-/* Copyright (C) 2006-2015 Free Software Foundation, Inc.
+/* Copyright (C) 2006-2019 Free Software Foundation, Inc.
This file 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
along with GCC; see the file COPYING3. If not see
<http://www.gnu.org/licenses/>. */
+#define IN_TARGET_CODE 1
+
#include "config.h"
#include "system.h"
#include "coretypes.h"
#include "backend.h"
-#include "cfghooks.h"
+#include "target.h"
+#include "rtl.h"
#include "tree.h"
#include "gimple.h"
-#include "rtl.h"
+#include "cfghooks.h"
+#include "cfgloop.h"
#include "df.h"
+#include "memmodel.h"
+#include "tm_p.h"
+#include "stringpool.h"
+#include "attribs.h"
+#include "expmed.h"
+#include "optabs.h"
#include "regs.h"
-#include "insn-config.h"
-#include "conditions.h"
-#include "insn-attr.h"
-#include "flags.h"
+#include "emit-rtl.h"
#include "recog.h"
+#include "diagnostic-core.h"
+#include "insn-attr.h"
#include "alias.h"
#include "fold-const.h"
-#include "stringpool.h"
#include "stor-layout.h"
#include "calls.h"
#include "varasm.h"
-#include "expmed.h"
-#include "dojump.h"
#include "explow.h"
-#include "emit-rtl.h"
-#include "stmt.h"
#include "expr.h"
-#include "insn-codes.h"
-#include "optabs.h"
-#include "except.h"
#include "output.h"
#include "cfgrtl.h"
-#include "cfganal.h"
-#include "lcm.h"
#include "cfgbuild.h"
-#include "cfgcleanup.h"
-#include "diagnostic-core.h"
-#include "tm_p.h"
-#include "target.h"
#include "langhooks.h"
#include "reload.h"
#include "sched-int.h"
#include "params.h"
-#include "internal-fn.h"
-#include "gimple-fold.h"
-#include "tree-eh.h"
#include "gimplify.h"
#include "tm-constrs.h"
#include "ddg.h"
-#include "timevar.h"
#include "dumpfile.h"
-#include "cfgloop.h"
#include "builtins.h"
#include "rtl-iter.h"
+#include "flags.h"
+#include "toplev.h"
/* This file should be included last. */
#include "target-def.h"
flag_omit_frame_pointer = 1;
/* Functions must be 8 byte aligned so we correctly handle dual issue */
- if (align_functions < 8)
- align_functions = 8;
+ parse_alignment_opts ();
+ if (align_functions.levels[0].get_value () < 8)
+ str_align_functions = "8";
spu_hint_dist = 8*4 - spu_max_nops*4;
if (spu_hint_dist < 0)
else if (strcmp (&spu_arch_string[0], "celledp") == 0)
spu_arch = PROCESSOR_CELLEDP;
else
- error ("bad value (%s) for -march= switch", spu_arch_string);
+ error ("bad value (%s) for %<-march=%> switch", spu_arch_string);
}
/* Determine processor to tune for. */
else if (strcmp (&spu_tune_string[0], "celledp") == 0)
spu_tune = PROCESSOR_CELLEDP;
else
- error ("bad value (%s) for -mtune= switch", spu_tune_string);
+ error ("bad value (%s) for %<-mtune=%> switch", spu_tune_string);
}
/* Change defaults according to the processor architecture. */
REAL_MODE_FORMAT (SFmode) = &spu_single_format;
}
\f
+/* Implement TARGET_HARD_REGNO_NREGS. */
+
+static unsigned int
+spu_hard_regno_nregs (unsigned int, machine_mode mode)
+{
+ return CEIL (GET_MODE_BITSIZE (mode), MAX_FIXED_MODE_SIZE);
+}
+
/* Handle an attribute requiring a FUNCTION_DECL; arguments as in
