Lines Matching defs:this
115 /* Variables which are this size or smaller are put in the sdata/sbss
410 /* ??? ABI doesn't allow us to define this. */
416 /* ??? ABI doesn't allow us to define this. */
715 /* Otherwise, this must be a constant, and that either 0 or 0.0 or 1.0. */
1106 /* We really should have taken care of this offset earlier. */
1277 but in this case it's more efficient to get rid
1379 This solution attempts to prevent this situation from occurring. When
1586 for this to be useful. */
1660 tricks to turn this into a signed comparison against 0. */
1940 /* ??? Should do this for functions known to bind local too. */
2029 That said, this means that we lack precise life info for
2030 whether ADDR is dead after this call. This is not terribly
2232 bits in GR0 to GR31 that have already been allocated by this routine.
2240 /* If this is a leaf function, first try an otherwise unused
2437 to be free, even if regs_ever_live is already set, because this is
2449 /* Emit a save of BR0 if we call other functions. Do this even
2450 if this function doesn't return, as EH depends on this to be
2566 size of this area up to keep things 16 byte aligned. */
2794 to be a post-modify. If we don't, then this copy may be
2795 eliminated by copyprop_hardreg_forward, which makes this
2797 propagate_one_insn. So mark this insn as legal to delete. */
2945 /* Set the local, input, and output register names. We need to do this
2967 there is a frame pointer. loc79 gets wasted in this case, as it is
3052 /* ??? At this point we must generate a magic insn that appears to
3084 /* Spill all varargs registers. Do this before spilling any GR registers,
3206 /* We don't know for sure yet if this is actually needed, since
3209 this save is dead. Allow it to go away. */
3281 /* ??? At this point we must generate a magic insn that appears to
3429 /* ??? At this point we must generate a magic insn that appears to
3540 /* Split this value after reload, now that we know where the return
3892 mode if this is contained within an aggregate. */
4015 If this is an SFmode aggregate, then it is possible to run out of
4019 /* Fill the FP regs. We do this always. We stop if we reach the end
4041 /* If this is an SFmode aggregate, then we might have some left over
4054 then this goes in a GR reg left adjusted/little endian, right
4056 /* ??? Currently this is handled wrong, because 4-byte hunks are
4061 multiple of 4 bytes in size, then this goes in a GR reg right
4173 /* Update CUM to point after this argument. This is patterned after
4208 If this is an SFmode aggregate, then it is possible to run out of
4212 /* Fill the FP regs. We do this always. We stop if we reach the end
4284 this is an indirect call. */
4468 /* ??? Do we need this? It gets used only for 'a' operands. We could perhaps
4469 also call this from ia64_print_operand for memory addresses. */
4652 /* If this operand is the constant zero, write it as register zero.
4847 we have to make this at least as expensive as MEMORY_MOVE_COST
4956 this.
5035 of this option is to mark the registers in the range from REG1 to
5216 may be written again by the complement of P (P^1) and when this happens,
5219 The result of this is that whenever an insn attempts to write a register
5243 /* Indicates whether this is the first instruction after a stop bit,
5244 in which case we don't need another stop bit. Without this,
5257 unsigned int is_sibcall : 1; /* Is this a sibling or normal call? */
5288 PRED. Update rws_insn and rws_sum arrays. Return 1 if this access creates
5320 /* The register has been written via a predicate. If this is
5361 by a branch instruction, handling this case is
5385 /* The register has been written via a predicate. If this is
5439 (1) The destination is (pc), in which case this is a branch,
5441 (2) The destination is ar.lc, in which case this is a
5443 (3) The destination is an fp register, in which case this is
5446 this is a check load.
5447 In all cases, nothing we do in this function applies. */
5458 /* Discover if this is a parallel comparison. We only handle
5470 /* Subroutine of rtx_needs_barrier; this function determines whether the
5473 for this insn. */
5489 /* ??? This seems redundant, as the caller sets this bit for
5519 PRED. Return 1 if this access creates a dependency with an earlier
5553 /* Avoid multiple register writes, in case this is a pattern with
5792 We force this by always returning true. */
5940 to get proper schedules. Undo this for dv analysis. */
6276 ??? When cycle display notes are implemented, update this. */
6370 /* We are about to being issuing insns for this clock cycle.
6458 /* We are about to being issuing insns for this clock cycle. Override
6484 ready queue that can be issued this cycle. */
6855 /* Offset to ld.sa patterns. For this patterns corresponding ld.c will
6936 /* We don't need recovery because even if this is ld.sa
6969 /* Set the desired check index. We add '1', because zero element in this
7137 As long as patterns are unique for each instruction, this can be
7503 /* If this is an insn that must be first in a group, then don't allow
7506 /* ??? The bundling automatons should handle this for us, but they do
7729 bundling in this function is implemented by using DFA
8354 /* ??? Is this bypass necessary for ld.c? */
8390 At present this assumes that we only consider predicate pairs to
8563 properly. Note that IA-64 differs from dwarf2 on this point. */
8627 the return address, so we force the value live this way. */
8667 /* Return true if this goes in small data/bss. */
8704 /* If this is an incomplete type with size 0, then we can't put it
8779 /* If this isn't the last block of the function, then we need to label the
8814 /* If this is the final destination for ar.pfs, then this must
9035 required to unwind this insn. */
9136 /* Under HPUX, this is a synonym for "long double". */
9297 /* ia64_expand_compare uses this. */
9355 /* For others, relax this so that relocations to local data goes in
9411 clear if this same convention should be used for other
9425 *(*this + vcall_offset) should be added to THIS. */
9432 rtx this, insn, funexp;
9452 /* Figure out whether "this" will be the first parameter (the
9462 this = gen_rtx_REG (Pmode, this_regno);
9469 emit_insn (gen_ptr_extend_plus_imm (this, tmp, GEN_INT (delta)));
9473 emit_insn (gen_ptr_extend (this, tmp));
9487 emit_insn (gen_adddi3 (this, this, delta_rtx));
9500 emit_move_insn (t, gen_rtx_MEM (ptr_mode, this));
9511 emit_move_insn (tmp, gen_rtx_MEM (Pmode, this));
9530 emit_insn (gen_adddi3 (this, this, tmp));