/******************************** -*- C -*- **************************** * * Support macros for the i386 math coprocessor * ***********************************************************************/ /*********************************************************************** * * Copyright 2000, 2001, 2002, 2004, 2008 Free Software Foundation, Inc. * Written by Paolo Bonzini. * * This file is part of GNU lightning. * * GNU lightning is free software; you can redistribute it and/or modify it * under the terms of the GNU Lesser General Public License as published * by the Free Software Foundation; either version 3, or (at your option) * any later version. * * GNU lightning 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 Lesser General Public * License for more details. * * You should have received a copy of the GNU Lesser General Public License * along with GNU lightning; see the file COPYING.LESSER; if not, write to the * Free Software Foundation, 51 Franklin Street, Fifth Floor, Boston, * MA 02110-1301, USA. * ***********************************************************************/ #ifndef __lightning_fp_h #define __lightning_fp_h /* We really must map the x87 stack onto a flat register file. In practice, we can provide something sensible and make it work on the x86 using the stack like a file of eight registers. We use six or seven registers so as to have some freedom for floor, ceil, round, (and log, tan, atn and exp). Not hard at all, basically play with FXCH. FXCH is mostly free, so the generated code is not bad. Of course we special case when one of the operands turns out to be ST0. Here are the macros that actually do the trick. */ #define JIT_FPR_NUM 6 #define JIT_FPRET 0 #define JIT_FPR(i) (i) #define jit_fxch(rs, op) (((rs) != 0 ? FXCHr(rs) : 0), \ op, ((rs) != 0 ? FXCHr(rs) : 0)) #define jit_fp_unary(rd, s1, op) \ ((rd) == (s1) ? jit_fxch ((rd), op) \ : (rd) == 0 ? (FSTPr (0), FLDr ((s1)-1), op) \ : (FLDr ((s1)), op, FSTPr ((rd) + 1))) #define jit_fp_binary(rd, s1, s2, op, opr) \ ((rd) == (s1) ? \ ((s2) == 0 ? opr(0, (rd)) \ : (s2) == (s1) ? jit_fxch((rd), op(0, 0)) \ : jit_fxch((rd), op((s2), 0))) \ : (rd) == (s2) ? \ ((s1) == 0 ? op(0, (rd)) \ : jit_fxch((rd), opr((s1), 0))) \ : (FLDr (s1), op((s2)+1, 0), FSTPr((rd)+1))) #define jit_addr_d(rd,s1,s2) jit_fp_binary((rd),(s1),(s2),FADDrr,FADDrr) #define jit_subr_d(rd,s1,s2) jit_fp_binary((rd),(s1),(s2),FSUBrr,FSUBRrr) #define jit_mulr_d(rd,s1,s2) jit_fp_binary((rd),(s1),(s2),FMULrr,FMULrr) #define jit_divr_d(rd,s1,s2) jit_fp_binary((rd),(s1),(s2),FDIVrr,FDIVRrr) #define jit_abs_d(rd,rs) jit_fp_unary ((rd), (rs), _OO (0xd9e1)) #define jit_negr_d(rd,rs) jit_fp_unary ((rd), (rs), _OO (0xd9e0)) #define jit_sqrt_d(rd,rs) jit_fp_unary ((rd), (rs), _OO (0xd9fa)) /* - moves: move FPR0 to FPR3 FST ST3 move FPR3 to FPR0 FXCH ST3 FST ST3 move FPR3 to FPR1 FLD ST3 FSTP ST2 Stack is rotated, so FPRn becomes STn+1 */ #define jit_movr_d(rd,s1) \ ((s1) == (rd) ? 