struct attribute_spec.handler. */
be manipulated in non-trivial ways. In particular, this means all
the arithmetic is supported. */
static bool
-spu_scalar_mode_supported_p (machine_mode mode)
+spu_scalar_mode_supported_p (scalar_mode mode)
{
switch (mode)
{
- case QImode:
- case HImode:
- case SImode:
- case SFmode:
- case DImode:
- case TImode:
- case DFmode:
+ case E_QImode:
+ case E_HImode:
+ case E_SImode:
+ case E_SFmode:
+ case E_DImode:
+ case E_TImode:
+ case E_DFmode:
return true;
default:
{
switch (mode)
{
- case V16QImode:
- case V8HImode:
- case V4SImode:
- case V2DImode:
- case V4SFmode:
- case V2DFmode:
+ case E_V16QImode:
+ case E_V8HImode:
+ case E_V4SImode:
+ case E_V2DImode:
+ case E_V4SFmode:
+ case E_V2DFmode:
return true;
default:
op_size = 32;
}
/* If it is not a MODE_INT (and/or it is smaller than SI) add a SUBREG. */
- mode = mode_for_size (op_size, MODE_INT, 0);
+ mode = int_mode_for_size (op_size, 0).require ();
if (mode != GET_MODE (op))
op = gen_rtx_SUBREG (mode, op, 0);
return op;
{
switch (dst_mode)
{
- case SImode:
+ case E_SImode:
emit_insn (gen_ashlsi3 (shift_reg, shift_reg, GEN_INT (shift)));
break;
- case DImode:
+ case E_DImode:
emit_insn (gen_ashldi3 (shift_reg, shift_reg, GEN_INT (shift)));
break;
- case TImode:
+ case E_TImode:
emit_insn (gen_ashlti3 (shift_reg, shift_reg, GEN_INT (shift)));
break;
default:
switch (op_mode)
{
- case QImode:
+ case E_QImode:
index = 0;
comp_mode = QImode;
break;
- case HImode:
+ case E_HImode:
index = 1;
comp_mode = HImode;
break;
- case SImode:
+ case E_SImode:
index = 2;
break;
- case DImode:
+ case E_DImode:
index = 3;
break;
- case TImode:
+ case E_TImode:
index = 4;
break;
- case SFmode:
+ case E_SFmode:
index = 5;
break;
- case DFmode:
+ case E_DFmode:
index = 6;
break;
- case V16QImode:
+ case E_V16QImode:
index = 7;
comp_mode = op_mode;
break;
- case V8HImode:
+ case E_V8HImode:
index = 8;
comp_mode = op_mode;
break;
- case V4SImode:
+ case E_V4SImode:
index = 9;
comp_mode = op_mode;
break;
- case V4SFmode:
+ case E_V4SFmode:
index = 10;
comp_mode = V4SImode;
break;
- case V2DFmode:
+ case E_V2DFmode:
index = 11;
comp_mode = V2DImode;
break;
- case V2DImode:
+ case E_V2DImode:
default:
abort ();
}
rtx target = operands[0];
int compare_size = GET_MODE_BITSIZE (comp_mode);
int target_size = GET_MODE_BITSIZE (GET_MODE (target));
- machine_mode mode = mode_for_size (target_size, MODE_INT, 0);
+ machine_mode mode = int_mode_for_size (target_size, 0).require ();
rtx select_mask;
rtx op_t = operands[2];
rtx op_f = operands[3];
/* Used in indirect function calls. */
fprintf (file, "%s", reg_names[REGNO (XEXP (x, 0))]);
else
- output_address (XEXP (x, 0));
+ output_address (GET_MODE (x), XEXP (x, 0));
}
return;
if (xcode == REG)
fprintf (file, "%s", reg_names[REGNO (x)]);
else if (xcode == MEM)
- output_address (XEXP (x, 0));
+ output_address (GET_MODE (x), XEXP (x, 0));
else if (xcode == CONST_VECTOR)
print_operand (file, CONST_VECTOR_ELT (x, 0), 0);
else
unsigned char arrlo[16];
rtx to, temp, hi, lo;
