blob: b1f8e1fc9e330a8d1b17b37ca5a3318b8a00201f [file] [log] [blame]
/* Definitions of target machine for GNU compiler, for IBM S/390
Copyright (C) 2002-2020 Free Software Foundation, Inc.
Contributed by Hartmut Penner (hpenner@de.ibm.com) and
Ulrich Weigand (uweigand@de.ibm.com).
This file is part of GCC.
GCC 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 3, or (at your option) any later
version.
GCC 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 GCC; see the file COPYING3. If not see
<http://www.gnu.org/licenses/>. */
/* 256-bit integer mode is needed for STACK_SAVEAREA_MODE. */
INT_MODE (OI, 32);
/* Define TFmode to work around reload problem PR 20927. */
FLOAT_MODE (TF, 16, ieee_quad_format);
/* Add any extra modes needed to represent the condition code. */
/*
Condition Codes
CC0 CC1 CC2 CC3
Check for zero
CCZ: EQ NE NE NE
CCZ1: EQ NE (CS)
Unsigned compares
CCU: EQ LTU GTU NE (CLG/R, CL/R/Y, CLM/Y, CLI/Y)
CCUR: EQ GTU LTU NE (CLGF/R)
Signed compares
CCS: EQ LT GT UNORDERED (LTGFR, LTGR, LTR, ICM/Y,
LTDBR, LTDR, LTEBR, LTER,
CG/R, C/R/Y, CGHI, CHI,
CDB/R, CD/R, CEB/R, CE/R,
ADB/R, AEB/R, SDB/R, SEB/R,
SRAG, SRA, SRDA)
CCSR: EQ GT LT UNORDERED (CGF/R, CH/Y)
CCSFPS: EQ LT GT UNORDERED (KEB/R, KDB/R, KXBR, KDTR,
KXTR, WFK)
Condition codes resulting from add with overflow
CCA: EQ LT GT Overflow
CCAP: EQ LT GT LT (AGHI, AHI)
CCAN: EQ LT GT GT (AGHI, AHI)
Condition codes for overflow checking resulting from signed adds/subs/mults
CCO: EQ EQ EQ NE (AGR, AGHI, SGR, MSC, ...)
Condition codes of unsigned adds and subs
CCL: EQ NE EQ NE (ALGF/R, ALG/R, AL/R/Y,
ALCG/R, ALC/R,
SLGF/R, SLG/R, SL/R/Y,
SLBG/R, SLB/R)
CCL1: GEU GEU LTU LTU (ALG/R, AL/R/Y)
CCL2: GTU GTU LEU LEU (SLG/R, SL/R/Y)
CCL3: EQ LTU EQ GTU (SLG/R, SL/R/Y)
Test under mask checks
CCT: EQ NE NE NE (ICM/Y, TML, CG/R, CGHI,
C/R/Y, CHI, NG/R, N/R/Y,
OG/R, O/R/Y, XG/R, X/R/Y)
CCT1: NE EQ NE NE (TMH, TML)
CCT2: NE NE EQ NE (TMH, TML)
CCT3: NE NE NE EQ (TMH, TML)
CCA and CCT modes are request only modes. These modes are never returned by
s390_select_cc_mode. They are only intended to match other modes.
Requested mode -> Destination CC register mode
CCS, CCU, CCT, CCSR, CCUR -> CCZ
CCA -> CCAP, CCAN
*** Comments ***
CCAP, CCAN
The CC obtained from add instruction usually can't be used for comparisons
because its coupling with overflow flag. In case of an overflow the
less than/greater than data are lost. Nevertheless a comparison can be done
whenever immediate values are involved because they are known at compile time.
If you know whether the used constant is positive or negative you can predict
the sign of the result even in case of an overflow.
CCO
This mode is used to check whether there was an overflow condition in
a signed add, sub, or mul operation. See (addv<mode>4, subv<mode>4,
mulv<mode>4 patterns).
CCT, CCT1, CCT2, CCT3
If bits of an integer masked with an AND instruction are checked, the test under
mask instructions turn out to be very handy for a set of special cases.
The simple cases are checks whether all masked bits are zero or ones:
int a;
if ((a & (16 + 128)) == 0) -> CCT/CCZ
if ((a & (16 + 128)) == 16 + 128) -> CCT3
Using two extra modes makes it possible to do complete checks on two bits of an
integer (This is possible on register operands only. TM does not provide the
information necessary for CCT1 and CCT2 modes.):
int a;
if ((a & (16 + 128)) == 16) -> CCT1
if ((a & (16 + 128)) == 128) -> CCT2
CCSR, CCUR
There are several instructions comparing 32 bit with 64-bit unsigned/signed
values. Such instructions can be considered to have a builtin zero/sign_extend.
The problem is that in the RTL (to be canonical) the zero/sign extended operand
has to be the first one but the machine instructions like it the other way
around. The following both modes can be considered as CCS and CCU modes with
exchanged operands.
