gcc/config/nvptx/nvptx.h

   1 /* Target Definitions for NVPTX.
   2    Copyright (C) 2014-2018 Free Software Foundation, Inc.
   3    Contributed by Bernd Schmidt <bernds@codesourcery.com>
   4
   5    This file is part of GCC.
   6
   7    GCC is free software; you can redistribute it and/or modify it
   8    under the terms of the GNU General Public License as published
   9    by the Free Software Foundation; either version 3, or (at your
  10    option) any later version.
  11
  12    GCC is distributed in the hope that it will be useful, but WITHOUT
  13    ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
  14    or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public
  15    License for more details.
  16
  17    You should have received a copy of the GNU General Public License
  18    along with GCC; see the file COPYING3.  If not see
  19    <http://www.gnu.org/licenses/>.  */
  20
  21 #ifndef GCC_NVPTX_H
  22 #define GCC_NVPTX_H
  23
  24 /* Run-time Target.  */
  25
  26 #define STARTFILE_SPEC "%{mmainkernel:crt0.o}"
  27
  28 #define TARGET_CPU_CPP_BUILTINS()               \
  29   do                                            \
  30     {                                           \
  31       builtin_assert ("machine=nvptx");         \
  32       builtin_assert ("cpu=nvptx");             \
  33       builtin_define ("__nvptx__");             \
  34       if (TARGET_SOFT_STACK)                    \
  35         builtin_define ("__nvptx_softstack__"); \
  36       if (TARGET_UNIFORM_SIMT)                  \
  37         builtin_define ("__nvptx_unisimt__");   \
  38     } while (0)
  39
  40 /* Avoid the default in ../../gcc.c, which adds "-pthread", which is not
  41    supported for nvptx.  */
  42 #define GOMP_SELF_SPECS ""
  43
  44 /* Storage Layout.  */
  45
  46 #define BITS_BIG_ENDIAN 0
  47 #define BYTES_BIG_ENDIAN 0
  48 #define WORDS_BIG_ENDIAN 0
  49
  50 /* Chosen such that we won't have to deal with multi-word subregs.  */
  51 #define UNITS_PER_WORD 8
  52
  53 /* Alignments in bits.  */
  54 #define PARM_BOUNDARY 32
  55 #define STACK_BOUNDARY 128
  56 #define FUNCTION_BOUNDARY 32
  57 #define BIGGEST_ALIGNMENT 128
  58 #define STRICT_ALIGNMENT 1
  59
  60 #define MAX_STACK_ALIGNMENT (1024 * 8)
  61
  62 #define DATA_ALIGNMENT nvptx_data_alignment
  63
  64 /* Copied from elf.h and other places.  We'd otherwise use
  65    BIGGEST_ALIGNMENT and fail a number of testcases.  */
  66 #define MAX_OFILE_ALIGNMENT (32768 * 8)
  67
  68 /* Type Layout.  */
  69
  70 #define DEFAULT_SIGNED_CHAR 1
  71
  72 #define SHORT_TYPE_SIZE 16
  73 #define INT_TYPE_SIZE 32
  74 #define LONG_TYPE_SIZE (TARGET_ABI64 ? 64 : 32)
  75 #define LONG_LONG_TYPE_SIZE 64
  76 #define FLOAT_TYPE_SIZE 32
  77 #define DOUBLE_TYPE_SIZE 64
  78 #define LONG_DOUBLE_TYPE_SIZE 64
  79 #define TARGET_SUPPORTS_WIDE_INT 1
  80
  81 #undef SIZE_TYPE
  82 #define SIZE_TYPE (TARGET_ABI64 ? "long unsigned int" : "unsigned int")
  83 #undef PTRDIFF_TYPE
  84 #define PTRDIFF_TYPE (TARGET_ABI64 ? "long int" : "int")
  85
  86 #define POINTER_SIZE (TARGET_ABI64 ? 64 : 32)
  87 #define Pmode (TARGET_ABI64 ? DImode : SImode)
  88 #define STACK_SIZE_MODE Pmode
  89
  90 /* Registers.  Since ptx is a virtual target, we just define a few
  91    hard registers for special purposes and leave pseudos unallocated.