int i;
- machine_mode imode = mode;
/* We need to do reals as ints because the constant used in the
IOR might not be a legitimate real constant. */
- imode = int_mode_for_mode (mode);
+ scalar_int_mode imode = int_mode_for_mode (mode).require ();
constant_to_array (mode, ops[1], arrhi);
if (imode != mode)
to = simplify_gen_subreg (imode, ops[0], mode, 0);
unsigned char arr_andbi[16];
rtx to, reg_fsmbi, reg_and;
int i;
- machine_mode imode = mode;
/* We need to do reals as ints because the constant used in the
* AND might not be a legitimate real constant. */
- imode = int_mode_for_mode (mode);
+ scalar_int_mode imode = int_mode_for_mode (mode).require ();
constant_to_array (mode, ops[1], arr_fsmbi);
if (imode != mode)
to = simplify_gen_subreg(imode, ops[0], GET_MODE (ops[0]), 0);
if (total_size > 0)
{
- if (flag_stack_check)
+ if (flag_stack_check || flag_stack_clash_protection)
{
/* We compare against total_size-1 because
($sp >= total_size) <=> ($sp > total_size-1) */
size_v4si = scratch_v4si;
}
emit_insn (gen_cgt_v4si (scratch_v4si, sp_v4si, size_v4si));
- emit_insn (gen_vec_extractv4si
+ emit_insn (gen_vec_extractv4sisi
(scratch_reg_0, scratch_v4si, GEN_INT (1)));
emit_insn (gen_spu_heq (scratch_reg_0, GEN_INT (0)));
}
spu_const (machine_mode mode, HOST_WIDE_INT val)
{
rtx inner;
- rtvec v;
- int units, i;
gcc_assert (GET_MODE_CLASS (mode) == MODE_INT
|| GET_MODE_CLASS (mode) == MODE_FLOAT
else
inner = hwint_to_const_double (GET_MODE_INNER (mode), val);
- units = GET_MODE_NUNITS (mode);
-
- v = rtvec_alloc (units);
-
- for (i = 0; i < units; ++i)
- RTVEC_ELT (v, i) = inner;
-
- return gen_rtx_CONST_VECTOR (mode, v);
+ return gen_const_vec_duplicate (mode, inner);
}
/* Create a MODE vector constant from 4 ints. */
for (; insn; insn = next_insn)
{
next_insn = next_active_insn (insn);
- if (INSN_CODE (insn) == CODE_FOR_iprefetch
- || INSN_CODE (insn) == CODE_FOR_hbr)
+ if (INSN_P (insn)
+ && (INSN_CODE (insn) == CODE_FOR_iprefetch
+ || INSN_CODE (insn) == CODE_FOR_hbr))
{
if (hbr_insn)
{
}
hbr_insn = insn;
}
- if (INSN_CODE (insn) == CODE_FOR_blockage && next_insn)
+ if (INSN_P (insn) && INSN_CODE (insn) == CODE_FOR_blockage && next_insn)
{
if (GET_MODE (insn) == TImode)
PUT_MODE (next_insn, TImode);
spu_emit_branch_hint (rtx_insn *before, rtx_insn *branch, rtx target,
int distance, sbitmap blocks)
{
- rtx branch_label = 0;
rtx_insn *hint;
rtx_insn *insn;
rtx_jump_table_data *table;
if (NOTE_INSN_BASIC_BLOCK_P (before))
before = NEXT_INSN (before);
- branch_label = gen_label_rtx ();
+ rtx_code_label *branch_label = gen_label_rtx ();
LABEL_NUSES (branch_label)++;
LABEL_PRESERVE_P (branch_label) = 1;
insn = emit_label_before (branch_label, branch);
- branch_label = gen_rtx_LABEL_REF (VOIDmode, branch_label);
+ rtx branch_label_ref = gen_rtx_LABEL_REF (VOIDmode, branch_label);
bitmap_set_bit (blocks, BLOCK_FOR_INSN (branch)->index);
- hint = emit_insn_before (gen_hbr (branch_label, target), before);
+ hint = emit_insn_before (gen_hbr (branch_label_ref, target), before);
