MahnoKropotkinvich/tinycc
1
0
Fork 0
mirror of git://repo.or.cz/tinycc.git synced 2026-06-17 15:44:18 +08:00
Code Issues Actions 1 Packages Projects Releases Wiki Activity

arm64-win32 support: symbolic op-code constants

Browse Source 
from: https://github.com/bold84/tinycc
Author: Benjamin Oldenburg <benjamin.oldenburg@ordis.co.th>  2026-04-04 16:29:28
Committer: Benjamin Oldenburg <benjamin.oldenburg@ordis.co.th>  2026-04-04 16:29:28
Branch: win_arm64_clean
Commit: 8b5ab1bb01

Also here: https://repo.or.cz/tinycc.git/shortlog/refs/mob/mob_bold84/win_arm64_clean

This and the previous commits on mob is selected parts
of that original branch. So it is not everything.

It is not, for example:
- unrelated whitespace changes in many files
- a "pin msvcrt.dll" feature in tccpe.c (why that)
- a native getenv() replacement in tcc.c (why that)
- larger changes to the win32/lib runtime and tccrun.c (not needed)
- a very gcc specific detail for struct alignent in tccgen.c
- a custom set/longjmp implementation/replacement (not needed)
- lots of rather basic test files in the win32 folder
- a 'tests/asm' folder with some files (one file renamed to 140_test...c)
- a .docs folder with one file
This commit is contained in:
Benjamin Oldenburg 2026-04-04 16:29:28 +02:00 committed by grischka
parent 44e6853cb1
commit 576cd2a923
2 changed files with 90 additions and 84 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

149
arm64-gen.c
View File

@ -181,35 +181,35 @@ static uint32_t arm64_movi(int r, uint64_t x)
uint64_t m = 0xffff; uint64_t m = 0xffff;
int e; int e;
if (!(x & ~m)) if (!(x & ~m))
return 0x52800000 | r | x << 5; // movz w(r),#(x) return ARM64_MOVZ | r | x << 5; // movz w(r),#(x)
if (!(x & ~(m << 16))) if (!(x & ~(m << 16)))
return 0x52a00000 | r | x >> 11; // movz w(r),#(x >> 16),lsl #16 return (ARM64_MOVZ | ARM64_HW(1) | r | x >> 11); // movz w(r),#(x >> 16),lsl #16
if (!(x & ~(m << 32))) if (!(x & ~(m << 32)))
return 0xd2c00000 | r | x >> 27; // movz x(r),#(x >> 32),lsl #32 return (ARM64_MOVZ64 | ARM64_HW(2) | r | x >> 27); // movz x(r),#(x >> 32),lsl #32
if (!(x & ~(m << 48))) if (!(x & ~(m << 48)))
return 0xd2e00000 | r | x >> 43; // movz x(r),#(x >> 48),lsl #48 return (ARM64_MOVZ64 | ARM64_HW(3) | r | x >> 43); // movz x(r),#(x >> 48),lsl #48
if ((x & ~m) == m << 16) if ((x & ~m) == m << 16)
return (0x12800000 | r | return (ARM64_MOVN | r |
(~x << 5 & 0x1fffe0)); // movn w(r),#(~x) (~x << 5 & 0x1fffe0)); // movn w(r),#(~x)
if ((x & ~(m << 16)) == m) if ((x & ~(m << 16)) == m)
return (0x12a00000 | r | return (ARM64_MOVN | ARM64_HW(1) | r |
(~x >> 11 & 0x1fffe0)); // movn w(r),#(~x >> 16),lsl #16 (~x >> 11 & 0x1fffe0)); // movn w(r),#(~x >> 16),lsl #16
if (!~(x | m)) if (!~(x | m))
return (0x92800000 | r | return (ARM64_MOVN64 | r |
(~x << 5 & 0x1fffe0)); // movn x(r),#(~x) (~x << 5 & 0x1fffe0)); // movn x(r),#(~x)
if (!~(x | m << 16)) if (!~(x | m << 16))
return (0x92a00000 | r | return (ARM64_MOVN64 | ARM64_HW(1) | r |
(~x >> 11 & 0x1fffe0)); // movn x(r),#(~x >> 16),lsl #16 (~x >> 11 & 0x1fffe0)); // movn x(r),#(~x >> 16),lsl #16
if (!~(x | m << 32)) if (!~(x | m << 32))
return (0x92c00000 | r | return (ARM64_MOVN64 | ARM64_HW(2) | r |
(~x >> 27 & 0x1fffe0)); // movn x(r),#(~x >> 32),lsl #32 (~x >> 27 & 0x1fffe0)); // movn x(r),#(~x >> 32),lsl #32
if (!~(x | m << 48)) if (!~(x | m << 48))
return (0x92e00000 | r | return (ARM64_MOVN64 | ARM64_HW(3) | r |
(~x >> 43 & 0x1fffe0)); // movn x(r),#(~x >> 32),lsl #32 (~x >> 43 & 0x1fffe0)); // movn x(r),#(~x >> 32),lsl #32
