old-cross-binutils/sim/aarch64/cpustate.c
Nick Clifton 7517e550ce Fix more bugs in AArch64 simulator.
* cpustate.c (aarch64_set_reg_s32): New function.
	(aarch64_set_reg_u32): New function.
	(aarch64_get_FP_half): Place half precision value into the correct
	slot of the union.
	(aarch64_set_FP_half): Likewise.
	* cpustate.h: Add prototypes for aarch64_set_reg_s32 and
	aarch64_set_reg_u32.
	* memory.c (FETCH_FUNC): Cast the read value to the access type
	before converting it to the return type.  Rename to FETCH_FUNC64.
	(FETCH_FUNC32): New macro.  Duplicates FETCH_FUNC64 but for 32-bit
	accesses.  Use for 32-bit memory access functions.
	* simulator.c (ldrsb_wb): Use sign extension not zero extension.
	(ldrb_scale_ext, ldrsh32_abs, ldrsh32_wb): Likewise.
	(ldrsh32_scale_ext, ldrsh_abs, ldrsh64_wb): Likewise.
	(ldrsh_scale_ext, ldrsw_abs): Likewise.
	(ldrh32_abs): Store 32 bit value not 64-bits.
	(ldrh32_wb, ldrh32_scale_ext): Likewise.
	(do_vec_MOV_immediate): Fix computation of val.
	(do_vec_MVNI): Likewise.
	(DO_VEC_WIDENING_MUL): New macro.
	(do_vec_mull): Use new macro.
	(do_vec_mul): Use new macro.
	(do_vec_MLA): Read values before writing.
	(do_vec_xtl): Likewise.
	(do_vec_SSHL): Select correct shift value.
	(do_vec_USHL): Likewise.
	(do_scalar_UCVTF): New function.
	(do_scalar_vec): Call new function.
	(store_pair_u64): Treat reads of SP as reads of XZR.
2016-03-30 10:29:04 +01:00

634 lines
14 KiB
C

/* cpustate.h -- Prototypes for AArch64 simulator functions.
Copyright (C) 2015-2016 Free Software Foundation, Inc.
Contributed by Red Hat.
This file is part of GDB.
This program is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation; either version 3 of the License, or
(at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License
along with this program. If not, see <http://www.gnu.org/licenses/>. */
#include <stdio.h>
#include "sim-main.h"
#include "cpustate.h"
#include "simulator.h"
/* Some operands are allowed to access the stack pointer (reg 31).
For others a read from r31 always returns 0, and a write to r31 is ignored. */
#define reg_num(reg) (((reg) == R31 && !r31_is_sp) ? 32 : (reg))
void
aarch64_set_reg_u64 (sim_cpu *cpu, GReg reg, int r31_is_sp, uint64_t val)
{
if (reg == R31 && ! r31_is_sp)
{
TRACE_REGISTER (cpu, "GR[31] NOT CHANGED!");
return;
}
if (val != cpu->gr[reg].u64)
TRACE_REGISTER (cpu,
"GR[%2d] changes from %16" PRIx64 " to %16" PRIx64,
reg, cpu->gr[reg].u64, val);
cpu->gr[reg].u64 = val;
}
void
aarch64_set_reg_s64 (sim_cpu *cpu, GReg reg, int r31_is_sp, int64_t val)
{
if (reg == R31 && ! r31_is_sp)
{
TRACE_REGISTER (cpu, "GR[31] NOT CHANGED!");
return;
}
if (val != cpu->gr[reg].s64)
TRACE_REGISTER (cpu,
"GR[%2d] changes from %16" PRIx64 " to %16" PRIx64,
reg, cpu->gr[reg].s64, val);
cpu->gr[reg].s64 = val;
}
uint64_t
aarch64_get_reg_u64 (sim_cpu *cpu, GReg reg, int r31_is_sp)
{
return cpu->gr[reg_num(reg)].u64;
}
int64_t
aarch64_get_reg_s64 (sim_cpu *cpu, GReg reg, int r31_is_sp)
{
return cpu->gr[reg_num(reg)].s64;
}
uint32_t
aarch64_get_reg_u32 (sim_cpu *cpu, GReg reg, int r31_is_sp)
{
return cpu->gr[reg_num(reg)].u32;
}
int32_t
aarch64_get_reg_s32 (sim_cpu *cpu, GReg reg, int r31_is_sp)
{
return cpu->gr[reg_num(reg)].s32;
}
void
aarch64_set_reg_s32 (sim_cpu *cpu, GReg reg, int r31_is_sp, int32_t val)
{
if (reg == R31 && ! r31_is_sp)
{
TRACE_REGISTER (cpu, "GR[31] NOT CHANGED!");
return;
}
if (val != cpu->gr[reg].s32)
TRACE_REGISTER (cpu, "GR[%2d] changes from %8x to %8x",
reg, cpu->gr[reg].s32, val);
/* The ARM ARM states that (C1.2.4):
When the data size is 32 bits, the lower 32 bits of the
register are used and the upper 32 bits are ignored on
a read and cleared to zero on a write.
