#ifndef GCC_ARM_H
#define GCC_ARM_H
+#include "config/vxworks-dummy.h"
+
/* The architecture define. */
extern char arm_arch_name[];
if (TARGET_VFP) \
builtin_define ("__VFP_FP__"); \
\
+ if (TARGET_NEON) \
+ builtin_define ("__ARM_NEON__"); \
+ \
/* Add a define for interworking. \
Needed when building libgcc.a. */ \
if (arm_cpp_interwork) \
/* 32-bit Thumb-2 code. */
#define TARGET_THUMB2 (TARGET_THUMB && arm_arch_thumb2)
+/* The following two macros concern the ability to execute coprocessor
+ instructions for VFPv3 or NEON. TARGET_VFP3 is currently only ever
+ tested when we know we are generating for VFP hardware; we need to
+ be more careful with TARGET_NEON as noted below. */
+
+/* FPU is VFPv3 (with twice the number of D registers). Setting the FPU to
+ Neon automatically enables VFPv3 too. */
+#define TARGET_VFP3 (arm_fp_model == ARM_FP_MODEL_VFP \
+ && (arm_fpu_arch == FPUTYPE_VFP3 \
+ || arm_fpu_arch == FPUTYPE_NEON))
+/* FPU supports Neon instructions. The setting of this macro gets
+ revealed via __ARM_NEON__ so we add extra guards upon TARGET_32BIT
+ and TARGET_HARD_FLOAT to ensure that NEON instructions are
+ available. */
+#define TARGET_NEON (TARGET_32BIT && TARGET_HARD_FLOAT \
+ && arm_fp_model == ARM_FP_MODEL_VFP \
+ && arm_fpu_arch == FPUTYPE_NEON)
+
/* "DSP" multiply instructions, eg. SMULxy. */
#define TARGET_DSP_MULTIPLY \
(TARGET_32BIT && arm_arch5e && arm_arch_notm)
/* Cirrus Maverick floating point co-processor. */
FPUTYPE_MAVERICK,
/* VFP. */
- FPUTYPE_VFP
+ FPUTYPE_VFP,
+ /* VFPv3. */
+ FPUTYPE_VFP3,
+ /* Neon. */
+ FPUTYPE_NEON
};
/* Recast the floating point class to be the floating point attribute. */
#define UNITS_PER_WORD 4
+/* Use the option -mvectorize-with-neon-quad to override the use of doubleword
+ registers when autovectorizing for Neon, at least until multiple vector
+ widths are supported properly by the middle-end. */
+#define UNITS_PER_SIMD_WORD \
+ (TARGET_NEON ? (TARGET_NEON_VECTORIZE_QUAD ? 16 : 8) : UNITS_PER_WORD)
+
/* True if natural alignment is used for doubleword types. */
#define ARM_DOUBLEWORD_ALIGN TARGET_AAPCS_BASED
1,1,1,1,1,1,1,1, \
1,1,1,1,1,1,1,1, \
1,1,1,1,1,1,1,1, \
+ 1,1,1,1,1,1,1,1, \
+ 1,1,1,1,1,1,1,1, \
+ 1,1,1,1,1,1,1,1, \
+ 1,1,1,1,1,1,1,1, \
1 \
}
1,1,1,1,1,1,1,1, \
1,1,1,1,1,1,1,1, \
1,1,1,1,1,1,1,1, \
+ 1,1,1,1,1,1,1,1, \
+ 1,1,1,1,1,1,1,1, \
+ 1,1,1,1,1,1,1,1, \
+ 1,1,1,1,1,1,1,1, \
1 \
}
} \
if (TARGET_VFP) \
{ \
+ /* VFPv3 registers are disabled when earlier VFP \
+ versions are selected due to the definition of \
+ LAST_VFP_REGNUM. */ \
for (regno = FIRST_VFP_REGNUM; \
regno <= LAST_VFP_REGNUM; ++ regno) \
{ \
fixed_regs[regno] = 0; \
- call_used_regs[regno] = regno < FIRST_VFP_REGNUM + 16; \
+ call_used_regs[regno] = regno < FIRST_VFP_REGNUM + 16 \
+ || regno >= FIRST_VFP_REGNUM + 32; \
} \
} \
} \
(((REGNUM) >= FIRST_CIRRUS_FP_REGNUM) && ((REGNUM) <= LAST_CIRRUS_FP_REGNUM))
#define FIRST_VFP_REGNUM 63
-#define LAST_VFP_REGNUM 94
+#define D7_VFP_REGNUM 78 /* Registers 77 and 78 == VFP reg D7. */
+#define LAST_VFP_REGNUM \
+ (TARGET_VFP3 ? LAST_HI_VFP_REGNUM : LAST_LO_VFP_REGNUM)
+
#define IS_VFP_REGNUM(REGNUM) \
(((REGNUM) >= FIRST_VFP_REGNUM) && ((REGNUM) <= LAST_VFP_REGNUM))
