diff options
author | Linus Torvalds <torvalds@ppc970.osdl.org> | 2005-04-16 15:20:36 -0700 |
---|---|---|
committer | Linus Torvalds <torvalds@ppc970.osdl.org> | 2005-04-16 15:20:36 -0700 |
commit | 1da177e4c3f41524e886b7f1b8a0c1fc7321cac2 (patch) | |
tree | 0bba044c4ce775e45a88a51686b5d9f90697ea9d /include/asm-m68k/math-emu.h | |
download | linux-1da177e4c3f41524e886b7f1b8a0c1fc7321cac2.tar.bz2 |
Linux-2.6.12-rc2v2.6.12-rc2
Initial git repository build. I'm not bothering with the full history,
even though we have it. We can create a separate "historical" git
archive of that later if we want to, and in the meantime it's about
3.2GB when imported into git - space that would just make the early
git days unnecessarily complicated, when we don't have a lot of good
infrastructure for it.
Let it rip!
Diffstat (limited to 'include/asm-m68k/math-emu.h')
-rw-r--r-- | include/asm-m68k/math-emu.h | 300 |
1 files changed, 300 insertions, 0 deletions
diff --git a/include/asm-m68k/math-emu.h b/include/asm-m68k/math-emu.h new file mode 100644 index 000000000000..7ac6259b68df --- /dev/null +++ b/include/asm-m68k/math-emu.h @@ -0,0 +1,300 @@ +#ifndef _ASM_M68K_SETUP_H +#define _ASM_M68K_SETUP_H + +#include <asm/setup.h> +#include <linux/linkage.h> + +/* Status Register bits */ + +/* accrued exception bits */ +#define FPSR_AEXC_INEX 3 +#define FPSR_AEXC_DZ 4 +#define FPSR_AEXC_UNFL 5 +#define FPSR_AEXC_OVFL 6 +#define FPSR_AEXC_IOP 7 + +/* exception status bits */ +#define FPSR_EXC_INEX1 8 +#define FPSR_EXC_INEX2 9 +#define FPSR_EXC_DZ 10 +#define FPSR_EXC_UNFL 11 +#define FPSR_EXC_OVFL 12 +#define FPSR_EXC_OPERR 13 +#define FPSR_EXC_SNAN 14 +#define FPSR_EXC_BSUN 15 + +/* quotient byte, assumes big-endian, of course */ +#define FPSR_QUOTIENT(fpsr) (*((signed char *) &(fpsr) + 1)) + +/* condition code bits */ +#define FPSR_CC_NAN 24 +#define FPSR_CC_INF 25 +#define FPSR_CC_Z 26 +#define FPSR_CC_NEG 27 + + +/* Control register bits */ + +/* rounding mode */ +#define FPCR_ROUND_RN 0 /* round to nearest/even */ +#define FPCR_ROUND_RZ 1 /* round to zero */ +#define FPCR_ROUND_RM 2 /* minus infinity */ +#define FPCR_ROUND_RP 3 /* plus infinity */ + +/* rounding precision */ +#define FPCR_PRECISION_X 0 /* long double */ +#define FPCR_PRECISION_S 1 /* double */ +#define FPCR_PRECISION_D 2 /* float */ + + +/* Flags to select the debugging output */ +#define PDECODE 0 +#define PEXECUTE 1 +#define PCONV 2 +#define PNORM 3 +#define PREGISTER 4 +#define PINSTR 5 +#define PUNIMPL 6 +#define PMOVEM 7 + +#define PMDECODE (1<<PDECODE) +#define PMEXECUTE (1<<PEXECUTE) +#define PMCONV (1<<PCONV) +#define PMNORM (1<<PNORM) +#define PMREGISTER (1<<PREGISTER) +#define PMINSTR (1<<PINSTR) +#define PMUNIMPL (1<<PUNIMPL) +#define PMMOVEM (1<<PMOVEM) + +#ifndef __ASSEMBLY__ + +#include <linux/kernel.h> +#include <linux/sched.h> + +union fp_mant64 { + unsigned long long m64; + unsigned long m32[2]; +}; + +union fp_mant128 { + unsigned long long m64[2]; + unsigned long m32[4]; +}; + +/* internal representation of extended fp numbers */ +struct fp_ext { + unsigned char lowmant; + unsigned char sign; + unsigned short exp; + union fp_mant64 mant; +}; + +/* C representation of FPU registers */ +/* NOTE: if you change this, you have to change the assembler offsets + below and the size in <asm/fpu.h>, too */ +struct fp_data { + struct fp_ext fpreg[8]; + unsigned int fpcr; + unsigned int fpsr; + unsigned int fpiar; + unsigned short prec; + unsigned short rnd; + struct fp_ext temp[2]; +}; + +#ifdef FPU_EMU_DEBUG +extern unsigned int fp_debugprint; + +#define dprint(bit, fmt, args...) ({ \ + if (fp_debugprint & (1 << (bit))) \ + printk(fmt, ## args); \ +}) +#else +#define dprint(bit, fmt, args...) +#endif + +#define uprint(str) ({ \ + static int __count = 3; \ + \ + if (__count > 0) { \ + printk("You just hit an unimplemented " \ + "fpu instruction (%s)\n", str); \ + printk("Please report this to ....