We simulate this by first clearing the whole 64-bits and
then writing to the 32-bit value in the GRegister union. */
cpu->gr[reg].s64 = 0;
cpu->gr[reg].s32 = val;
}
void
aarch64_set_reg_u32 (sim_cpu *cpu, GReg reg, int r31_is_sp, uint32_t val)
{
if (reg == R31 && ! r31_is_sp)
{
TRACE_REGISTER (cpu, "GR[31] NOT CHANGED!");
return;
}
if (val != cpu->gr[reg].u32)
TRACE_REGISTER (cpu, "GR[%2d] changes from %8x to %8x",
reg, cpu->gr[reg].u32, val);
cpu->gr[reg].u64 = 0;
cpu->gr[reg].u32 = val;
}
uint32_t
aarch64_get_reg_u16 (sim_cpu *cpu, GReg reg, int r31_is_sp)
{
return cpu->gr[reg_num(reg)].u16;
}
int32_t
aarch64_get_reg_s16 (sim_cpu *cpu, GReg reg, int r31_is_sp)
{
return cpu->gr[reg_num(reg)].s16;
}
uint32_t
aarch64_get_reg_u8 (sim_cpu *cpu, GReg reg, int r31_is_sp)
{
return cpu->gr[reg_num(reg)].u8;
}
int32_t
aarch64_get_reg_s8 (sim_cpu *cpu, GReg reg, int r31_is_sp)
{
return cpu->gr[reg_num(reg)].s8;
}
uint64_t
aarch64_get_PC (sim_cpu *cpu)
{
return cpu->pc;
}
uint64_t
aarch64_get_next_PC (sim_cpu *cpu)
{
return cpu->nextpc;
}
void
aarch64_set_next_PC (sim_cpu *cpu, uint64_t next)
{
if (next != cpu->nextpc + 4)
TRACE_REGISTER (cpu,
"NextPC changes from %16" PRIx64 " to %16" PRIx64,
cpu->nextpc, next);
cpu->nextpc = next;
}
void
aarch64_set_next_PC_by_offset (sim_cpu *cpu, int64_t offset)
{
if (cpu->pc + offset != cpu->nextpc + 4)
TRACE_REGISTER (cpu,
"NextPC changes from %16" PRIx64 " to %16" PRIx64,
cpu->nextpc, cpu->pc + offset);
cpu->nextpc = cpu->pc + offset;
}
/* Install nextpc as current pc. */
void
aarch64_update_PC (sim_cpu *cpu)
{
cpu->pc = cpu->nextpc;
/* Rezero the register we hand out when asked for ZR just in case it
was used as the destination for a write by the previous
instruction. */
cpu->gr[32].u64 = 0UL;
}
/* This instruction can be used to save the next PC to LR
just before installing a branch PC. */
void
aarch64_save_LR (sim_cpu *cpu)
{
if (cpu->gr[LR].u64 != cpu->nextpc)
TRACE_REGISTER (cpu,
"LR changes from %16" PRIx64 " to %16" PRIx64,
cpu->gr[LR].u64, cpu->nextpc);
cpu->gr[LR].u64 = cpu->nextpc;
}
static const char *
decode_cpsr (FlagMask flags)
{
switch (flags & CPSR_ALL_FLAGS)
{
default:
case 0: return "----";
case 1: return "---V";
case 2: return "--C-";
case 3: return "--CV";
case 4: return "-Z--";
case 5: return "-Z-V";
case 6: return "-ZC-";
case 7: return "-ZCV";
case 8: return "N---";
case 9: return "N--V";
case 10: return "N-C-";
case 11: return "N-CV";
case 12: return "NZ--";
case 13: return "NZ-V";
case 14: return "NZC-";