+/* VFP registers are split into two types: those defined by VFP versions < 3
+ have D registers overlaid on consecutive pairs of S registers. VFP version 3
+ defines 16 new D registers (d16-d31) which, for simplicity and correctness
+ in various parts of the backend, we implement as "fake" single-precision
+ registers (which would be S32-S63, but cannot be used in that way). The
+ following macros define these ranges of registers. */
+#define LAST_LO_VFP_REGNUM 94
+#define FIRST_HI_VFP_REGNUM 95
+#define LAST_HI_VFP_REGNUM 126
+
+#define VFP_REGNO_OK_FOR_SINGLE(REGNUM) \
+ ((REGNUM) <= LAST_LO_VFP_REGNUM)
+
+/* DFmode values are only valid in even register pairs. */
+#define VFP_REGNO_OK_FOR_DOUBLE(REGNUM) \
+ ((((REGNUM) - FIRST_VFP_REGNUM) & 1) == 0)
+
+/* Neon Quad values must start at a multiple of four registers. */
+#define NEON_REGNO_OK_FOR_QUAD(REGNUM) \
+ ((((REGNUM) - FIRST_VFP_REGNUM) & 3) == 0)
+
+/* Neon structures of vectors must be in even register pairs and there
+ must be enough registers available. Because of various patterns
+ requiring quad registers, we require them to start at a multiple of
+ four. */
+#define NEON_REGNO_OK_FOR_NREGS(REGNUM, N) \
+ ((((REGNUM) - FIRST_VFP_REGNUM) & 3) == 0 \
+ && (LAST_VFP_REGNUM - (REGNUM) >= 2 * (N) - 1))
+
/* The number of hard registers is 16 ARM + 8 FPA + 1 CC + 1 SFP + 1 AFP. */
/* + 16 Cirrus registers take us up to 43. */
/* Intel Wireless MMX Technology registers add 16 + 4 more. */
-/* VFP adds 32 + 1 more. */
-#define FIRST_PSEUDO_REGISTER 96
+/* VFP (VFP3) adds 32 (64) + 1 more. */
+#define FIRST_PSEUDO_REGISTER 128
#define DBX_REGISTER_NUMBER(REGNO) arm_dbx_register_number (REGNO)
#define VALID_IWMMXT_REG_MODE(MODE) \
(arm_vector_mode_supported_p (MODE) || (MODE) == DImode)
+/* Modes valid for Neon D registers. */
+#define VALID_NEON_DREG_MODE(MODE) \
+ ((MODE) == V2SImode || (MODE) == V4HImode || (MODE) == V8QImode \
+ || (MODE) == V2SFmode || (MODE) == DImode)
+
+/* Modes valid for Neon Q registers. */
+#define VALID_NEON_QREG_MODE(MODE) \
+ ((MODE) == V4SImode || (MODE) == V8HImode || (MODE) == V16QImode \
+ || (MODE) == V4SFmode || (MODE) == V2DImode)
+
+/* Structure modes valid for Neon registers. */
+#define VALID_NEON_STRUCT_MODE(MODE) \
+ ((MODE) == TImode || (MODE) == EImode || (MODE) == OImode \
+ || (MODE) == CImode || (MODE) == XImode)
+
/* The order in which register should be allocated. It is good to use ip
since no saving is required (though calls clobber it) and it never contains
function parameters. It is quite good to use lr since other calls may
clobber it anyway. Allocate r0 through r3 in reverse order since r3 is
least likely to contain a function parameter; in addition results are
- returned in r0. */
-
-#define REG_ALLOC_ORDER \
-{ \
- 3, 2, 1, 0, 12, 14, 4, 5, \
- 6, 7, 8, 10, 9, 11, 13, 15, \
- 16, 17, 18, 19, 20, 21, 22, 23, \
- 27, 28, 29, 30, 31, 32, 33, 34, \
- 35, 36, 37, 38, 39, 40, 41, 42, \
- 43, 44, 45, 46, 47, 48, 49, 50, \
- 51, 52, 53, 54, 55, 56, 57, 58, \
- 59, 60, 61, 62, \
- 24, 25, 26, \
- 78, 77, 76, 75, 74, 73, 72, 71, \
- 70, 69, 68, 67, 66, 65, 64, 63, \
- 79, 80, 81, 82, 83, 84, 85, 86, \
- 87, 88, 89, 90, 91, 92, 93, 94, \
- 95 \
+ returned in r0.
