| Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1 | #include <asm/ppc_asm.h> |
| 2 | #include <asm/processor.h> |
| 3 | |
| 4 | /* |
| 5 | * The routines below are in assembler so we can closely control the |
| 6 | * usage of floating-point registers. These routines must be called |
| 7 | * with preempt disabled. |
| 8 | */ |
| 9 | .section ".toc","aw" |
| 10 | fpzero: |
| 11 | .tc FD_0_0[TC],0 |
| 12 | fpone: |
| 13 | .tc FD_3ff00000_0[TC],0x3ff0000000000000 /* 1.0 */ |
| 14 | fphalf: |
| 15 | .tc FD_3fe00000_0[TC],0x3fe0000000000000 /* 0.5 */ |
| 16 | |
| 17 | .text |
| 18 | /* |
| 19 | * Internal routine to enable floating point and set FPSCR to 0. |
| 20 | * Don't call it from C; it doesn't use the normal calling convention. |
| 21 | */ |
| 22 | fpenable: |
| 23 | mfmsr r10 |
| 24 | ori r11,r10,MSR_FP |
| 25 | mtmsr r11 |
| 26 | isync |
| 27 | stfd fr31,-8(r1) |
| 28 | stfd fr0,-16(r1) |
| 29 | stfd fr1,-24(r1) |
| 30 | mffs fr31 |
| 31 | lfd fr1,fpzero@toc(r2) |
| 32 | mtfsf 0xff,fr1 |
| 33 | blr |
| 34 | |
| 35 | fpdisable: |
| 36 | mtlr r12 |
| 37 | mtfsf 0xff,fr31 |
| 38 | lfd fr1,-24(r1) |
| 39 | lfd fr0,-16(r1) |
| 40 | lfd fr31,-8(r1) |
| 41 | mtmsr r10 |
| 42 | isync |
| 43 | blr |
| 44 | |
| 45 | /* |
| 46 | * Vector add, floating point. |
| 47 | */ |
| 48 | _GLOBAL(vaddfp) |
| 49 | mflr r12 |
| 50 | bl fpenable |
| 51 | li r0,4 |
| 52 | mtctr r0 |
| 53 | li r6,0 |
| 54 | 1: lfsx fr0,r4,r6 |
| 55 | lfsx fr1,r5,r6 |
| 56 | fadds fr0,fr0,fr1 |
| 57 | stfsx fr0,r3,r6 |
| 58 | addi r6,r6,4 |
| 59 | bdnz 1b |
| 60 | b fpdisable |
| 61 | |
| 62 | /* |
| 63 | * Vector subtract, floating point. |
| 64 | */ |
| 65 | _GLOBAL(vsubfp) |
| 66 | mflr r12 |
| 67 | bl fpenable |
| 68 | li r0,4 |
| 69 | mtctr r0 |
| 70 | li r6,0 |
| 71 | 1: lfsx fr0,r4,r6 |
| 72 | lfsx fr1,r5,r6 |
| 73 | fsubs fr0,fr0,fr1 |
| 74 | stfsx fr0,r3,r6 |
| 75 | addi r6,r6,4 |
| 76 | bdnz 1b |
| 77 | b fpdisable |
| 78 | |
| 79 | /* |
| 80 | * Vector multiply and add, floating point. |
| 81 | */ |
| 82 | _GLOBAL(vmaddfp) |
| 83 | mflr r12 |
| 84 | bl fpenable |
| 85 | stfd fr2,-32(r1) |
| 86 | li r0,4 |
| 87 | mtctr r0 |
| 88 | li r7,0 |
| 89 | 1: lfsx fr0,r4,r7 |
| 90 | lfsx fr1,r5,r7 |
| 91 | lfsx fr2,r6,r7 |
| 92 | fmadds fr0,fr0,fr2,fr1 |
| 93 | stfsx fr0,r3,r7 |