if (!(x >> 32) && (e = arm64_encode_bimm64(x | x << 32)) >= 0) if (!(x >> 32) && (e = arm64_encode_bimm64(x | x << 32)) >= 0)
return 0x320003e0 | r | (uint32_t)e << 10; // movi w(r),#(x) return (ARM64_ORR_IMM | r | (uint32_t)e << 10); // movi w(r),#(x)
if ((e = arm64_encode_bimm64(x)) >= 0) if ((e = arm64_encode_bimm64(x)) >= 0)
return 0xb20003e0 | r | (uint32_t)e << 10; // movi x(r),#(x) return (ARM64_ORR_IMM | ARM64_SF(1) | r | (uint32_t)e << 10); // movi x(r),#(x)
return 0; return 0;
} }
@ -221,7 +221,7 @@ static void arm64_movimm(int r, uint64_t x)
else { else {
// MOVZ/MOVN and 1-3 MOVKs // MOVZ/MOVN and 1-3 MOVKs
int z = 0, m = 0; int z = 0, m = 0;
uint32_t mov1 = 0xd2800000; // movz uint32_t mov1 = ARM64_MOVZ64; // movz
uint64_t x1 = x; uint64_t x1 = x;
for (i = 0; i < 64; i += 16) { for (i = 0; i < 64; i += 16) {
z += !(x >> i & 0xffff); z += !(x >> i & 0xffff);
@ -229,7 +229,7 @@ static void arm64_movimm(int r, uint64_t x)
} }
if (m > z) { if (m > z) {
x1 = ~x; x1 = ~x;
mov1 = 0x92800000; // movn mov1 = ARM64_MOVN64; // movn
} }
for (i = 0; i < 64; i += 16) for (i = 0; i < 64; i += 16)
if (x1 >> i & 0xffff) { if (x1 >> i & 0xffff) {
@ -239,7 +239,7 @@ static void arm64_movimm(int r, uint64_t x)
} }
for (i += 16; i < 64; i += 16) for (i += 16; i < 64; i += 16)
if (x1 >> i & 0xffff) if (x1 >> i & 0xffff)
o(0xf2800000 | r | (x >> i & 0xffff) << 5 | i << 17); o(ARM64_MOVK | ARM64_SF(1) | r | (x >> i & 0xffff) << 5 | i << 17);
// movk x(r),#(*),lsl #(i) // movk x(r),#(*),lsl #(i)
} }
} }
@ -254,8 +254,8 @@ ST_FUNC void gsym_addr(int t_, int a_)
uint32_t next = read32le(ptr); uint32_t next = read32le(ptr);
if (a - t + 0x8000000 >= 0x10000000) if (a - t + 0x8000000 >= 0x10000000)
tcc_error("branch out of range"); tcc_error("branch out of range");
write32le(ptr, (a - t == 4 ? 0xd503201f : // nop write32le(ptr, (a - t == 4 ? ARM64_NOP :
0x14000000 | ((a - t) >> 2 & 0x3ffffff))); // b ARM64_B | ((a - t) >> 2 & 0x3ffffff)));
t = next; t = next;
} }
} }
@ -290,11 +290,10 @@ static void arm64_spoff(int reg, uint64_t off)
if (sub) if (sub)
off = -off; off = -off;
if (off < 4096) if (off < 4096)
o(0x910003e0 | sub << 30 | reg | off << 10); o(ARM64_ADD_IMM | ARM64_SF(1) | ARM64_RN(31) | ARM64_RD(reg) | ARM64_IMM12(off));
// (add|sub) x(reg),sp,#(off)
else { else {
arm64_movimm(30, off); // use x30 for offset arm64_movimm(30, off);
o(0x8b3e63e0 | sub << 30 | reg); // (add|sub) x(reg),sp,x30 o(ARM64_ADD_REG | ARM64_SF(1) | ARM64_RM(30) | ARM64_RN(31) | ARM64_RD(reg) | (sub << 30));
} }
} }
@ -323,14 +322,14 @@ static void arm64_ldrx(int sg, int sz_, int dst, int bas, uint64_t off)
if (sz >= 2) if (sz >= 2)
sg = 0; sg = 0;
if (!(off & ~scaled_mask)) if (!(off & ~scaled_mask))
o(0x39400000 | dst | bas << 5 | off << (10 - sz) | o(ARM64_LDR_B | dst | bas << 5 | off << (10 - sz) |
(uint32_t)!!sg << 23 | sz << 30); // ldr(*) x(dst),[x(bas),#(off)] (uint32_t)!!sg << 23 | sz << 30); // ldr(*) x(dst),[x(bas),#(off)]
else if (off < 256 || -off <= 256) else if (off < 256 || -off <= 256)
o(0x38400000 | dst | bas << 5 | (off & 511) << 12 | o(ARM64_LDUR_B | dst | bas << 5 | (off & 511) << 12 |
(uint32_t)!!sg << 23 | sz << 30); // ldur(*) x(dst),[x(bas),#(off)] (uint32_t)!!sg << 23 | sz << 30); // ldur(*) x(dst),[x(bas),#(off)]
else { else {
arm64_movimm(30, off); // use x30 for offset arm64_movimm(30, off); // use x30 for offset
o(0x38206800 | dst | bas << 5 | (uint32_t)30 << 16 | o(ARM64_LDR_B_REG | dst | bas << 5 | (uint32_t)30 << 16 |