CCSFPS
This mode is used for signaling rtxes: LT, LE, GT, GE and LTGT.
CCL1, CCL2
These modes represent the result of overflow checks.
if (a + b < a) -> CCL1 state of the carry bit (CC2 | CC3)
if (a - b > a) -> CCL2 state of the borrow bit (CC0 | CC1)
They are used when multi word numbers are computed dealing one SImode part after
another or whenever manual overflow checks like the examples above are
compiled.
CCL3
A logical subtract instruction sets the borrow bit in case of an overflow.
The resulting condition code of those instructions is represented by the
CCL3 mode. Together with the CCU mode this mode is used for jumpless
implementations of several if-constructs - see s390_expand_addcc for more
details.
CCZ1
The compare and swap instructions sets the condition code to 0/1 if the
operands were equal/unequal. The CCZ1 mode ensures the result can be
effectively placed into a register.
CCVIH, CCVIHU, CCVFH, CCVFHE
These are condition code modes used in instructions setting the
condition code. The mode determines which comparison to perform (H -
high, HU - high unsigned, HE - high or equal) and whether it is a
floating point comparison or not (I - int, F - float).
The comparison operation to be performed needs to be encoded into the
condition code mode since the comparison operator is not available in
compare style patterns (set cc (compare (op0) (op1))). So the
condition code mode is the only information to determine the
instruction to be used.
CCVIALL, CCVIANY, CCVFALL, CCVFANY
These modes are used in instructions reading the condition code.
Opposed to the CC producer patterns the comparison operator is
available. Hence the comparison operation does not need to be part of
the CC mode. However, we still need to know whether CC has been
generated by a float or an integer comparison in order to be able to
invert the condition correctly (int: GT -> LE, float: GT -> UNLE).
The ALL and ANY variants differ only in the usage of CC1 which
indicates a mixed result across the vector elements. Be aware that
depending on the comparison code the ALL and ANY variants might
actually refer to their opposite meaning. I.e. while inverting the
comparison in (EQ (reg:CCVIALL 33) (const_int 0)) results in (NE
(reg:CCVIALL 33) (const_int 0)) it in fact describes an ANY comparison
(inverting "all equal" should be "any not equal") However, the
middle-end does invert only the comparison operator without touching
the mode.
Hence, the ALL/ANY in the mode names refer to the meaning in the
context of EQ, GT, GE while for the inverted codes it actually means
ANY/ALL.
CCRAW
The cc mode generated by a non-compare instruction. The condition
code mask for the CC consumer is determined by the comparison operator
(only EQ and NE allowed) and the immediate value given as second
operand to the operator. For the other CC modes this value used to be
0.
*/
CC_MODE (CCZ);
CC_MODE (CCZ1);
CC_MODE (CCA);
CC_MODE (CCAP);
CC_MODE (CCAN);
CC_MODE (CCO);
CC_MODE (CCL);
CC_MODE (CCL1);
CC_MODE (CCL2);
CC_MODE (CCL3);
CC_MODE (CCU);
CC_MODE (CCUR);
CC_MODE (CCS);
CC_MODE (CCSR);
CC_MODE (CCSFPS);
CC_MODE (CCT);
CC_MODE (CCT1);
CC_MODE (CCT2);
CC_MODE (CCT3);
CC_MODE (CCRAW);
CC_MODE (CCVEQ);
CC_MODE (CCVIH);
CC_MODE (CCVIHU);
CC_MODE (CCVFH);
CC_MODE (CCVFHE);
CC_MODE (CCVIALL);
CC_MODE (CCVIANY);
CC_MODE (CCVFALL);
CC_MODE (CCVFANY);
/* Vector modes. */
VECTOR_MODES (INT, 2); /* V2QI */
VECTOR_MODES (INT, 4); /* V4QI V2HI */
VECTOR_MODES (INT, 8); /* V8QI V4HI V2SI */
VECTOR_MODES (INT, 16); /* V16QI V8HI V4SI V2DI */
VECTOR_MODE (FLOAT, SF, 2); /* V2SF */
VECTOR_MODE (FLOAT, SF, 4); /* V4SF */
VECTOR_MODE (FLOAT, DF, 2); /* V2DF */
VECTOR_MODE (INT, QI, 1); /* V1QI */
VECTOR_MODE (INT, HI, 1); /* V1HI */
VECTOR_MODE (INT, SI, 1); /* V1SI */
VECTOR_MODE (INT, DI, 1); /* V1DI */
VECTOR_MODE (INT, TI, 1); /* V1TI */
VECTOR_MODE (FLOAT, SF, 1); /* V1SF */
VECTOR_MODE (FLOAT, DF, 1); /* V1DF */
VECTOR_MODE (FLOAT, TF, 1); /* V1TF */