+ For VFP/VFPv3, allocate D16-D31 first, then caller-saved registers (D0-D7),
+ then D8-D15. The reason for doing this is to attempt to reduce register
+ pressure when both single- and double-precision registers are used in a
+ function. */
+
+#define REG_ALLOC_ORDER \
+{ \
+ 3, 2, 1, 0, 12, 14, 4, 5, \
+ 6, 7, 8, 10, 9, 11, 13, 15, \
+ 16, 17, 18, 19, 20, 21, 22, 23, \
+ 27, 28, 29, 30, 31, 32, 33, 34, \
+ 35, 36, 37, 38, 39, 40, 41, 42, \
+ 43, 44, 45, 46, 47, 48, 49, 50, \
+ 51, 52, 53, 54, 55, 56, 57, 58, \
+ 59, 60, 61, 62, \
+ 24, 25, 26, \
+ 95, 96, 97, 98, 99, 100, 101, 102, \
+ 103, 104, 105, 106, 107, 108, 109, 110, \
+ 111, 112, 113, 114, 115, 116, 117, 118, \
+ 119, 120, 121, 122, 123, 124, 125, 126, \
+ 78, 77, 76, 75, 74, 73, 72, 71, \
+ 70, 69, 68, 67, 66, 65, 64, 63, \
+ 79, 80, 81, 82, 83, 84, 85, 86, \
+ 87, 88, 89, 90, 91, 92, 93, 94, \
+ 127 \
}
/* Interrupt functions can only use registers that have already been
call-clobbered. */
#define HARD_REGNO_RENAME_OK(SRC, DST) \
(! IS_INTERRUPT (cfun->machine->func_type) || \
- regs_ever_live[DST])
+ df_regs_ever_live_p (DST))
\f
/* Register and constant classes. */
NO_REGS,
FPA_REGS,
CIRRUS_REGS,
+ VFP_D0_D7_REGS,
+ VFP_LO_REGS,
+ VFP_HI_REGS,
VFP_REGS,
IWMMXT_GR_REGS,
IWMMXT_REGS,
"NO_REGS", \
"FPA_REGS", \
"CIRRUS_REGS", \
+ "VFP_D0_D7_REGS", \
+ "VFP_LO_REGS", \
+ "VFP_HI_REGS", \
"VFP_REGS", \
"IWMMXT_GR_REGS", \
"IWMMXT_REGS", \
/* Define which registers fit in which classes.