(uint32_t)(!!sg + 1) << 22 | sz << 30); // ldr(*) x(dst),[x(bas),x30] (uint32_t)(!!sg + 1) << 22 | sz << 30); // ldr(*) x(dst),[x(bas),x30]
} }
} }
@ -341,14 +340,14 @@ static void arm64_ldrv(int sz_, int dst, int bas, uint64_t off)
uint64_t scaled_mask = 0xffful << sz; uint64_t scaled_mask = 0xffful << sz;
if (!(off & ~scaled_mask)) if (!(off & ~scaled_mask))
o(0x3d400000 | dst | bas << 5 | off << (10 - sz) | o(ARM64_LDR_SCALAR | dst | bas << 5 | off << (10 - sz) |
(sz & 4) << 21 | (sz & 3) << 30); // ldr (s|d|q)(dst),[x(bas),#(off)] (sz & 4) << 21 | (sz & 3) << 30); // ldr (s|d|q)(dst),[x(bas),#(off)]
else if (off < 256 || -off <= 256) else if (off < 256 || -off <= 256)
o(0x3c400000 | dst | bas << 5 | (off & 511) << 12 | o(ARM64_LDUR_Q_SIMD | dst | bas << 5 | (off & 511) << 12 |
(sz & 4) << 21 | (sz & 3) << 30); // ldur (s|d|q)(dst),[x(bas),#(off)] (sz & 4) << 21 | (sz & 3) << 30); // ldur (s|d|q)(dst),[x(bas),#(off)]
else { else {
arm64_movimm(30, off); // use x30 for offset arm64_movimm(30, off); // use x30 for offset
o(0x3c606800 | dst | bas << 5 | (uint32_t)30 << 16 | o(ARM64_LDR_Q_REG | dst | bas << 5 | (uint32_t)30 << 16 |
sz << 30 | (sz & 4) << 21); // ldr (s|d|q)(dst),[x(bas),x30] sz << 30 | (sz & 4) << 21); // ldr (s|d|q)(dst),[x(bas),x30]
} }
} }
@ -480,25 +479,27 @@ static void arm64_sym(int r, Sym *sym, unsigned long addend)
o(ARM64_ADD_IMM | ARM64_SF(1) | ARM64_RN(r) | r); // add xr, xr, #sym o(ARM64_ADD_IMM | ARM64_SF(1) | ARM64_RN(r) | r); // add xr, xr, #sym
#else #else
greloca(cur_text_section, sym, ind, R_AARCH64_ADR_GOT_PAGE, 0); greloca(cur_text_section, sym, ind, R_AARCH64_ADR_GOT_PAGE, 0);
o(0x90000000 | r); // adrp xr, #sym o(ARM64_ADRP | r); // adrp xr, #sym
greloca(cur_text_section, sym, ind, R_AARCH64_LD64_GOT_LO12_NC, 0); greloca(cur_text_section, sym, ind, R_AARCH64_LD64_GOT_LO12_NC, 0);
o(0xf9400000 | r | (r << 5)); // ld xr,[xr, #sym] o(ARM64_LDR_X | ARM64_RN(r) | r); // ld xr,[xr, #sym]
#endif #endif
if (addend) { if (addend) {
// add xr, xr, #addend // add xr, xr, #addend
if (addend & 0xffful) if (addend & 0xffful)
o(0x91000000 | r | r << 5 | (addend & 0xfff) << 10); o(ARM64_ADD_IMM | ARM64_SF(1) | ARM64_RN(r) | r |
(addend & 0xfff) << 10);
if (addend > 0xffful) { if (addend > 0xffful) {
// add xr, xr, #addend, lsl #12 // add xr, xr, #addend, lsl #12
if (addend & 0xfff000ul) if (addend & 0xfff000ul)
o(0x91400000 | r | r << 5 | ((addend >> 12) & 0xfff) << 10); o(ARM64_ADD_IMM | ARM64_SF(1) | ARM64_SH(1) |
ARM64_RN(r) | r | ((addend >> 12) & 0xfff) << 10);
if (addend > 0xfffffful) { if (addend > 0xfffffful) {
/* very unlikely */ /* very unlikely */
int t = r ? 0 : 1; int t = r ? 0 : 1;
o(0xf81f0fe0 | t); /* str xt, [sp, #-16]! */ o(ARM64_STR_X_PRE | 0x001F0FE0U | t); /* str xt, [sp, #-16]! */
arm64_movimm(t, addend & ~0xfffffful); // use xt for addent arm64_movimm(t, addend & ~0xfffffful); // use xt for addent
o(0x8B000000 | (t << 16) | (r << 5) | r); /* add xr, xr, xt */ o(ARM64_ADD_REG | ARM64_SF(1) | ARM64_RM(t) | ARM64_RN(r) | r); /* add xr, xr, xt */
o(0xf84107e0 | t); /* ldr xt, [sp], #16 */ o(ARM64_LDR_X_POST | 0x000107E0U | t); /* ldr xt, [sp], #16 */
} }
} }
} }
@ -512,7 +513,7 @@ ST_FUNC void load(int r, SValue *sv)
int svr = sv->r & ~(VT_BOUNDED | VT_NONCONST); int svr = sv->r & ~(VT_BOUNDED | VT_NONCONST);
int svrv = svr & VT_VALMASK; int svrv = svr & VT_VALMASK;
uint64_t svcul = sv->c.i; uint64_t svcul = sv->c.i;
uint64_t svcoff = (int32_t)sv->c.i; uint64_t svcoff = (uint64_t)(int64_t)(int32_t)sv->c.i;
