tcg/riscv: Introduce prepare_host_addr

Merge tcg_out_tlb_load, add_qemu_ldst_label, tcg_out_test_alignment,
and some code that lived in both tcg_out_qemu_ld and tcg_out_qemu_st
into one function that returns TCGReg and TCGLabelQemuLdst.

Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
master
Richard Henderson 2023-04-23 21:06:38 +01:00
parent 7069e03699
commit 001dddfe0e
1 changed files with 114 additions and 139 deletions

View File

@ -899,10 +899,6 @@ static void * const qemu_st_helpers[MO_SIZE + 1] = {
#endif
};
/* We expect to use a 12-bit negative offset from ENV. */
QEMU_BUILD_BUG_ON(TLB_MASK_TABLE_OFS(0) > 0);
QEMU_BUILD_BUG_ON(TLB_MASK_TABLE_OFS(0) < -(1 << 11));
static void tcg_out_goto(TCGContext *s, const tcg_insn_unit *target)
{
tcg_out_opc_jump(s, OPC_JAL, TCG_REG_ZERO, 0);
@ -910,76 +906,6 @@ static void tcg_out_goto(TCGContext *s, const tcg_insn_unit *target)
tcg_debug_assert(ok);
}
static TCGReg tcg_out_tlb_load(TCGContext *s, TCGReg addr, MemOpIdx oi,
tcg_insn_unit **label_ptr, bool is_load)
{
MemOp opc = get_memop(oi);
unsigned s_bits = opc & MO_SIZE;
unsigned a_bits = get_alignment_bits(opc);
tcg_target_long compare_mask;
int mem_index = get_mmuidx(oi);
int fast_ofs = TLB_MASK_TABLE_OFS(mem_index);
int mask_ofs = fast_ofs + offsetof(CPUTLBDescFast, mask);
int table_ofs = fast_ofs + offsetof(CPUTLBDescFast, table);
TCGReg mask_base = TCG_AREG0, table_base = TCG_AREG0;
tcg_out_ld(s, TCG_TYPE_PTR, TCG_REG_TMP0, mask_base, mask_ofs);
tcg_out_ld(s, TCG_TYPE_PTR, TCG_REG_TMP1, table_base, table_ofs);
tcg_out_opc_imm(s, OPC_SRLI, TCG_REG_TMP2, addr,
TARGET_PAGE_BITS - CPU_TLB_ENTRY_BITS);
tcg_out_opc_reg(s, OPC_AND, TCG_REG_TMP2, TCG_REG_TMP2, TCG_REG_TMP0);
tcg_out_opc_reg(s, OPC_ADD, TCG_REG_TMP2, TCG_REG_TMP2, TCG_REG_TMP1);
/* Load the tlb comparator and the addend. */
tcg_out_ld(s, TCG_TYPE_TL, TCG_REG_TMP0, TCG_REG_TMP2,
is_load ? offsetof(CPUTLBEntry, addr_read)
: offsetof(CPUTLBEntry, addr_write));
tcg_out_ld(s, TCG_TYPE_PTR, TCG_REG_TMP2, TCG_REG_TMP2,
offsetof(CPUTLBEntry, addend));
/* We don't support unaligned accesses. */
if (a_bits < s_bits) {
a_bits = s_bits;
}
/* Clear the non-page, non-alignment bits from the address. */
compare_mask = (tcg_target_long)TARGET_PAGE_MASK | ((1 << a_bits) - 1);
if (compare_mask == sextreg(compare_mask, 0, 12)) {
tcg_out_opc_imm(s, OPC_ANDI, TCG_REG_TMP1, addr, compare_mask);
} else {
tcg_out_movi(s, TCG_TYPE_TL, TCG_REG_TMP1, compare_mask);
tcg_out_opc_reg(s, OPC_AND, TCG_REG_TMP1, TCG_REG_TMP1, addr);
}
/* Compare masked address with the TLB entry. */
label_ptr[0] = s->code_ptr;
tcg_out_opc_branch(s, OPC_BNE, TCG_REG_TMP0, TCG_REG_TMP1, 0);
/* TLB Hit - translate address using addend. */
if (TARGET_LONG_BITS == 32) {
tcg_out_ext32u(s, TCG_REG_TMP0, addr);
addr = TCG_REG_TMP0;
}
tcg_out_opc_reg(s, OPC_ADD, TCG_REG_TMP0, TCG_REG_TMP2, addr);
return TCG_REG_TMP0;
}
static void add_qemu_ldst_label(TCGContext *s, int is_ld, MemOpIdx oi,
TCGType data_type, TCGReg data_reg,
TCGReg addr_reg, void *raddr,
tcg_insn_unit **label_ptr)
{
TCGLabelQemuLdst *label = new_ldst_label(s);
label->is_ld = is_ld;
label->oi = oi;
label->type = data_type;
label->datalo_reg = data_reg;
label->addrlo_reg = addr_reg;
label->raddr = tcg_splitwx_to_rx(raddr);
