View | Details | Raw Unified | Return to bug 233790
Collapse All | Expand All

(-)files/patch-configure.ac (+11 lines)
Line 0 Link Here
1 
--- configure.ac.orig	2018-12-04 23:39:19 UTC
2 
+++ configure.ac
3 
@@ -195,7 +195,7 @@ else
4 
                        src/dotprod/src/dotprod_rrrf.av.o \
5 
                        src/dotprod/src/dotprod_crcf.av.o \
6 
                        src/dotprod/src/sumsq.o"
7 
-        ARCH_OPTION="-fno-common -faltivec";;
8 
+        ARCH_OPTION="-fno-common -maltivec";;
9 
     armv1*|armv2*|armv3*|armv4*|armv5*|armv6*)
10 
         # assume neon instructions are NOT available
11 
         MLIBS_DOTPROD="src/dotprod/src/dotprod_cccf.o \
(-)files/patch-src_dotprod_src_dotprod__crcf.av.c (+39 lines)
Line 0 Link Here
1 
--- src/dotprod/src/dotprod_crcf.av.c.orig	2018-12-04 23:40:11 UTC
2 
+++ src/dotprod/src/dotprod_crcf.av.c
3 
@@ -26,6 +26,7 @@
4
5 
 #include <stdio.h>
6 
 #include <stdlib.h>
7 
+#include <altivec.h>
8
9 
 #include "liquid.internal.h"
10
11 
@@ -168,8 +169,8 @@ void dotprod_crcf_execute(dotprod_crcf    _q,
12 
     union { vector float v; float w[4];} s;
13 
     unsigned int nblocks;
14
15 
-    ar = (vector float*)( (int)_x & ~15);
16 
-    al = ((int)_x & 15)/sizeof(float);
17 
+    ar = (vector float*)( (uintptr_t)_x & ~15);
18 
+    al = ((uintptr_t)_x & 15)/sizeof(float);
19
20 
     d = (vector float*)_q->h[al];
21
22 
@@ -179,7 +180,7 @@ void dotprod_crcf_execute(dotprod_crcf    _q,
23 
     // split into four vectors each with four 32-bit
24 
     // partial sums.  Effectively each loop iteration
25 
     // operates on 16 input samples at a time.
26 
-    s0 = s1 = s2 = s3 = (vector float)(0);
27 
+    s0 = s1 = s2 = s3 = (vector float){0,0,0,0};
28 
     while (nblocks >= 4) {
29 
         s0 = vec_madd(ar[nblocks-1],d[nblocks-1],s0);
30 
         s1 = vec_madd(ar[nblocks-2],d[nblocks-2],s1);
31 
@@ -200,7 +201,7 @@ void dotprod_crcf_execute(dotprod_crcf    _q,
32 
     // move the result into the union s (effetively,
33 
     // this loads the four 32-bit values in s0 into
34 
     // the array w).
35 
-    s.v = vec_add(s0,(vector float)(0));
36 
+    s.v = vec_add(s0,(vector float){0,0,0,0});
37
38 
     // sum the resulting array
39 
     //*_r = s.w[0] + s.w[1] + s.w[2] + s.w[3];
(-)files/patch-src_dotprod_src_dotprod__rrrf.av.c (+39 lines)
Line 0 Link Here
1 
--- src/dotprod/src/dotprod_rrrf.av.c.orig	2018-12-04 23:41:32 UTC
2 
+++ src/dotprod/src/dotprod_rrrf.av.c
3 
@@ -26,6 +26,7 @@
4
5 
 #include <stdio.h>
6 
 #include <stdlib.h>
7 
+#include <altivec.h>
8
9 
 #include "liquid.internal.h"
10
11 
@@ -163,8 +164,8 @@ void dotprod_rrrf_execute(dotprod_rrrf _q,
12 
     union { vector float v; float w[4];} s;
13 
     unsigned int nblocks;
14
15 
-    ar = (vector float*)( (int)_x & ~15);
16 
-    al = ((int)_x & 15)/sizeof(float);
17 
+    ar = (vector float*)( (uintptr_t)_x & ~15);
18 
+    al = ((uintptr_t)_x & 15)/sizeof(float);
19
20 
     d = (vector float*)_q->h[al];
21
22 
@@ -173,7 +174,7 @@ void dotprod_rrrf_execute(dotprod_rrrf _q,
23 
     // split into four vectors each with four 32-bit
24 
     // partial sums.  Effectively each loop iteration
25 
     // operates on 16 input samples at a time.
26 
-    s0 = s1 = s2 = s3 = (vector float)(0);
27 
+    s0 = s1 = s2 = s3 = (vector float){0,0,0,0};
28 
     while (nblocks >= 4) {
29 
         s0 = vec_madd(ar[nblocks-1],d[nblocks-1],s0);
30 
         s1 = vec_madd(ar[nblocks-2],d[nblocks-2],s1);
31 
@@ -194,7 +195,7 @@ void dotprod_rrrf_execute(dotprod_rrrf _q,
32 
     // move the result into the union s (effetively,
33 
     // this loads the four 32-bit values in s0 into
34 
     // the array w).
35 
-    s.v = vec_add(s0,(vector float)(0));
36 
+    s.v = vec_add(s0,(vector float){0,0,0,0});
37
38 
     // sum the resulting array
39 
     *_r = s.w[0] + s.w[1] + s.w[2] + s.w[3];

Return to bug 233790