  92    We have to have some available hard registers, to keep gcc setup
  93    happy.  */
  94 #define FIRST_PSEUDO_REGISTER 16
  95 #define FIXED_REGISTERS     { 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0 }
  96 #define CALL_USED_REGISTERS { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 }
  97
  98 /* Register Classes.  */
  99 enum reg_class             {  NO_REGS,    ALL_REGS,     LIM_REG_CLASSES };
 100 #define REG_CLASS_NAMES    { "NO_REGS",  "ALL_REGS" }
 101 #define REG_CLASS_CONTENTS { { 0x0000 }, { 0xFFFF } }
 102 #define N_REG_CLASSES (int) LIM_REG_CLASSES
 103
 104 #define GENERAL_REGS ALL_REGS
 105 #define REGNO_REG_CLASS(R) ((void)(R), ALL_REGS)
 106 #define BASE_REG_CLASS ALL_REGS
 107 #define INDEX_REG_CLASS NO_REGS
 108
 109 #define REGNO_OK_FOR_BASE_P(X) true
 110 #define REGNO_OK_FOR_INDEX_P(X) false
 111
 112 #define CLASS_MAX_NREGS(class, mode) \
 113   ((GET_MODE_SIZE (mode) + UNITS_PER_WORD - 1) / UNITS_PER_WORD)
 114
 115 #define PROMOTE_MODE(MODE, UNSIGNEDP, TYPE)             \
 116   if ((MODE) == QImode || (MODE) == HImode)             \
 117     {                                                   \
 118       (MODE) = SImode;                                  \
 119       (void)(UNSIGNEDP);                                \
 120       (void)(TYPE);                                     \
 121     }
 122
 123 /* Stack and Calling.  */
 124
 125 #define FRAME_GROWS_DOWNWARD 0
 126 #define STACK_GROWS_DOWNWARD 1
 127
 128 #define NVPTX_RETURN_REGNUM 0
 129 #define STACK_POINTER_REGNUM 1
 130 #define FRAME_POINTER_REGNUM 2
 131 #define ARG_POINTER_REGNUM 3
 132 #define STATIC_CHAIN_REGNUM 4
 133 /* This register points to the shared memory location with the current warp's
 134    soft stack pointer (__nvptx_stacks[tid.y]).  */
 135 #define SOFTSTACK_SLOT_REGNUM 5
 136 /* This register is used to save the previous value of the soft stack pointer
 137    in the prologue and restore it when returning.  */
 138 #define SOFTSTACK_PREV_REGNUM 6
 139
 140 #define REGISTER_NAMES                                                  \
 141   {                                                                     \
 142     "%value", "%stack", "%frame", "%args",                              \
 143     "%chain", "%sspslot", "%sspprev", "%hr7",                           \
 144     "%hr8", "%hr9", "%hr10", "%hr11", "%hr12", "%hr13", "%hr14", "%hr15" \
 145   }
 146
 147 #define FIRST_PARM_OFFSET(FNDECL) ((void)(FNDECL), 0)
 148 #define PUSH_ARGS_REVERSED 1
 149 #define ACCUMULATE_OUTGOING_ARGS 1
 150
 151 /* Avoid using the argument pointer for frame-related things.  */
 152 #define FRAME_POINTER_CFA_OFFSET(FNDECL) ((void)(FNDECL), 0)
 153
 154 #ifdef HOST_WIDE_INT
 155 struct nvptx_args {
 156   tree fntype;
 157   /* Number of arguments passed in registers so far.  */
 158   int count;
 159 };
 160 #endif
 161
 162 #define CUMULATIVE_ARGS struct nvptx_args
 163
 164 #define INIT_CUMULATIVE_ARGS(CUM, FNTYPE, LIBNAME, FNDECL, N_NAMED_ARGS) \