if (svr == (VT_LOCAL | VT_LVAL)) { if (svr == (VT_LOCAL | VT_LVAL)) {
if (IS_FREG(r)) if (IS_FREG(r))
@ -553,13 +554,13 @@ ST_FUNC void load(int r, SValue *sv)
if (svr == (VT_CONST | VT_LVAL | VT_SYM)) { if (svr == (VT_CONST | VT_LVAL | VT_SYM)) {
arm64_sym(30, sv->sym, // use x30 for address arm64_sym(30, sv->sym, // use x30 for address
arm64_check_offset(0, arm64_type_size(svtt), svcul)); arm64_check_offset(0, arm64_type_size(svtt), svcoff));
if (IS_FREG(r)) if (IS_FREG(r))
arm64_ldrv(arm64_type_size(svtt), fltr(r), 30, arm64_ldrv(arm64_type_size(svtt), fltr(r), 30,
arm64_check_offset(1, arm64_type_size(svtt), svcul)); arm64_check_offset(1, arm64_type_size(svtt), svcoff));
else else
arm64_ldrx(!(svtt&VT_UNSIGNED), arm64_type_size(svtt), intr(r), 30, arm64_ldrx(!(svtt&VT_UNSIGNED), arm64_type_size(svtt), intr(r), 30,
arm64_check_offset(1, arm64_type_size(svtt), svcul)); arm64_check_offset(1, arm64_type_size(svtt), svcoff));
return; return;
} }
@ -578,12 +579,12 @@ ST_FUNC void load(int r, SValue *sv)
if (svr < VT_CONST) { if (svr < VT_CONST) {
if (IS_FREG(r) && IS_FREG(svr)) if (IS_FREG(r) && IS_FREG(svr))
if (svtt == VT_LDOUBLE) if (svtt == VT_LDOUBLE)
o(0x4ea01c00 | fltr(r) | fltr(svr) << 5); o(ARM64_MOV_V16B | fltr(r) | fltr(svr) << 5);
// mov v(r).16b,v(svr).16b // mov v(r).16b,v(svr).16b
else else
o(0x1e604000 | fltr(r) | fltr(svr) << 5); // fmov d(r),d(svr) o(ARM64_FMOV_SCALAR | fltr(r) | fltr(svr) << 5); // fmov d(r),d(svr)
else if (!IS_FREG(r) && !IS_FREG(svr)) else if (!IS_FREG(r) && !IS_FREG(svr))
o(0xaa0003e0 | intr(r) | intr(svr) << 16); // mov x(r),x(svr) o(ARM64_MOV_REG | ARM64_SF(1) | intr(r) | intr(svr) << 16); // mov x(r),x(svr)
else else
assert(0); assert(0);
return; return;
@ -602,7 +603,7 @@ ST_FUNC void load(int r, SValue *sv)
if (svr == VT_JMP || svr == VT_JMPI) { if (svr == VT_JMP || svr == VT_JMPI) {
int t = (svr == VT_JMPI); int t = (svr == VT_JMPI);
arm64_movimm(intr(r), t); arm64_movimm(intr(r), t);
o(0x14000002); // b .+8 o(ARM64_B | 2); // b .+8
gsym(svcul); gsym(svcul);
arm64_movimm(intr(r), t ^ 1); arm64_movimm(intr(r), t ^ 1);
return; return;
@ -623,7 +624,7 @@ ST_FUNC void load(int r, SValue *sv)
return; return;
} }
printf("load(%x, (%x, %x, %llx))\n", r, svtt, sv->r, (long long)svcul); printf("load(%x, (%x, %x, %lx))\n", r, svtt, sv->r, (long)svcul);
assert(0); assert(0);
} }
@ -632,7 +633,7 @@ ST_FUNC void store(int r, SValue *sv)
int svtt = sv->type.t; int svtt = sv->type.t;
int svr = sv->r & ~VT_BOUNDED; int svr = sv->r & ~VT_BOUNDED;
int svrv = svr & VT_VALMASK; int svrv = svr & VT_VALMASK;
uint64_t svcoff = (int32_t)sv->c.i; uint64_t svcoff = (uint64_t)(int64_t)(int32_t)sv->c.i;
if (svr == (VT_LOCAL | VT_LVAL)) { if (svr == (VT_LOCAL | VT_LVAL)) {
if (IS_FREG(r)) if (IS_FREG(r))
@ -679,7 +680,7 @@ ST_FUNC void store(int r, SValue *sv)
return; return;
} }
printf("store(%x, (%x, %x, %llx))\n", r, svtt, sv->r, (long long)svcoff); printf("store(%x, (%x, %x, %lx))\n", r, svtt, sv->r, (long)svcoff);
assert(0); assert(0);
} }
@ -688,13 +689,13 @@ static void arm64_gen_bl_or_b(int b)
if ((vtop->r & (VT_VALMASK | VT_LVAL)) == VT_CONST && (vtop->r & VT_SYM)) { if ((vtop->r & (VT_VALMASK | VT_LVAL)) == VT_CONST && (vtop->r & VT_SYM)) {
greloca(cur_text_section, vtop->sym, ind, greloca(cur_text_section, vtop->sym, ind,
b ? R_AARCH64_JUMP26 : R_AARCH64_CALL26, 0); b ? R_AARCH64_JUMP26 : R_AARCH64_CALL26, 0);
o(0x14000000 | (uint32_t)!b << 31); // b/bl . o(b ? ARM64_B : ARM64_BL); // b/bl .
