shithub: qk1

Info  •  Files  •  Log

turn DotProduct into a C function; provide arm64 neon implementation

--- a/cmd.c

+++ b/cmd.c

@@ -621,7 +621,7 @@

 // check cvars

 	if (!Cvar_Command ())

-		Con_Printf ("Unknown command \"%s\"\n", Cmd_Argv(0));

+		Con_Printf ("Unknown command \"%s\": %s\n", Cmd_Argv(0), text);

 }

--- /dev/null

+++ b/dotproduct.c

@@ -1,0 +1,5 @@

+float

+DotProduct(float v1[3], float v2[3])

+{

+	return v1[0]*v2[0] + v1[1]*v2[1] + v1[2]*v2[2];

+}

--- /dev/null

+++ b/dotproduct_arm64.s

@@ -1,0 +1,8 @@

+TEXT DotProduct+0(SB), 1, $-4

+	MOV	v2+8(FP), R1

+	WORD $0x0d40a000 // ld3 {v0.s, v1.s, v2.s}[0], [x0]

+	WORD $0x0d40a023 // ld3 {v3.s, v4.s, v5.s}[0], [x1]

+	WORD $0x1e230800 // fmul s0, s0, s3

+	WORD $0x1f040020 // fmadd s0, s1, s4, s0

+	WORD $0x1f050040 // fmadd s0, s2, s5, s0

+	RETURN

--- a/mathlib.c

+++ b/mathlib.c

@@ -293,12 +293,6 @@

 	vecc[2] = veca[2] + scale*vecb[2];

 }

-

-vec_t _DotProduct (vec3_t v1, vec3_t v2)

-{

-	return v1[0]*v2[0] + v1[1]*v2[1] + v1[2]*v2[2];

-}

-

 void _VectorSubtract (vec3_t veca, vec3_t vecb, vec3_t out)

 {

 	out[0] = veca[0]-vecb[0];

--- a/mathlib.h

+++ b/mathlib.h

@@ -21,7 +21,7 @@

 #define	IS_NAN(x) (((*(int *)&x)&nanmask)==nanmask)

-#define DotProduct(x,y) (x[0]*y[0]+x[1]*y[1]+x[2]*y[2])

+#define DotProduct_(x,y) (x[0]*y[0]+x[1]*y[1]+x[2]*y[2])

 #define DotProductDouble(x,y) ((double)x[0]*y[0]+(double)x[1]*y[1]+(double)x[2]*y[2])

 #define VectorSubtract(a,b,c) {c[0]=a[0]-b[0];c[1]=a[1]-b[1];c[2]=a[2]-b[2];}

 #define VectorAdd(a,b,c) {c[0]=a[0]+b[0];c[1]=a[1]+b[1];c[2]=a[2]+b[2];}

@@ -29,7 +29,7 @@

 void VectorMA (vec3_t veca, float scale, vec3_t vecb, vec3_t vecc);

-vec_t _DotProduct (vec3_t v1, vec3_t v2);

+vec_t DotProduct (vec3_t v1, vec3_t v2);

 void _VectorSubtract (vec3_t veca, vec3_t vecb, vec3_t out);

 void _VectorAdd (vec3_t veca, vec3_t vecb, vec3_t out);

 void _VectorCopy (vec3_t in, vec3_t out);

--- a/mkfile

+++ b/mkfile

@@ -8,6 +8,7 @@

 	pal`{test -f pal_$objtype.s && echo -n _$objtype}.$O\

 	span`{test -f span_$objtype.s && echo -n _$objtype}.$O\

 	span_alpha.$O\

+	dotproduct`{test -f span_$objtype.s && echo -n _$objtype}.$O\

 	cd.$O\

 	cl_demo.$O\

 	cl_input.$O\