 165   ((CUM).fntype = (FNTYPE), (CUM).count = 0, (void)0)
 166
 167 #define FUNCTION_ARG_REGNO_P(r) 0
 168
 169 #define DEFAULT_PCC_STRUCT_RETURN 0
 170
 171 #define FUNCTION_PROFILER(file, labelno) \
 172   fatal_error (input_location, \
 173                "profiling is not yet implemented for this architecture")
 174
 175 #define TRAMPOLINE_SIZE 32
 176 #define TRAMPOLINE_ALIGNMENT 256
 177 \f
 178 /* We don't run reload, so this isn't actually used, but it still needs to be
 179    defined.  Showing an argp->fp elimination also stops
 180    expand_builtin_setjmp_receiver from generating invalid insns.  */
 181 #define ELIMINABLE_REGS                                 \
 182   {                                                     \
 183     { ARG_POINTER_REGNUM, FRAME_POINTER_REGNUM} \
 184   }
 185
 186 /* Define the offset between two registers, one to be eliminated, and the other
 187    its replacement, at the start of a routine.  */
 188
 189 #define INITIAL_ELIMINATION_OFFSET(FROM, TO, OFFSET) \
 190   ((OFFSET) = 0)
 191 \f
 192 /* Addressing Modes.  */
 193
 194 #define MAX_REGS_PER_ADDRESS 1
 195
 196 #define LEGITIMATE_PIC_OPERAND_P(X) 1
 197 \f
 198
 199 #if defined HOST_WIDE_INT
 200 struct GTY(()) machine_function
 201 {
 202   rtx_expr_list *call_args;  /* Arg list for the current call.  */
 203   bool doing_call; /* Within a CALL_ARGS ... CALL_ARGS_END sequence.  */
 204   bool is_varadic;  /* This call is varadic  */
 205   bool has_varadic;  /* Current function has a varadic call.  */
 206   bool has_chain; /* Current function has outgoing static chain.  */
 207   bool has_softstack; /* Current function has a soft stack frame.  */
 208   bool has_simtreg; /* Current function has an OpenMP SIMD region.  */
 209   int num_args; /* Number of args of current call.  */
 210   int return_mode; /* Return mode of current fn.
 211                       (machine_mode not defined yet.) */
 212   rtx axis_predicate[2]; /* Neutering predicates.  */
 213   rtx unisimt_master; /* 'Master lane index' for -muniform-simt.  */
 214   rtx unisimt_predicate; /* Predicate for -muniform-simt.  */
 215   rtx unisimt_location; /* Mask location for -muniform-simt.  */
 216   /* The following two fields hold the maximum size resp. alignment required
 217      for per-lane storage in OpenMP SIMD regions.  */
 218   unsigned HOST_WIDE_INT simt_stack_size;
 219   unsigned HOST_WIDE_INT simt_stack_align;
 220 };
 221 #endif
 222 \f
 223 /* Costs.  */
 224
 225 #define NO_FUNCTION_CSE 1
 226 #define SLOW_BYTE_ACCESS 0
 227 #define BRANCH_COST(speed_p, predictable_p) 6
 228 \f
 229 /* Assembler Format.  */
 230
 231 #undef ASM_DECLARE_FUNCTION_NAME
 232 #define ASM_DECLARE_FUNCTION_NAME(FILE, NAME, DECL)             \
 233   nvptx_declare_function_name (FILE, NAME, DECL)
 234
 235 #undef ASM_DECLARE_FUNCTION_SIZE
 236 #define ASM_DECLARE_FUNCTION_SIZE(STREAM, NAME, DECL) \
 237   nvptx_function_end (STREAM)
 238
 239 #define DWARF2_ASM_LINE_DEBUG_INFO 1
 240
 241 #undef ASM_APP_ON
