tcg-arm: Improve GUEST_BASE qemu_ld/st

If we pull the code to emit the actual load/store into a subroutine,
we can share the reg+reg addressing mode code between softmmu and
usermode.  This lets us load GUEST_BASE into a temporary register
rather than attempting to add it piece-wise to the address.

Which lets us use movw+movt for armv7, rather than (up to) 4 adds.
Code size for pre-armv7 stays the same.

Signed-off-by: Richard Henderson <rth@twiddle.net>
This commit is contained in:
Richard Henderson 2013-09-12 15:06:23 -07:00
parent 15ecf6e394
commit 091d567771

View File

@ -1367,33 +1367,11 @@ static void tcg_out_qemu_st_slow_path(TCGContext *s, TCGLabelQemuLdst *lb)
}
#endif /* SOFTMMU */
static void tcg_out_qemu_ld(TCGContext *s, const TCGArg *args, bool is64)
static inline void tcg_out_qemu_ld_index(TCGContext *s, TCGMemOp opc,
TCGReg datalo, TCGReg datahi,
TCGReg addrlo, TCGReg addend)
{
TCGReg addrlo, datalo, datahi, addrhi __attribute__((unused));
TCGMemOp opc, bswap;
#ifdef CONFIG_SOFTMMU
TCGMemOp s_bits;
int mem_index;
TCGReg addend;
uint8_t *label_ptr;
#endif
datalo = *args++;
datahi = (is64 ? *args++ : 0);
addrlo = *args++;
addrhi = (TARGET_LONG_BITS == 64 ? *args++ : 0);
opc = *args++;
bswap = opc & MO_BSWAP;
#ifdef CONFIG_SOFTMMU
s_bits = opc & MO_SIZE;
mem_index = *args;
addend = tcg_out_tlb_read(s, addrlo, addrhi, s_bits, mem_index, 1);
/* This a conditional BL only to load a pointer within this opcode into LR
for the slow path. We will not be using the value for a tail call. */
label_ptr = s->code_ptr;
tcg_out_bl_noaddr(s, COND_NE);
TCGMemOp bswap = opc & MO_BSWAP;
switch (opc & MO_SSIZE) {
case MO_UB:
@ -1425,8 +1403,6 @@ static void tcg_out_qemu_ld(TCGContext *s, const TCGArg *args, bool is64)
break;
case MO_Q:
{
/* Be careful not to modify datalo and datahi
for the slow path below. */
TCGReg dl = (bswap ? datahi : datalo);
TCGReg dh = (bswap ? datalo : datahi);
@ -1442,30 +1418,20 @@ static void tcg_out_qemu_ld(TCGContext *s, const TCGArg *args, bool is64)
tcg_out_ld32_12(s, COND_AL, dh, TCG_REG_TMP, 4);
}
if (bswap) {
tcg_out_bswap32(s, COND_AL, dh, dh);
tcg_out_bswap32(s, COND_AL, dl, dl);
tcg_out_bswap32(s, COND_AL, dh, dh);
}
}
break;
}
}
add_qemu_ldst_label(s, 1, opc, datalo, datahi, addrlo, addrhi,
mem_index, s->code_ptr, label_ptr);
#else /* !CONFIG_SOFTMMU */
if (GUEST_BASE) {
uint32_t offset = GUEST_BASE;
int i, rot;
static inline void tcg_out_qemu_ld_direct(TCGContext *s, TCGMemOp opc,
TCGReg datalo, TCGReg datahi,
TCGReg addrlo)
{
TCGMemOp bswap = opc & MO_BSWAP;
while (offset) {
i = ctz32(offset) & ~1;
rot = ((32 - i) << 7) & 0xf00;
tcg_out_dat_imm(s, COND_AL, ARITH_ADD, TCG_REG_TMP, addrlo,
((offset >> i) & 0xff) | rot);
addrlo = TCG_REG_TMP;
offset &= ~(0xff << i);
}
}
switch (opc & MO_SSIZE) {
case MO_UB:
tcg_out_ld8_12(s, COND_AL, datalo, addrlo, 0);
@ -1495,32 +1461,32 @@ static void tcg_out_qemu_ld(TCGContext *s, const TCGArg *args, bool is64)
}
break;
case MO_Q:
if (use_armv6_instructions && !bswap
&& (datalo & 1) == 0 && datahi == datalo + 1) {
tcg_out_ldrd_8(s, COND_AL, datalo, addrlo, 0);
