Overview
Comment: | OFMatrix4x4: Use 3DNow! to transform vectors |
---|---|
Downloads: | Tarball | ZIP archive | SQL archive |
Timelines: | family | ancestors | descendants | both | trunk |
Files: | files | file ages | folders |
SHA3-256: |
1ac0583aaed9bf64d07fcb9e1b2e286e |
User & Date: | js on 2023-10-30 23:58:08 |
Other Links: | manifest | tags |
Context
2023-10-31
| ||
20:25 | Don't use -masm=intel check-in: d9af65de97 user: js tags: trunk | |
2023-10-30
| ||
23:58 | OFMatrix4x4: Use 3DNow! to transform vectors check-in: 1ac0583aae user: js tags: trunk | |
23:31 | OFMatrix4x4: Use 3DNow! for multiplication check-in: 5b213166ee user: js tags: trunk | |
Changes
Modified src/OFMatrix4x4.m from [5d349ae57d] to [9f0079818c].
︙ | ︙ | |||
56 57 58 59 60 61 62 63 64 65 66 67 68 69 | } } __asm__ ("femms"); memcpy(self->_values, result, sizeof(result)); } + (void)initialize { if (self != [OFMatrix4x4 class]) return; if ([OFSystemInfo supports3DNow]) { | > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > | | > > > > > > | 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 | } } __asm__ ("femms"); memcpy(self->_values, result, sizeof(result)); } static OFVector4D transformedVector_3DNow(OFMatrix4x4 *self, SEL _cmd, OFVector4D vector) { OFVector4D result; __asm__ ( "movq mm0, [%2]\n\t" "movq mm1, [%2 + 8]\n" "\n\t" "movq mm2, mm0\n\t" "movq mm3, mm1\n\t" "pfmul mm2, [%1]\n\t" "pfmul mm3, [%1 + 8]\n\t" "pfadd mm2, mm3\n\t" "movq mm3, mm2\n\t" "psrlq mm3, 32\n\t" "pfadd mm2, mm3\n" "\n\t" "movq mm3, mm0\n\t" "movq mm4, mm1\n\t" "pfmul mm3, [%1 + 16]\n\t" "pfmul mm4, [%1 + 24]\n\t" "pfadd mm3, mm4\n\t" "movq mm4, mm3\n\t" "psrlq mm4, 32\n\t" "pfadd mm3, mm4\n" "\n\t" "punpckldq mm2, mm3\n\t" "movq [%0], mm2\n" "\n\t" "movq mm2, mm0\n\t" "movq mm3, mm1\n\t" "pfmul mm2, [%1 + 32]\n\t" "pfmul mm3, [%1 + 40]\n\t" "pfadd mm2, mm3\n\t" "movq mm3, mm2\n\t" "psrlq mm3, 32\n\t" "pfadd mm2, mm3\n" "\n\t" "pfmul mm0, [%1 + 48]\n\t" "pfmul mm1, [%1 + 56]\n\t" "pfadd mm0, mm1\n\t" "movq mm1, mm0\n\t" "psrlq mm1, 32\n\t" "pfadd mm0, mm1\n" "\n\t" "punpckldq mm2, mm0\n\t" "movq [%0 + 8], mm2\n" "\n\t" "femms" :: "r"(&result), "r"(&self->_values), "r"(&vector) : "mm0", "mm1", "mm2", "mm3", "mm4", "memory" ); return result; } + (void)initialize { if (self != [OFMatrix4x4 class]) return; if ([OFSystemInfo supports3DNow]) { SEL selector; const char *typeEncoding; selector = @selector(multiplyWithMatrix:); typeEncoding = method_getTypeEncoding( class_getInstanceMethod(self, selector)); class_replaceMethod(self, selector, (IMP)multiplyWithMatrix_3DNow, typeEncoding); selector = @selector(transformedVector:); typeEncoding = method_getTypeEncoding( class_getInstanceMethod(self, selector)); class_replaceMethod(self, selector, (IMP)transformedVector_3DNow, typeEncoding); } } #endif + (OFMatrix4x4 *)identityMatrix { return [[[OFMatrix4x4 alloc] |
︙ | ︙ |