recog_memoized (hint);
INSN_LOCATION (hint) = INSN_LOCATION (branch);
HINTED_P (branch) = 1;
{
/* If the more probable case is not a fall through, then
try a branch hint. */
- int prob = XINT (note, 0);
+ int prob = profile_probability::from_reg_br_prob_note
+ (XINT (note, 0)).to_reg_br_prob_base ();
if (prob > (REG_BR_PROB_BASE * 6 / 10)
&& GET_CODE (XEXP (src, 1)) != PC)
lab = XEXP (src, 1);
spu_sched_init (FILE *file ATTRIBUTE_UNUSED, int verbose ATTRIBUTE_UNUSED,
int max_ready ATTRIBUTE_UNUSED)
{
- if (align_labels > 4 || align_loops > 4 || align_jumps > 4)
+ if (align_labels.levels[0].get_value () > 4
+ || align_loops.levels[0].get_value () > 4
+ || align_jumps.levels[0].get_value () > 4)
{
/* When any block might be at least 8-byte aligned, assume they
will all be at least 8-byte aligned to make sure dual issue
case TYPE_LOAD:
case TYPE_STORE:
pipe_ls = i;
+ /* FALLTHRU */
case TYPE_LNOP:
case TYPE_SHUF:
case TYPE_BR:
/* INSN is dependent on DEP_INSN. */
static int
-spu_sched_adjust_cost (rtx_insn *insn, rtx link, rtx_insn *dep_insn, int cost)
+spu_sched_adjust_cost (rtx_insn *insn, int dep_type, rtx_insn *dep_insn,
+ int cost, unsigned int)
{
rtx set;
scheduler makes every insn in a block anti-dependent on the final
jump_insn. We adjust here so higher cost insns will get scheduled
earlier. */
- if (JUMP_P (insn) && REG_NOTE_KIND (link) == REG_DEP_ANTI)
- return insn_cost (dep_insn) - 3;
+ if (JUMP_P (insn) && dep_type == REG_DEP_ANTI)
+ return insn_sched_cost (dep_insn) - 3;
return cost;
}
constant_to_array (mode, op, arr);
bytes = GET_MODE_UNIT_SIZE (mode);
- mode = mode_for_size (GET_MODE_UNIT_BITSIZE (mode), MODE_INT, 0);
+ mode = int_mode_for_mode (GET_MODE_INNER (mode)).require ();
/* Check that bytes are repeated. */
for (i = bytes; i < 16; i += bytes)
mode = GET_MODE_INNER (mode);
bytes = GET_MODE_SIZE (mode);
- int_mode = mode_for_size (GET_MODE_BITSIZE (mode), MODE_INT, 0);
+ int_mode = int_mode_for_mode (mode).require ();
/* Check that bytes are repeated. */
for (i = bytes; i < 16; i += bytes)
case SUBREG:
x = XEXP (x, 0);
- if (REG_P (x))
+ if (!REG_P (x))
return 0;
+ /* FALLTHRU */
case REG:
return INT_REG_OK_FOR_BASE_P (x, reg_ok_strict);
unsigned_p = TYPE_UNSIGNED (type);
switch (mode)
{
- case DImode:
+ case E_DImode:
result = (unsigned_p ? unsigned_V2DI_type_node : V2DI_type_node);
break;
- case SImode:
+ case E_SImode:
result = (unsigned_p ? unsigned_V4SI_type_node : V4SI_type_node);
break;
- case HImode:
+ case E_HImode:
result = (unsigned_p ? unsigned_V8HI_type_node : V8HI_type_node);
break;
- case QImode:
+ case E_QImode:
result = (unsigned_p ? unsigned_V16QI_type_node : V16QI_type_node);
break;
- case SFmode:
+ case E_SFmode:
result = V4SF_type_node;
break;
- case DFmode:
+ case E_DFmode:
result = V2DF_type_node;
break;
default:
{
if (byte_size < 4)
byte_size = 4;
- smode =
- smallest_mode_for_size (byte_size * BITS_PER_UNIT, MODE_INT);
+ smode = smallest_int_mode_for_size (byte_size * BITS_PER_UNIT);
RTVEC_ELT (v, n) =
gen_rtx_EXPR_LIST (VOIDmode,