} }
else { else {
#ifdef CONFIG_TCC_BCHECK #ifdef CONFIG_TCC_BCHECK
vtop->r &= ~VT_MUSTBOUND; vtop->r &= ~VT_MUSTBOUND;
#endif #endif
o(0xd61f0000 | (uint32_t)!b << 21 | intr(gv(RC_R30)) << 5); // br/blr o((b ? ARM64_BR : ARM64_BLR) | intr(gv(RC_R30)) << 5); // br/blr
} }
} }
@ -705,7 +706,7 @@ static void gen_bounds_call(int v)
Sym *sym = external_helper_sym(v); Sym *sym = external_helper_sym(v);
greloca(cur_text_section, sym, ind, R_AARCH64_CALL26, 0); greloca(cur_text_section, sym, ind, R_AARCH64_CALL26, 0);
o(0x94000000); // bl o(ARM64_BL); // bl
} }
static void gen_bounds_prolog(void) static void gen_bounds_prolog(void)
@ -714,10 +715,10 @@ static void gen_bounds_prolog(void)
func_bound_offset = lbounds_section->data_offset; func_bound_offset = lbounds_section->data_offset;
func_bound_ind = ind; func_bound_ind = ind;
func_bound_add_epilog = 0; func_bound_add_epilog = 0;
o(0xd503201f); /* nop -> mov x0, lbound section pointer */ o(ARM64_NOP); /* nop -> mov x0, lbound section pointer */
o(0xd503201f); o(ARM64_NOP);
o(0xd503201f); o(ARM64_NOP);
o(0xd503201f); /* nop -> call __bound_local_new */ o(ARM64_NOP); /* nop -> call __bound_local_new */
} }
static void gen_bounds_epilog(void) static void gen_bounds_epilog(void)
@ -1222,8 +1223,8 @@ ST_FUNC void gfunc_call(int nb_args)
for (j = 0; j < n; j++) for (j = 0; j < n; j++)
o(0x3d000100 | o(0x3d000100 |
(sz & 16) << 19 | -(sz & 8) << 27 | (sz & 4) << 29 | (sz & 16) << 19 | -(sz & 8) << 27 | (sz & 4) << 29 |
(a[i] / 2 - 8 + j) | (fltr(REG_FRET) + j) |
j << 10); // str ([sdq])(*),[x8,#(j * sz)] j << 10); // str ([sdq])(j),[x8,#(j * sz)]
} }
} }
} }
@ -1432,7 +1433,7 @@ ST_FUNC void gen_va_arg(CType *t)
uint32_t r0, r1; uint32_t r0, r1;
#ifdef TCC_TARGET_PE #ifdef TCC_TARGET_PE
int indirect = 0, slot = (size + 7) & -8; int indirect = 0, slot = size + 7 & -8;
if (size > 16) if (size > 16)
indirect = 1, slot = 8; indirect = 1, slot = 8;
@ -1443,7 +1444,7 @@ ST_FUNC void gen_va_arg(CType *t)
vtop[0].r = r1 | VT_LVAL; vtop[0].r = r1 | VT_LVAL;
r1 = intr(r1); r1 = intr(r1);
o(0xF9400000 | r0 << 5 | r1); // ldr x(r1),[x(r0)] // ap o(ARM64_LDR_X | ARM64_RN(r0) | r1); // ldr x(r1),[x(r0)] // ap
if (slot) { if (slot) {
if (slot == 16) { if (slot == 16) {
o(0x910363be); // add x30,x29,#216 o(0x910363be); // add x30,x29,#216
@ -1460,7 +1461,7 @@ ST_FUNC void gen_va_arg(CType *t)
} }
if (indirect) if (indirect)
o(0xF9400000 | ARM64_RN(r1) | r1); // ldr x(r1),[x(r1)] o(ARM64_LDR_X | ARM64_RN(r1) | r1); // ldr x(r1),[x(r1)]
#else /* !PE */ #else /* !PE */
unsigned fsize = size, hfa = 1; unsigned fsize = size, hfa = 1;
@ -1476,6 +1477,7 @@ ST_FUNC void gen_va_arg(CType *t)
if (!hfa) { if (!hfa) {
uint32_t n = size > 16 ? 8 : (size + 7) & -8; uint32_t n = size > 16 ? 8 : (size + 7) & -8;
#if !defined(TCC_TARGET_MACHO) #if !defined(TCC_TARGET_MACHO)
o(0xb940181e | r0 << 5); // ldr w30,[x(r0),#24] // __gr_offs o(0xb940181e | r0 << 5); // ldr w30,[x(r0),#24] // __gr_offs
if (align == 16) { if (align == 16) {
@ -1486,21 +1488,24 @@ ST_FUNC void gen_va_arg(CType *t)
o(0x310003c0 | r1 | n << 10); // adds w(r1),w30,#(n) o(0x310003c0 | r1 | n << 10); // adds w(r1),w30,#(n)
o(0x540000ad); // b.le .+20 o(0x540000ad); // b.le .+20
#endif #endif
o(0xf9400000 | r1 | r0 << 5); // ldr x(r1),[x(r0)] // __stack
o(ARM64_LDR_X | ARM64_RN(r0) | r1); // ldr x(r1),[x(r0)] // __stack
