diff options
| author | Andrew Lee <alee14498@protonmail.com> | 2021-08-15 00:34:05 -0400 |
|---|---|---|
| committer | Andrew Lee <alee14498@protonmail.com> | 2021-08-15 00:34:05 -0400 |
| commit | 60cc83bf91bfc9bb02f6304b5d6c8234ba6d210f (patch) | |
| tree | fdc0be85a1ca35e34c3ae2c805fe9b718e3c1091 /gcc-1.40/config/out-sparc.c | |
| parent | dd8dfab51b832a654365ed00c06bf802ff628bfa (diff) | |
| download | linux-0.01-distro-master.tar.gz linux-0.01-distro-master.tar.bz2 linux-0.01-distro-master.zip | |
Diffstat (limited to 'gcc-1.40/config/out-sparc.c')
| -rw-r--r-- | gcc-1.40/config/out-sparc.c | 2092 |
1 files changed, 2092 insertions, 0 deletions
diff --git a/gcc-1.40/config/out-sparc.c b/gcc-1.40/config/out-sparc.c new file mode 100644 index 0000000..d180b1f --- /dev/null +++ b/gcc-1.40/config/out-sparc.c @@ -0,0 +1,2092 @@ +/* Subroutines for insn-output.c for Sun SPARC. + Copyright (C) 1987, 1988, 1989 Free Software Foundation, Inc. + Contributed by Michael Tiemann (tiemann@mcc.com) + +This file is part of GNU CC. + +GNU CC 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 1, or (at your option) +any later version. + +GNU CC 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 GNU CC; see the file COPYING. If not, write to +the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA. */ + +/* Global variables for machine-dependend things. */ + +/* This should go away if we pass floats to regs via + the stack instead of the frame, and if we learn how + to renumber all the registers when we don't do a save (hard!). */ +extern int frame_pointer_needed; + +static rtx find_addr_reg (); + +rtx next_real_insn_no_labels (); + +/* Return non-zero only if OP is a register of mode MODE, + or const0_rtx. */ +int +reg_or_0_operand (op, mode) + rtx op; + enum machine_mode mode; +{ + return (op == const0_rtx || register_operand (op, mode)); +} + +/* Return non-zero if INSN is a conditional insn with a predicate + valid after an addcc or subcc instruction. */ + +int +ignore_overflow_conditional_p (insn) + rtx insn; +{ + rtx x = SET_SRC (PATTERN (insn)); + RTX_CODE code; + if (GET_CODE (x) == IF_THEN_ELSE) + x = XEXP (x, 0); + code = GET_CODE (x); + return code == EQ || code == NE || code == GE || code == LT; +} + +/* Return non-zero if this pattern, can be evaluated safely, even if it + was not asked for. */ +int +safe_insn_src_p (op, mode) + rtx op; + enum machine_mode mode; +{ + /* Just experimenting. */ + + /* No floating point src is safe if it contains an arithmetic + operation, since that operation may trap. */ + switch (GET_CODE (op)) + { + case CONST_INT: + case LABEL_REF: + case SYMBOL_REF: + case CONST: + return 1; + + case REG: + return 1; + + case MEM: + return CONSTANT_ADDRESS_P (XEXP (op, 0)); + + /* We never need to negate or complement constants. */ + case NEG: + return (mode != SFmode && mode != DFmode); + case NOT: + return 1; + + case COMPARE: + case MINUS: + case PLUS: + return (mode != SFmode && mode != DFmode); + case AND: + case IOR: + case XOR: + case LSHIFT: + case ASHIFT: + case ASHIFTRT: + case LSHIFTRT: + if ((GET_CODE (XEXP (op, 0)) == CONST_INT && ! SMALL_INT (XEXP (op, 0))) + || (GET_CODE (XEXP (op, 1)) == CONST_INT && ! SMALL_INT (XEXP (op, 1)))) + return 0; + return 1; + + default: + return 0; + } +} + +/* Return 1 if REG is clobbered in IN. + Return 0 if REG is used in IN (other than being clobbered). + Return 2 if REG does not appear in IN. */ + +static int +reg_clobbered_p (reg, in) + rtx reg; + rtx in; +{ + register char *fmt; + register int i, result = 0; + + register enum rtx_code code; + + if (in == 0) + return 2; + + code = GET_CODE (in); + + switch (code) + { + /* Let these fail out quickly. */ + case CONST_INT: + case SYMBOL_REF: + case CONST: + return 2; + + case SUBREG: + if (SUBREG_WORD (in) != 0) + in = gen_rtx (REG, SImode, REGNO (SUBREG_REG (in)) + SUBREG_WORD (in)); + else + in = SUBREG_REG (in); + + case REG: + if (in == reg + || refers_to_regno_p (REGNO (reg), + REGNO (reg) + HARD_REGNO_NREGS (reg, GET_MODE (reg)), + in, 0)) + return 0; + return 2; + + case SET: + if (SET_SRC (in) == reg + || refers_to_regno_p (REGNO (reg), + REGNO (reg) + HARD_REGNO_NREGS (reg, GET_MODE (reg)), + SET_SRC (in), 0)) + return 0; + + if (SET_DEST (in) == reg) + return 1; + + if (refers_to_regno_p (REGNO (reg), + REGNO (reg) + HARD_REGNO_NREGS (reg, GET_MODE (reg)), + SET_DEST (in), 0)) + if (GET_CODE (SET_DEST (in)) == REG + || GET_CODE (SET_DEST (in)) == SUBREG) + return 1; + else + return 0; + return 2; + + case USE: + if (XEXP (in, 0) == reg + || refers_to_regno_p (REGNO (reg), + REGNO (reg) + HARD_REGNO_NREGS (reg, GET_MODE (reg)), + XEXP (in, 0), 0)) + return 0; + return 2; + + case CLOBBER: + if (XEXP (in, 0) == reg) + return 1; + /* If the CLOBBER expression is a SUBREG, accept that as a + clobber. But if it is some expression based on this register, + that is like a USE as far as this register is concerned, + so we won't take it. */ + if (refers_to_regno_p (REGNO (reg), + REGNO (reg) + HARD_REGNO_NREGS (reg, GET_MODE (reg)), + XEXP (in, 0), 0)) + if (GET_CODE (XEXP (in, 0)) == REG + || GET_CODE (XEXP (in, 0)) == SUBREG) + return 1; + else + return 0; + return 2; + } + + fmt = GET_RTX_FORMAT (code); + + result = 2; + + for (i = GET_RTX_LENGTH (code) - 1; i >= 0; i--) + { + if (fmt[i] == 'E') + { + register int j; + for (j = XVECLEN (in, i) - 1; j >= 0; j--) + switch (reg_clobbered_p (reg, XVECEXP (in, i, j))) + { + case 0: + return 0; + case 2: + continue; + case 1: + result = 1; + break; + } + } + else if (fmt[i] == 'e') + switch (reg_clobbered_p (reg, XEXP (in, i))) + { + case 0: + return 0; + case 2: + continue; + case 1: + result = 1; + break; + } + } + return result; +} + +/* Return non-zero if OP can be written to without screwing up + GCC's model of what's going on. It is assumed that this operand + appears in the dest position of a SET insn in a conditional + branch's delay slot. AFTER is the label to start looking from. */ +int +operand_clobbered_before_used_after (op, after) + rtx op; + rtx after; +{ + extern char call_used_regs[]; + + /* Just experimenting. */ + if (GET_CODE (op) == CC0) + return 1; + if (GET_CODE (op) == REG) + { + rtx insn; + + if (op == stack_pointer_rtx) + return 0; + + for (insn = NEXT_INSN (after); insn; insn = NEXT_INSN (insn)) + { + if (GET_CODE (insn) == NOTE) + continue; + if (GET_CODE (insn) == INSN + || GET_CODE (insn) == JUMP_INSN + || GET_CODE (insn) == CALL_INSN) + { + switch (reg_clobbered_p (op, PATTERN (insn))) + { + case 0: + return 0; + case 2: + break; + case 1: + return 1; + } + if (dead_or_set_p (insn, op)) + return 1; + } + else if (GET_CODE (insn) == CODE_LABEL) + return 0; + if (GET_CODE (insn) == JUMP_INSN) + { + if (condjump_p (insn)) + return 0; + /* This is a jump insn which has already + been mangled. We can't tell what it does. */ + if (GET_CODE (PATTERN (insn)) == PARALLEL) + return 0; + if (! JUMP_LABEL (insn)) + return 0; + /* Keep following jumps. */ + insn = JUMP_LABEL (insn); + } + } + return 1; + } + + /* In both of these cases, the first insn executed + for this op will be a sethi %hi(whatever),%g1, + which is tolerable. */ + if (GET_CODE (op) == MEM) + return (CONSTANT_ADDRESS_P (XEXP (op, 0))); + + return 0; +} + +/* Return non-zero if this pattern, as a source to a "SET", + is known to yield an instruction of unit size. */ +int +single_insn_src_p (op, mode) + rtx op; + enum machine_mode mode; +{ + switch (GET_CODE (op)) + { + case CONST_INT: +#if 1 + /* This is not always a single insn src, technically, + but output_delayed_branch knows how to deal with it. */ + return 1; +#else + if (SMALL_INT (op)) + return 1; + /* We can put this set insn into delay slot, because this is one + insn; 'sethi'. */ + if ((INTVAL (op) & 0x3ff) == 0) + return 1; + + /* This is not a single insn src, technically, + but output_delayed_branch knows how to deal with it. */ + return 1; +#endif + +#if 1 + case SYMBOL_REF: + /* This is not a single insn src, technically, + but output_delayed_branch knows how to deal with it. */ + return 1; +#else + return 0; +#endif + + case REG: + return 1; + + case MEM: +#if 0 + /* This is not a single insn src, technically, + but output_delayed_branch knows how to deal with it. */ + if (GET_CODE (XEXP (op, 0)) == SYMBOL_REF) + return 0; +#endif + return 1; + + /* We never need to negate or complement constants. */ + case NEG: + return (mode != DFmode); + case NOT: + return 1; + + case COMPARE: + case MINUS: + /* If the target is cc0, then these insns will take + two insns (one being a nop). */ + return (mode != SFmode && mode != DFmode); + case PLUS: + case AND: + case IOR: + case XOR: + case LSHIFT: + case ASHIFT: + case ASHIFTRT: + case LSHIFTRT: + if ((GET_CODE (XEXP (op, 0)) == CONST_INT && ! SMALL_INT (XEXP (op, 0))) + || (GET_CODE (XEXP (op, 1)) == CONST_INT && ! SMALL_INT (XEXP (op, 1)))) + return 0; + return 1; + + case SUBREG: + if (SUBREG_WORD (op) != 0) + return 0; + return single_insn_src_p (SUBREG_REG (op), mode); + + case SIGN_EXTEND: + case ZERO_EXTEND: + /* Lazy... could check for more cases. */ + if (GET_CODE (XEXP (op, 0)) == MEM + && ! CONSTANT_ADDRESS_P (XEXP (XEXP (op, 0), 0))) + return 1; + return 0; + + /* Not doing floating point, since they probably + take longer than the branch slot they might fill. */ + case FLOAT_EXTEND: + case FLOAT_TRUNCATE: + case FLOAT: + case FIX: + case UNSIGNED_FLOAT: + case UNSIGNED_FIX: + return 0; + + default: + return 0; + } +} + +/* This extra test must be done to verify that a move insn + really is just one assembler insn. */ + +int +single_insn_extra_test (dest, src) + rtx dest, src; +{ + /* Moves between FP regs and CPU regs are two insns. */ + return (!(GET_CODE (src) == REG + && GET_CODE (dest) == REG + && (FP_REG_P (src) != FP_REG_P (dest)))); +} + +/* Nonzero only if this *really* is a single insn operand. */ +int +strict_single_insn_op_p (op, mode) + rtx op; + enum machine_mode mode; +{ + if (mode == VOIDmode) + mode = GET_MODE (op); + + switch (GET_CODE (op)) + { + case CC0: + return 1; + + case CONST_INT: + if (SMALL_INT (op)) + return 1; + /* We can put this set insn into delay slot, because this is one + insn; 'sethi'. */ + if ((INTVAL (op) & 0x3ff) == 0) + return 1; + return 0; + + case SYMBOL_REF: + return 0; + + case REG: + return (mode != DFmode && mode != DImode); + + case MEM: + if (! CONSTANT_ADDRESS_P (XEXP (op, 0))) + return (mode != DFmode && mode != DImode); + return 0; + + /* We never need to negate or complement constants. */ + case NEG: + return (mode != DFmode); + case NOT: + return 1; + + case COMPARE: + case MINUS: + /* If the target is cc0, then these insns will take + two insns (one being a nop). */ + return (mode != SFmode && mode != DFmode); + case PLUS: + case AND: + case IOR: + case XOR: + case LSHIFT: + case ASHIFT: + case ASHIFTRT: + case LSHIFTRT: + if ((GET_CODE (XEXP (op, 0)) == CONST_INT && ! SMALL_INT (XEXP (op, 0))) + || (GET_CODE (XEXP (op, 1)) == CONST_INT && ! SMALL_INT (XEXP (op, 1)))) + return 0; + return 1; + + case SUBREG: + if (SUBREG_WORD (op) != 0) + return 0; + return strict_single_insn_op_p (SUBREG_REG (op), mode); + + case SIGN_EXTEND: + case ZERO_EXTEND: + if (GET_CODE (XEXP (op, 0)) == MEM + && ! CONSTANT_ADDRESS_P (XEXP (XEXP (op, 0), 0))) + return 1; + return 0; + + /* Not doing floating point, since they probably + take longer than the branch slot they might fill. */ + case FLOAT_EXTEND: + case FLOAT_TRUNCATE: + case FLOAT: + case FIX: + case UNSIGNED_FLOAT: + case UNSIGNED_FIX: + return 0; + + default: + return 0; + } +} + +/* Return truth value of whether OP is a relational operator. */ +int +relop (op, mode) + rtx op; + enum machine_mode mode; +{ + switch (GET_CODE (op)) + { + case EQ: + case NE: + case GT: + case GE: + case LT: + case LE: + case GTU: + case GEU: + case LTU: + case LEU: + return 1; + } + return 0; +} + +/* Return truth value of wheterh OP is EQ or NE. */ +int +eq_or_neq (op, mode) + rtx op; + enum machine_mode mode; +{ + return (GET_CODE (op) == EQ || GET_CODE (op) == NE); +} + +/* Return truth value of whether OP can be used as an operands in a three + address arithmetic insn (such as add %o1,7,%l2) of mode MODE. */ + +int +arith_operand (op, mode) + rtx op; + enum machine_mode mode; +{ + return (register_operand (op, mode) + || (GET_CODE (op) == CONST_INT && SMALL_INT (op))); +} + +/* Return truth value of whether OP can be used as an operand in a two + address arithmetic insn (such as set 123456,%o4) of mode MODE. */ + +int +arith32_operand (op, mode) + rtx op; + enum machine_mode mode; +{ + return (register_operand (op, mode) || GET_CODE (op) == CONST_INT); +} + +/* Return truth value of whether OP is a integer which fits the + range constraining immediate operands in three-address insns. */ + +int +small_int (op, mode) + rtx op; + enum machine_mode mode; +{ + return (GET_CODE (op) == CONST_INT && SMALL_INT (op)); +} + +/* Return the best assembler insn template + for moving operands[1] into operands[0] as a fullword. */ + +static char * +singlemove_string (operands) + rtx *operands; +{ + if (GET_CODE (operands[0]) == MEM) + { + if (GET_CODE (operands[1]) != MEM) + if (CONSTANT_ADDRESS_P (XEXP (operands[0], 0))) + { + if (! ((cc_prev_status.flags & CC_KNOW_HI_G1) + && cc_prev_status.mdep == XEXP (operands[0], 0))) + output_asm_insn ("sethi %%hi(%m0),%%g1", operands); + cc_status.flags |= CC_KNOW_HI_G1; + cc_status.mdep = XEXP (operands[0], 0); + return "st %1,[%%lo(%m0)+%%g1]"; + } + else + return "st %r1,%0"; + else + { + rtx xoperands[2]; + + cc_status.flags &= ~CC_F0_IS_0; + xoperands[0] = gen_rtx (REG, SFmode, 32); + xoperands[1] = operands[1]; + output_asm_insn (singlemove_string (xoperands), xoperands); + xoperands[1] = xoperands[0]; + xoperands[0] = operands[0]; + output_asm_insn (singlemove_string (xoperands), xoperands); + return ""; + } + } + if (GET_CODE (operands[1]) == MEM) + { + if (CONSTANT_ADDRESS_P (XEXP (operands[1], 0))) + { + if (! ((cc_prev_status.flags & CC_KNOW_HI_G1) + && cc_prev_status.mdep == XEXP (operands[1], 0))) + output_asm_insn ("sethi %%hi(%m1),%%g1", operands); + cc_status.flags |= CC_KNOW_HI_G1; + cc_status.mdep = XEXP (operands[1], 0); + return "ld [%%lo(%m1)+%%g1],%0"; + } + return "ld %1,%0"; + } + return "mov %1,%0"; +} + +/* Output assembler code to perform a doubleword move insn + with operands OPERANDS. */ + +char * +output_move_double (operands) + rtx *operands; +{ + enum { REGOP, OFFSOP, MEMOP, PUSHOP, POPOP, CNSTOP, RNDOP } optype0, optype1; + rtx latehalf[2]; + rtx addreg0 = 0, addreg1 = 0; + + /* First classify both operands. */ + + if (REG_P (operands[0])) + optype0 = REGOP; + else if (offsettable_memref_p (operands[0])) + optype0 = OFFSOP; + else if (GET_CODE (operands[0]) == MEM) + optype0 = MEMOP; + else + optype0 = RNDOP; + + if (REG_P (operands[1])) + optype1 = REGOP; + else if (CONSTANT_P (operands[1]) + || GET_CODE (operands[1]) == CONST_DOUBLE) + optype1 = CNSTOP; + else if (offsettable_memref_p (operands[1])) + optype1 = OFFSOP; + else if (GET_CODE (operands[1]) == MEM) + optype1 = MEMOP; + else + optype1 = RNDOP; + + /* Check for the cases that the operand constraints are not + supposed to allow to happen. Abort if we get one, + because generating code for these cases is painful. */ + + if (optype0 == RNDOP || optype1 == RNDOP) + abort (); + + /* If an operand is an unoffsettable memory ref, find a register + we can increment temporarily to make it refer to the second word. */ + + if (optype0 == MEMOP) + addreg0 = find_addr_reg (XEXP (operands[0], 0)); + + if (optype1 == MEMOP) + addreg1 = find_addr_reg (XEXP (operands[1], 0)); + + /* Ok, we can do one word at a time. + Normally we do the low-numbered word first, + but if either operand is autodecrementing then we + do the high-numbered word first. + + In either case, set up in LATEHALF the operands to use + for the high-numbered word and in some cases alter the + operands in OPERANDS to be suitable for the low-numbered word. */ + + if (optype0 == REGOP) + latehalf[0] = gen_rtx (REG, SImode, REGNO (operands[0]) + 1); + else if (optype0 == OFFSOP) + latehalf[0] = adj_offsettable_operand (operands[0], 4); + else + latehalf[0] = operands[0]; + + if (optype1 == REGOP) + latehalf[1] = gen_rtx (REG, SImode, REGNO (operands[1]) + 1); + else if (optype1 == OFFSOP) + latehalf[1] = adj_offsettable_operand (operands[1], 4); + else if (optype1 == CNSTOP) + { + if (CONSTANT_P (operands[1])) + latehalf[1] = const0_rtx; + else if (GET_CODE (operands[1]) == CONST_DOUBLE) + { + latehalf[1] = gen_rtx (CONST_INT, VOIDmode, + CONST_DOUBLE_HIGH (operands[1])); + operands[1] = gen_rtx (CONST_INT, VOIDmode, + CONST_DOUBLE_LOW (operands[1])); + } + } + else + latehalf[1] = operands[1]; + + /* If the first move would clobber the source of the second one, + do them in the other order. + + RMS says "This happens only for registers; + such overlap can't happen in memory unless the user explicitly + sets it up, and that is an undefined circumstance." + + but it happens on the sparc when loading parameter registers, + so I am going to define that circumstance, and make it work + as expected. */ + + /* Easy case: try moving both words at once. */ + /* First check for moving between an even/odd register pair + and a memory location. */ + if ((optype0 == REGOP && optype1 != REGOP && optype1 != CNSTOP + && (REGNO (operands[0]) & 1) == 0) + || (optype0 != REGOP && optype1 != CNSTOP && optype1 == REGOP + && (REGNO (operands[1]) & 1) == 0)) + { + rtx op1, op2; + rtx base = 0, offset = const0_rtx; + + /* OP1 gets the register pair, and OP2 gets the memory address. */ + if (optype0 == REGOP) + op1 = operands[0], op2 = XEXP (operands[1], 0); + else + op1 = operands[1], op2 = XEXP (operands[0], 0); + + /* Now see if we can trust the address to be 8-byte aligned. */ + /* Trust global variables. */ + if (CONSTANT_ADDRESS_P (op2)) + { + operands[0] = op1; + operands[1] = op2; + if (! ((cc_prev_status.flags & CC_KNOW_HI_G1) + && cc_prev_status.mdep == op2)) + output_asm_insn ("sethi %%hi(%1),%%g1", operands); + cc_status.flags |= CC_KNOW_HI_G1; + cc_status.mdep = op2; + if (op1 == operands[0]) + return "ldd [%%lo(%1)+%%g1],%0"; + else + return "std [%%lo(%1)+%%g1],%0"; + } + + if (GET_CODE (op2) == PLUS) + { + if (GET_CODE (XEXP (op2, 0)) == REG) + base = XEXP (op2, 0), offset = XEXP (op2, 1); + else if (GET_CODE (XEXP (op2, 1)) == REG) + base = XEXP (op2, 1), offset = XEXP (op2, 0); + } + + /* Trust round enough offsets from the stack or frame pointer. */ + if (base + && (REGNO (base) == FRAME_POINTER_REGNUM + || REGNO (base) == STACK_POINTER_REGNUM)) + { + if (GET_CODE (offset) == CONST_INT + && (INTVAL (offset) & 0x7) == 0) + { + if (op1 == operands[0]) + return "ldd %1,%0"; + else + return "std %1,%0"; + } + } + else + { + /* We know structs not on the stack are properly aligned. + Since a double asks for 8-byte alignment, + we know it must have got that if it is in a struct. + But a DImode need not be 8-byte aligned, because it could be a + struct containing two ints or pointers. */ + + /* Sun fucks us here. We cannot trust references + to doubles via varying addresses. It might be on the stack + even if we don't know that it is; and then it might not be + double-word aligned. */ +#if 0 + if (GET_CODE (operands[1]) == MEM && GET_MODE (operands[1]) == DFmode + && MEM_IN_STRUCT_P (operands[1])) + return "ldd %1,%0"; + else if (GET_CODE (operands[0]) == MEM + && GET_MODE (operands[0]) == DFmode + && MEM_IN_STRUCT_P (operands[0])) + return "std %1,%0"; +#endif + } + } + + if (optype0 == REGOP && optype1 == REGOP + && REGNO (operands[0]) == REGNO (latehalf[1])) + { + /* Make any unoffsettable addresses point at high-numbered word. */ + if (addreg0) + output_asm_insn ("add %0,0x4,%0", &addreg0); + if (addreg1) + output_asm_insn ("add %0,0x4,%0", &addreg1); + + /* Do that word. */ + output_asm_insn (singlemove_string (latehalf), latehalf); + + /* Undo the adds we just did. */ + if (addreg0) + output_asm_insn ("add %0,-0x4,%0", &addreg0); + if (addreg1) + output_asm_insn ("add %0,-0x4,%0", &addreg0); + + /* Do low-numbered word. */ + return singlemove_string (operands); + } + else if (optype0 == REGOP && optype1 != REGOP + && reg_overlap_mentioned_p (operands[0], operands[1])) + { + /* Do the late half first. */ + output_asm_insn (singlemove_string (latehalf), latehalf); + /* Then clobber. */ + return singlemove_string (operands); + } + + /* Normal case: do the two words, low-numbered first. */ + + output_asm_insn (singlemove_string (operands), operands); + + /* Make any unoffsettable addresses point at high-numbered word. */ + if (addreg0) + output_asm_insn ("add %0,0x4,%0", &addreg0); + if (addreg1) + output_asm_insn ("add %0,0x4,%0", &addreg1); + + /* Do that word. */ + output_asm_insn (singlemove_string (latehalf), latehalf); + + /* Undo the adds we just did. */ + if (addreg0) + output_asm_insn ("add %0,-0x4,%0", &addreg0); + if (addreg1) + output_asm_insn ("add %0,-0x4,%0", &addreg1); + + return ""; +} + +static char * +output_fp_move_double (operands) + rtx *operands; +{ + if (FP_REG_P (operands[0])) + { + if (FP_REG_P (operands[1])) + { + output_asm_insn ("fmovs %1,%0", operands); + operands[0] = gen_rtx (REG, VOIDmode, REGNO (operands[0]) + 1); + operands[1] = gen_rtx (REG, VOIDmode, REGNO (operands[1]) + 1); + return "fmovs %1,%0"; + } + if (GET_CODE (operands[1]) == REG) + { + if ((REGNO (operands[1]) & 1) == 0) + return "std %1,[%%fp-8]\n\tldd [%%fp-8],%0"; + else + { + rtx xoperands[3]; + xoperands[0] = operands[0]; + xoperands[1] = operands[1]; + xoperands[2] = gen_rtx (REG, SImode, REGNO (operands[1]) + 1); + output_asm_insn ("st %2,[%%fp-4]\n\tst %1,[%%fp-8]\n\tldd [%%fp-8],%0", xoperands); + return ""; + } + } + /* Use ldd if known to be aligned. */ + if (GET_CODE (XEXP (operands[1], 0)) == PLUS + && (((XEXP (XEXP (operands[1], 0), 0) == frame_pointer_rtx + || XEXP (XEXP (operands[1], 0), 0) == stack_pointer_rtx) + && GET_CODE (XEXP (XEXP (operands[1], 0), 1)) == CONST_INT + && (INTVAL (XEXP (XEXP (operands[1], 0), 1)) & 0x7) == 0) +#if 0 /* An array in a structure that is a parm need not be aligned! */ + /* Arrays are known to be aligned, + and reg+reg addresses are used (on this machine) + only for array accesses. */ + || (REG_P (XEXP (XEXP (operands[1], 0), 0)) + && REG_P (XEXP (XEXP (operands[1], 0), 1))) +#endif + )) + return "ldd %1,%0"; + if (CONSTANT_ADDRESS_P (XEXP (operands[1], 0))) + { + if (! ((cc_prev_status.flags & CC_KNOW_HI_G1) + && cc_prev_status.mdep == XEXP (operands[1], 0))) + output_asm_insn ("sethi %%hi(%m1),%%g1", operands); + cc_status.flags |= CC_KNOW_HI_G1; + cc_status.mdep = XEXP (operands[1], 0); + return "ldd [%%lo(%m1)+%%g1],%0"; + } + /* Otherwise use two ld insns. */ + { + rtx xoperands[2]; + output_asm_insn ("ld %1,%0", operands); + xoperands[0] = gen_rtx (REG, GET_MODE (operands[0]), + REGNO (operands[0]) + 1); + if (GET_CODE (XEXP (operands[1], 0)) == PLUS + && offsettable_address_p (1, GET_MODE (operands[1]), + XEXP (operands[1], 0))) + { + xoperands[1] = adj_offsettable_operand (operands[1], 4); + output_asm_insn ("ld %1,%0", xoperands); + } + else if (GET_CODE (XEXP (operands[1], 0)) == PLUS) + { + rtx memref = operands[1]; + rtx inc_reg = XEXP (XEXP (operands[1], 0), 0); + if (inc_reg == frame_pointer_rtx + && GET_CODE (XEXP (XEXP (operands[1], 0), 1)) == REG + && XEXP (XEXP (operands[1], 0), 1) != frame_pointer_rtx) + inc_reg = XEXP (XEXP (operands[1], 0), 1); + if (inc_reg == frame_pointer_rtx) + { + output_asm_insn ("mov %%fp,%%g1", xoperands); + inc_reg = gen_rtx (REG, SImode, 1); + memref = gen_rtx (GET_CODE (operands[1]), + GET_MODE (operands[1]), + gen_rtx (PLUS, GET_MODE (XEXP (operands[1], 0)), + inc_reg, + XEXP (XEXP (operands[1], 0), 1))); + } + xoperands[1] = inc_reg; + output_asm_insn ("add 4,%1,%1", xoperands); + xoperands[1] = memref; + output_asm_insn ("ld %1,%0", xoperands); + xoperands[1] = inc_reg; + output_asm_insn ("add -4,%1,%1", xoperands); + } + else + { + xoperands[1] = gen_rtx (MEM, GET_MODE (operands[1]), + plus_constant (XEXP (operands[1], 0), 4)); + output_asm_insn ("ld %1,%0", xoperands); + } + return ""; + } + } + else if (FP_REG_P (operands[1])) + { + if (GET_CODE (operands[0]) == REG) + { + if ((REGNO (operands[0]) & 1) == 0) + return "std %1,[%%fp-8]\n\tldd [%%fp-8],%0"; + else + { + rtx xoperands[3]; + xoperands[2] = operands[1]; + xoperands[1] = gen_rtx (REG, SImode, REGNO (operands[0]) + 1); + xoperands[0] = operands[0]; + output_asm_insn ("std %2,[%%fp-8]\n\tld [%%fp-4],%1\n\tld [%%fp-8],%0", xoperands); + return ""; + } + } + /* Use std if we can be sure it is well-aligned. */ + if (GET_CODE (XEXP (operands[0], 0)) == PLUS + && (((XEXP (XEXP (operands[0], 0), 0) == frame_pointer_rtx + || XEXP (XEXP (operands[0], 0), 0) == stack_pointer_rtx) + && GET_CODE (XEXP (XEXP (operands[0], 0), 1)) == CONST_INT + && (INTVAL (XEXP (XEXP (operands[0], 0), 1)) & 0x7) == 0) +#if 0 /* An array in a structure that is a parm need not be aligned! */ + /* Arrays are known to be aligned, + and reg+reg addresses are used (on this machine) + only for array accesses. */ + || (REG_P (XEXP (XEXP (operands[0], 0), 0)) + && REG_P (XEXP (XEXP (operands[0], 0), 1))) +#endif + )) + return "std %1,%0"; + if (CONSTANT_ADDRESS_P (XEXP (operands[0], 0))) + { + if (! ((cc_prev_status.flags & CC_KNOW_HI_G1) + && cc_prev_status.mdep == XEXP (operands[0], 0))) + output_asm_insn ("sethi %%hi(%m0),%%g1", operands); + cc_status.flags |= CC_KNOW_HI_G1; + cc_status.mdep = XEXP (operands[0], 0); + return "std %1,[%%lo(%m0)+%%g1]"; + } + /* Otherwise use two st insns. */ + { + rtx xoperands[2]; + output_asm_insn ("st %r1,%0", operands); + xoperands[1] = gen_rtx (REG, GET_MODE (operands[1]), + REGNO (operands[1]) + 1); + if (GET_CODE (XEXP (operands[0], 0)) == PLUS + && offsettable_address_p (1, GET_MODE (operands[0]), + XEXP (operands[0], 0))) + { + xoperands[0] = adj_offsettable_operand (operands[0], 4); + output_asm_insn ("st %r1,%0", xoperands); + } + else if (GET_CODE (XEXP (operands[0], 0)) == PLUS) + { + rtx memref = operands[0]; + rtx inc_reg = XEXP (XEXP (operands[0], 0), 0); + if (inc_reg == frame_pointer_rtx + && GET_CODE (XEXP (XEXP (operands[0], 0), 1)) == REG + && XEXP (XEXP (operands[0], 0), 1) != frame_pointer_rtx) + inc_reg = XEXP (XEXP (operands[0], 0), 1); + if (inc_reg == frame_pointer_rtx) + { + output_asm_insn ("mov %%fp,%%g1", xoperands); + inc_reg = gen_rtx (REG, SImode, 1); + memref = gen_rtx (GET_CODE (operands[0]), + GET_MODE (operands[0]), + gen_rtx (PLUS, GET_MODE (XEXP (operands[0], 0)), + inc_reg, + XEXP (XEXP (operands[0], 0), 1))); + } + xoperands[0] = inc_reg; + output_asm_insn ("add 4,%0,%0", xoperands); + xoperands[0] = memref; + output_asm_insn ("st %r1,%0", xoperands); + xoperands[0] = inc_reg; + output_asm_insn ("add -4,%0,%0", xoperands); + } + else + { + xoperands[0] = gen_rtx (MEM, GET_MODE (operands[0]), + plus_constant (XEXP (operands[0], 0), 4)); + output_asm_insn ("st %r1,%0", xoperands); + } + return ""; + } + } + else abort (); +} + +/* Return a REG that occurs in ADDR with coefficient 1. + ADDR can be effectively incremented by incrementing REG. */ + +static rtx +find_addr_reg (addr) + rtx addr; +{ + while (GET_CODE (addr) == PLUS) + { + if (GET_CODE (XEXP (addr, 0)) == REG + && !