@@ -1941,15 +1941,21 @@ aarch64_expand_mov_immediate (rtx dest, rtx imm)
}
/* Add DELTA to REGNUM in mode MODE. SCRATCHREG can be used to held
- intermediate value if necessary.
+ intermediate value if necessary. FRAME_RELATED_P should be true if
+ the RTX_FRAME_RELATED flag should be set and CFA adjustments added
+ to the generated instructions. If SCRATCHREG is known to hold
+ abs (delta), EMIT_MOVE_IMM can be set to false to avoid emitting the
+ immediate again.
- This function is sometimes used to adjust the stack pointer, so we must
- ensure that it can never cause transient stack deallocation by writing an
- invalid value into REGNUM. */
+ Since this function may be used to adjust the stack pointer, we must
+ ensure that it cannot cause transient stack deallocation (for example
+ by first incrementing SP and then decrementing when adjusting by a
+ large immediate). */
static void
-aarch64_add_constant (machine_mode mode, int regnum, int scratchreg,
- HOST_WIDE_INT delta, bool frame_related_p)
+aarch64_add_constant_internal (machine_mode mode, int regnum, int scratchreg,
+ HOST_WIDE_INT delta, bool frame_related_p,
+ bool emit_move_imm)
{
HOST_WIDE_INT mdelta = abs_hwi (delta);
rtx this_rtx = gen_rtx_REG (mode, regnum);
@@ -1967,11 +1973,11 @@ aarch64_add_constant (machine_mode mode, int regnum, int scratchreg,
return;
}
- /* We need two add/sub instructions, each one performing part of the
- calculation. Don't do this if the addend can be loaded into register with
- a single instruction, in that case we prefer a move to a scratch register
- following by an addition. */
- if (mdelta < 0x1000000 && !aarch64_move_imm (delta, mode))
+ /* We need two add/sub instructions, each one perform part of the
+ addition/subtraction, but don't this if the addend can be loaded into
+ register by single instruction, in that case we prefer a move to scratch
+ register following by addition. */
+ if (mdelta < 0x1000000 && !aarch64_move_imm (mdelta, mode))
{
HOST_WIDE_INT low_off = mdelta & 0xfff;
@@ -1985,8 +1991,10 @@ aarch64_add_constant (machine_mode mode, int regnum, int scratchreg,
/* Otherwise use generic function to handle all other situations. */
rtx scratch_rtx = gen_rtx_REG (mode, scratchreg);
- aarch64_internal_mov_immediate (scratch_rtx, GEN_INT (delta), true, mode);
- insn = emit_insn (gen_add2_insn (this_rtx, scratch_rtx));
+ if (emit_move_imm)
+ aarch64_internal_mov_immediate (scratch_rtx, GEN_INT (mdelta), true, mode);
+ insn = emit_insn (delta < 0 ? gen_sub2_insn (this_rtx, scratch_rtx)
+ : gen_add2_insn (this_rtx, scratch_rtx));
if (frame_related_p)
{
RTX_FRAME_RELATED_P (insn) = frame_related_p;
@@ -1995,6 +2003,27 @@ aarch64_add_constant (machine_mode mode, int regnum, int scratchreg,
}
}
+static inline void
+aarch64_add_constant (machine_mode mode, int regnum, int scratchreg,
+ HOST_WIDE_INT delta)
+{
+ aarch64_add_constant_internal (mode, regnum, scratchreg, delta, false, true);
+}
+