if (align == 16) { if (align == 16) {
o(0x91003c00 | r1 | r1 << 5); // add x(r1),x(r1),#15 o(0x91003c00 | r1 | r1 << 5); // add x(r1),x(r1),#15
o(0x927cec00 | r1 | r1 << 5); // and x(r1),x(r1),#-16 o(0x927cec00 | r1 | r1 << 5); // and x(r1),x(r1),#-16
} }
o(0x9100001e | r1 << 5 | n << 10); // add x30,x(r1),#(n) o(0x9100001e | r1 << 5 | n << 10); // add x30,x(r1),#(n)
o(0xf900001e | r0 << 5); // str x30,[x(r0)] // __stack o(0xf900001e | r0 << 5); // str x30,[x(r0)] // __stack
#if !defined(TCC_TARGET_MACHO) #if !defined(TCC_TARGET_MACHO)
o(0x14000004); // b .+16 o(ARM64_B | 4); // b .+16
o(0xb9001800 | r1 | r0 << 5); // str w(r1),[x(r0),#24] // __gr_offs o(0xb9001800 | r1 | r0 << 5); // str w(r1),[x(r0),#24] // __gr_offs
o(0xf9400400 | r1 | r0 << 5); // ldr x(r1),[x(r0),#8] // __gr_top o(0xf9400400 | r1 | r0 << 5); // ldr x(r1),[x(r0),#8] // __gr_top
o(0x8b3ec000 | r1 | r1 << 5); // add x(r1),x(r1),w30,sxtw o(0x8b3ec000 | r1 | r1 << 5); // add x(r1),x(r1),w30,sxtw
#endif #endif
if (size > 16) if (size > 16)
o(0xf9400000 | r1 | r1 << 5); // ldr x(r1),[x(r1)] o(ARM64_LDR_X | ARM64_RN(r1) | r1); // ldr x(r1),[x(r1)]
} }
else { else {
uint32_t ssz = (size + 7) & -(uint32_t)8; uint32_t ssz = (size + 7) & -(uint32_t)8;
@ -1511,7 +1516,7 @@ ST_FUNC void gen_va_arg(CType *t)
o(0x310003c0 | r1 | rsz << 10); // adds w(r1),w30,#(rsz) o(0x310003c0 | r1 | rsz << 10); // adds w(r1),w30,#(rsz)
b1 = ind; o(0x5400000d); // b.le lab1 b1 = ind; o(0x5400000d); // b.le lab1
#endif #endif
o(0xf9400000 | r1 | r0 << 5); // ldr x(r1),[x(r0)] // __stack o(ARM64_LDR_X | ARM64_RN(r0) | r1); // ldr x(r1),[x(r0)] // __stack
if (fsize == 16) { if (fsize == 16) {
o(0x91003c00 | r1 | r1 << 5); // add x(r1),x(r1),#15 o(0x91003c00 | r1 | r1 << 5); // add x(r1),x(r1),#15
o(0x927cec00 | r1 | r1 << 5); // and x(r1),x(r1),#-16 o(0x927cec00 | r1 | r1 << 5); // and x(r1),x(r1),#-16
@ -1519,7 +1524,7 @@ ST_FUNC void gen_va_arg(CType *t)
o(0x9100001e | r1 << 5 | ssz << 10); // add x30,x(r1),#(ssz) o(0x9100001e | r1 << 5 | ssz << 10); // add x30,x(r1),#(ssz)
o(0xf900001e | r0 << 5); // str x30,[x(r0)] // __stack o(0xf900001e | r0 << 5); // str x30,[x(r0)] // __stack
#if !defined(TCC_TARGET_MACHO) #if !defined(TCC_TARGET_MACHO)
b2 = ind; o(0x14000000); // b lab2 b2 = ind; o(ARM64_B); // b lab2
// lab1: // lab1:
write32le(cur_text_section->data + b1, 0x5400000d | (ind - b1) << 3); write32le(cur_text_section->data + b1, 0x5400000d | (ind - b1) << 3);
o(0xb9001c00 | r1 | r0 << 5); // str w(r1),[x(r0),#28] // __vr_offs o(0xb9001c00 | r1 | r0 << 5); // str w(r1),[x(r0),#28] // __vr_offs
@ -1541,7 +1546,7 @@ ST_FUNC void gen_va_arg(CType *t)
(uint32_t)(hfa != 3) << 21); // st(hfa) {v28.(s|d),...}[0],[x(r1)] (uint32_t)(hfa != 3) << 21); // st(hfa) {v28.(s|d),...}[0],[x(r1)]
} }
// lab2: // lab2:
write32le(cur_text_section->data + b2, 0x14000000 | (ind - b2) >> 2); write32le(cur_text_section->data + b2, ARM64_B | ((ind - b2) >> 2));
#endif #endif
} }
#endif /* not pe */ #endif /* not pe */
@ -1590,7 +1595,7 @@ ST_FUNC void gfunc_return(CType *func_type)
for (j = 0; j < n; j++) for (j = 0; j < n; j++)
o(0x3d400000 | o(0x3d400000 |
(sz & 16) << 19 | -(sz & 8) << 27 | (sz & 4) << 29 | (sz & 16) << 19 | -(sz & 8) << 27 | (sz & 4) << 29 |
j | j << 10); // ldr ([sdq])(*),[x0,#(j * sz)] (fltr(REG_FRET) + j) | j << 10); // ldr ([sdq])(j),[x0,#(j * sz)]
} }
else else
gv(RC_FRET); gv(RC_FRET);