(GET_CODE (XEXP (addr, 1)) == REG + && XEXP (addr, 0) == frame_pointer_rtx)) + addr = XEXP (addr, 0); + else if (GET_CODE (XEXP (addr, 1)) == REG) + addr = XEXP (addr, 1); + else if (CONSTANT_P (XEXP (addr, 0))) + addr = XEXP (addr, 1); + else if (CONSTANT_P (XEXP (addr, 1))) + addr = XEXP (addr, 0); + else + abort (); + } + if (GET_CODE (addr) == REG) + return addr; + abort (); +} + +void +output_sized_memop (opname, mode) + char *opname; + enum machine_mode mode; +{ + extern struct _iobuf *asm_out_file; + + static char *ld_size_suffix[] = { "ub", "uh", "", "?", "d" }; + static char *st_size_suffix[] = { "b", "h", "", "?", "d" }; + char *modename + = (opname[0] == 'l' ? ld_size_suffix : st_size_suffix)[GET_MODE_SIZE (mode) >> 1]; + + fprintf (asm_out_file, "\t%s%s", opname, modename); +} + +/* Output a store-in-memory whose operands are OPERANDS[0,1]. + OPERANDS[0] is a MEM, and OPERANDS[1] is a reg or zero. */ + +char * +output_store (operands) + rtx *operands; +{ + enum machine_mode mode = GET_MODE (operands[0]); + rtx address = XEXP (operands[0], 0); + + cc_status.flags |= CC_KNOW_HI_G1; + cc_status.mdep = address; + + if (! ((cc_prev_status.flags & CC_KNOW_HI_G1) + && address == cc_prev_status.mdep)) + { + output_asm_insn ("sethi %%hi(%m0),%%g1", operands); + cc_prev_status.mdep = address; + } + + /* Store zero in two parts when appropriate. */ + if (mode == DFmode && operands[1] == dconst0_rtx) + { + /* We can't cross a page boundary here because the + SYMBOL_REF must be double word aligned, and for this + to be the case, SYMBOL_REF+4 cannot cross. */ + output_sized_memop ("st", SImode); + output_asm_insn ("%r1,[%%g1+%%lo(%m0)]", operands); + output_sized_memop ("st", SImode); + return "%r1,[%%g1+%%lo(%m0)+4]"; + } + + /* Code below isn't smart enough to move a doubleword in two parts, + so use output_move_double to do that in the cases that require it. */ + if ((mode == DImode || mode == DFmode) + && (GET_CODE (operands[1]) == REG + && (REGNO (operands[1]) & 1))) + return output_move_double (operands); + + output_sized_memop ("st", mode); + return "%r1,[%%g1+%%lo(%m0)]"; +} + +/* Output a fixed-point load-from-memory whose operands are OPERANDS[0,1]. + OPERANDS[0] is a reg, and OPERANDS[1] is a mem. */ + +char * +output_load_fixed (operands) + rtx *operands; +{ + enum machine_mode mode = GET_MODE (operands[0]); + rtx address = XEXP (operands[1], 0); + + /* We don't bother trying to see if we know %hi(address). + This is because we are doing a load, and if we know the + %hi value, we probably also know that value in memory. */ + cc_status.flags |= CC_KNOW_HI_G1; + cc_status.mdep = address; + + if (! ((cc_prev_status.flags & CC_KNOW_HI_G1) + && address == cc_prev_status.mdep + && cc_prev_status.mdep == cc_status.mdep)) + { + output_asm_insn ("sethi %%hi(%m1),%%g1", operands); + cc_prev_status.mdep = address; + } + + /* Code below isn't smart enough to do a doubleword in two parts. + So handle that case the slow way. */ + if (mode == DImode + && GET_CODE (operands[0]) == REG /* Moving to nonaligned reg pair */ + && (REGNO (operands[0]) & 1)) + return output_move_double (operands); + + output_sized_memop ("ld", mode); + if (GET_CODE (operands[0]) == REG) + return "[%%g1+%%lo(%m1)],%0"; + abort (); +} + +/* Output a floating-point load-from-memory whose operands are OPERANDS[0,1]. + OPERANDS[0] is a reg, and OPERANDS[1] is a mem. + We also handle the case where OPERANDS[0] is a mem. */ + +char * +output_load_floating (operands) + rtx *operands; +{ + enum machine_mode mode = GET_MODE (operands[0]); + rtx address = XEXP (operands[1], 0); + + /* We don't bother trying to see if we know %hi(address). + This is because we are doing a load, and if we know the + %hi value, we probably also know that value in memory. */ + cc_status.flags |= CC_KNOW_HI_G1; + cc_status.mdep = address; + + if (! ((cc_prev_status.flags & CC_KNOW_HI_G1) + && address == cc_prev_status.mdep + && cc_prev_status.mdep == cc_status.mdep)) + { + output_asm_insn ("sethi %%hi(%m1),%%g1", operands); + cc_prev_status.mdep = address; + } + + if (mode == DFmode) + { + if (REG_P (operands[0])) + { + if (REGNO (operands[0]) & 1) + return output_move_double (operands); + else + return "ldd [%%g1+%%lo(%m1)],%0"; + } + cc_status.flags &= ~(CC_F0_IS_0|CC_F1_IS_0); + output_asm_insn ("ldd [%%g1+%%lo(%m1)],%%f0", operands); + operands[1] = gen_rtx (REG, DFmode, 32); + return output_fp_move_double (operands); + } + + if (GET_CODE (operands[0]) == MEM) + { + cc_status.flags &= ~CC_F1_IS_0; + output_asm_insn ("ld [%%g1+%%lo(%1)],%%f1", operands); + if (CONSTANT_ADDRESS_P (XEXP (operands[0], 0))) + { + cc_status.mdep = XEXP (operands[0], 0); + return "sethi %%hi(%m0),%%g1\n\tst %%f1,[%%g1+%%lo(%m0)]"; + } + else + return "st %%f1,%0"; + } + return "ld [%%g1+%%lo(%m1)],%0"; +} + +/* Load the address specified by OPERANDS[3] into the register + specified by OPERANDS[0]. + + OPERANDS[3] may be the result of a sum, hence it could either be: + + (1) CONST + (2) REG + (2) REG + CONST_INT + (3) REG + REG + CONST_INT + (4) REG + REG (special case of 3). + + Note that (3) is not a legitimate address. + All cases are handled here. */ + +void +output_load_address (operands) + rtx *operands; +{ + rtx base, offset; + + if (CONSTANT_P (operands[3])) + { + output_asm_insn ("set %3,%0", operands); + return; + } + + if (REG_P (operands[3])) + { + if (REGNO (operands[0]) != REGNO (operands[3])) + output_asm_insn ("mov %3,%0", operands); + return; + } + + if (GET_CODE (operands[3]) != PLUS) + abort (); + + base = XEXP (operands[3], 0); + offset = XEXP (operands[3], 1); + + if (GET_CODE (base) == CONST_INT) + { + rtx tmp = base; + base = offset; + offset = tmp; + } + + if (GET_CODE (offset) != CONST_INT) + { + /* Operand is (PLUS (REG) (REG)). */ + base = operands[3]; + offset = const0_rtx; + } + + if (REG_P (base)) + { + operands[6] = base; + operands[7] = offset; + if (SMALL_INT (offset)) + output_asm_insn ("add %6,%7,%0", operands); + else + output_asm_insn ("set %7,%0\n\tadd %0,%6,%0", operands); + } + else if (GET_CODE (base) == PLUS) + { + operands[6] = XEXP (base, 0); + operands[7] = XEXP (base, 1); + operands[8] = offset; + + if (SMALL_INT (offset)) + output_asm_insn ("add %6,%7,%0\n\tadd %0,%8,%0", operands); + else + output_asm_insn ("set %8,%0\n\tadd %0,%6,%0\n\tadd %0,%7,%0", operands); + } + else + abort (); +} + +/* Output code to place a size count SIZE in register REG. + ALIGN is the size of the unit of transfer. + + Because block moves are pipelined, we don't include the + first element in the transfer of SIZE to REG. */ + +static void +output_size_for_block_move (size, reg, align) + rtx size, reg; + rtx align; +{ + rtx xoperands[3]; + + xoperands[0] = reg; + xoperands[1] = size; + xoperands[2] = align; + if (GET_CODE (size) == REG) + output_asm_insn ("sub %1,%2,%0", xoperands); + else + { + xoperands[1] + = gen_rtx (CONST_INT, VOIDmode, INTVAL (size) - INTVAL (align)); + cc_status.flags &= ~ CC_KNOW_HI_G1; + output_asm_insn ("set %1,%0", xoperands); + } +} + +/* Emit code to perform a block move. + + OPERANDS[0] is the destination. + OPERANDS[1] is the source. + OPERANDS[2] is the size. + OPERANDS[3] is the alignment safe to use. + OPERANDS[4] is a register we can safely clobber as a temp. */ + +char * +output_block_move (operands) + rtx *operands; +{ + /* A vector for our computed operands. Note that load_output_address + makes use of (and can clobber) up to the 8th element of this vector. */ + rtx xoperands[10]; + rtx zoperands[10]; + static int movstrsi_label = 0; + int i, j; + rtx temp1 = operands[4]; + rtx alignrtx = operands[3]; + int align = INTVAL (alignrtx); + + xoperands[0] = operands[0]; + xoperands[1] = operands[1]; + xoperands[2] = temp1; + + /* We can't move more than four bytes at a time + because we have only one register to move them through. */ + if (align > 4) + { + align = 4; + alignrtx = gen_rtx (CONST_INT, VOIDmode, 4); + } + + /* Since we clobber untold things, nix the condition codes. */ + CC_STATUS_INIT; + + /* Recognize special cases of block moves. These occur + when GNU C++ is forced to treat something as BLKmode + to keep it in memory, when its mode could be represented + with something smaller. + + We cannot do this for global variables, since we don't know + what pages they don't cross. Sigh. */ + if (GET_CODE (operands[2]) == CONST_INT + && INTVAL (operands[2]) <= 16 + && ! CONSTANT_ADDRESS_P (operands[0]) + && ! CONSTANT_ADDRESS_P (operands[1])) + { + int size = INTVAL (operands[2]); + + cc_status.flags &= ~CC_KNOW_HI_G1; + if (align == 1) + { + if (memory_address_p (QImode, plus_constant (xoperands[0], size)) + && memory_address_p (QImode, plus_constant (xoperands[1], size))) + { + /* We will store different integers into this particular RTX. */ + xoperands[2] = gen_rtx (CONST_INT, VOIDmode, 13); + for (i = size-1; i >= 0; i--) + { + INTVAL (xoperands[2]) = i; + output_asm_insn ("ldub [%a1+%2],%%g1\n\tstb %%g1,[%a0+%2]", + xoperands); + } + return ""; + } + } + else if (align == 2) + { + if (memory_address_p (HImode, plus_constant (xoperands[0], size)) + && memory_address_p (HImode, plus_constant (xoperands[1], size))) + { + /* We will store different integers into this particular RTX. */ + xoperands[2] = gen_rtx (CONST_INT, VOIDmode, 13); + for (i = (size>>1)-1; i >= 0; i--) + { + INTVAL (xoperands[2]) = i<<1; + output_asm_insn ("lduh [%a1+%2],%%g1\n\tsth %%g1,[%a0+%2]", + xoperands); + } + return ""; + } + } + else + { + if (memory_address_p (SImode, plus_constant (xoperands[0], size)) + && memory_address_p (SImode, plus_constant (xoperands[1], size))) + { + /* We will store different integers into this particular RTX. */ + xoperands[2] = gen_rtx (CONST_INT, VOIDmode, 13); + for (i = (size>>2)-1; i >= 0; i--) + { + INTVAL (xoperands[2]) = i<<2; + output_asm_insn ("ld [%a1+%2],%%g1\n\tst %%g1,[%a0+%2]", + xoperands); + } + return ""; + } + } + } + + /* This is the size of the transfer. + Either use the register which already contains the size, + or use a free register (used by no operands). + Also emit code to decrement the size value by ALIGN. */ + output_size_for_block_move (operands[2], temp1, alignrtx); + + zoperands[0] = operands[0]; + zoperands[3] = plus_constant (operands[0], align); + output_load_address (zoperands); + + xoperands[3] = gen_rtx (CONST_INT, VOIDmode, movstrsi_label++); + xoperands[4] = gen_rtx (CONST_INT, VOIDmode, align); + +#ifdef NO_UNDERSCORES + if (align == 1) + output_asm_insn ("\n.Lm%3:\n\tldub [%1+%2],%%g1\n\tsubcc %2,%4,%2\n\tbge .Lm%3\n\tstb %%g1,[%0+%2]", xoperands); + else if (align == 2) + output_asm_insn ("\n.Lm%3:\n\tlduh [%1+%2],%%g1\n\tsubcc %2,%4,%2\n\tbge .Lm%3\n\tsth %%g1,[%0+%2]", xoperands); + else + output_asm_insn ("\n.Lm%3:\n\tld [%1+%2],%%g1\n\tsubcc %2,%4,%2\n\tbge .Lm%3\n\tst %%g1,[%0+%2]", xoperands); +#else + if (align == 1) + output_asm_insn ("\nLm%3:\n\tldub [%1+%2],%%g1\n\tsubcc %2,%4,%2\n\tbge Lm%3\n\tstb %%g1,[%0+%2]", xoperands); + else if (align == 2) + output_asm_insn ("\nLm%3:\n\tlduh [%1+%2],%%g1\n\tsubcc %2,%4,%2\n\tbge Lm%3\n\tsth %%g1,[%0+%2]", xoperands); + else + output_asm_insn ("\nLm%3:\n\tld [%1+%2],%%g1\n\tsubcc %2,%4,%2\n\tbge Lm%3\n\tst %%g1,[%0+%2]", xoperands); +#endif + return ""; +} + +/* What the sparc lacks in hardware, make up for in software. + Compute a fairly good sequence of shift and add insns + to make a multiply happen. */ + +#define ABS(x) ((x) < 0 ? -(x) : x) + +char * +output_mul_by_constant (insn, operands, unsignedp) + rtx insn; + rtx *operands; + int unsignedp; +{ + int c; /* Size of constant */ + int shifts[BITS_PER_WORD]; /* Table of shifts */ + unsigned int p, log; /* A power of two, and its log */ + int d1, d2; /* Differences of c and p */ + int first = 1; /* True if dst has unknown data in it */ + int i; + + CC_STATUS_INIT; + + c = INTVAL (operands[2]); + if (c == 0) + { + /* Does happen, at least when not optimizing. */ + if (GET_CODE (operands[0]) == MEM) + return "st %%g0,%0"; + return "mov %%g0,%0"; + } + + output_asm_insn ("! start open coded multiply"); + + /* Clear out the table of shifts. */ + for (i = 0; i < BITS_PER_WORD; ++i) + shifts[i] = 0; + + while (c) + { + /* Find the power of two nearest ABS(c) */ + p = 1, log = 0; + do + { + d1 = ABS(c) - p; + p *= 2; + ++log; + } + while (p < ABS(c)); + d2 = p - ABS(c); + + /* Make an appropriate entry in shifts for p. */ + if (d2 < d1) + { + shifts[log] = c < 0 ? -1 : 1; + c = c < 0 ? d2 : -d2; + } + else + { + shifts[log - 1] = c < 0 ? -1 : 1; + c = c < 0 ? -d1 : d1; + } + } + + /* Take care of the first insn in sequence. + We know we have at least one. */ + + /* A value of -1 in shifts says to subtract that power of two, and a value + of 1 says to add that power of two. */ + for (i = 0; ; i++) + if (shifts[i]) + { + if (i) + { + operands[2] = gen_rtx (CONST_INT, VOIDmode, i); + output_asm_insn ("sll %1,%2,%%g1", operands); + } + else output_asm_insn ("mov %1,%%g1", operands); + + log = i; + if (shifts[i] < 0) + output_asm_insn ("sub %%g0,%%g1,%0", operands); + else + output_asm_insn ("mov %%g1,%0", operands); + break; + } + + /* A value of -1 in shifts says to subtract that power of two, and a value + of 1 says to add that power of two--continued. */ + for (i += 1; i < BITS_PER_WORD; ++i) + if (shifts[i]) + { + if (i - log > 0) + { + operands[2] = gen_rtx (CONST_INT, VOIDmode, i - log); + output_asm_insn ("sll %%g1,%2,%%g1", operands); + } + else + { + operands[2] = gen_rtx (CONST_INT, VOIDmode, log - i); + output_asm_insn ("sra %%g1,%2,%%g1", operands); + } + log = i; + if (shifts[i] < 0) + output_asm_insn ("sub %0,%%g1,%0", operands); + else + output_asm_insn ("add %0,%%g1,%0", operands); + } + + output_asm_insn ("! end open coded multiply"); + + return ""; +} + +char * +output_mul_insn (operands, unsignedp) + rtx *operands; + int unsignedp; +{ + int lucky1 = ((unsigned)REGNO (operands[1]) - 8) <= 1; + int lucky2 = ((unsigned)REGNO (operands[2]) - 8) <= 1; + + CC_STATUS_INIT; + + if (lucky1) + { + if (lucky2) + { + if (REGNO (operands[1]) == REGNO (operands[2])) + { + if (REGNO (operands[1]) == 8) + output_asm_insn ("mov %%o0,%%o1"); + else + output_asm_insn ("mov %%o1,%%o0"); + } + output_asm_insn ("call .mul,2\n\tnop", operands); + } + else + { + rtx xoperands[2]; + xoperands[0] = gen_rtx (REG, SImode, + 8 ^ (REGNO (operands[1]) == 8)); + xoperands[1] = operands[2]; + output_asm_insn ("call .mul,2\n\tmov %1,%0", xoperands); + } + } + else if (lucky2) + { + rtx xoperands[2]; + xoperands[0] = gen_rtx (REG, SImode, + 8 ^ (REGNO (operands[2]) == 8)); + xoperands[1] = operands[1]; + output_asm_insn ("call .mul,2\n\tmov %1,%0", xoperands); + } + else + { + output_asm_insn ("mov %1,%%o0\n\tcall .mul,2\n\tmov %2,%%o1", + operands); + } + + if (REGNO (operands[0]) == 8) + return ""; + return "mov %%o0,%0"; +} + +/* Make floating point register f0 contain 0. + SIZE is the number of registers (including f0) + which should contain 0. */ + +void +make_f0_contain_0 (size) + int size; +{ + if (size == 1) + { + if ((cc_status.flags & (CC_F0_IS_0)) == 0) + output_asm_insn ("ld [%%fp-16],%%f0", 0); + cc_status.flags |= CC_F0_IS_0; + } + else if (size == 2) + { + if ((cc_status.flags & CC_F0_IS_0) == 0) + output_asm_insn ("ld [%%fp-16],%%f0", 0); + if ((cc_status.flags & (CC_F1_IS_0)) == 0) + output_asm_insn ("ld [%%fp-12],%%f1", 0); + cc_status.flags |= CC_F0_IS_0 | CC_F1_IS_0; + } +} + +/* Since condition codes don't have logical links, we need to keep + their setting and use together for set-cc insns. */ +void +gen_scc_insn (code, mode, operands) + enum rtx_code code; + enum machine_mode mode; + rtx *operands; +{ + extern rtx sequence_stack; + rtx last_insn = XEXP (XEXP (sequence_stack, 1), 0); + rtx last_pat; + + /* Skip back over the CLOBBERs that may precede this insn. */ + while (last_insn && GET_CODE (last_insn) == INSN + && GET_CODE (PATTERN (last_insn)) == CLOBBER) + last_insn = PREV_INSN (last_insn); + /* We should have found the preceding compare. */ + if (last_insn == 0 || GET_CODE (last_insn) != INSN) + abort (); + last_pat = PATTERN (last_insn); + if (GET_CODE (last_pat) != SET + || GET_CODE (SET_DEST (last_pat)) != CC0) + abort (); + + /* Turn off that previous insn, now that we have got the data out of it. */ + PUT_CODE (last_insn, NOTE); + NOTE_LINE_NUMBER (last_insn) = NOTE_INSN_DELETED; + + /* Emit one replacement insn to compare operands and store result. */ + emit_insn (gen_rtx (SET, VOIDmode, operands[0], + gen_rtx (code, mode, SET_SRC (last_pat), const0_rtx))); +} + +/* Output reasonable peephole for set-on-condition-code insns. + Note that these insns assume a particular way of defining + labels. Therefore, *both* tm-sparc.h and this function must + be changed if a new syntax is needed. */ + +char * +output_scc_insn (code, operand) + enum rtx_code code; + rtx operand; +{ + rtx xoperands[2]; + rtx label = gen_label_rtx (); + int cc_in_fccr = cc_status.flags & CC_IN_FCCR; + int antisymmetric = 0; + + xoperands[0] = operand; + xoperands[1] = label; + + switch (code) + { + case NE: + if (cc_in_fccr) + output_asm_insn ("fbne,a %l0", &label); + else + output_asm_insn ("bne,a %l0", &label); + break; + case EQ: + if (cc_in_fccr) + output_asm_insn ("fbe,a %l0", &label); + else + output_asm_insn ("be,a %l0", &label); + break; + case GE: + if (cc_in_fccr) + output_asm_insn ("fbge,a %l0", &label); + else + output_asm_insn ("bge,a %l0", &label); + antisymmetric = 1; + break; + case GT: + if (cc_in_fccr) + output_asm_insn ("fbg,a %l0", &label); + else + output_asm_insn ("bg,a %l0", &label); + antisymmetric = 1; + break; + case LE: + if (cc_in_fccr) + output_asm_insn ("fble,a %l0", &label); + else + output_asm_insn ("ble,a %l0", &label); + antisymmetric = 1; + break; + case LT: + if (cc_in_fccr) + output_asm_insn ("fbl,a %l0", &label); + else + output_asm_insn ("bl,a %l0", &label); + antisymmetric = 1; + break; + case GEU: + if (cc_in_fccr) + abort (); + else + output_asm_insn ("bgeu,a %l0", &label); + antisymmetric = 1; + break; + case GTU: + if (cc_in_fccr) + abort (); + else + output_asm_insn ("bgu,a %l0", &label); + antisymmetric = 1; + break; + case LEU: + if (cc_in_fccr) + abort (); + else + output_asm_insn ("bleu,a %l0", &label); + antisymmetric = 1; + break; + case LTU: + if (cc_in_fccr) + abort (); + else + output_asm_insn ("blu,a %l0", &label); + antisymmetric = 1; + break; + default: + abort (); + } + + if (antisymmetric + && (cc_status.flags & CC_REVERSED)) + output_asm_insn ("orcc %%g0,0,%0\n\torcc %%g0,1,%0\n%l1:", xoperands); + else + output_asm_insn ("orcc %%g0,1,%0\n\torcc %%g0,0,%0\n%l1:", xoperands); + cc_status.flags &= ~CC_IN_FCCR; + + return ""; +} + +/* Output a delayed branch insn with the delay insn in its + branch slot. The delayed branch insn template is in TEMPLATE, + with operands OPERANDS. The insn in its delay slot is INSN. + + As a special case, since we know that all memory transfers are via + ld/st insns, if we see a (MEM (SYMBOL_REF ...)) we divide the memory + reference around the branch as + + sethi %hi(x),%%g1 + b ... + ld/st [%g1+%lo(x)],... + + As another special case, we handle loading (SYMBOL_REF ...) and + other large constants around branches as well: + + sethi %hi(x),%0 + b ... + or %0,%lo(x),%1 + + */ + +char * +output_delayed_branch (template, operands, insn) + char *template; + rtx *operands; + rtx insn; +{ + extern rtx recog_operand[]; + rtx src = XVECEXP (PATTERN (insn), 0, 1); + rtx dest = XVECEXP (PATTERN (insn), 0, 0); + + if (GET_CODE (src) == SYMBOL_REF + || (GET_CODE (src) == CONST_INT + && !(SMALL_INT (src) || (INTVAL (src) & 0x3ff) == 0))) + { + rtx xoperands[2]; + xoperands[0] = dest; + xoperands[1] = src; + + /* Output the `sethi' insn. */ + output_asm_insn ("sethi %%hi(%1),%0", xoperands); + + /* Output the branch instruction next. */ + output_asm_insn (template, operands); + + /* Now output the `or' insn. */ + output_asm_insn ("or %0,%%lo(%1),%0", xoperands); + } + else if ((GET_CODE (src) == MEM + && CONSTANT_ADDRESS_P (XEXP (src, 0))) + || (GET_CODE (dest) == MEM + && CONSTANT_ADDRESS_P (XEXP (dest, 0)))) + { + rtx xoperands[2]; + char *split_template; + xoperands[0] = dest; + xoperands[1] = src; + + /* Output the `sethi' insn. */ + if (GET_CODE (src) == MEM) + { + if (! ((cc_prev_status.flags & CC_KNOW_HI_G1) + && cc_prev_status.mdep == XEXP (operands[1], 0))) + output_asm_insn ("sethi %%hi(%m1),%%g1", xoperands); + split_template = "ld [%%g1+%%lo(%m1)],%0"; + } + else + { + if (! ((cc_prev_status.flags & CC_KNOW_HI_G1) + && cc_prev_status.mdep == XEXP (operands[0], 0))) + output_asm_insn ("sethi %%hi(%m0),%%g1", xoperands); + split_template = "st %r1,[%%g1+%%lo(%m0)]"; + } + + /* Output the branch instruction next. */ + output_asm_insn (template, operands); + + /* Now output the load or store. + No need to do a CC_STATUS_INIT, because we are branching anyway. */ + output_asm_insn (split_template, xoperands); + } + else + { + extern char *insn_template[]; + extern char *(*insn_outfun[])(); + int insn_code_number; + rtx pat = gen_rtx (SET, VOIDmode, dest, src); + rtx delay_insn = gen_rtx (INSN, VOIDmode, 0, 0, 