ObjFW  Check-in [f949f7775b]

Overview
Comment:OFMatrix4x4: Use __asm__ __volatile__

The outputs aren't consumed as they are used as inputs that can be
modified.

Downloads: Tarball | ZIP archive | SQL archive
Timelines: family | ancestors | descendants | both | trunk
Files: files | file ages | folders
SHA3-256: f949f7775b0110c5eea187548c0911ca213630640d66f58d69bf27b65e7d5e5b
User & Date: js on 2023-11-01 21:22:10
Other Links: manifest | tags
Context
2023-11-02
22:48
OFMatrix4x4: Move __asm__ out of loop check-in: 0d671245d4 user: js tags: trunk
2023-11-01
21:22
OFMatrix4x4: Use __asm__ __volatile__ check-in: f949f7775b user: js tags: trunk
20:54
Add -[OFMatrix4x4 transformVectors:count:] check-in: ad6b3e6442 user: js tags: trunk
Changes
Hide Diffs Side-by-Side Diffs Ignore Whitespace Patch

Modified src/OFMatrix4x4.m from [43a6889a80] to [b42a0641ea].

︙
38
39
40
41
42
43
44
45

46
47
48
49
50
51
52
53
54
55
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

38
39
40
41
42
43
44

45
46
47
48
49
50
51
52
53
54
55
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








-
+

















-
+










-
+


















-
+








-
+








multiplyWithMatrix_enhanced3DNow(OFMatrix4x4 *self, SEL _cmd,
    OFMatrix4x4 *matrix)
{
	float result[4][4];

	for (uint_fast8_t i = 0; i < 4; i++) {
		for (uint_fast8_t j = 0; j < 4; j++) {
			__asm__ (
			__asm__ __volatile__ (
			    "movd	(%2), %%mm0\n\t"
			    "punpckldq	16(%2), %%mm0\n\t"
			    "pfmul	(%1), %%mm0\n\t"
			    "movd	32(%2), %%mm1\n\t"
			    "punpckldq	48(%2), %%mm1\n\t"
			    "pfmul	8(%1), %%mm1\n\t"
			    "pfadd	%%mm1, %%mm0\n\t"
			    "pswapd	%%mm0, %%mm1\n\t"
			    "pfadd	%%mm1, %%mm0\n\t"
			    "movd	%%mm0, %0"
			    :: "m"(result[i][j]), "r"(&matrix->_values[i][0]),
			       "r"(&self->_values[0][j])
			    : "mm0", "mm1", "memory"
			);
		}
	}

	__asm__ ("femms");
	__asm__ __volatile__ ("femms");

	memcpy(self->_values, result, sizeof(result));
}
static void
multiplyWithMatrix_3DNow(OFMatrix4x4 *self, SEL _cmd, OFMatrix4x4 *matrix)
{
	float result[4][4];

	for (uint_fast8_t i = 0; i < 4; i++) {
		for (uint_fast8_t j = 0; j < 4; j++) {
			__asm__ (
			__asm__ __volatile__ (
			    "movd	(%2), %%mm0\n\t"
			    "punpckldq	16(%2), %%mm0\n\t"
			    "pfmul	(%1), %%mm0\n\t"
			    "movd	32(%2), %%mm1\n\t"
			    "punpckldq	48(%2), %%mm1\n\t"
			    "pfmul	8(%1), %%mm1\n\t"
			    "pfadd	%%mm1, %%mm0\n\t"
			    "movq	%%mm0, %%mm1\n\t"
			    "psrlq	$32, %%mm1\n\t"
			    "pfadd	%%mm1, %%mm0\n\t"
			    "movd	%%mm0, %0"
			    :: "m"(result[i][j]), "r"(&matrix->_values[i][0]),
			       "r"(&self->_values[0][j])
			    : "mm0", "mm1", "memory"
			);
		}
	}

	__asm__ ("femms");
	__asm__ __volatile__ ("femms");

	memcpy(self->_values, result, sizeof(result));
}

static void
transformVectors_enhanced3DNow(OFMatrix4x4 *self, SEL _cmd, OFVector4D *vectors,
    size_t count)
{
	__asm__ (
	__asm__ __volatile__ (
	    "0:\n\t"
	    "test	%0, %0\n\t"
	    "jz		0f\n"
	    "\n\t"
	    "movq	(%1), %%mm0\n\t"
	    "movq	8(%1), %%mm1\n"
	    "\n\t"
︙
155
156
157
158
159
160
161
162

163
164
165
166
167
168
169

155
156
157
158
159
160
161

162
163
164
165
166
167
168
169








-
+








	);
}

static void
transformVectors_3DNow(OFMatrix4x4 *self, SEL _cmd, OFVector4D *vectors,
    size_t count)
{
	__asm__ (
	__asm__ __volatile__ (
	    "0:\n\t"
	    "test	%0, %0\n\t"
	    "jz		0f\n"
	    "\n\t"
	    "movq	(%1), %%mm0\n\t"
	    "movq	8(%1), %%mm1\n"
	    "\n\t"
︙