@ -1635,7 +1640,7 @@ ST_FUNC void gen_fill_nops(int bytes)
if ((bytes & 3)) if ((bytes & 3))
tcc_error("alignment of code section not multiple of 4"); tcc_error("alignment of code section not multiple of 4");
while (bytes > 0) { while (bytes > 0) {
o(0xd503201f); // nop o(ARM64_NOP); // nop
bytes -= 4; bytes -= 4;
} }
} }
@ -1654,7 +1659,7 @@ ST_FUNC int gjmp(int t)
ST_FUNC void gjmp_addr(int a) ST_FUNC void gjmp_addr(int a)
{ {
assert(a - ind + 0x8000000 < 0x10000000); assert(a - ind + 0x8000000 < 0x10000000);
o(0x14000000 | ((a - ind) >> 2 & 0x3ffffff)); o(ARM64_B | (((a - ind) >> 2) & 0x3ffffff));
} }
ST_FUNC int gjmp_append(int n, int t) ST_FUNC int gjmp_append(int n, int t)
@ -1759,7 +1764,7 @@ static int arm64_gen_opic(int op, uint32_t l, int rev, uint64_t val,
val = l ? val : (uint32_t)val; val = l ? val : (uint32_t)val;
if (!(val & ~0xffful)) if (!(val & ~0xffful))
o(0x11000000 | l << 31 | s << 30 | x | a << 5 | val << 10); o(0x11000000 | l << 31 | s << 30 | x | a << 5 | val << 10);
else if (!(val & ~(uint64_t)0xfff000)) else if (!(val & ~0xfff000ul))
o(0x11400000 | l << 31 | s << 30 | x | a << 5 | val >> 12 << 10); o(0x11400000 | l << 31 | s << 30 | x | a << 5 | val >> 12 << 10);
else { else {
arm64_movimm(30, val); // use x30 arm64_movimm(30, val); // use x30
@ -2199,7 +2204,7 @@ ST_FUNC void gen_increment_tcov (SValue *sv)
vtop->r = r1 = get_reg(RC_INT); vtop->r = r1 = get_reg(RC_INT);
r2 = get_reg(RC_INT); r2 = get_reg(RC_INT);
arm64_sym(r1, sv->sym, 0); arm64_sym(r1, sv->sym, 0);
o(0xf9400000 | (intr(r1)<<5) | intr(r2)); // ldr r2, [r1] o(ARM64_LDR_X | ARM64_RN(intr(r1)) | intr(r2)); // ldr r2, [r1]
o(0x91000400 | (intr(r2)<<5) | intr(r2)); // add r2, r2, #1 o(0x91000400 | (intr(r2)<<5) | intr(r2)); // add r2, r2, #1
o(0xf9000000 | (intr(r1)<<5) | intr(r2)); // str r2, [r1] o(0xf9000000 | (intr(r1)<<5) | intr(r2)); // str r2, [r1]
vpop(); vpop();
@ -2236,21 +2241,21 @@ ST_FUNC void gen_clear_cache(void)
o(0x1ac02000 | isz | p << 5 | isz << 16); // lsl w(isz),w(p),w(isz) o(0x1ac02000 | isz | p << 5 | isz << 16); // lsl w(isz),w(p),w(isz)
o(0x51000400 | p | dsz << 5); // sub w(p),w(dsz),#1 o(0x51000400 | p | dsz << 5); // sub w(p),w(dsz),#1
o(0x8a240004 | p | beg << 5 | p << 16); // bic x(p),x(beg),x(p) o(0x8a240004 | p | beg << 5 | p << 16); // bic x(p),x(beg),x(p)
b1 = ind; o(0x14000000); // b b1 = ind; o(ARM64_B); // b
lab1 = ind; lab1 = ind;
o(0xd50b7b20 | p); // dc cvau,x(p) o(0xd50b7b20 | p); // dc cvau,x(p)
o(0x8b000000 | p | p << 5 | dsz << 16); // add x(p),x(p),x(dsz) o(0x8b000000 | p | p << 5 | dsz << 16); // add x(p),x(p),x(dsz)
write32le(cur_text_section->data + b1, 0x14000000 | (ind - b1) >> 2); write32le(cur_text_section->data + b1, ARM64_B | ((ind - b1) >> 2));
o(0xeb00001f | p << 5 | end << 16); // cmp x(p),x(end) o(0xeb00001f | p << 5 | end << 16); // cmp x(p),x(end)
o(0x54ffffa3 | ((lab1 - ind) << 3 & 0xffffe0)); // b.cc lab1 o(0x54ffffa3 | ((lab1 - ind) << 3 & 0xffffe0)); // b.cc lab1
o(0xd5033b9f); // dsb ish o(0xd5033b9f); // dsb ish
o(0x51000400 | p | isz << 5); // sub w(p),w(isz),#1 o(0x51000400 | p | isz << 5); // sub w(p),w(isz),#1
o(0x8a240004 | p | beg << 5 | p << 16); // bic x(p),x(beg),x(p) o(0x8a240004 | p | beg << 5 | p << 16); // bic x(p),x(beg),x(p)
b1 = ind; o(0x14000000); // b b1 = ind; o(ARM64_B); // b
lab1 = ind; lab1 = ind;