0, pat, -1, 0, 0); + int i; + extern rtx alter_subreg(); + extern int insn_n_operands[]; + + /* Output the branch instruction first. */ + output_asm_insn (template, operands); + + /* Now recognize the insn which we put in its delay slot. + We must do this after outputing the branch insn, + since operands may just be a pointer to `recog_operand'. */ + insn_code_number = recog (pat, delay_insn); + if (insn_code_number == -1) + abort (); + + for (i = 0; i < insn_n_operands[insn_code_number]; i++) + { + if (GET_CODE (recog_operand[i]) == SUBREG) + recog_operand[i] = alter_subreg (recog_operand[i]); + } + + /* Now get the template for what this insn would + have been, without the branch. Its operands are + exactly the same as they would be, so we don't + need to do an insn_extract. */ + template = insn_template[insn_code_number]; + if (template == 0) + template = (*insn_outfun[insn_code_number]) (recog_operand, delay_insn); + output_asm_insn (template, recog_operand); + } + CC_STATUS_INIT; + return ""; +} + +/* Output a newly constructed insn DELAY_INSN. */ +char * +output_delay_insn (delay_insn) + rtx delay_insn; +{ + char *template; + extern rtx recog_operand[]; + extern char call_used_regs[]; + extern char *insn_template[]; + extern int insn_n_operands[]; + extern char *(*insn_outfun[])(); + extern rtx alter_subreg(); + int insn_code_number; + extern int insn_n_operands[]; + int i; + + /* Now recognize the insn which we put in its delay slot. + We must do this after outputing the branch insn, + since operands may just be a pointer to `recog_operand'. */ + insn_code_number = recog_memoized (delay_insn); + if (insn_code_number == -1) + abort (); + + /* Extract the operands of this delay insn. */ + INSN_CODE (delay_insn) = insn_code_number; + insn_extract (delay_insn); + + /* It is possible that this insn has not been properly scaned by final + yet. If this insn's operands don't appear in the peephole's + actual operands, then they won't be fixed up by final, so we + make sure they get fixed up here. -- This is a kludge. */ + for (i = 0; i < insn_n_operands[insn_code_number]; i++) + { + if (GET_CODE (recog_operand[i]) == SUBREG) + recog_operand[i] = alter_subreg (recog_operand[i]); + } + +#ifdef REGISTER_CONSTRAINTS + if (! constrain_operands (insn_code_number)) + abort (); +#endif + + cc_prev_status = cc_status; + + /* Update `cc_status' for this instruction. + The instruction's output routine may change it further. + If the output routine for a jump insn needs to depend + on the cc status, it should look at cc_prev_status. */ + + NOTICE_UPDATE_CC (PATTERN (delay_insn), delay_insn); + + /* Now get the template for what this insn would + have been, without the branch. */ + + template = insn_template[insn_code_number]; + if (template == 0) + template = (*insn_outfun[insn_code_number]) (recog_operand, delay_insn); + output_asm_insn (template, recog_operand); + return ""; +} + +/* Output the insn HEAD, keeping OPERANDS protected (wherever they are). + HEAD comes from the target of some branch, so before we output it, + we delete it from the target, lest we execute it twice. The caller + of this function promises that such code motion is permissable. */ +char * +output_eager_then_insn (head, operands) + rtx head; + rtx *operands; +{ + extern rtx alter_subreg (); + extern int insn_n_operands[]; + extern rtx recog_operand[]; + rtx xoperands[MAX_RECOG_OPERANDS]; + int insn_code_number, i, nbytes; + rtx nhead; + + /* Micro-hack: run peephole on head if it looks like a good idea. + Right now there's only one such case worth doing... + + This could be made smarter if the peephole for ``2-insn combine'' + were also made smarter. */ + if (GET_CODE (PATTERN (head)) == SET + && REG_P (SET_SRC (PATTERN (head))) + && REG_P (SET_DEST (PATTERN (head))) + && (nhead = next_real_insn_no_labels (head)) + && GET_CODE (nhead) == INSN + && GET_CODE (PATTERN (nhead)) == SET + && GET_CODE (SET_DEST (PATTERN (nhead))) == CC0 + && (SET_SRC (PATTERN (nhead)) == SET_SRC (PATTERN (head)) + || SET_SRC (PATTERN (nhead)) == SET_DEST (PATTERN (head)))) + /* Something's wrong if this does not fly. */ + if (! peephole (head)) + abort (); + + /* Save our contents of `operands', since output_delay_insn sets them. */ + insn_code_number = recog_memoized (head); + nbytes = insn_n_operands[insn_code_number] * sizeof (rtx); + bcopy (operands, xoperands, nbytes); + + /* Output the delay insn, and prevent duplication later. */ + delete_insn (head); + output_delay_insn (head); + + /* Restore this insn's operands. */ + bcopy (xoperands, operands, nbytes); +} + +/* Return the next INSN, CALL_INSN or JUMP_INSN after LABEL; + or 0, if there is none. Also return 0 if we cross a label. */ + +rtx +next_real_insn_no_labels (label) + rtx label; +{ + register rtx insn = NEXT_INSN (label); + register RTX_CODE code; + + while (insn) + { + code = GET_CODE (insn); + if (code == INSN) + { + if (GET_CODE (PATTERN (insn)) != CLOBBER + && GET_CODE (PATTERN (insn)) != USE) + return insn; + } + if (code == CALL_INSN || code == JUMP_INSN) + return insn; + if (code == CODE_LABEL) + return 0; + insn = NEXT_INSN (insn); + } + + return 0; +} + +int +operands_satisfy_eager_branch_peephole (operands, conditional) + rtx *operands; + int conditional; +{ + rtx label; + + if (conditional) + { + if (GET_CODE (operands[0]) != IF_THEN_ELSE) + return 0; + + if (GET_CODE (XEXP (operands[0], 1)) == LABEL_REF) + label = XEXP (XEXP (operands[0], 1), 0); + else if (GET_CODE (XEXP (operands[0], 2)) == LABEL_REF) + label = XEXP (XEXP (operands[0], 2), 0); + else return 0; + } + else + { + label = operands[0]; + } + + if (LABEL_NUSES (label) == 1) + { + rtx prev = PREV_INSN (label); + while (prev && GET_CODE (prev) == NOTE) + prev = PREV_INSN (prev); + if (prev == 0 + || GET_CODE (prev) == BARRIER) + { + rtx head = next_real_insn_no_labels (label); + + if (head + && ! INSN_DELETED_P (head) + && GET_CODE (head) == INSN + && GET_CODE (PATTERN (head)) == SET + && strict_single_insn_op_p (SET_SRC (PATTERN (head)), + GET_MODE (SET_DEST (PATTERN (head)))) + && strict_single_insn_op_p (SET_DEST (PATTERN (head)), + GET_MODE (SET_DEST (PATTERN (head)))) + /* Moves between FP regs and CPU regs are two insns. */ + && !(GET_CODE (SET_SRC (PATTERN (head))) == REG + && GET_CODE (SET_DEST (PATTERN (head))) == REG + && (FP_REG_P (SET_SRC (PATTERN (head))) + != FP_REG_P (SET_DEST (PATTERN (head)))))) + { + if (conditional == 2) + return (GET_CODE (operands[1]) != PC + && safe_insn_src_p (operands[2], VOIDmode) + && strict_single_insn_op_p (operands[2], VOIDmode) + && operand_clobbered_before_used_after (operands[1], label)); + return 1; + } + } + } + + if (conditional == 1 + && GET_CODE (operands[1]) != PC + && safe_insn_src_p (operands[2], VOIDmode) + && strict_single_insn_op_p (operands[2], VOIDmode) + && operand_clobbered_before_used_after (operands[1], label)) + return 1; + + return 0; +} + |