0 \ : (s1) == 0 ? FSTr ((rd)) \ : (rd) == 0 ? (FXCHr ((s1)), FSTr ((s1))) \ : (FLDr ((s1)), FSTPr ((rd)+1))) /* - loads: load into FPR0 FSTP ST0 FLD [FUBAR] load into FPR3 FSTP ST3 Save old st0 into destination register FLD [FUBAR] FXCH ST3 Get back old st0 (and similarly for immediates, using the stack) */ #define jit_movi_f(rd,immf) \ (_O (0x68), \ *((float *) _jit.x.pc) = (float) immf, \ _jit.x.uc_pc += sizeof (float), \ jit_ldr_f((rd), _ESP), \ ADDLir(4, _ESP)) union jit_double_imm { double d; int i[2]; }; #define jit_movi_d(rd,immd) \ (_O (0x68), \ _jit.x.uc_pc[4] = 0x68, \ ((union jit_double_imm *) (_jit.x.uc_pc + 5))->d = (double) immd, \ *((int *) _jit.x.uc_pc) = ((union jit_double_imm *) (_jit.x.uc_pc + 5))->i[1], \ _jit.x.uc_pc += 9, \ jit_ldr_d((rd), _ESP), \ ADDLir(8, _ESP)) #define jit_ldi_f(rd, is) \ ((rd) == 0 ? (FSTPr (0), FLDSm((is), 0, 0, 0)) \ : (FLDSm((is), 0, 0, 0), FSTPr ((rd) + 1))) #define jit_ldi_d(rd, is) \ ((rd) == 0 ? (FSTPr (0), FLDLm((is), 0, 0, 0)) \ : (FLDLm((is), 0, 0, 0), FSTPr ((rd) + 1))) #define jit_ldr_f(rd, rs) \ ((rd) == 0 ? (FSTPr (0), FLDSm(0, (rs), 0, 0)) \ : (FLDSm(0, (rs), 0, 0), FSTPr ((rd) + 1))) #define jit_ldr_d(rd, rs) \ ((rd) == 0 ? (FSTPr (0), FLDLm(0, (rs), 0, 0)) \ : (FLDLm(0, (rs), 0, 0), FSTPr ((rd) + 1))) #define jit_ldxi_f(rd, rs, is) \ ((rd) == 0 ? (FSTPr (0), FLDSm((is), (rs), 0, 0)) \ : (FLDSm((is), (rs), 0, 0), FSTPr ((rd) + 1))) #define jit_ldxi_d(rd, rs, is) \ ((rd) == 0 ? (FSTPr (0), FLDLm((is), (rs), 0, 0)) \ : (FLDLm((is), (rs), 0, 0), FSTPr ((rd) + 1))) #define jit_ldxr_f(rd, s1, s2) \ ((rd) == 0 ? (FSTPr (0), FLDSm(0, (s1), (s2), 1)) \ : (FLDSm(0, (s1), (s2), 1), FSTPr ((rd) + 1))) #define jit_ldxr_d(rd, s1, s2) \ ((rd) == 0 ? (FSTPr (0), FLDLm(0, (s1), (s2), 1)) \ : (FLDLm(0, (s1), (s2), 1), FSTPr ((rd) + 1))) #define jit_extr_i_d(rd, rs) (PUSHLr((rs)), \ ((rd) == 0 ? (FSTPr (0), FILDLm(0, _ESP, 0, 0)) \ : (FILDLm(0, _ESP, 0, 0), FSTPr ((rd) + 1))), \ POPLr((rs))) #define jit_stxi_f(id, rd, rs) jit_fxch ((rs), FSTSm((id), (rd), 0, 0)) #define jit_stxr_f(d1, d2, rs) jit_fxch ((rs), FSTSm(0, (d1), (d2), 1)) #define jit_stxi_d(id, rd, rs) jit_fxch ((rs), FSTLm((id), (rd), 0, 0)) #define jit_stxr_d(d1, d2, rs) jit_fxch ((rs), FSTLm(0, (d1), (d2), 1)) #define jit_sti_f(id, rs) jit_fxch ((rs), FSTSm((id), 0, 0, 0)) #define jit_str_f(rd, rs) jit_fxch ((rs), FSTSm(0, (rd), 0, 0)) #define jit_sti_d(id, rs) jit_fxch ((rs), FSTLm((id), 0, 0, 0)) #define jit_str_d(rd, rs) jit_fxch ((rs), FSTLm(0, (rd), 0, 0)) /* ABI */ #define jit_retval_d(rd) FSTPr((rd) + 1) /* Assume round to near mode */ #define jit_floorr_d_i(rd, rs) \ (FLDr (rs), jit_floor2((rd), ((rd) == _EDX ? _EAX : _EDX))) #define jit_ceilr_d_i(rd, rs) \ (FLDr (rs), jit_ceil2((rd), ((rd) == _EDX ? _EAX : _EDX))) #define jit_truncr_d_i(rd, rs) \ (FLDr (rs), jit_trunc2((rd), ((rd) == _EDX ? _EAX : _EDX))) #define jit_calc_diff(ofs) \ FISTLm(ofs, _ESP, 0, 0), \ FILDLm(ofs, _ESP, 0, 0), \ FSUBRPr(1), \ FSTPSm(4+ofs, _ESP, 0, 0) \ /* The real meat */ #define jit_floor2(rd, aux) \ (PUSHLr(aux), \ SUBLir(8, _ESP), \ jit_calc_diff(0), \ POPLr(rd), /* floor in rd */ \ POPLr(aux), /* x-round(x) in aux */ \ ADDLir(0x7FFFFFFF, aux), /* carry if x-round(x) < -0 */ \ SBBLir(0, rd), /* subtract 1 if carry */ \ POPLr(aux)) #define jit_ceil2(rd, aux) \ (PUSHLr(aux), \ SUBLir(8, _ESP), \ jit_calc_diff(0), \ POPLr(rd), /* floor in rd */ \ POPLr(aux), /* x-round(x) in aux */ \ TESTLrr(aux, aux), \ SETGr(jit_reg8(aux)), \ SHRLir(1, aux), \ ADCLir(0, rd), \ POPLr(aux)) /* a mingling of the two above */ #define jit_trunc2(rd, aux) \ (PUSHLr(aux), \ SUBLir(12, _ESP), \ FSTSm(0, _ESP, 0, 0), \ jit_calc_diff(4), \ POPLr(aux), \ POPLr(rd), \ TESTLrr(aux, aux), \ POPLr(aux), \ JSSm(_jit.x.pc + 11), \ ADDLir(0x7FFFFFFF, aux), /* 6 */ \ SBBLir(0, rd), /* 3 */ \ JMPSm(_jit.x.pc + 10), /* 2 */ \ TESTLrr(aux, aux), /* 2 */ \ SETGr(jit_reg8(aux)), /* 3 */ \ SHRLir(1, aux), /* 2 */ \ ADCLir(0, rd), /* 3 */ \ POPLr(aux)) /* the easy one */ #define jit_roundr_d_i(rd, rs) \ (PUSHLr(_EAX), \ jit_fxch ((rs), FISTLm(0, _ESP, 0, 0)), \ POPLr((rd))) #define jit_fp_test(d, s1, s2, n, _and, res) \ (((s1) == 0 ? FUCOMr((s2)) : (FLDr((s1)), FUCOMPr((s2) + 1))), \ ((d) != _EAX ? MOVLrr(_EAX, (d)) : 0), \ FNSTSWr(_EAX), \ SHRLir(n, _EAX), \ ((_and) ? ANDLir((_and), _EAX) : MOVLir(0, _EAX)), \ res, \ ((d) != _EAX ? _O (0x90 + ((d) & 7)) : 0)) /* xchg */ #define jit_fp_btest(d, s1, s2, n, _and, cmp, res) \ (((s1) == 0 ? FUCOMr((s2)) : (FLDr((s1)), FUCOMPr((s2) + 1))), \ PUSHLr(_EAX), \ FNSTSWr(_EAX), \ SHRLir(n, _EAX), \ ((_and) ? ANDLir ((_and), _EAX) : 0), \ ((cmp) ? CMPLir ((cmp), _EAX) : 0), \ POPLr(_EAX), \ res ((d)), \ _jit.x.pc) #define jit_nothing_needed(x) /* After FNSTSW we have 1 if <, 40 if =, 0 if >, 45 if unordered. Here is how to map the values of the status word's high byte to the conditions. < = > unord valid values condition gt no no yes no 0 STSW & 45 == 0 lt yes no no no 1 STSW & 45 == 1 eq no yes no no 40 STSW & 45 == 40 unord no no no yes 45 bit 2 == 1 ge no yes no no 0, 40 bit 0 == 0 unlt yes no no yes 1, 45 bit 0 == 1 ltgt yes no yes no 0, 1 bit 6 == 0 uneq no yes no yes 40, 45 bit 6 == 1 le yes yes no no 1, 40 odd parity for STSW & 41 ungt no no yes yes 0, 45 even parity for STSW & 41 unle yes yes no yes 1, 40, 45 STSW & 45 != 0 unge no yes yes yes 0, 40, 45 STSW & 45 != 1 ne yes no yes yes 0, 1, 45 STSW & 45 != 40 ord yes yes yes no 0, 1, 40 bit 2 == 0 lt, le, ungt, unge are actually computed as gt, ge, unlt, unle with the operands swapped; it is more efficient this way. */ #define jit_gtr_d(d, s1, s2) jit_fp_test((d), (s1), (s2), 8, 0x45, SETZr (_AL)) #define jit_ger_d(d, s1, s2) jit_fp_test((d), (s1), (s2), 9, 0, SBBBir (-1, _AL)) #define jit_unler_d(d, s1, s2) jit_fp_test((d), (s1), (s2), 8, 0x45, SETNZr (_AL)) #define jit_unltr_d(d, s1, s2) jit_fp_test((d), (s1), (s2), 9, 0, ADCBir (0, _AL)) #define jit_ltr_d(d, s1, s2) jit_fp_test((d), (s2), (s1), 8, 0x45, SETZr (_AL)) #define