o(0xd50b7520 | p); // ic ivau,x(p) o(0xd50b7520 | p); // ic ivau,x(p)
o(0x8b000000 | p | p << 5 | isz << 16); // add x(p),x(p),x(isz) o(0x8b000000 | p | p << 5 | isz << 16); // add x(p),x(p),x(isz)
write32le(cur_text_section->data + b1, 0x14000000 | (ind - b1) >> 2); write32le(cur_text_section->data + b1, ARM64_B | ((ind - b1) >> 2));
o(0xeb00001f | p << 5 | end << 16); // cmp x(p),x(end) o(0xeb00001f | p << 5 | end << 16); // cmp x(p),x(end)
o(0x54ffffa3 | ((lab1 - ind) << 3 & 0xffffe0)); // b.cc lab1 o(0x54ffffa3 | ((lab1 - ind) << 3 & 0xffffe0)); // b.cc lab1
o(0xd5033b9f); // dsb ish o(0xd5033b9f); // dsb ish

25
arm64-link.c
View File

@ -131,17 +131,18 @@ ST_FUNC void relocate_plt(TCCState *s1)
uint64_t off = (got >> 12) - (plt >> 12); uint64_t off = (got >> 12) - (plt >> 12);
if ((off + ((uint32_t)1 << 20)) >> 21) if ((off + ((uint32_t)1 << 20)) >> 21)
tcc_error_noabort("Failed relocating PLT (off=0x%lx, got=0x%lx, plt=0x%lx)", (long)off, (long)got, (long)plt); tcc_error_noabort("Failed relocating PLT (off=0x%lx, got=0x%lx, plt=0x%lx)", (long)off, (long)got, (long)plt);
write32le(p, 0xa9bf7bf0); // stp x16,x30,[sp,#-16]! write32le(p, ARM64_STP_X_PRE | ARM64_RT(16) | ARM64_RT2(30) |
write32le(p + 4, (0x90000010 | // adrp x16,... ARM64_RN(31) | ARM64_IMM7(-2)); // stp x16,x30,[sp,#-16]!
write32le(p + 4, (ARM64_ADRP | ARM64_RD(16) | // adrp x16,...
(off & 0x1ffffc) << 3 | (off & 3) << 29)); (off & 0x1ffffc) << 3 | (off & 3) << 29));
write32le(p + 8, (0xf9400211 | // ldr x17,[x16,#...] write32le(p + 8, (ARM64_LDR_X | ARM64_RT(17) | ARM64_RN(16) | // ldr x17,[x16,#...]
(got & 0xff8) << 7)); (got & 0xff8) << 7));
write32le(p + 12, (0x91000210 | // add x16,x16,#... write32le(p + 12, (ARM64_ADD_IMM | ARM64_SF(1) | ARM64_RD(16) | ARM64_RN(16) | // add x16,x16,#...
(got & 0xfff) << 10)); (got & 0xfff) << 10));
write32le(p + 16, 0xd61f0220); // br x17 write32le(p + 16, ARM64_BR | ARM64_RN(17)); // br x17
write32le(p + 20, 0xd503201f); // nop write32le(p + 20, ARM64_NOP); // nop
write32le(p + 24, 0xd503201f); // nop write32le(p + 24, ARM64_NOP); // nop
write32le(p + 28, 0xd503201f); // nop write32le(p + 28, ARM64_NOP); // nop
p += 32; p += 32;
got = s1->got->sh_addr; got = s1->got->sh_addr;
while (p < p_end) { while (p < p_end) {
@ -150,13 +151,13 @@ ST_FUNC void relocate_plt(TCCState *s1)
uint64_t off = (addr >> 12) - (pc >> 12); uint64_t off = (addr >> 12) - (pc >> 12);
if ((off + ((uint32_t)1 << 20)) >> 21) if ((off + ((uint32_t)1 << 20)) >> 21)
tcc_error_noabort("Failed relocating PLT (off=0x%lx, addr=0x%lx, pc=0x%lx)", (long)off, (long)addr, (long)pc); tcc_error_noabort("Failed relocating PLT (off=0x%lx, addr=0x%lx, pc=0x%lx)", (long)off, (long)addr, (long)pc);
write32le(p, (0x90000010 | // adrp x16,... write32le(p, (ARM64_ADRP | ARM64_RD(16) | // adrp x16,...
(off & 0x1ffffc) << 3 | (off & 3) << 29)); (off & 0x1ffffc) << 3 | (off & 3) << 29));
write32le(p + 4, (0xf9400211 | // ldr x17,[x16,#...] write32le(p + 4, (ARM64_LDR_X | ARM64_RT(17) | ARM64_RN(16) | // ldr x17,[x16,#...]
(addr & 0xff8) << 7)); (addr & 0xff8) << 7));
write32le(p + 8, (0x91000210 | // add x16,x16,#... write32le(p + 8, (ARM64_ADD_IMM | ARM64_SF(1) | ARM64_RD(16) | ARM64_RN(16) | // add x16,x16,#...
(addr & 0xfff) << 10)); (addr & 0xfff) << 10));
write32le(p + 12, 0xd61f0220); // br x17 write32le(p + 12, ARM64_BR | ARM64_RN(17)); // br x17
p += 16; p += 16;
} }
} }