tcg/loongarch64: Use tcg_use_softmmu

Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
This commit is contained in:
Richard Henderson 2023-10-01 18:05:27 +00:00
parent 915e1d52e2
commit 10e1fd2784

View File

@ -165,10 +165,7 @@ static TCGReg tcg_target_call_oarg_reg(TCGCallReturnKind kind, int slot)
return TCG_REG_A0 + slot;
}
#ifndef CONFIG_SOFTMMU
#define USE_GUEST_BASE (guest_base != 0)
#define TCG_GUEST_BASE_REG TCG_REG_S1
#endif
#define TCG_CT_CONST_ZERO 0x100
#define TCG_CT_CONST_S12 0x200
@ -908,76 +905,77 @@ static TCGLabelQemuLdst *prepare_host_addr(TCGContext *s, HostAddress *h,
h->aa = atom_and_align_for_opc(s, opc, MO_ATOM_IFALIGN, false);
a_bits = h->aa.align;
#ifdef CONFIG_SOFTMMU
unsigned s_bits = opc & MO_SIZE;
int mem_index = get_mmuidx(oi);
int fast_ofs = tlb_mask_table_ofs(s, mem_index);
int mask_ofs = fast_ofs + offsetof(CPUTLBDescFast, mask);
int table_ofs = fast_ofs + offsetof(CPUTLBDescFast, table);
if (tcg_use_softmmu) {
unsigned s_bits = opc & MO_SIZE;
int mem_index = get_mmuidx(oi);
int fast_ofs = tlb_mask_table_ofs(s, mem_index);
int mask_ofs = fast_ofs + offsetof(CPUTLBDescFast, mask);
int table_ofs = fast_ofs + offsetof(CPUTLBDescFast, table);
ldst = new_ldst_label(s);
ldst->is_ld = is_ld;
ldst->oi = oi;
ldst->addrlo_reg = addr_reg;
tcg_out_ld(s, TCG_TYPE_PTR, TCG_REG_TMP0, TCG_AREG0, mask_ofs);
tcg_out_ld(s, TCG_TYPE_PTR, TCG_REG_TMP1, TCG_AREG0, table_ofs);
tcg_out_opc_srli_d(s, TCG_REG_TMP2, addr_reg,
s->page_bits - CPU_TLB_ENTRY_BITS);
tcg_out_opc_and(s, TCG_REG_TMP2, TCG_REG_TMP2, TCG_REG_TMP0);
tcg_out_opc_add_d(s, TCG_REG_TMP2, TCG_REG_TMP2, TCG_REG_TMP1);
/* Load the tlb comparator and the addend. */
QEMU_BUILD_BUG_ON(HOST_BIG_ENDIAN);
tcg_out_ld(s, addr_type, TCG_REG_TMP0, TCG_REG_TMP2,
is_ld ? offsetof(CPUTLBEntry, addr_read)
: offsetof(CPUTLBEntry, addr_write));
tcg_out_ld(s, TCG_TYPE_PTR, TCG_REG_TMP2, TCG_REG_TMP2,
offsetof(CPUTLBEntry, addend));
/*
* For aligned accesses, we check the first byte and include the alignment
* bits within the address. For unaligned access, we check that we don't
* cross pages using the address of the last byte of the access.
*/
if (a_bits < s_bits) {
unsigned a_mask = (1u << a_bits) - 1;
unsigned s_mask = (1u << s_bits) - 1;
tcg_out_addi(s, addr_type, TCG_REG_TMP1, addr_reg, s_mask - a_mask);
} else {
tcg_out_mov(s, addr_type, TCG_REG_TMP1, addr_reg);
}
tcg_out_opc_bstrins_d(s, TCG_REG_TMP1, TCG_REG_ZERO,
a_bits, s->page_bits - 1);
/* Compare masked address with the TLB entry. */
ldst->label_ptr[0] = s->code_ptr;
tcg_out_opc_bne(s, TCG_REG_TMP0, TCG_REG_TMP1, 0);
h->index = TCG_REG_TMP2;
#else
if (a_bits) {
ldst = new_ldst_label(s);
ldst->is_ld = is_ld;
ldst->oi = oi;
ldst->addrlo_reg = addr_reg;
