(MIPS_MACH_HAS_MULT_HILO_HAZARD, MIPS_MACH_HAS_DIV_HILO_HAZARD): New.
* mips.igen (check_mt_hilo, check_mult_hilo, check_div_hilo): Provide
separate implementations for mipsIV and mipsV. Use new macros to
determine whether the restrictions apply.
* mips.igen (check_mf_cycles, check_mt_hilo, check_mf_hilo)
(check_mult_hilo): Improve comments.
(check_div_hilo): Likewise. Also, fork off a new version
to handle mips32/mips64 (since there are no hazards to check
in MIPS32/MIPS64).
* mips.igen (do_load_double, do_store_double): New functions.
(LDC1, SDC1): Rename to...
(LDC1b, SDC1b): respectively.
(LDC1a, SDC1a): New instructions for MIPS II and MIPS32 support.
Ed Satterthwaite <ehs@broadcom.com>
* mips3d.igen: New file which contains MIPS-3D ASE instructions.
* Makefile.in (IGEN_INCLUDE): Add mips3d.igen.
* mips.igen: Include mips3d.igen.
(mips3d): New model name for MIPS-3D ASE instructions.
(CVT.W.fmt): Don't use this instruction for word (source) format
instructions.
* cp1.c (fp_binary_r, fp_add_r, fp_mul_r, fpu_inv1, fpu_inv1_32)
(fpu_inv1_64, fp_recip1, fp_recip2, fpu_inv_sqrt1, fpu_inv_sqrt1_32)
(fpu_inv_sqrt1_64, fp_rsqrt1, fp_rsqrt2): New functions.
(NR_FRAC_GUARD, IMPLICIT_1): New macros.
* sim-main.h (fmt_pw, CompareAbs, AddR, MultiplyR, Recip1, Recip2)
(RSquareRoot1, RSquareRoot2): New macros.
(fp_add_r, fp_mul_r, fp_recip1, fp_recip2, fp_rsqrt1)
(fp_rsqrt2): New functions.
* configure.in: Add MIPS-3D support to mipsisa64 simulator.
* configure: Regenerate.
* cp1.c (FP_PS_upper, FP_PS_lower, FP_PS_cat, FPQNaN_PS): New macros.
(value_fpr, store_fpr, fp_cmp, fp_unary, fp_binary, fp_mac)
(fp_inv_sqrt, fpu_format_name): Add paired-single support.
(convert): Note that this function is not used for paired-single
format conversions.
(ps_lower, ps_upper, pack_ps, convert_ps): New functions.
* mips.igen (FMT, MOVtf.fmt): Add paired-single support.
(check_fmt_p): Enable paired-single support.
(ALNV.PS, CVT.PS.S, CVT.S.PL, CVT.S.PU, PLL.PS, PLU.PS, PUL.PS)
(PUU.PS): New instructions.
(CVT.S.fmt): Don't use this instruction for paired-single format
destinations.
* sim-main.h (FP_formats): New value 'fmt_ps.'
(ps_lower, ps_upper, pack_ps, convert_ps): New prototypes.
(PSLower, PSUpper, PackPS, ConvertPS): New macros.
* mips.igen (MOVN, MOVZ): Trace result.
(TNEI): Print "tnei" as the opcode name in traces.
(CEIL.W): Add disassembly string for traces.
(RSQRT.fmt): Make location of disassembly string consistent
with other instructions.
* sim-main.h (FGRIDX): Remove, replace all uses with...
(FGR_BASE): New macro.
(FP0_REGNUM, FCRCS_REGNUM, FCRIR_REGNUM): New macros.
(_sim_cpu): Move 'fgr' member to be right before 'fpr_state' member.
(NR_FGR, FGR): Likewise.
* interp.c: Replace all uses of FGRIDX with FGR_BASE.
* mips.igen: Likewise.
Ed Satterthwaite <ehs@broadcom.com>
* configure.in (mipsisa64sb1*-*-*): New target for supporting
Broadcom SiByte SB-1 processor configurations.
* configure: Regenerate.
* sb1.igen: New file.
* mips.igen: Include sb1.igen.
(sb1): New model.
* Makefile.in (IGEN_INCLUDE): Add sb1.igen.
* mdmx.igen: Add "sb1" model to all appropriate functions and
instructions.
* mdmx.c (AbsDiffOB, AvgOB, AccAbsDiffOB): New functions.
(ob_func, ob_acc): Reference the above.
(qh_acc): Adjust to keep the same size as ob_acc.
* sim-main.h (status_SBX, MX_VECT_ABSD, MX_VECT_AVG, MX_AbsDiff)
(MX_Avg, MX_VECT_ABSDA, MX_AbsDiffC): New macros.
* cp1.c (store_fpr): Remove #ifdef'd out call to UndefinedResult
which wouldn't compile anyway.
* sim-main.h (unpredictable_action): New function prototype.
(Unpredictable): Define to call igen function unpredictable().
(NotWordValue): New macro to call igen function not_word_value().
(UndefinedResult): Remove.
* interp.c (undefined_result): Remove.
(unpredictable_action): New function.
* mips.igen (not_word_value, unpredictable): New functions.