This is an initializer for a vector of HARD_REG_SET
of length N_REG_CLASSES. */
-#define REG_CLASS_CONTENTS \
-{ \
- { 0x00000000, 0x00000000, 0x00000000 }, /* NO_REGS */ \
- { 0x00FF0000, 0x00000000, 0x00000000 }, /* FPA_REGS */ \
- { 0xF8000000, 0x000007FF, 0x00000000 }, /* CIRRUS_REGS */ \
- { 0x00000000, 0x80000000, 0x7FFFFFFF }, /* VFP_REGS */ \
- { 0x00000000, 0x00007800, 0x00000000 }, /* IWMMXT_GR_REGS */ \
- { 0x00000000, 0x7FFF8000, 0x00000000 }, /* IWMMXT_REGS */ \
- { 0x000000FF, 0x00000000, 0x00000000 }, /* LO_REGS */ \
- { 0x00002000, 0x00000000, 0x00000000 }, /* STACK_REG */ \
- { 0x000020FF, 0x00000000, 0x00000000 }, /* BASE_REGS */ \
- { 0x0000FF00, 0x00000000, 0x00000000 }, /* HI_REGS */ \
- { 0x01000000, 0x00000000, 0x00000000 }, /* CC_REG */ \
- { 0x00000000, 0x00000000, 0x80000000 }, /* VFPCC_REG */ \
- { 0x0200FFFF, 0x00000000, 0x00000000 }, /* GENERAL_REGS */ \
- { 0xFAFFFFFF, 0xFFFFFFFF, 0x7FFFFFFF } /* ALL_REGS */ \
+#define REG_CLASS_CONTENTS \
+{ \
+ { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, /* NO_REGS */ \
+ { 0x00FF0000, 0x00000000, 0x00000000, 0x00000000 }, /* FPA_REGS */ \
+ { 0xF8000000, 0x000007FF, 0x00000000, 0x00000000 }, /* CIRRUS_REGS */ \
+ { 0x00000000, 0x80000000, 0x00007FFF, 0x00000000 }, /* VFP_D0_D7_REGS */ \
+ { 0x00000000, 0x80000000, 0x7FFFFFFF, 0x00000000 }, /* VFP_LO_REGS */ \
+ { 0x00000000, 0x00000000, 0x80000000, 0x7FFFFFFF }, /* VFP_HI_REGS */ \
+ { 0x00000000, 0x80000000, 0xFFFFFFFF, 0x7FFFFFFF }, /* VFP_REGS */ \
+ { 0x00000000, 0x00007800, 0x00000000, 0x00000000 }, /* IWMMXT_GR_REGS */ \
+ { 0x00000000, 0x7FFF8000, 0x00000000, 0x00000000 }, /* IWMMXT_REGS */ \
+ { 0x000000FF, 0x00000000, 0x00000000, 0x00000000 }, /* LO_REGS */ \
+ { 0x00002000, 0x00000000, 0x00000000, 0x00000000 }, /* STACK_REG */ \
+ { 0x000020FF, 0x00000000, 0x00000000, 0x00000000 }, /* BASE_REGS */ \
+ { 0x0000FF00, 0x00000000, 0x00000000, 0x00000000 }, /* HI_REGS */ \
+ { 0x01000000, 0x00000000, 0x00000000, 0x00000000 }, /* CC_REG */ \
+ { 0x00000000, 0x00000000, 0x00000000, 0x80000000 }, /* VFPCC_REG */ \
+ { 0x0200FFFF, 0x00000000, 0x00000000, 0x00000000 }, /* GENERAL_REGS */ \
+ { 0xFAFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0x7FFFFFFF } /* ALL_REGS */ \
}
+/* Any of the VFP register classes. */
+#define IS_VFP_CLASS(X) \
+ ((X) == VFP_D0_D7_REGS || (X) == VFP_LO_REGS \
+ || (X) == VFP_HI_REGS || (X) == VFP_REGS)
+
/* The same information, inverted:
Return the class number of the smallest class containing
reg number REGNO. This could be a conditional expression
or out of a register in CLASS in MODE. If it can be done directly,
NO_REGS is returned. */
#define SECONDARY_OUTPUT_RELOAD_CLASS(CLASS, MODE, X) \
- /* Restrict which direct reloads are allowed for VFP regs. */ \
+ /* Restrict which direct reloads are allowed for VFP/iWMMXt regs. */ \