tcg_out_ld(s, TCG_TYPE_PTR, TCG_REG_TMP0, TCG_AREG0, mask_ofs);
tcg_out_ld(s, TCG_TYPE_PTR, TCG_REG_TMP1, TCG_AREG0, table_ofs);
tcg_out_opc_srli_d(s, TCG_REG_TMP2, addr_reg,
s->page_bits - CPU_TLB_ENTRY_BITS);
tcg_out_opc_and(s, TCG_REG_TMP2, TCG_REG_TMP2, TCG_REG_TMP0);
tcg_out_opc_add_d(s, TCG_REG_TMP2, TCG_REG_TMP2, TCG_REG_TMP1);
/* Load the tlb comparator and the addend. */
QEMU_BUILD_BUG_ON(HOST_BIG_ENDIAN);
tcg_out_ld(s, addr_type, TCG_REG_TMP0, TCG_REG_TMP2,
is_ld ? offsetof(CPUTLBEntry, addr_read)
: offsetof(CPUTLBEntry, addr_write));
tcg_out_ld(s, TCG_TYPE_PTR, TCG_REG_TMP2, TCG_REG_TMP2,
offsetof(CPUTLBEntry, addend));
/*
* Without micro-architecture details, we don't know which of
* bstrpick or andi is faster, so use bstrpick as it's not
* constrained by imm field width. Not to say alignments >= 2^12
* are going to happen any time soon.
* For aligned accesses, we check the first byte and include the
* alignment bits within the address. For unaligned access, we
* check that we don't cross pages using the address of the last
* byte of the access.
*/
tcg_out_opc_bstrpick_d(s, TCG_REG_TMP1, addr_reg, 0, a_bits - 1);
if (a_bits < s_bits) {
unsigned a_mask = (1u << a_bits) - 1;
unsigned s_mask = (1u << s_bits) - 1;
tcg_out_addi(s, addr_type, TCG_REG_TMP1, addr_reg, s_mask - a_mask);
} else {
tcg_out_mov(s, addr_type, TCG_REG_TMP1, addr_reg);
}
tcg_out_opc_bstrins_d(s, TCG_REG_TMP1, TCG_REG_ZERO,
a_bits, s->page_bits - 1);
/* Compare masked address with the TLB entry. */
ldst->label_ptr[0] = s->code_ptr;
tcg_out_opc_bne(s, TCG_REG_TMP1, TCG_REG_ZERO, 0);
}
tcg_out_opc_bne(s, TCG_REG_TMP0, TCG_REG_TMP1, 0);
h->index = USE_GUEST_BASE ? TCG_GUEST_BASE_REG : TCG_REG_ZERO;
#endif
h->index = TCG_REG_TMP2;
} else {
if (a_bits) {
ldst = new_ldst_label(s);
ldst->is_ld = is_ld;
ldst->oi = oi;
ldst->addrlo_reg = addr_reg;
/*
* Without micro-architecture details, we don't know which of
* bstrpick or andi is faster, so use bstrpick as it's not
* constrained by imm field width. Not to say alignments >= 2^12
* are going to happen any time soon.
*/
tcg_out_opc_bstrpick_d(s, TCG_REG_TMP1, addr_reg, 0, a_bits - 1);
ldst->label_ptr[0] = s->code_ptr;
tcg_out_opc_bne(s, TCG_REG_TMP1, TCG_REG_ZERO, 0);
}
h->index = guest_base ? TCG_GUEST_BASE_REG : TCG_REG_ZERO;
}
if (addr_type == TCG_TYPE_I32) {
h->base = TCG_REG_TMP0;
@ -2272,12 +2270,10 @@ static void tcg_target_qemu_prologue(TCGContext *s)
TCG_REG_SP, SAVE_OFS + i * REG_SIZE);
}
#if !defined(CONFIG_SOFTMMU)
if (USE_GUEST_BASE) {
if (!tcg_use_softmmu && guest_base) {
tcg_out_movi(s, TCG_TYPE_PTR, TCG_GUEST_BASE_REG, guest_base);
tcg_regset_set_reg(s->reserved_regs, TCG_GUEST_BASE_REG);
}
#endif
/* Call generated code */
tcg_out_mov(s, TCG_TYPE_PTR, TCG_AREG0, tcg_target_call_iarg_regs[0]);