gen_rtx_REG (smode, FIRST_RETURN_REGNUM + n),
rtx gr_reg;
if (byte_size < 4)
byte_size = 4;
- smode = smallest_mode_for_size (byte_size * BITS_PER_UNIT, MODE_INT);
+ smode = smallest_int_mode_for_size (byte_size * BITS_PER_UNIT);
gr_reg = gen_rtx_EXPR_LIST (VOIDmode,
gen_rtx_REG (smode, FIRST_ARG_REGNUM + *cum),
const0_rtx);
? ((int_size_in_bytes (type) + 15) / 16)
: mode == VOIDmode
? 1
- : HARD_REGNO_NREGS (cum, mode));
+ : spu_hard_regno_nregs (FIRST_ARG_REGNUM, mode));
+}
+
+/* Implement TARGET_FUNCTION_ARG_OFFSET. The SPU ABI wants 32/64-bit
+ types at offset 0 in the quad-word on the stack. 8/16-bit types
+ should be at offsets 3/2 respectively. */
+
+static HOST_WIDE_INT
+spu_function_arg_offset (machine_mode mode, const_tree type)
+{
+ if (type && INTEGRAL_TYPE_P (type) && GET_MODE_SIZE (mode) < 4)
+ return 4 - GET_MODE_SIZE (mode);
+ return 0;
+}
+
+/* Implement TARGET_FUNCTION_ARG_PADDING. */
+
+static pad_direction
+spu_function_arg_padding (machine_mode, const_tree)
+{
+ return PAD_UPWARD;
}
/* Variable sized types are passed by reference. */
FIELD_DECL, get_identifier ("__skip"), ptr_type_node);
DECL_FIELD_CONTEXT (f_args) = record;
- DECL_ALIGN (f_args) = 128;
+ SET_DECL_ALIGN (f_args, 128);
DECL_USER_ALIGN (f_args) = 1;
DECL_FIELD_CONTEXT (f_skip) = record;
- DECL_ALIGN (f_skip) = 128;
+ SET_DECL_ALIGN (f_skip, 128);
DECL_USER_ALIGN (f_skip) = 1;
TYPE_STUB_DECL (record) = type_decl;
which is both 16-byte aligned and padded to a 16-byte boundary. This
would make it safe to store with a single instruction.
We guarantee the alignment and padding for static objects by aligning
- all of them to 16-bytes. (DATA_ALIGNMENT and CONSTANT_ALIGNMENT.)
+ all of them to 16-bytes. (DATA_ALIGNMENT and TARGET_CONSTANT_ALIGNMENT.)
FIXME: We currently cannot guarantee this for objects on the stack
because assign_parm_setup_stack calls assign_stack_local with the
alignment of the parameter mode and in that case the alignment never
if (!cache_fetch_dirty)
cache_fetch_dirty = init_one_libfunc ("__cache_fetch_dirty");
emit_library_call_value (cache_fetch_dirty, data_addr, LCT_NORMAL, Pmode,
- 2, ea_addr, EAmode, ndirty, SImode);
+ ea_addr, EAmode, ndirty, SImode);
}
else
{
if (!cache_fetch)
cache_fetch = init_one_libfunc ("__cache_fetch");
emit_library_call_value (cache_fetch, data_addr, LCT_NORMAL, Pmode,
- 1, ea_addr, EAmode);
+ ea_addr, EAmode);
}
}
gen_rtx_IF_THEN_ELSE (VOIDmode, bcomp,
hit_ref, pc_rtx)));
/* Say that this branch is very likely to happen. */
- v = REG_BR_PROB_BASE - REG_BR_PROB_BASE / 100 - 1;
- add_int_reg_note (insn, REG_BR_PROB, v);
+ add_reg_br_prob_note (insn, profile_probability::very_likely ());
ea_load_store (mem, is_store, ea_addr, data_addr);
cont_label = gen_label_rtx ();
if (GET_CODE (ops[1]) == SUBREG && !valid_subreg (ops[1]))
{
rtx from = SUBREG_REG (ops[1]);
- machine_mode imode = int_mode_for_mode (GET_MODE (from));
+ scalar_int_mode imode = int_mode_for_mode (GET_MODE (from)).require ();
gcc_assert (GET_MODE_CLASS (mode) == MODE_INT