jit_ler_d(d, s1, s2) jit_fp_test((d), (s2), (s1), 9, 0, SBBBir (-1, _AL)) #define jit_unger_d(d, s1, s2) jit_fp_test((d), (s2), (s1), 8, 0x45, SETNZr (_AL)) #define jit_ungtr_d(d, s1, s2) jit_fp_test((d), (s2), (s1), 9, 0, ADCBir (0, _AL)) #define jit_eqr_d(d, s1, s2) jit_fp_test((d), (s1), (s2), 8, 0x45, (CMPBir (0x40, _AL), SETEr (_AL))) #define jit_ner_d(d, s1, s2) jit_fp_test((d), (s1), (s2), 8, 0x45, (CMPBir (0x40, _AL), SETNEr (_AL))) #define jit_ltgtr_d(d, s1, s2) jit_fp_test((d), (s1), (s2), 15, 0, SBBBir (-1, _AL)) #define jit_uneqr_d(d, s1, s2) jit_fp_test((d), (s1), (s2), 15, 0, ADCBir (0, _AL)) #define jit_ordr_d(d, s1, s2) jit_fp_test((d), (s1), (s2), 11, 0, SBBBir (-1, _AL)) #define jit_unordr_d(d, s1, s2) jit_fp_test((d), (s1), (s2), 11, 0, ADCBir (0, _AL)) #define jit_bgtr_d(d, s1, s2) jit_fp_btest((d), (s1), (s2), 8, 0x45, 0, JZm) #define jit_bger_d(d, s1, s2) jit_fp_btest((d), (s1), (s2), 9, 0, 0, JNCm) #define jit_bunler_d(d, s1, s2) jit_fp_btest((d), (s1), (s2), 8, 0x45, 0, JNZm) #define jit_bunltr_d(d, s1, s2) jit_fp_btest((d), (s1), (s2), 9, 0, 0, JCm) #define jit_bltr_d(d, s1, s2) jit_fp_btest((d), (s2), (s1), 8, 0x45, 0, JZm) #define jit_bler_d(d, s1, s2) jit_fp_btest((d), (s2), (s1), 9, 0, 0, JNCm) #define jit_bunger_d(d, s1, s2) jit_fp_btest((d), (s2), (s1), 8, 0x45, 0, JNZm) #define jit_bungtr_d(d, s1, s2) jit_fp_btest((d), (s2), (s1), 9, 0, 0, JCm) #define jit_beqr_d(d, s1, s2) jit_fp_btest((d), (s1), (s2), 8, 0x45, 0x40, JZm) #define jit_bner_d(d, s1, s2) jit_fp_btest((d), (s1), (s2), 8, 0x45, 0x40, JNZm) #define jit_bltgtr_d(d, s1, s2) jit_fp_btest((d), (s1), (s2), 15, 0, 0, JNCm) #define jit_buneqr_d(d, s1, s2) jit_fp_btest((d), (s1), (s2), 15, 0, 0, JCm) #define jit_bordr_d(d, s1, s2) jit_fp_btest((d), (s1), (s2), 11, 0, 0, JNCm) #define jit_bunordr_d(d, s1, s2) jit_fp_btest((d), (s1), (s2), 11, 0, 0, JCm) #define jit_pusharg_d(rs) (jit_subi_i(JIT_SP,JIT_SP,sizeof(double)), jit_str_d(JIT_SP,(rs))) #define jit_pusharg_f(rs) (jit_subi_i(JIT_SP,JIT_SP,sizeof(float)), jit_str_f(JIT_SP,(rs))) #if 0 #define jit_sin() _OO(0xd9fe) /* fsin */ #define jit_cos() _OO(0xd9ff) /* fcos */ #define jit_tan() (_OO(0xd9f2), /* fptan */ \ FSTPr(0)) /* fstp st */ #define jit_atn() (_OO(0xd9e8), /* fld1 */ \ _OO(0xd9f3)) /* fpatan */ #define jit_exp() (_OO(0xd9ea), /* fldl2e */ \ FMULPr(1), /* fmulp */ \ _OO(0xd9c0), /* fld st */ \ _OO(0xd9fc), /* frndint */ \ _OO(0xdce9), /* fsubr */ \ FXCHr(1), /* fxch st(1) */ \ _OO(0xd9f0), /* f2xm1 */ \ _OO(0xd9e8), /* fld1 */ \ _OO(0xdec1), /* faddp */ \ _OO(0xd9fd), /* fscale */ \ FSTPr(1)) /* fstp st(1) */ #define jit_log() (_OO(0xd9ed), /* fldln2 */ \ FXCHr(1), /* fxch st(1) */ \ _OO(0xd9f1)) /* fyl2x */ #endif #define jit_prepare_f(nf) (_jitl.argssize += (nf)) #define jit_prepare_d(nd) (_jitl.argssize += 2 * (nd)) #define jit_arg_f() ((_jitl.framesize += sizeof(float)) - sizeof(float)) #define jit_arg_d() ((_jitl.framesize += sizeof(double)) - sizeof(double)) #endif /* __lightning_fp_h */