label->label_ptr[0] = label_ptr[0];
}
static bool tcg_out_qemu_ld_slow_path(TCGContext *s, TCGLabelQemuLdst *l)
{
MemOpIdx oi = l->oi;
@ -1037,26 +963,6 @@ static bool tcg_out_qemu_st_slow_path(TCGContext *s, TCGLabelQemuLdst *l)
return true;
}
#else
static void tcg_out_test_alignment(TCGContext *s, bool is_ld, TCGReg addr_reg,
unsigned a_bits)
{
unsigned a_mask = (1 << a_bits) - 1;
TCGLabelQemuLdst *l = new_ldst_label(s);
l->is_ld = is_ld;
l->addrlo_reg = addr_reg;
/* We are expecting a_bits to max out at 7, so we can always use andi. */
tcg_debug_assert(a_bits < 12);
tcg_out_opc_imm(s, OPC_ANDI, TCG_REG_TMP1, addr_reg, a_mask);
l->label_ptr[0] = s->code_ptr;
tcg_out_opc_branch(s, OPC_BNE, TCG_REG_TMP1, TCG_REG_ZERO, 0);
l->raddr = tcg_splitwx_to_rx(s->code_ptr);
}
static bool tcg_out_fail_alignment(TCGContext *s, TCGLabelQemuLdst *l)
{
/* resolve label address */
@ -1083,9 +989,108 @@ static bool tcg_out_qemu_st_slow_path(TCGContext *s, TCGLabelQemuLdst *l)
{
return tcg_out_fail_alignment(s, l);
}
#endif /* CONFIG_SOFTMMU */
/*
* For softmmu, perform the TLB load and compare.
* For useronly, perform any required alignment tests.
* In both cases, return a TCGLabelQemuLdst structure if the slow path
* is required and fill in @h with the host address for the fast path.
*/
static TCGLabelQemuLdst *prepare_host_addr(TCGContext *s, TCGReg *pbase,
TCGReg addr_reg, MemOpIdx oi,
bool is_ld)
{
TCGLabelQemuLdst *ldst = NULL;
MemOp opc = get_memop(oi);
unsigned a_bits = get_alignment_bits(opc);
unsigned a_mask = (1u << a_bits) - 1;
#ifdef CONFIG_SOFTMMU
unsigned s_bits = opc & MO_SIZE;
int mem_index = get_mmuidx(oi);
int fast_ofs = TLB_MASK_TABLE_OFS(mem_index);
int mask_ofs = fast_ofs + offsetof(CPUTLBDescFast, mask);
int table_ofs = fast_ofs + offsetof(CPUTLBDescFast, table);
TCGReg mask_base = TCG_AREG0, table_base = TCG_AREG0;
tcg_target_long compare_mask;
ldst = new_ldst_label(s);
ldst->is_ld = is_ld;
ldst->oi = oi;
ldst->addrlo_reg = addr_reg;
QEMU_BUILD_BUG_ON(TLB_MASK_TABLE_OFS(0) > 0);
QEMU_BUILD_BUG_ON(TLB_MASK_TABLE_OFS(0) < -(1 << 11));
tcg_out_ld(s, TCG_TYPE_PTR, TCG_REG_TMP0, mask_base, mask_ofs);
tcg_out_ld(s, TCG_TYPE_PTR, TCG_REG_TMP1, table_base, table_ofs);
tcg_out_opc_imm(s, OPC_SRLI, TCG_REG_TMP2, addr_reg,
TARGET_PAGE_BITS - CPU_TLB_ENTRY_BITS);
tcg_out_opc_reg(s, OPC_AND, TCG_REG_TMP2, TCG_REG_TMP2, TCG_REG_TMP0);
tcg_out_opc_reg(s, OPC_ADD, TCG_REG_TMP2, TCG_REG_TMP2, TCG_REG_TMP1);
/* Load the tlb comparator and the addend. */
tcg_out_ld(s, TCG_TYPE_TL, 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));
/* We don't support unaligned accesses. */
if (a_bits < s_bits) {
a_bits = s_bits;
}
/* Clear the non-page, non-alignment bits from the address. */
compare_mask = (tcg_target_long)TARGET_PAGE_MASK | a_mask;
if (compare_mask == sextreg(compare_mask, 0, 12)) {
tcg_out_opc_imm(s, OPC_ANDI, TCG_REG_TMP1, addr_reg, compare_mask);
} else {
tcg_out_movi(s, TCG_TYPE_TL, TCG_REG_TMP1, compare_mask);
tcg_out_opc_reg(s, OPC_AND, TCG_REG_TMP1, TCG_REG_TMP1, addr_reg);
}
/* Compare masked address with the TLB entry. */
ldst->label_ptr[0] = s->code_ptr;
tcg_out_opc_branch(s, OPC_BNE, TCG_REG_TMP0, TCG_REG_TMP1, 0);
/* TLB Hit - translate address using addend. */
if (TARGET_LONG_BITS == 32) {
tcg_out_ext32u(s, TCG_REG_TMP0, addr_reg);