} else if (use_armv6_instructions && bswap
&& (datahi & 1) == 0 && datalo == datahi + 1) {
tcg_out_ldrd_8(s, COND_AL, datahi, addrlo, 0);
} else if (datalo == addrlo) {
tcg_out_ld32_12(s, COND_AL, datahi, addrlo, bswap ? 0 : 4);
tcg_out_ld32_12(s, COND_AL, datalo, addrlo, bswap ? 4 : 0);
} else {
tcg_out_ld32_12(s, COND_AL, datalo, addrlo, bswap ? 4 : 0);
tcg_out_ld32_12(s, COND_AL, datahi, addrlo, bswap ? 0 : 4);
}
if (bswap) {
tcg_out_bswap32(s, COND_AL, datalo, datalo);
tcg_out_bswap32(s, COND_AL, datahi, datahi);
{
TCGReg dl = (bswap ? datahi : datalo);
TCGReg dh = (bswap ? datalo : datahi);
if (use_armv6_instructions && (dl & 1) == 0 && dh == dl + 1) {
tcg_out_ldrd_8(s, COND_AL, dl, addrlo, 0);
} else if (dl == addrlo) {
tcg_out_ld32_12(s, COND_AL, dh, addrlo, bswap ? 0 : 4);
tcg_out_ld32_12(s, COND_AL, dl, addrlo, bswap ? 4 : 0);
} else {
tcg_out_ld32_12(s, COND_AL, dl, addrlo, bswap ? 4 : 0);
tcg_out_ld32_12(s, COND_AL, dh, addrlo, bswap ? 0 : 4);
}
if (bswap) {
tcg_out_bswap32(s, COND_AL, dl, dl);
tcg_out_bswap32(s, COND_AL, dh, dh);
}
}
break;
}
#endif
}
static void tcg_out_qemu_st(TCGContext *s, const TCGArg *args, bool is64)
static void tcg_out_qemu_ld(TCGContext *s, const TCGArg *args, bool is64)
{
TCGReg addrlo, datalo, datahi, addrhi __attribute__((unused));
TCGMemOp opc, bswap, s_bits;
TCGMemOp opc;
#ifdef CONFIG_SOFTMMU
int mem_index;
TCGReg addend;
@ -1532,73 +1498,81 @@ static void tcg_out_qemu_st(TCGContext *s, const TCGArg *args, bool is64)
addrlo = *args++;
addrhi = (TARGET_LONG_BITS == 64 ? *args++ : 0);
opc = *args++;
bswap = opc & MO_BSWAP;
s_bits = opc & MO_SIZE;
#ifdef CONFIG_SOFTMMU
mem_index = *args;
addend = tcg_out_tlb_read(s, addrlo, addrhi, s_bits, mem_index, 0);
addend = tcg_out_tlb_read(s, addrlo, addrhi, opc & MO_SIZE, mem_index, 1);
switch (s_bits) {
/* This a conditional BL only to load a pointer within this opcode into LR
for the slow path. We will not be using the value for a tail call. */
label_ptr = s->code_ptr;
tcg_out_bl_noaddr(s, COND_NE);
tcg_out_qemu_ld_index(s, opc, datalo, datahi, addrlo, addend);
add_qemu_ldst_label(s, 1, opc, datalo, datahi, addrlo, addrhi,
mem_index, s->code_ptr, label_ptr);
#else /* !CONFIG_SOFTMMU */
if (GUEST_BASE) {
tcg_out_movi(s, TCG_TYPE_PTR, TCG_REG_TMP, GUEST_BASE);
tcg_out_qemu_ld_index(s, opc, datalo, datahi, addrlo, TCG_REG_TMP);
} else {
tcg_out_qemu_ld_direct(s, opc, datalo, datahi, addrlo);
}
#endif
}
static inline void tcg_out_qemu_st_index(TCGContext *s, int cond, TCGMemOp opc,
TCGReg datalo, TCGReg datahi,
TCGReg addrlo, TCGReg addend)
{
TCGMemOp bswap = opc & MO_BSWAP;
switch (opc & MO_SIZE) {
case MO_8:
tcg_out_st8_r(s, COND_EQ, datalo, addrlo, addend);
tcg_out_st8_r(s, cond, datalo, addrlo, addend);
break;
case MO_16:
if (bswap) {
tcg_out_bswap16st(s, COND_EQ, TCG_REG_R0, datalo);
tcg_out_st16_r(s, COND_EQ, TCG_REG_R0, addrlo, addend);
tcg_out_bswap16st(s, cond, TCG_REG_R0, datalo);
tcg_out_st16_r(s, cond, TCG_REG_R0, addrlo, addend);
} else {
tcg_out_st16_r(s, COND_EQ, datalo, addrlo, addend);
tcg_out_st16_r(s, cond, datalo, addrlo, addend);
}
break;
case MO_32:
default:
