fp.isa revision 5568
1// -*- mode:c++ -*- 2 3// Copyright (c) 2003-2005 The Regents of The University of Michigan 4// All rights reserved. 5// 6// Redistribution and use in source and binary forms, with or without 7// modification, are permitted provided that the following conditions are 8// met: redistributions of source code must retain the above copyright 9// notice, this list of conditions and the following disclaimer; 10// redistributions in binary form must reproduce the above copyright 11// notice, this list of conditions and the following disclaimer in the 12// documentation and/or other materials provided with the distribution; 13// neither the name of the copyright holders nor the names of its 14// contributors may be used to endorse or promote products derived from 15// this software without specific prior written permission. 16// 17// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS 18// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT 19// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR 20// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT 21// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, 22// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT 23// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 24// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 25// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 26// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE 27// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 28// 29// Authors: Steve Reinhardt 30 31//////////////////////////////////////////////////////////////////// 32// 33// Floating-point instructions 34// 35// Note that many FP-type instructions which do not support all the 36// various rounding & trapping modes use the simpler format 37// BasicOperateWithNopCheck. 38// 39 40output exec {{ 41 /// Check "FP enabled" machine status bit. Called when executing any FP 42 /// instruction in full-system mode. 43 /// @retval Full-system mode: NoFault if FP is enabled, FenFault 44 /// if not. Non-full-system mode: always returns NoFault. 45#if FULL_SYSTEM 46 inline Fault checkFpEnableFault(%(CPU_exec_context)s *xc) 47 { 48 Fault fault = NoFault; // dummy... this ipr access should not fault 49 if (!ICSR_FPE(xc->readMiscReg(IPR_ICSR))) { 50 fault = new FloatEnableFault; 51 } 52 return fault; 53 } 54#else 55 inline Fault checkFpEnableFault(%(CPU_exec_context)s *xc) 56 { 57 return NoFault; 58 } 59#endif 60}}; 61 62output header {{ 63 /** 64 * Base class for general floating-point instructions. Includes 65 * support for various Alpha rounding and trapping modes. Only FP 66 * instructions that require this support are derived from this 67 * class; the rest derive directly from AlphaStaticInst. 68 */ 69 class AlphaFP : public AlphaStaticInst 70 { 71 public: 72 /// Alpha FP rounding modes. 73 enum RoundingMode { 74 Chopped = 0, ///< round toward zero 75 Minus_Infinity = 1, ///< round toward minus infinity 76 Normal = 2, ///< round to nearest (default) 77 Dynamic = 3, ///< use FPCR setting (in instruction) 78 Plus_Infinity = 3 ///< round to plus inifinity (in FPCR) 79 }; 80 81 /// Alpha FP trapping modes. 82 /// For instructions that produce integer results, the 83 /// "Underflow Enable" modes really mean "Overflow Enable", and 84 /// the assembly modifier is V rather than U. 85 enum TrappingMode { 86 /// default: nothing enabled 87 Imprecise = 0, ///< no modifier 88 /// underflow/overflow traps enabled, inexact disabled 89 Underflow_Imprecise = 1, ///< /U or /V 90 Underflow_Precise = 5, ///< /SU or /SV 91 /// underflow/overflow and inexact traps enabled 92 Underflow_Inexact_Precise = 7 ///< /SUI or /SVI 93 }; 94 95 protected: 96 /// Map Alpha rounding mode to C99 constants from <fenv.h>. 97 static const int alphaToC99RoundingMode[]; 98 99 /// Map enum RoundingMode values to disassembly suffixes. 100 static const char *roundingModeSuffix[]; 101 /// Map enum TrappingMode values to FP disassembly suffixes. 102 static const char *fpTrappingModeSuffix[]; 103 /// Map enum TrappingMode values to integer disassembly suffixes. 104 static const char *intTrappingModeSuffix[]; 105 106 /// This instruction's rounding mode. 107 RoundingMode roundingMode; 108 /// This instruction's trapping mode. 109 TrappingMode trappingMode; 110 111 /// Have we warned about this instruction's unsupported 112 /// rounding mode (if applicable)? 113 mutable bool warnedOnRounding; 114 115 /// Have we warned about this instruction's unsupported 116 /// trapping mode (if applicable)? 117 mutable bool warnedOnTrapping; 118 119 /// Constructor 120 AlphaFP(const char *mnem, ExtMachInst _machInst, OpClass __opClass) 121 : AlphaStaticInst(mnem, _machInst, __opClass), 122 roundingMode((enum RoundingMode)FP_ROUNDMODE), 123 trappingMode((enum TrappingMode)FP_TRAPMODE), 124 warnedOnRounding(false), 125 warnedOnTrapping(false) 126 { 127 } 128 129 int getC99RoundingMode(uint64_t fpcr_val) const; 130 131 // This differs from the AlphaStaticInst version only in 132 // printing suffixes for non-default rounding & trapping modes. 