addr_reg = TCG_REG_TMP0;
}
tcg_out_opc_reg(s, OPC_ADD, TCG_REG_TMP0, TCG_REG_TMP2, addr_reg);
*pbase = TCG_REG_TMP0;
#else
if (a_mask) {
ldst = new_ldst_label(s);
ldst->is_ld = is_ld;
ldst->oi = oi;
ldst->addrlo_reg = addr_reg;
/* We are expecting a_bits max 7, so we can always use andi. */
tcg_debug_assert(a_bits < 12);
tcg_out_opc_imm(s, OPC_ANDI, TCG_REG_TMP1, addr_reg, a_mask);
ldst->label_ptr[0] = s->code_ptr;
tcg_out_opc_branch(s, OPC_BNE, TCG_REG_TMP1, TCG_REG_ZERO, 0);
}
TCGReg base = addr_reg;
if (TARGET_LONG_BITS == 32) {
tcg_out_ext32u(s, TCG_REG_TMP0, base);
base = TCG_REG_TMP0;
}
if (guest_base != 0) {
tcg_out_opc_reg(s, OPC_ADD, TCG_REG_TMP0, TCG_GUEST_BASE_REG, base);
base = TCG_REG_TMP0;
}
*pbase = base;
#endif
return ldst;
}
static void tcg_out_qemu_ld_direct(TCGContext *s, TCGReg val,
TCGReg base, MemOp opc, TCGType type)
{
@ -1125,32 +1130,17 @@ static void tcg_out_qemu_ld_direct(TCGContext *s, TCGReg val,
static void tcg_out_qemu_ld(TCGContext *s, TCGReg data_reg, TCGReg addr_reg,
MemOpIdx oi, TCGType data_type)
{
MemOp opc = get_memop(oi);
TCGLabelQemuLdst *ldst;
TCGReg base;
#if defined(CONFIG_SOFTMMU)
tcg_insn_unit *label_ptr[1];
ldst = prepare_host_addr(s, &base, addr_reg, oi, true);
tcg_out_qemu_ld_direct(s, data_reg, base, get_memop(oi), data_type);
base = tcg_out_tlb_load(s, addr_reg, oi, label_ptr, 1);
tcg_out_qemu_ld_direct(s, data_reg, base, opc, data_type);
add_qemu_ldst_label(s, true, oi, data_type, data_reg, addr_reg,
s->code_ptr, label_ptr);
#else
unsigned a_bits = get_alignment_bits(opc);
if (a_bits) {
tcg_out_test_alignment(s, true, addr_reg, a_bits);
if (ldst) {
ldst->type = data_type;
ldst->datalo_reg = data_reg;
ldst->raddr = tcg_splitwx_to_rx(s->code_ptr);
}
base = addr_reg;
if (TARGET_LONG_BITS == 32) {
tcg_out_ext32u(s, TCG_REG_TMP0, base);
base = TCG_REG_TMP0;
}
if (guest_base != 0) {
tcg_out_opc_reg(s, OPC_ADD, TCG_REG_TMP0, TCG_GUEST_BASE_REG, base);
base = TCG_REG_TMP0;
}
tcg_out_qemu_ld_direct(s, data_reg, base, opc, data_type);
#endif
}
static void tcg_out_qemu_st_direct(TCGContext *s, TCGReg val,
@ -1180,32 +1170,17 @@ static void tcg_out_qemu_st_direct(TCGContext *s, TCGReg val,
static void tcg_out_qemu_st(TCGContext *s, TCGReg data_reg, TCGReg addr_reg,
MemOpIdx oi, TCGType data_type)
{
MemOp opc = get_memop(oi);
TCGLabelQemuLdst *ldst;
TCGReg base;
#if defined(CONFIG_SOFTMMU)
tcg_insn_unit *label_ptr[1];
ldst = prepare_host_addr(s, &base, addr_reg, oi, false);
tcg_out_qemu_st_direct(s, data_reg, base, get_memop(oi));
base = tcg_out_tlb_load(s, addr_reg, oi, label_ptr, 0);
tcg_out_qemu_st_direct(s, data_reg, base, opc);
add_qemu_ldst_label(s, false, oi, data_type, data_reg, addr_reg,
s->code_ptr, label_ptr);
#else
unsigned a_bits = get_alignment_bits(opc);
if (a_bits) {
tcg_out_test_alignment(s, false, addr_reg, a_bits);
if (ldst) {
ldst->type = data_type;
ldst->datalo_reg = data_reg;
ldst->raddr = tcg_splitwx_to_rx(s->code_ptr);
}
base = addr_reg;
if (TARGET_LONG_BITS == 32) {
tcg_out_ext32u(s, TCG_REG_TMP0, base);
base = TCG_REG_TMP0;
}
if (guest_base != 0) {
tcg_out_opc_reg(s, OPC_ADD, TCG_REG_TMP0, TCG_GUEST_BASE_REG, base);
base = TCG_REG_TMP0;
}
tcg_out_qemu_st_direct(s, data_reg, base, opc);
#endif
}
static const tcg_insn_unit *tb_ret_addr;