1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
|
--- src/dotprod/src/dotprod_rrrf.av.c.orig 2024-06-19 07:25:01 UTC
+++ src/dotprod/src/dotprod_rrrf.av.c
@@ -26,6 +26,7 @@
#include <stdio.h>
#include <stdlib.h>
+#include <altivec.h>
#include "liquid.internal.h"
@@ -117,7 +118,7 @@ dotprod_rrrf dotprod_rrrf_create_opt(float * _h,
for (i=0; i<4; i++) {
q->h[i] = calloc(1+(q->n+i-1)/4,sizeof(vector float));
for (j=0; j<q->n; j++)
- e->h[i][j+i] = _h[_rev ? q->n-j-1 : j];
+ q->h[i][j+i] = _h[_rev ? q->n-j-1 : j];
}
return q;
@@ -190,8 +191,8 @@ int dotprod_rrrf_execute(dotprod_rrrf _q,
union { vector float v; float w[4];} s;
unsigned int nblocks;
- ar = (vector float*)( (int)_x & ~15);
- al = ((int)_x & 15)/sizeof(float);
+ ar = (vector float*)( (uintptr_t)_x & ~15);
+ al = ((uintptr_t)_x & 15)/sizeof(float);
d = (vector float*)_q->h[al];
@@ -200,7 +201,7 @@ int dotprod_rrrf_execute(dotprod_rrrf _q,
// split into four vectors each with four 32-bit
// partial sums. Effectively each loop iteration
// operates on 16 input samples at a time.
- s0 = s1 = s2 = s3 = (vector float)(0);
+ s0 = s1 = s2 = s3 = (vector float){0,0,0,0};
while (nblocks >= 4) {
s0 = vec_madd(ar[nblocks-1],d[nblocks-1],s0);
s1 = vec_madd(ar[nblocks-2],d[nblocks-2],s1);
@@ -221,7 +222,7 @@ int dotprod_rrrf_execute(dotprod_rrrf _q,
// move the result into the union s (effetively,
// this loads the four 32-bit values in s0 into
// the array w).
- s.v = vec_add(s0,(vector float)(0));
+ s.v = vec_add(s0,(vector float){0,0,0,0});
// sum the resulting array
*_r = s.w[0] + s.w[1] + s.w[2] + s.w[3];
|