 242 #define ASM_APP_ON "\t// #APP \n"
 243 #undef ASM_APP_OFF
 244 #define ASM_APP_OFF "\t// #NO_APP \n"
 245
 246 #define DBX_REGISTER_NUMBER(N) N
 247
 248 #define TEXT_SECTION_ASM_OP ""
 249 #define DATA_SECTION_ASM_OP ""
 250
 251 #undef  ASM_GENERATE_INTERNAL_LABEL
 252 #define ASM_GENERATE_INTERNAL_LABEL(LABEL, PREFIX, NUM)         \
 253   do                                                            \
 254     {                                                           \
 255       char *__p;                                                \
 256       __p = stpcpy (&(LABEL)[1], PREFIX);                       \
 257       (LABEL)[0] = '$';                                         \
 258       sprint_ul (__p, (unsigned long) (NUM));                   \
 259     }                                                           \
 260   while (0)
 261
 262 #define ASM_OUTPUT_ALIGN(FILE, POWER)           \
 263   do                                            \
 264     {                                           \
 265       (void) (FILE);                            \
 266       (void) (POWER);                           \
 267     }                                           \
 268   while (0)
 269
 270 #define ASM_OUTPUT_SKIP(FILE, N)                \
 271   nvptx_output_skip (FILE, N)
 272
 273 #undef  ASM_OUTPUT_ASCII
 274 #define ASM_OUTPUT_ASCII(FILE, STR, LENGTH)                     \
 275   nvptx_output_ascii (FILE, STR, LENGTH);
 276
 277 #define ASM_DECLARE_OBJECT_NAME(FILE, NAME, DECL)       \
 278   nvptx_declare_object_name (FILE, NAME, DECL)
 279
 280 #undef  ASM_OUTPUT_ALIGNED_DECL_COMMON
 281 #define ASM_OUTPUT_ALIGNED_DECL_COMMON(FILE, DECL, NAME, SIZE, ALIGN)   \
 282   nvptx_output_aligned_decl (FILE, NAME, DECL, SIZE, ALIGN)
 283
 284 #undef  ASM_OUTPUT_ALIGNED_DECL_LOCAL
 285 #define ASM_OUTPUT_ALIGNED_DECL_LOCAL(FILE, DECL, NAME, SIZE, ALIGN)    \
 286   nvptx_output_aligned_decl (FILE, NAME, DECL, SIZE, ALIGN)
 287
 288 #define CASE_VECTOR_PC_RELATIVE flag_pic
 289 #define JUMP_TABLES_IN_TEXT_SECTION flag_pic
 290
 291 #define ADDR_VEC_ALIGN(VEC) (JUMP_TABLES_IN_TEXT_SECTION ? 5 : 2)
 292
 293 /* Misc.  */
 294
 295 #define DWARF2_LINENO_DEBUGGING_INFO 1
 296
 297 #define CLZ_DEFINED_VALUE_AT_ZERO(MODE, VALUE) \
 298   ((VALUE) = GET_MODE_BITSIZE ((MODE)), 2)
 299 #define CTZ_DEFINED_VALUE_AT_ZERO(MODE, VALUE) \
 300   ((VALUE) = GET_MODE_BITSIZE ((MODE)), 2)
 301
 302 #define SUPPORTS_WEAK 1
 303 #define NO_DOT_IN_LABEL
 304 #define ASM_COMMENT_START "//"
 305
 306 #define STORE_FLAG_VALUE -1
 307 #define FLOAT_STORE_FLAG_VALUE(MODE) REAL_VALUE_ATOF("1.0", (MODE))
 308
 309 #define CASE_VECTOR_MODE SImode
 310 #define MOVE_MAX 8
 311 #define MOVE_RATIO(SPEED) 4
 312 #define FUNCTION_MODE QImode
 313 #define HAS_INIT_SECTION 1
 314
 315 /* The C++ front end insists to link against libstdc++ -- which we don't build.
 316    Tell it to instead link against the innocuous libgcc.  */
 317 #define LIBSTDCXX "gcc"
 318
 319 #endif /* GCC_NVPTX_H */