| 94 | addi r7,r7,4 |
| 95 | bdnz 1b |
| 96 | lfd fr2,-32(r1) |
| 97 | b fpdisable |
| 98 | |
| 99 | /* |
| 100 | * Vector negative multiply and subtract, floating point. |
| 101 | */ |
| 102 | _GLOBAL(vnmsubfp) |
| 103 | mflr r12 |
| 104 | bl fpenable |
| 105 | stfd fr2,-32(r1) |
| 106 | li r0,4 |
| 107 | mtctr r0 |
| 108 | li r7,0 |
| 109 | 1: lfsx fr0,r4,r7 |
| 110 | lfsx fr1,r5,r7 |
| 111 | lfsx fr2,r6,r7 |
| 112 | fnmsubs fr0,fr0,fr2,fr1 |
| 113 | stfsx fr0,r3,r7 |
| 114 | addi r7,r7,4 |
| 115 | bdnz 1b |
| 116 | lfd fr2,-32(r1) |
| 117 | b fpdisable |
| 118 | |
| 119 | /* |
| 120 | * Vector reciprocal estimate. We just compute 1.0/x. |
| 121 | * r3 -> destination, r4 -> source. |
| 122 | */ |
| 123 | _GLOBAL(vrefp) |
| 124 | mflr r12 |
| 125 | bl fpenable |
| 126 | li r0,4 |
| 127 | lfd fr1,fpone@toc(r2) |
| 128 | mtctr r0 |
| 129 | li r6,0 |
| 130 | 1: lfsx fr0,r4,r6 |
| 131 | fdivs fr0,fr1,fr0 |
| 132 | stfsx fr0,r3,r6 |
| 133 | addi r6,r6,4 |
| 134 | bdnz 1b |
| 135 | b fpdisable |
| 136 | |
| 137 | /* |
| 138 | * Vector reciprocal square-root estimate, floating point. |
| 139 | * We use the frsqrte instruction for the initial estimate followed |
| 140 | * by 2 iterations of Newton-Raphson to get sufficient accuracy. |
| 141 | * r3 -> destination, r4 -> source. |
| 142 | */ |
| 143 | _GLOBAL(vrsqrtefp) |
| 144 | mflr r12 |
| 145 | bl fpenable |
| 146 | stfd fr2,-32(r1) |
| 147 | stfd fr3,-40(r1) |
| 148 | stfd fr4,-48(r1) |
| 149 | stfd fr5,-56(r1) |
| 150 | li r0,4 |
| 151 | lfd fr4,fpone@toc(r2) |
| 152 | lfd fr5,fphalf@toc(r2) |
| 153 | mtctr r0 |
| 154 | li r6,0 |
| 155 | 1: lfsx fr0,r4,r6 |
| 156 | frsqrte fr1,fr0 /* r = frsqrte(s) */ |
| 157 | fmuls fr3,fr1,fr0 /* r * s */ |
| 158 | fmuls fr2,fr1,fr5 /* r * 0.5 */ |
| 159 | fnmsubs fr3,fr1,fr3,fr4 /* 1 - s * r * r */ |
| 160 | fmadds fr1,fr2,fr3,fr1 /* r = r + 0.5 * r * (1 - s * r * r) */ |
| 161 | fmuls fr3,fr1,fr0 /* r * s */ |
| 162 | fmuls fr2,fr1,fr5 /* r * 0.5 */ |
| 163 | fnmsubs fr3,fr1,fr3,fr4 /* 1 - s * r * r */ |
| 164 | fmadds fr1,fr2,fr3,fr1 /* r = r + 0.5 * r * (1 - s * r * r) */ |
| 165 | stfsx fr1,r3,r6 |
| 166 | addi r6,r6,4 |
| 167 | bdnz 1b |
| 168 | lfd fr5,-56(r1) |
| 169 | lfd fr4,-48(r1) |
| 170 | lfd fr3,-40(r1) |
| 171 | lfd fr2,-32(r1) |
| 172 | b fpdisable |