\n"); \ + __count--; \ + } \ +}) + +#define FPDATA ((struct fp_data *)current->thread.fp) + +#else /* __ASSEMBLY__ */ + +#define FPDATA %a2 + +/* offsets from the base register to the floating point data in the task struct */ +#define FPD_FPREG (TASK_THREAD+THREAD_FPREG+0) +#define FPD_FPCR (TASK_THREAD+THREAD_FPREG+96) +#define FPD_FPSR (TASK_THREAD+THREAD_FPREG+100) +#define FPD_FPIAR (TASK_THREAD+THREAD_FPREG+104) +#define FPD_PREC (TASK_THREAD+THREAD_FPREG+108) +#define FPD_RND (TASK_THREAD+THREAD_FPREG+110) +#define FPD_TEMPFP1 (TASK_THREAD+THREAD_FPREG+112) +#define FPD_TEMPFP2 (TASK_THREAD+THREAD_FPREG+124) +#define FPD_SIZEOF (TASK_THREAD+THREAD_FPREG+136) + +/* offsets on the stack to access saved registers, + * these are only used during instruction decoding + * where we always know how deep we're on the stack. + */ +#define FPS_DO (PT_D0) +#define FPS_D1 (PT_D1) +#define FPS_D2 (PT_D2) +#define FPS_A0 (PT_A0) +#define FPS_A1 (PT_A1) +#define FPS_A2 (PT_A2) +#define FPS_SR (PT_SR) +#define FPS_PC (PT_PC) +#define FPS_EA (PT_PC+6) +#define FPS_PC2 (PT_PC+10) + +.macro fp_get_fp_reg + lea (FPD_FPREG,FPDATA,%d0.w*4),%a0 + lea (%a0,%d0.w*8),%a0 +.endm + +/* Macros used to get/put the current program counter. + * 020/030 use a different stack frame then 040/060, for the + * 040/060 the return pc points already to the next location, + * so this only needs to be modified for jump instructions. + */ +.macro fp_get_pc dest + move.l (FPS_PC+4,%sp),\dest +.endm + +.macro fp_put_pc src,jump=0 + move.l \src,(FPS_PC+4,%sp) +.endm + +.macro fp_get_instr_data f,s,dest,label + getuser \f,%sp@(FPS_PC+4)@(0),\dest,\label,%sp@(FPS_PC+4) + addq.l #\s,%sp@(FPS_PC+4) +.endm + +.macro fp_get_instr_word dest,label,addr + fp_get_instr_data w,2,\dest,\label,\addr +.endm + +.macro fp_get_instr_long dest,label,addr + fp_get_instr_data l,4,\dest,\label,\addr +.endm + +/* These macros are used to read from/write to user space + * on error we jump to the fixup section, load the fault + * address into %a0 and jump to the exit. + * (derived from <asm/uaccess.h>) + */ +.macro getuser size,src,dest,label,addr +| printf ,"[\size<%08x]",1,\addr +.Lu1\@: moves\size \src,\dest + + .section .fixup,"ax" + .even +.Lu2\@: move.l \addr,%a0 + jra \label + .previous + + .section __ex_table,"a" + .align 4 + .long .Lu1\@,.Lu2\@ + .previous +.endm + +.macro putuser size,src,dest,label,addr +| printf ,"[\size>%08x]",1,\addr +.Lu1\@: moves\size \src,\dest +.Lu2\@: + + .section .fixup,"ax" + .even +.Lu3\@: move.l \addr,%a0 + jra \label + .previous + + .section __ex_table,"a" + .align 4 + .long .Lu1\@,.Lu3\@ + .long .Lu2\@,.Lu3\@ + .previous +.endm + + +.macro movestack nr,arg1,arg2,arg3,arg4,arg5 + .if \nr + movestack (\nr-1),\arg2,\arg3,\arg4,\arg5 + move.l \arg1,-(%sp) + .endif +.endm + +.macro printf bit=-1,string,nr=0,arg1,arg2,arg3,arg4,arg5 +#ifdef FPU_EMU_DEBUG + .data +.Lpdata\@: + .string "\string" + .previous + + movem.l %d0/%d1/%a0/%a1,-(%sp) + .if \bit+1 +#if 0 + moveq #\bit,%d0 + andw #7,%d0 + btst %d0,fp_debugprint+((31-\bit)/8) +#else + btst #\bit,fp_debugprint+((31-\bit)/8) +#endif + jeq .Lpskip\@ + .endif + movestack \nr,\arg1,\arg2,\arg3,\arg4,\arg5 + pea .Lpdata\@ + jsr printk + lea ((\nr+1)*4,%sp),%sp +.Lpskip\@: + movem.l (%sp)+,%d0/%d1/%a0/%a1 +#endif +.endm + +.macro printx bit,fp +#ifdef FPU_EMU_DEBUG + movem.l %d0/%a0,-(%sp) + lea \fp,%a0 +#if 0 + moveq #'+',%d0 + tst.w (%a0) + jeq .Lx1\@ + moveq #'-',%d0 +.Lx1\@: printf \bit," %c",1,%d0 + move.l (4,%a0),%d0 + bclr #31,%d0 + jne .Lx2\@ + printf \bit,"0." + jra .Lx3\@ +.Lx2\@: printf \bit,"1." +.Lx3\@: printf \bit,"%08x%08x",2,%d0,%a0@(8) + move.w (2,%a0),%d0 + ext.l %d0 + printf \bit,"E%04x",1,%d0 +#else + printf \bit," %08x%08x%08x",3,%a0@,%a0@(4),%a0@(8) +#endif + movem.l (%sp)+,%d0/%a0 +#endif +.endm + +.macro debug instr,args +#ifdef FPU_EMU_DEBUG + \instr \args +#endif +.endm + + +#endif /* __ASSEMBLY__ */ + +#endif /* _ASM_M68K_SETUP_H */ |