((TARGET_VFP && TARGET_HARD_FLOAT \
- && (CLASS) == VFP_REGS) \
- ? vfp_secondary_reload_class (MODE, X) \
+ && IS_VFP_CLASS (CLASS)) \
+ ? coproc_secondary_reload_class (MODE, X, FALSE) \
+ : (TARGET_IWMMXT && (CLASS) == IWMMXT_REGS) \
+ ? coproc_secondary_reload_class (MODE, X, TRUE) \
: TARGET_32BIT \
? (((MODE) == HImode && ! arm_arch4 && true_regnum (X) == -1) \
? GENERAL_REGS : NO_REGS) \
/* If we need to load shorts byte-at-a-time, then we need a scratch. */
#define SECONDARY_INPUT_RELOAD_CLASS(CLASS, MODE, X) \
- /* Restrict which direct reloads are allowed for VFP regs. */ \
+ /* Restrict which direct reloads are allowed for VFP/iWMMXt regs. */ \
((TARGET_VFP && TARGET_HARD_FLOAT \
- && (CLASS) == VFP_REGS) \
- ? vfp_secondary_reload_class (MODE, X) : \
+ && IS_VFP_CLASS (CLASS)) \
+ ? coproc_secondary_reload_class (MODE, X, FALSE) : \
+ (TARGET_IWMMXT && (CLASS) == IWMMXT_REGS) ? \
+ coproc_secondary_reload_class (MODE, X, TRUE) : \
/* Cannot load constants into Cirrus registers. */ \
(TARGET_MAVERICK && TARGET_HARD_FLOAT \
&& (CLASS) == CIRRUS_REGS \
(TARGET_32BIT ? \
((FROM) == FPA_REGS && (TO) != FPA_REGS ? 20 : \
(FROM) != FPA_REGS && (TO) == FPA_REGS ? 20 : \
- (FROM) == VFP_REGS && (TO) != VFP_REGS ? 10 : \
- (FROM) != VFP_REGS && (TO) == VFP_REGS ? 10 : \
+ IS_VFP_CLASS (FROM) && !IS_VFP_CLASS (TO) ? 10 : \
+ !IS_VFP_CLASS (FROM) && IS_VFP_CLASS (TO) ? 10 : \
(FROM) == IWMMXT_REGS && (TO) != IWMMXT_REGS ? 4 : \
(FROM) != IWMMXT_REGS && (TO) == IWMMXT_REGS ? 4 : \
(FROM) == IWMMXT_GR_REGS || (TO) == IWMMXT_GR_REGS ? 20 : \
than a word, or if they contain elements offset from zero in the struct. */
#define DEFAULT_PCC_STRUCT_RETURN 0
-/* Flags for the call/call_value rtl operations set up by function_arg. */
-#define CALL_NORMAL 0x00000000 /* No special processing. */
-#define CALL_LONG 0x00000001 /* Always call indirect. */
-#define CALL_SHORT 0x00000002 /* Never call indirect. */
-
/* These bits describe the different types of function supported
by the ARM backend. They are exclusive. i.e. a function cannot be both a
normal function and an interworked function, for example. Knowing the
int iwmmxt_nregs;
int named_count;
int nargs;
- /* One of CALL_NORMAL, CALL_LONG or CALL_SHORT. */
- int call_cookie;
int can_split;
} CUMULATIVE_ARGS;
frame. */
#define EXIT_IGNORE_STACK 1
-#define EPILOGUE_USES(REGNO) (reload_completed && (REGNO) == LR_REGNUM)
+#define EPILOGUE_USES(REGNO) ((REGNO) == LR_REGNUM)
/* Determine if the epilogue should be output as RTL.
You should override this if you define FUNCTION_EXTRA_EPILOGUE. */
#endif /* AOF_ASSEMBLER */
+/* True if SYMBOL + OFFSET constants must refer to something within
+ SYMBOL's section. */
+#define ARM_OFFSETS_MUST_BE_WITHIN_SECTIONS_P 0
+
/* Nonzero if the constant value X is a legitimate general operand.