&& GET_MODE_CLASS (imode) == MODE_INT
spu_convert_move (rtx dst, rtx src)
{
machine_mode mode = GET_MODE (dst);
- machine_mode int_mode = mode_for_size (GET_MODE_BITSIZE (mode), MODE_INT, 0);
+ machine_mode int_mode = int_mode_for_mode (mode).require ();
rtx reg;
gcc_assert (GET_MODE (src) == TImode);
reg = int_mode != mode ? gen_reg_rtx (int_mode) : dst;
dash = strchr (str, '-');
if (!dash)
{
- warning (0, "value of -mfixed-range must have form REG1-REG2");
+ warning (0, "value of %<-mfixed-range%> must have form REG1-REG2");
return;
}
*dash = '\0';
return true;
}
-static machine_mode
+static scalar_int_mode
spu_unwind_word_mode (void)
{
return SImode;
emit_insn (gen_spu_convert (sp, stack_pointer_rtx));
emit_insn (gen_subv4si3 (sp, sp, splatted));
- if (flag_stack_check)
+ if (flag_stack_check || flag_stack_clash_protection)
{
rtx avail = gen_reg_rtx(SImode);
rtx result = gen_reg_rtx(SImode);
- emit_insn (gen_vec_extractv4si (avail, sp, GEN_INT (1)));
+ emit_insn (gen_vec_extractv4sisi (avail, sp, GEN_INT (1)));
emit_insn (gen_cgt_si(result, avail, GEN_INT (-1)));
emit_insn (gen_spu_heq (result, GEN_INT(0) ));
}
ops[1] = force_reg (GET_MODE_INNER (mode), ops[1]);
switch (mode)
{
- case V2DImode:
- case V2DFmode:
+ case E_V2DImode:
+ case E_V2DFmode:
shuf =
immed_double_const (0x0001020304050607ll, 0x1011121314151617ll,
TImode);
break;
- case V4SImode:
- case V4SFmode:
+ case E_V4SImode:
+ case E_V4SFmode:
shuf =
immed_double_const (0x0001020300010203ll, 0x0001020300010203ll,
TImode);
break;
- case V8HImode:
+ case E_V8HImode:
shuf =
immed_double_const (0x0203020302030203ll, 0x0203020302030203ll,
TImode);
break;
- case V16QImode:
+ case E_V16QImode:
shuf =
immed_double_const (0x0303030303030303ll, 0x0303030303030303ll,
TImode);
{
switch (mode)
{
- case V16QImode:
- emit_insn (gen_vec_extractv16qi (ops[0], ops[1], ops[2]));
+ case E_V16QImode:
+ emit_insn (gen_vec_extractv16qiqi (ops[0], ops[1], ops[2]));
break;
- case V8HImode:
- emit_insn (gen_vec_extractv8hi (ops[0], ops[1], ops[2]));
+ case E_V8HImode:
+ emit_insn (gen_vec_extractv8hihi (ops[0], ops[1], ops[2]));
break;
- case V4SFmode:
- emit_insn (gen_vec_extractv4sf (ops[0], ops[1], ops[2]));
+ case E_V4SFmode:
+ emit_insn (gen_vec_extractv4sfsf (ops[0], ops[1], ops[2]));
break;
- case V4SImode:
- emit_insn (gen_vec_extractv4si (ops[0], ops[1], ops[2]));
+ case E_V4SImode:
+ emit_insn (gen_vec_extractv4sisi (ops[0], ops[1], ops[2]));
break;
- case V2DImode:
- emit_insn (gen_vec_extractv2di (ops[0], ops[1], ops[2]));
+ case E_V2DImode:
+ emit_insn (gen_vec_extractv2didi (ops[0], ops[1], ops[2]));
break;
- case V2DFmode:
- emit_insn (gen_vec_extractv2df (ops[0], ops[1], ops[2]));
+ case E_V2DFmode:
+ emit_insn (gen_vec_extractv2dfdf (ops[0], ops[1], ops[2]));
break;
default:
abort ();
switch (mode)
{
- case V16QImode:
+ case E_V16QImode:
emit_insn (gen_addsi3 (tmp, ops[2], GEN_INT (-3)));
break;
- case V8HImode:
+ case E_V8HImode:
emit_insn (gen_addsi3 (tmp, ops[2], ops[2]));
emit_insn (gen_addsi3 (tmp, tmp, GEN_INT (-2)));
break;