case 15: return "NZCV";
}
}
/* Retrieve the CPSR register as an int. */
uint32_t
aarch64_get_CPSR (sim_cpu *cpu)
{
return cpu->CPSR;
}
/* Set the CPSR register as an int. */
void
aarch64_set_CPSR (sim_cpu *cpu, uint32_t new_flags)
{
if (TRACE_REGISTER_P (cpu))
{
if (cpu->CPSR != new_flags)
TRACE_REGISTER (cpu,
"CPSR changes from %s to %s",
decode_cpsr (cpu->CPSR), decode_cpsr (new_flags));
else
TRACE_REGISTER (cpu,
"CPSR stays at %s", decode_cpsr (cpu->CPSR));
}
cpu->CPSR = new_flags & CPSR_ALL_FLAGS;
}
/* Read a specific subset of the CPSR as a bit pattern. */
uint32_t
aarch64_get_CPSR_bits (sim_cpu *cpu, FlagMask mask)
{
return cpu->CPSR & mask;
}
/* Assign a specific subset of the CPSR as a bit pattern. */
void
aarch64_set_CPSR_bits (sim_cpu *cpu, uint32_t mask, uint32_t value)
{
uint32_t old_flags = cpu->CPSR;
mask &= CPSR_ALL_FLAGS;
cpu->CPSR &= ~ mask;
cpu->CPSR |= (value & mask);
if (old_flags != cpu->CPSR)
TRACE_REGISTER (cpu,
"CPSR changes from %s to %s",
decode_cpsr (old_flags), decode_cpsr (cpu->CPSR));
}
/* Test the value of a single CPSR returned as non-zero or zero. */
uint32_t
aarch64_test_CPSR_bit (sim_cpu *cpu, FlagMask bit)
{
return cpu->CPSR & bit;
}
/* Set a single flag in the CPSR. */
void
aarch64_set_CPSR_bit (sim_cpu *cpu, FlagMask bit)
{
uint32_t old_flags = cpu->CPSR;
cpu->CPSR |= (bit & CPSR_ALL_FLAGS);
if (old_flags != cpu->CPSR)
TRACE_REGISTER (cpu,
"CPSR changes from %s to %s",
decode_cpsr (old_flags), decode_cpsr (cpu->CPSR));
}
/* Clear a single flag in the CPSR. */
void
aarch64_clear_CPSR_bit (sim_cpu *cpu, FlagMask bit)
{
uint32_t old_flags = cpu->CPSR;
cpu->CPSR &= ~(bit & CPSR_ALL_FLAGS);
if (old_flags != cpu->CPSR)
TRACE_REGISTER (cpu,
"CPSR changes from %s to %s",
decode_cpsr (old_flags), decode_cpsr (cpu->CPSR));
}
float
aarch64_get_FP_half (sim_cpu *cpu, VReg reg)
{
union
{
uint16_t h[2];
float f;
} u;
u.h[0] = 0;
u.h[1] = cpu->fr[reg].h[0];
return u.f;
}
float
aarch64_get_FP_float (sim_cpu *cpu, VReg reg)
{
return cpu->fr[reg].s;
}
double
aarch64_get_FP_double (sim_cpu *cpu, VReg reg)
{
return cpu->fr[reg].d;
}
void
aarch64_get_FP_long_double (sim_cpu *cpu, VReg reg, FRegister *a)
{
a->v[0] = cpu->fr[reg].v[0];
a->v[1] = cpu->fr[reg].v[1];
}
void
aarch64_set_FP_half (sim_cpu *cpu, VReg reg, float val)
{
union
{
uint16_t h[2];
float f;
} u;
u.f = val;
cpu->fr[reg].h[0] = u.h[1];
cpu->fr[reg].h[1] = 0;
}
void
aarch64_set_FP_float (sim_cpu *cpu, VReg reg, float val)
{
if (val != cpu->fr[reg].s)
{
FRegister v;
v.s = val;
TRACE_REGISTER (cpu,