It is given that X satisfies CONSTANT_P or is a CONST_DOUBLE.
|| flag_pic)
#define LEGITIMATE_CONSTANT_P(X) \
- (!arm_tls_referenced_p (X) \
+ (!arm_cannot_force_const_mem (X) \
&& (TARGET_32BIT ? ARM_LEGITIMATE_CONSTANT_P (X) \
: THUMB_LEGITIMATE_CONSTANT_P (X)))
-/* Special characters prefixed to function names
- in order to encode attribute like information.
- Note, '@' and '*' have already been taken. */
-#define SHORT_CALL_FLAG_CHAR '^'
-#define LONG_CALL_FLAG_CHAR '#'
-
-#define ENCODED_SHORT_CALL_ATTR_P(SYMBOL_NAME) \
- (*(SYMBOL_NAME) == SHORT_CALL_FLAG_CHAR)
-
-#define ENCODED_LONG_CALL_ATTR_P(SYMBOL_NAME) \
- (*(SYMBOL_NAME) == LONG_CALL_FLAG_CHAR)
-
#ifndef SUBTARGET_NAME_ENCODING_LENGTHS
#define SUBTARGET_NAME_ENCODING_LENGTHS
#endif
be stripped from the start of a function's name, if that
name starts with the indicated character. */
#define ARM_NAME_ENCODING_LENGTHS \
- case SHORT_CALL_FLAG_CHAR: return 1; \
- case LONG_CALL_FLAG_CHAR: return 1; \
case '*': return 1; \
SUBTARGET_NAME_ENCODING_LENGTHS
#define ASM_OUTPUT_LABELREF(FILE, NAME) \
arm_asm_output_labelref (FILE, NAME)
-/* Output IT instructions for conditonally executed Thumb-2 instructions. */
+/* Output IT instructions for conditionally executed Thumb-2 instructions. */
#define ASM_OUTPUT_OPCODE(STREAM, PTR) \
if (TARGET_THUMB2) \
thumb2_asm_output_opcode (STREAM);
#define TARGET_ARM_DYNAMIC_VAGUE_LINKAGE_P true
#endif
-/* Set the short-call flag for any function compiled in the current
- compilation unit. We skip this for functions with the section
- attribute when long-calls are in effect as this tells the compiler
- that the section might be placed a long way from the caller.
- See arm_is_longcall_p() for more information. */
-#define ARM_DECLARE_FUNCTION_SIZE(STREAM, NAME, DECL) \
- if (!TARGET_LONG_CALLS || ! DECL_SECTION_NAME (DECL)) \
- arm_encode_call_attribute (DECL, SHORT_CALL_FLAG_CHAR)
-
#define ARM_OUTPUT_FN_UNWIND(F, PROLOGUE) arm_output_fn_unwind (F, PROLOGUE)
#ifdef TARGET_UNWIND_INFO
The usual definition accepts all pseudo regs; the other rejects
them unless they have been allocated suitable hard regs.
The symbol REG_OK_STRICT causes the latter definition to be used.
- Thumb-2 has the same restictions as arm. */
+ Thumb-2 has the same restrictions as arm. */
#ifndef REG_OK_STRICT
#define ARM_REG_OK_FOR_BASE_P(X) \
#define PRINT_OPERAND_PUNCT_VALID_P(CODE) \
(CODE == '@' || CODE == '|' || CODE == '.' \
- || CODE == '(' || CODE == ')' \
+ || CODE == '(' || CODE == ')' || CODE == '#' \
|| (TARGET_32BIT && (CODE == '?')) \
|| (TARGET_THUMB2 && (CODE == '!')) \
|| (TARGET_THUMB && (CODE == '_')))
: arm_gen_return_addr_mask ())
\f
+/* Neon defines builtins from ARM_BUILTIN_MAX upwards, though they don't have
+ symbolic names defined here (which would require too much duplication).
+ FIXME? */
enum arm_builtins
{
ARM_BUILTIN_GETWCX,
ARM_BUILTIN_THREAD_POINTER,
- ARM_BUILTIN_MAX
+ ARM_BUILTIN_NEON_BASE,
+
+ ARM_BUILTIN_MAX = ARM_BUILTIN_NEON_BASE /* FIXME: Wrong! */
};
/* Do not emit .note.GNU-stack by default. */
OSDN Git Service