- case V4SFmode:
- case V4SImode:
+ case E_V4SFmode:
+ case E_V4SImode:
emit_insn (gen_ashlsi3 (tmp, ops[2], GEN_INT (2)));
break;
- case V2DImode:
- case V2DFmode:
+ case E_V2DImode:
+ case E_V2DFmode:
emit_insn (gen_ashlsi3 (tmp, ops[2], GEN_INT (3)));
break;
default:
offset = gen_reg_rtx (SImode);
switch (mode)
{
- case V16QImode:
+ case E_V16QImode:
emit_insn (gen_subsi3 (offset, GEN_INT (3), ops[2]));
break;
- case V8HImode:
+ case E_V8HImode:
emit_insn (gen_subsi3 (offset, GEN_INT (1), ops[2]));
emit_insn (gen_addsi3 (offset, offset, offset));
break;
- case V4SFmode:
- case V4SImode:
+ case E_V4SFmode:
+ case E_V4SImode:
emit_insn (gen_subsi3 (offset, GEN_INT (0), ops[2]));
emit_insn (gen_ashlsi3 (offset, offset, GEN_INT (2)));
break;
- case V2DImode:
- case V2DFmode:
+ case E_V2DImode:
+ case E_V2DFmode:
emit_insn (gen_ashlsi3 (offset, ops[2], GEN_INT (3)));
break;
default:
arr[i] = 0x10;
switch (GET_MODE (ops[1]))
{
- case HImode:
+ case E_HImode:
sign = gen_reg_rtx (SImode);
emit_insn (gen_extendhisi2 (sign, ops[1]));
arr[last] = 0x03;
arr[last - 1] = 0x02;
break;
- case SImode:
+ case E_SImode:
sign = gen_reg_rtx (SImode);
emit_insn (gen_ashrsi3 (sign, ops[1], GEN_INT (31)));
for (i = 0; i < 4; i++)
arr[last - i] = 3 - i;
break;
- case DImode:
+ case E_DImode:
sign = gen_reg_rtx (SImode);
c = gen_reg_rtx (SImode);
emit_insn (gen_spu_convert (c, ops[1]));
return 2;
case unaligned_load:
+ case vector_gather_load:
+ case vector_scatter_store:
return 2;
case cond_branch_taken:
}
/* Return the appropriate mode for a named address pointer. */
-static machine_mode
+static scalar_int_mode
spu_addr_space_pointer_mode (addr_space_t addrspace)
{
switch (addrspace)
}
/* Return the appropriate mode for a named address address. */
-static machine_mode
+static scalar_int_mode
spu_addr_space_address_mode (addr_space_t addrspace)
{
switch (addrspace)
}
}
-static machine_mode
+static scalar_int_mode
spu_libgcc_cmp_return_mode (void)
{
return SImode;
}
-static machine_mode
+static scalar_int_mode
spu_libgcc_shift_count_mode (void)
{
/* For SPU word mode is TI mode so it is better to use SImode
emit_move_insn (orig_after, after);
}
+/* Implement TARGET_MODES_TIEABLE_P. */
+
+static bool
+spu_modes_tieable_p (machine_mode mode1, machine_mode mode2)
+{
+ return (GET_MODE_BITSIZE (mode1) <= MAX_FIXED_MODE_SIZE
+ && GET_MODE_BITSIZE (mode2) <= MAX_FIXED_MODE_SIZE);
+}
+
+/* Implement TARGET_CAN_CHANGE_MODE_CLASS. GCC assumes that modes are
+ in the lowpart of a register, which is only true for SPU. */
+
+static bool
+spu_can_change_mode_class (machine_mode from, machine_mode to, reg_class_t)
+{
+ return (GET_MODE_SIZE (from) == GET_MODE_SIZE (to)
+ || (GET_MODE_SIZE (from) <= 4 && GET_MODE_SIZE (to) <= 4)
+ || (GET_MODE_SIZE (from) >= 16 && GET_MODE_SIZE (to) >= 16));
+}
+
+/* Implement TARGET_TRULY_NOOP_TRUNCATION. */
+
+static bool
+spu_truly_noop_truncation (poly_uint64 outprec, poly_uint64 inprec)
+{
+ return inprec <= 32 && outprec <= inprec;
+}
+
+/* Implement TARGET_STATIC_RTX_ALIGNMENT.
