 .cfi_rel_offset \x
, 0
1075 stp \x
, \y
, [sp
, #-16]!
1076 .cfi_adjust_cfa_offset
+16
1077 .cfi_rel_offset \x
, 0
1078 .cfi_rel_offset \y
, 8
1082 .macro popreg x
, y
=nil
1086 .cfi_adjust_cfa_offset
-16
1088 ldp \x
, \y
, [sp
], #16
1091 .cfi_adjust_cfa_offset
-16
1095 .macro savereg x
, y
, z
=nil
1098 .cfi_rel_offset \x
, \y
1100 stp \x
, \y
, [sp
, #\z]
1101 .cfi_rel_offset \x
, \z
1102 .cfi_rel_offset \y
, \z
+ 8
1106 .macro rstrreg x
, y
, z
=nil
1111 ldp \x
, \y
, [sp
, #\z]
1120 // cmov RD, RN, CC: set RD to RN if CC is satisfied, otherwise do nothing
1121 .macro cmov rd
, rn
, cc
1122 csel
\rd
, \rn
, \rd
, \cc
1125 // Notational improvement: write `csel.CC' etc., rather than `csel ..., CC'.
1127 _(eq) _(ne) _(cs) _(cc) _(vs) _(vc) _(mi) _(pl) \
1128 _(ge) _(lt) _(gt) _(le) _(hi) _(ls) _(al) _(nv) \
1133 _(csinc) _(cinc) _(cset) \
1135 _(csinv) _(cinv) _(csetm)
1136 #define _CONDVAR(cc) _definstvar cc;
1137 #define _INSTVARS(inst) \
1138 .macro _definstvar cc; \
1139 .macro inst.\cc args:vararg; inst \args, \cc; .endm; \
1142 .purgem _definstvar;
1149 // Flag bits for `ccmp' and friends.
1155 // Flag settings for satisfying conditions.
1156 #define CCMP_MI CCMP_N
1158 #define CCMP_EQ CCMP_Z
1160 #define CCMP_CS CCMP_C
1161 #define CCMP_HS CCMP_C
1164 #define CCMP_VS CCMP_V
1166 #define CCMP_HI CCMP_C
1168 #define CCMP_LT CCMP_N
1170 #define CCMP_LE CCMP_N
1175 ///--------------------------------------------------------------------------
1178 // Default values for the various hooks.
1179 #ifndef FUNC_PREHOOK
1180 # define FUNC_PREHOOK(_)
1182 #ifndef FUNC_POSTHOOK
1183 # define FUNC_POSTHOOK(_)
1185 #ifndef ENDFUNC_HOOK
1186 # define ENDFUNC_HOOK(_)
1191 # define F(name) _##name
1193 # define F(name) name
1198 # define TYPE_FUNC(name)
1201 # define TYPE_OBJ(name)
1204 # define SIZE_OBJ(name)
1207 #if __ELF__ && !defined(FORCE_EXECUTABLE_STACK)
1208 .pushsection
.note
.GNU
-stack
, "", _SECTTY(progbits
)
1212 ///----- That's all, folks --------------------------------------------------