+static inline void
+aarch64_add_sp (int scratchreg, HOST_WIDE_INT delta, bool emit_move_imm)
+{
+ aarch64_add_constant_internal (Pmode, SP_REGNUM, scratchreg, delta,
+ true, emit_move_imm);
+}
+
+static inline void
+aarch64_sub_sp (int scratchreg, HOST_WIDE_INT delta, bool frame_related_p)
+{
+ aarch64_add_constant_internal (Pmode, SP_REGNUM, scratchreg, -delta,
+ frame_related_p, true);
+}
+
static bool
aarch64_function_ok_for_sibcall (tree decl ATTRIBUTE_UNUSED,
tree exp ATTRIBUTE_UNUSED)
@@ -3201,7 +3230,7 @@ aarch64_expand_prologue (void)
aarch64_emit_probe_stack_range (STACK_CHECK_PROTECT, frame_size);
}
- aarch64_add_constant (Pmode, SP_REGNUM, IP0_REGNUM, -initial_adjust, true);
+ aarch64_sub_sp (IP0_REGNUM, initial_adjust, true);
if (callee_adjust != 0)
aarch64_push_regs (reg1, reg2, callee_adjust);
@@ -3222,8 +3251,7 @@ aarch64_expand_prologue (void)
callee_adjust != 0 || frame_pointer_needed);
aarch64_save_callee_saves (DFmode, callee_offset, V0_REGNUM, V31_REGNUM,
callee_adjust != 0 || frame_pointer_needed);
- aarch64_add_constant (Pmode, SP_REGNUM, IP1_REGNUM, -final_adjust,
- !frame_pointer_needed);
+ aarch64_sub_sp (IP1_REGNUM, final_adjust, !frame_pointer_needed);
}
/* Return TRUE if we can use a simple_return insn.
@@ -3288,7 +3316,7 @@ aarch64_expand_epilogue (bool for_sibcall)
RTX_FRAME_RELATED_P (insn) = callee_adjust == 0;
}
else
- aarch64_add_constant (Pmode, SP_REGNUM, IP1_REGNUM, final_adjust, true);
+ aarch64_add_sp (IP1_REGNUM, final_adjust, df_regs_ever_live_p (IP1_REGNUM));
aarch64_restore_callee_saves (DImode, callee_offset, R0_REGNUM, R30_REGNUM,
callee_adjust != 0, &cfi_ops);
@@ -3311,7 +3339,7 @@ aarch64_expand_epilogue (bool for_sibcall)
cfi_ops = NULL;
}
- aarch64_add_constant (Pmode, SP_REGNUM, IP0_REGNUM, initial_adjust, true);
+ aarch64_add_sp (IP0_REGNUM, initial_adjust, df_regs_ever_live_p (IP0_REGNUM));
if (cfi_ops)
{
@@ -3406,7 +3434,7 @@ aarch64_output_mi_thunk (FILE *file, tree thunk ATTRIBUTE_UNUSED,
emit_note (NOTE_INSN_PROLOGUE_END);
if (vcall_offset == 0)
- aarch64_add_constant (Pmode, this_regno, IP1_REGNUM, delta, false);
+ aarch64_add_constant (Pmode, this_regno, IP1_REGNUM, delta);
else
{
gcc_assert ((vcall_offset & (POINTER_BYTES - 1)) == 0);
@@ -3422,7 +3450,7 @@ aarch64_output_mi_thunk (FILE *file, tree thunk ATTRIBUTE_UNUSED,
addr = gen_rtx_PRE_MODIFY (Pmode, this_rtx,
plus_constant (Pmode, this_rtx, delta));
else
- aarch64_add_constant (Pmode, this_regno, IP1_REGNUM, delta, false);
+ aarch64_add_constant (Pmode, this_regno, IP1_REGNUM, delta);
}
if (Pmode == ptr_mode)
new file mode 100644
@@ -0,0 +1,21 @@
+/* { dg-do compile } */
+/* { dg-options "-O2 --save-temps" } */
+
+/* Test reuse of stack adjustment temporaries. */
+
+void foo ();
+
+int reuse_mov (int i)
+{
+ int arr[1025];
+ return arr[i];
+}
+
+int no_reuse_mov (int i)
+{
+ int arr[1025];
+ foo ();
+ return arr[i];
+}
+
+/* { dg-final { scan-assembler-times "mov\tx16, \[0-9\]+" 3 } } */