+
+ Make all static objects 16-byte aligned. This allows us to assume
+ they are also padded to 16 bytes, which means we can use a single
+ load or store instruction to access them. */
+
+static HOST_WIDE_INT
+spu_static_rtx_alignment (machine_mode mode)
+{
+ return MAX (GET_MODE_ALIGNMENT (mode), 128);
+}
+
+/* Implement TARGET_CONSTANT_ALIGNMENT.
+
+ Make all static objects 16-byte aligned. This allows us to assume
+ they are also padded to 16 bytes, which means we can use a single
+ load or store instruction to access them. */
+
+static HOST_WIDE_INT
+spu_constant_alignment (const_tree, HOST_WIDE_INT align)
+{
+ return MAX (align, 128);
+}
\f
/* Table of machine attributes. */
static const struct attribute_spec spu_attribute_table[] =
{
- /* { name, min_len, max_len, decl_req, type_req, fn_type_req, handler,
- affects_type_identity } */
- { "naked", 0, 0, true, false, false, spu_handle_fndecl_attribute,
- false },
- { "spu_vector", 0, 0, false, true, false, spu_handle_vector_attribute,
- false },
- { NULL, 0, 0, false, false, false, NULL, false }
+ /* { name, min_len, max_len, decl_req, type_req, fn_type_req,
+ affects_type_identity, handler, exclude } */
+ { "naked", 0, 0, true, false, false, false,
+ spu_handle_fndecl_attribute, NULL },
+ { "spu_vector", 0, 0, false, true, false, false,
+ spu_handle_vector_attribute, NULL },
+ { NULL, 0, 0, false, false, false, false, NULL, NULL }
};
/* TARGET overrides. */
+#undef TARGET_LRA_P
+#define TARGET_LRA_P hook_bool_void_false
+
#undef TARGET_ADDR_SPACE_POINTER_MODE
#define TARGET_ADDR_SPACE_POINTER_MODE spu_addr_space_pointer_mode
#undef TARGET_FUNCTION_ARG_ADVANCE
#define TARGET_FUNCTION_ARG_ADVANCE spu_function_arg_advance
+#undef TARGET_FUNCTION_ARG_OFFSET
+#define TARGET_FUNCTION_ARG_OFFSET spu_function_arg_offset
+
+#undef TARGET_FUNCTION_ARG_PADDING
+#define TARGET_FUNCTION_ARG_PADDING spu_function_arg_padding
+
#undef TARGET_MUST_PASS_IN_STACK
#define TARGET_MUST_PASS_IN_STACK must_pass_in_stack_var_size
#undef TARGET_CAN_USE_DOLOOP_P
#define TARGET_CAN_USE_DOLOOP_P can_use_doloop_if_innermost
+#undef TARGET_MODES_TIEABLE_P
+#define TARGET_MODES_TIEABLE_P spu_modes_tieable_p
+
+#undef TARGET_HARD_REGNO_NREGS
+#define TARGET_HARD_REGNO_NREGS spu_hard_regno_nregs
+
+#undef TARGET_CAN_CHANGE_MODE_CLASS
+#define TARGET_CAN_CHANGE_MODE_CLASS spu_can_change_mode_class
+
+#undef TARGET_TRULY_NOOP_TRUNCATION
+#define TARGET_TRULY_NOOP_TRUNCATION spu_truly_noop_truncation
+
+#undef TARGET_STATIC_RTX_ALIGNMENT
+#define TARGET_STATIC_RTX_ALIGNMENT spu_static_rtx_alignment
+#undef TARGET_CONSTANT_ALIGNMENT
+#define TARGET_CONSTANT_ALIGNMENT spu_constant_alignment
+
+#undef TARGET_HAVE_SPECULATION_SAFE_VALUE
+#define TARGET_HAVE_SPECULATION_SAFE_VALUE speculation_safe_value_not_needed
+
struct gcc_target targetm = TARGET_INITIALIZER;
#include "gt-spu.h"