133 std::string 134 generateDisassembly(Addr pc, const SymbolTable *symtab) const; 135 }; 136 137}}; 138 139 140output decoder {{ 141 int 142 AlphaFP::getC99RoundingMode(uint64_t fpcr_val) const 143 { 144 if (roundingMode == Dynamic) { 145 return alphaToC99RoundingMode[bits(fpcr_val, 59, 58)]; 146 } 147 else { 148 return alphaToC99RoundingMode[roundingMode]; 149 } 150 } 151 152 std::string 153 AlphaFP::generateDisassembly(Addr pc, const SymbolTable *symtab) const 154 { 155 std::string mnem_str(mnemonic); 156 157#ifndef SS_COMPATIBLE_DISASSEMBLY 158 std::string suffix(""); 159 suffix += ((_destRegIdx[0] >= FP_Base_DepTag) 160 ? fpTrappingModeSuffix[trappingMode] 161 : intTrappingModeSuffix[trappingMode]); 162 suffix += roundingModeSuffix[roundingMode]; 163 164 if (suffix != "") { 165 mnem_str = csprintf("%s/%s", mnemonic, suffix); 166 } 167#endif 168 169 std::stringstream ss; 170 ccprintf(ss, "%-10s ", mnem_str.c_str()); 171 172 // just print the first two source regs... if there's 173 // a third one, it's a read-modify-write dest (Rc), 174 // e.g. for CMOVxx 175 if (_numSrcRegs > 0) { 176 printReg(ss, _srcRegIdx[0]); 177 } 178 if (_numSrcRegs > 1) { 179 ss << ","; 180 printReg(ss, _srcRegIdx[1]); 181 } 182 183 // just print the first dest... if there's a second one, 184 // it's generally implicit 185 if (_numDestRegs > 0) { 186 if (_numSrcRegs > 0) 187 ss << ","; 188 printReg(ss, _destRegIdx[0]); 189 } 190 191 return ss.str(); 192 } 193 194 const int AlphaFP::alphaToC99RoundingMode[] = { 195 M5_FE_TOWARDZERO, // Chopped 196 M5_FE_DOWNWARD, // Minus_Infinity 197 M5_FE_TONEAREST, // Normal 198 M5_FE_UPWARD // Dynamic in inst, Plus_Infinity in FPCR 199 }; 200 201 const char *AlphaFP::roundingModeSuffix[] = { "c", "m", "", "d" }; 202 // mark invalid trapping modes, but don't fail on them, because 203 // you could decode anything on a misspeculated path 204 const char *AlphaFP::fpTrappingModeSuffix[] = 205 { "", "u", "INVTM2", "INVTM3", "INVTM4", "su", "INVTM6", "sui" }; 206 const char *AlphaFP::intTrappingModeSuffix[] = 207 { "", "v", "INVTM2", "INVTM3", "INVTM4", "sv", "INVTM6", "svi" }; 208}}; 209 210// FP instruction class execute method template. Handles non-standard 211// rounding modes. 212def template FloatingPointExecute {{ 213 Fault %(class_name)s::execute(%(CPU_exec_context)s *xc, 214 Trace::InstRecord *traceData) const 215 { 216 if (trappingMode != Imprecise && !warnedOnTrapping) { 217 warn("%s: non-standard trapping mode not supported", 218 generateDisassembly(0, NULL)); 219 warnedOnTrapping = true; 220 } 221 222 Fault fault = NoFault; 223 224 %(fp_enable_check)s; 225 %(op_decl)s; 226 %(op_rd)s; 227#if USE_FENV 228 if (roundingMode == Normal) { 229 %(code)s; 230 } else { 231 m5_fesetround(getC99RoundingMode( 232 xc->readMiscRegNoEffect(MISCREG_FPCR))); 233 %(code)s; 234 m5_fesetround(M5_FE_TONEAREST); 235 } 236#else 237 if (roundingMode != Normal && !warnedOnRounding) { 238 warn("%s: non-standard rounding mode not supported", 239 generateDisassembly(0, NULL)); 240 warnedOnRounding = true; 241 } 242 %(code)s; 243#endif 244 245 if (fault == NoFault) { 246 %(op_wb)s; 247 } 248 249 return fault; 250 } 251}}; 252 253// FP instruction class execute method template where no dynamic 254// rounding mode control is needed. Like BasicExecute, but includes 255// check & warning for non-standard trapping mode. 256def template FPFixedRoundingExecute {{ 257 Fault %(class_name)s::execute(%(CPU_exec_context)s *xc, 258 Trace::InstRecord *traceData) const 259 { 260 if (trappingMode != Imprecise && !warnedOnTrapping) { 261 warn("%s: non-standard trapping mode not supported", 262 generateDisassembly(0, NULL)); 263 warnedOnTrapping = true; 264 } 265 266 Fault fault = NoFault; 267 268 %(fp_enable_check)s; 269 %(op_decl)s; 270 %(op_rd)s; 271 %(code)s; 272 273 if (fault == NoFault) { 274 %(op_wb)s; 275 } 276 277 return fault; 278 } 279}}; 280 281def template FloatingPointDecode {{ 282 { 283 AlphaStaticInst *i = new %(class_name)s(machInst); 284 if (FC == 31) { 285 i = makeNop(i); 286 } 287 return i; 288 } 289}}; 290 291// General format for floating-point operate instructions: 292// - Checks trapping and rounding mode flags. Trapping modes 293// currently unimplemented (will fail). 294// - Generates NOP if FC == 31. 295def format FloatingPointOperate(code, *opt_args) {{ 296 iop = InstObjParams(name, Name, 'AlphaFP', code, opt_args) 297 decode_block = FloatingPointDecode.subst(iop) 298 header_output = BasicDeclare.subst(iop) 299 decoder_output = BasicConstructor.subst(iop) 300 exec_output = FloatingPointExecute.subst(iop) 301}}; 302 303// Special format for cvttq where rounding mode is pre-decoded 304def format FPFixedRounding(code, class_suffix, *opt_args) {{ 305 Name += class_suffix 306 iop = InstObjParams(name, Name, 'AlphaFP', code, opt_args) 307 decode_block = FloatingPointDecode.subst(iop) 308 header_output = BasicDeclare.subst(iop) 309 decoder_output = BasicConstructor.subst(iop) 310 exec_output = FPFixedRoundingExecute.subst(iop) 311}}; 312 313