if (bswap) {
tcg_out_bswap32(s, COND_EQ, TCG_REG_R0, datalo);
tcg_out_st32_r(s, COND_EQ, TCG_REG_R0, addrlo, addend);
tcg_out_bswap32(s, cond, TCG_REG_R0, datalo);
tcg_out_st32_r(s, cond, TCG_REG_R0, addrlo, addend);
} else {
tcg_out_st32_r(s, COND_EQ, datalo, addrlo, addend);
tcg_out_st32_r(s, cond, datalo, addrlo, addend);
}
break;
case MO_64:
if (bswap) {
tcg_out_bswap32(s, COND_EQ, TCG_REG_R0, datahi);
tcg_out_st32_rwb(s, COND_EQ, TCG_REG_R0, addend, addrlo);
tcg_out_bswap32(s, COND_EQ, TCG_REG_R0, datalo);
tcg_out_st32_12(s, COND_EQ, TCG_REG_R0, addend, 4);
tcg_out_bswap32(s, cond, TCG_REG_R0, datahi);
tcg_out_st32_rwb(s, cond, TCG_REG_R0, addend, addrlo);
tcg_out_bswap32(s, cond, TCG_REG_R0, datalo);
tcg_out_st32_12(s, cond, TCG_REG_R0, addend, 4);
} else if (use_armv6_instructions
&& (datalo & 1) == 0 && datahi == datalo + 1) {
tcg_out_strd_r(s, COND_EQ, datalo, addrlo, addend);
tcg_out_strd_r(s, cond, datalo, addrlo, addend);
} else {
tcg_out_st32_rwb(s, COND_EQ, datalo, addend, addrlo);
tcg_out_st32_12(s, COND_EQ, datahi, addend, 4);
tcg_out_st32_rwb(s, cond, datalo, addend, addrlo);
tcg_out_st32_12(s, cond, datahi, addend, 4);
}
break;
}
}
/* The conditional call must come last, as we're going to return here. */
label_ptr = s->code_ptr;
tcg_out_bl_noaddr(s, COND_NE);
static inline void tcg_out_qemu_st_direct(TCGContext *s, TCGMemOp opc,
TCGReg datalo, TCGReg datahi,
TCGReg addrlo)
{
TCGMemOp bswap = opc & MO_BSWAP;
add_qemu_ldst_label(s, 0, opc, datalo, datahi, addrlo, addrhi,
mem_index, s->code_ptr, label_ptr);
#else /* !CONFIG_SOFTMMU */
if (GUEST_BASE) {
uint32_t offset = GUEST_BASE;
int i;
int rot;
while (offset) {
i = ctz32(offset) & ~1;
rot = ((32 - i) << 7) & 0xf00;
tcg_out_dat_imm(s, COND_AL, ARITH_ADD, TCG_REG_R1, addrlo,
((offset >> i) & 0xff) | rot);
addrlo = TCG_REG_R1;
offset &= ~(0xff << i);
}
}
switch (s_bits) {
switch (opc & MO_SIZE) {
case MO_8:
tcg_out_st8_12(s, COND_AL, datalo, addrlo, 0);
break;
@ -1634,6 +1608,44 @@ static void tcg_out_qemu_st(TCGContext *s, const TCGArg *args, bool is64)
}
break;
}
}
static void tcg_out_qemu_st(TCGContext *s, const TCGArg *args, bool is64)
{
TCGReg addrlo, datalo, datahi, addrhi __attribute__((unused));
TCGMemOp opc;
#ifdef CONFIG_SOFTMMU
int mem_index;
TCGReg addend;
uint8_t *label_ptr;
#endif
datalo = *args++;
datahi = (is64 ? *args++ : 0);
addrlo = *args++;
addrhi = (TARGET_LONG_BITS == 64 ? *args++ : 0);
opc = *args++;
#ifdef CONFIG_SOFTMMU
mem_index = *args;
addend = tcg_out_tlb_read(s, addrlo, addrhi, opc & MO_SIZE, mem_index, 0);
tcg_out_qemu_st_index(s, COND_EQ, opc, datalo, datahi, addrlo, addend);
/* The conditional call must come last, as we're going to return here. */
label_ptr = s->code_ptr;
tcg_out_bl_noaddr(s, COND_NE);
add_qemu_ldst_label(s, 0, opc, datalo, datahi, addrlo, addrhi,
mem_index, s->code_ptr, label_ptr);
#else /* !CONFIG_SOFTMMU */
if (GUEST_BASE) {
tcg_out_movi(s, TCG_TYPE_PTR, TCG_REG_TMP, GUEST_BASE);
tcg_out_qemu_st_index(s, COND_AL, opc, datalo,
datahi, addrlo, TCG_REG_TMP);
} else {
tcg_out_qemu_st_direct(s, opc, datalo, datahi, addrlo);
}
#endif
}