(ADD, ADDI, do_addiu, do_addu, BGEZAL, BGEZALL, BLTZAL, BLTZALL)
(CLO, CLZ, MADD, MADDU, MSUB, MSUBU, MUL, do_mult, do_multu)
(do_sra, do_srav, do_srl, do_srlv, SUB, do_subu): Invoke
NotWordValue() to check for unpredictable inputs, then
Unpredictable() to handle them.
* configure.in (mipsisa32*-*-*, mipsisa64*-*-*): New targets.
* mips.igen (mips32, mips64): New models, add to all instructions
and functions as appropriate.
(loadstore_ea, check_u64): New variant for model mips64.
(check_fmt_p): New variant for models mipsV and mips64, remove
mipsV model marking fro other variant.
(SLL) Rename to...
(SLLa) this.
(CLO, CLZ, MADD, MADDU, MSUB, MSUBU, MUL, SLLb): New instructions
for mips32 and mips64.
(DCLO, DCLZ): New instructions for mips64.
* mips.igen (BREAK, LUI, ORI, SYSCALL, XORI): Print
immediate or code as a hex value with the "%#lx" format.
(ANDI): Likewise, and fix printed instruction name.
* mips.igen (check_fpu): Enable check for coprocessor 1 usability.
* sim-main.h (COP_Usable): Define, but for now coprocessor 1
is always enabled.
(SignalExceptionCoProcessorUnusable): Take as argument the
unusable coprocessor number.
* mips.igen (check_fmt, check_fmt_p): New functions to check
whether specific floating point formats are usable.
(ABS.fmt, ADD.fmt, CEIL.L.fmt, CEIL.W, DIV.fmt, FLOOR.L.fmt)
(FLOOR.W.fmt, MOV.fmt, MUL.fmt, NEG.fmt, RECIP.fmt, ROUND.L.fmt)
(ROUND.W.fmt, RSQRT.fmt, SQRT.fmt, SUB.fmt, TRUNC.L.fmt, TRUNC.W):
Use the new functions.
(do_c_cond_fmt): Remove format checks...
(C.cond.fmta, C.cond.fmtb): And move them into all callers.
* mips.igen (loadstore_ea): New function to do effective
address calculations.
(do_load, do_load_left, do_load_right, LL, LDD, PREF, do_store,
do_store_left, do_store_right, SC, SCD, PREFX, SWC1, SWXC1,
CACHE): Use loadstore_ea to do effective address computations.
* mips.igen (PREFX): Tweak instruction opcode fields (i.e.,
add a comma) so that it more closely match the MIPS ISA
documentation opcode partitioning.
(PREF): Put useful names on opcode fields, and include
instruction-printing string.
* mips.igen (check_u64): New function which in the future will
check whether 64-bit instructions are usable and signal an
exception if not. Currently a no-op.
(DADD, DADDI, DADDIU, DADDU, DDIV, DDIVU, DMULT, DMULTU, DSLL,
DSLL32, DSLLV, DSRA, DSRA32, DSRAV, DSRL, DSRL32, DSRLV, DSUB,
DSUBU, LD, LDL, LDR, LLD, LWU, SCD, SD, SDL, SDR, DMxC1, LDXC1,
LWXC1, SDXC1, SWXC1, DMFC0, DMTC0): Use check_u64.
* mips.igen (check_fpu): New function which in the future will
check whether FPU instructions are usable and signal an exception
if not. Currently a no-op.
(ABS.fmt, ADD.fmt, BC1a, BC1b, C.cond.fmta, C.cond.fmtb,
CEIL.L.fmt, CEIL.W, CxC1, CVT.D.fmt, CVT.L.fmt, CVT.S.fmt,
CVT.W.fmt, DIV.fmt, DMxC1, DMxC1, FLOOR.L.fmt, FLOOR.W.fmt, LDC1,
LDXC1, LWC1, LWXC1, MADD.D, MADD.S, MxC1, MOV.fmt, MOVtf,
MOVtf.fmt, MOVN.fmt, MOVZ.fmt, MSUB.D, MSUB.S, MUL.fmt, NEG.fmt,
NMADD.D, NMADD.S, NMSUB.D, NMSUB.S, RECIP.fmt, ROUND.L.fmt,
ROUND.W.fmt, RSQRT.fmt, SDC1, SDXC1, SQRT.fmt, SUB.fmt, SWC1,
SWXC1, TRUNC.L.fmt, TRUNC.W): Use check_fpu.
* mips.igen (do_load_left, do_load_right): Move to be immediately
following do_load.
(do_store_left, do_store_right): Move to be immediately following
do_store.
* mips.igen: Add some additional comments about supported
models, and about which instructions go where.
(BC1b, MFC0, MTC0, RFE): Sort supported models in the same
order as is used in the rest of the file.
* mips.igen (ADD, ADDI, DADDI, DSUB, SUB): Add comment
indicating that ALU32_END or ALU64_END are there to check
for overflow.
(DADD): Likewise, but also remove previous comment about
overflow checking.
* mips.igen (ADDI): Print immediate value.
(BREAK): Print code.
(DADDIU, DSRAV, DSRLV): Print correct instruction name.
(SLL): Print "nop" specially, and don't run the code
that does the shift for the "nop" case.