"FR[%d].s changes from %f to %f [hex: %0lx]",
reg, cpu->fr[reg].s, val, v.v[0]);
}
cpu->fr[reg].s = val;
}
void
aarch64_set_FP_double (sim_cpu *cpu, VReg reg, double val)
{
if (val != cpu->fr[reg].d)
{
FRegister v;
v.d = val;
TRACE_REGISTER (cpu,
"FR[%d].d changes from %f to %f [hex: %0lx]",
reg, cpu->fr[reg].d, val, v.v[0]);
}
cpu->fr[reg].d = val;
}
void
aarch64_set_FP_long_double (sim_cpu *cpu, VReg reg, FRegister a)
{
if (cpu->fr[reg].v[0] != a.v[0]
|| cpu->fr[reg].v[1] != a.v[1])
TRACE_REGISTER (cpu,
"FR[%d].q changes from [%0lx %0lx] to [%0lx %0lx] ",
reg,
cpu->fr[reg].v[0], cpu->fr[reg].v[1],
a.v[0], a.v[1]);
cpu->fr[reg].v[0] = a.v[0];
cpu->fr[reg].v[1] = a.v[1];
}
#define GET_VEC_ELEMENT(REG, ELEMENT, FIELD) \
do \
{ \
if (ELEMENT >= ARRAY_SIZE (cpu->fr[0].FIELD)) \
{ \
TRACE_REGISTER (cpu, \
"Internal SIM error: invalid element number: %d ",\
ELEMENT); \
sim_engine_halt (CPU_STATE (cpu), cpu, NULL, aarch64_get_PC (cpu), \
sim_stopped, SIM_SIGBUS); \
} \
return cpu->fr[REG].FIELD [ELEMENT]; \
} \
while (0)
uint64_t
aarch64_get_vec_u64 (sim_cpu *cpu, VReg reg, unsigned element)
{
GET_VEC_ELEMENT (reg, element, v);
}
uint32_t
aarch64_get_vec_u32 (sim_cpu *cpu, VReg reg, unsigned element)
{
GET_VEC_ELEMENT (reg, element, w);
}
uint16_t
aarch64_get_vec_u16 (sim_cpu *cpu, VReg reg, unsigned element)
{
GET_VEC_ELEMENT (reg, element, h);
}
uint8_t
aarch64_get_vec_u8 (sim_cpu *cpu, VReg reg, unsigned element)
{
GET_VEC_ELEMENT (reg, element, b);
}
int64_t
aarch64_get_vec_s64 (sim_cpu *cpu, VReg reg, unsigned element)
{
GET_VEC_ELEMENT (reg, element, V);
}
int32_t
aarch64_get_vec_s32 (sim_cpu *cpu, VReg reg, unsigned element)
{
GET_VEC_ELEMENT (reg, element, W);
}
int16_t
aarch64_get_vec_s16 (sim_cpu *cpu, VReg reg, unsigned element)
{
GET_VEC_ELEMENT (reg, element, H);
}
int8_t
aarch64_get_vec_s8 (sim_cpu *cpu, VReg reg, unsigned element)
{
GET_VEC_ELEMENT (reg, element, B);
}
float
aarch64_get_vec_float (sim_cpu *cpu, VReg reg, unsigned element)
{
GET_VEC_ELEMENT (reg, element, S);
}
double
aarch64_get_vec_double (sim_cpu *cpu, VReg reg, unsigned element)
{
GET_VEC_ELEMENT (reg, element, D);
}
#define SET_VEC_ELEMENT(REG, ELEMENT, VAL, FIELD, PRINTER) \
do \
{ \
if (ELEMENT >= ARRAY_SIZE (cpu->fr[0].FIELD)) \
{ \
TRACE_REGISTER (cpu, \
"Internal SIM error: invalid element number: %d ",\
ELEMENT); \
sim_engine_halt (CPU_STATE (cpu), cpu, NULL, aarch64_get_PC (cpu), \
sim_stopped, SIM_SIGBUS); \
} \
if (VAL != cpu->fr[REG].FIELD [ELEMENT]) \
TRACE_REGISTER (cpu, \
"VR[%2d]." #FIELD " [%d] changes from " PRINTER \
" to " PRINTER , REG, \
ELEMENT, cpu->fr[REG].FIELD [ELEMENT], VAL); \
\
cpu->fr[REG].FIELD [ELEMENT] = VAL; \
} \
while (0)
void
aarch64_set_vec_u64 (sim_cpu *cpu, VReg reg, unsigned element, uint64_t val)
{
SET_VEC_ELEMENT (reg, element, val, v, "%16lx");
}
void
aarch64_set_vec_u32 (sim_cpu *cpu, VReg reg, unsigned element, uint32_t val)
{
SET_VEC_ELEMENT (reg, element, val, w, "%8x");
}
void
aarch64_set_vec_u16 (sim_cpu *cpu, VReg reg, unsigned element, uint16_t val)
{
SET_VEC_ELEMENT (reg, element, val, h, "%4x");
}
void
aarch64_set_vec_u8 (sim_cpu *cpu, VReg reg, unsigned element, uint8_t val)
{
SET_VEC_ELEMENT (reg, element, val, b, "%x");
}
void
aarch64_set_vec_s64 (sim_cpu *cpu, VReg reg, unsigned element, int64_t val)
{
SET_VEC_ELEMENT (reg, element, val, V, "%16lx");
}
void
aarch64_set_vec_s32 (sim_cpu *cpu, VReg reg, unsigned element, int32_t val)
{
SET_VEC_ELEMENT (reg, element, val, W, "%8x");
}
void
aarch64_set_vec_s16 (sim_cpu *cpu, VReg reg, unsigned element, int16_t val)
{
SET_VEC_ELEMENT (reg, element, val, H, "%4x");
}
void
aarch64_set_vec_s8 (sim_cpu *cpu, VReg reg, unsigned element, int8_t val)
{
SET_VEC_ELEMENT (reg, element, val, B, "%x");
}
void
aarch64_set_vec_float (sim_cpu *cpu, VReg reg, unsigned element, float val)
{
SET_VEC_ELEMENT (reg, element, val, S, "%f");
}
void
aarch64_set_vec_double (sim_cpu *cpu, VReg reg, unsigned element, double val)
{
SET_VEC_ELEMENT (reg, element, val, D, "%f");
}
void
aarch64_set_FPSR (sim_cpu *cpu, uint32_t value)
{
if (cpu->FPSR != value)
TRACE_REGISTER (cpu,
"FPSR changes from %x to %x", cpu->FPSR, value);
cpu->FPSR = value & FPSR_ALL_FPSRS;
}
uint32_t
aarch64_get_FPSR (sim_cpu *cpu)
{
return cpu->FPSR;
}
void
aarch64_set_FPSR_bits (sim_cpu *cpu, uint32_t mask, uint32_t value)
{
uint32_t old_FPSR = cpu->FPSR;
mask &= FPSR_ALL_FPSRS;
cpu->FPSR &= ~mask;
cpu->FPSR |= (value & mask);
if (cpu->FPSR != old_FPSR)
TRACE_REGISTER (cpu,
"FPSR changes from %x to %x", old_FPSR, cpu->FPSR);
}
uint32_t
aarch64_get_FPSR_bits (sim_cpu *cpu, uint32_t mask)
{
mask &= FPSR_ALL_FPSRS;
return cpu->FPSR & mask;
}
int
aarch64_test_FPSR_bit (sim_cpu *cpu, FPSRMask flag)
{
return cpu->FPSR & flag;
}
uint64_t
aarch64_get_thread_id (sim_cpu *cpu)
{
return cpu->tpidr;
}
uint32_t
aarch64_get_FPCR (sim_cpu *cpu)
{
return cpu->FPCR;
}
void
aarch64_set_FPCR (sim_cpu *cpu, uint32_t val)
{
if (cpu->FPCR != val)
TRACE_REGISTER (cpu,
"FPCR changes from %x to %x", cpu->FPCR, val);
cpu->FPCR = val;
}