use shorter types vec_"type" instead of the too long vector "type"

part 1 of h264 luma interpolation 8x8 for altivec contributed by
Mauricio Alvarez % lokifo A gmail P com %
Original thread:
Date: Jun 26, 2007 8:15 PM
Subject: Re: [FFmpeg-devel] [PATCH] h264 luma interpolation 8x8 for altivec

Originally committed as revision 10090 to svn://svn.ffmpeg.org/ffmpeg/trunk

1 files changed, 168 insertions, 218 deletions

diff --git a/libavcodec/ppc/h264_template_altivec.c b/libavcodec/ppc/h264_template_altivec.c
index 7638687de2..c427ce419d 100644
--- a/libavcodec/ppc/h264_template_altivec.c
+++ b/libavcodec/ppc/h264_template_altivec.c
@@ -27,37 +27,37 @@ void PREFIX_h264_chroma_mc8_altivec(uint8_t * dst, uint8_t * src, int stride, in
                           ((8 - x) * (y)),
                           ((x) * (y))};
     register int i;
-    vector unsigned char fperm;
-    const vector signed int vABCD = vec_ld(0, ABCD);
-    const vector signed short vA = vec_splat((vector signed short)vABCD, 1);
-    const vector signed short vB = vec_splat((vector signed short)vABCD, 3);
-    const vector signed short vC = vec_splat((vector signed short)vABCD, 5);
-    const vector signed short vD = vec_splat((vector signed short)vABCD, 7);
-    const vector signed int vzero = vec_splat_s32(0);
-    const vector signed short v32ss = vec_sl(vec_splat_s16(1),vec_splat_u16(5));
-    const vector unsigned short v6us = vec_splat_u16(6);
+    vec_u8_t fperm;
+    const vec_s32_t vABCD = vec_ld(0, ABCD);
+    const vec_s16_t vA = vec_splat((vec_s16_t)vABCD, 1);
+    const vec_s16_t vB = vec_splat((vec_s16_t)vABCD, 3);
+    const vec_s16_t vC = vec_splat((vec_s16_t)vABCD, 5);
+    const vec_s16_t vD = vec_splat((vec_s16_t)vABCD, 7);
+    LOAD_ZERO;
+    const vec_s16_t v32ss = vec_sl(vec_splat_s16(1),vec_splat_u16(5));
+    const vec_u16_t v6us = vec_splat_u16(6);
     register int loadSecond = (((unsigned long)src) % 16) <= 7 ? 0 : 1;
     register int reallyBadAlign = (((unsigned long)src) % 16) == 15 ? 1 : 0;
 
-    vector unsigned char vsrcAuc, vsrcBuc, vsrcperm0, vsrcperm1;
-    vector unsigned char vsrc0uc, vsrc1uc;
-    vector signed short vsrc0ssH, vsrc1ssH;
-    vector unsigned char vsrcCuc, vsrc2uc, vsrc3uc;
-    vector signed short vsrc2ssH, vsrc3ssH, psum;
-    vector unsigned char vdst, ppsum, vfdst, fsum;
+    vec_u8_t vsrcAuc, vsrcBuc, vsrcperm0, vsrcperm1;
+    vec_u8_t vsrc0uc, vsrc1uc;
+    vec_s16_t vsrc0ssH, vsrc1ssH;
+    vec_u8_t vsrcCuc, vsrc2uc, vsrc3uc;
+    vec_s16_t vsrc2ssH, vsrc3ssH, psum;
+    vec_u8_t vdst, ppsum, vfdst, fsum;
 
   POWERPC_PERF_START_COUNT(PREFIX_h264_chroma_mc8_num, 1);
 
     if (((unsigned long)dst) % 16 == 0) {
-      fperm = (vector unsigned char)AVV(0x10, 0x11, 0x12, 0x13,
-                                        0x14, 0x15, 0x16, 0x17,
-                                        0x08, 0x09, 0x0A, 0x0B,
-                                        0x0C, 0x0D, 0x0E, 0x0F);
+      fperm = (vec_u8_t)AVV(0x10, 0x11, 0x12, 0x13,
+                            0x14, 0x15, 0x16, 0x17,
+                            0x08, 0x09, 0x0A, 0x0B,
+                            0x0C, 0x0D, 0x0E, 0x0F);
     } else {
-      fperm = (vector unsigned char)AVV(0x00, 0x01, 0x02, 0x03,
-                                        0x04, 0x05, 0x06, 0x07,
-                                        0x18, 0x19, 0x1A, 0x1B,
-                                        0x1C, 0x1D, 0x1E, 0x1F);
+      fperm = (vec_u8_t)AVV(0x00, 0x01, 0x02, 0x03,
+                            0x04, 0x05, 0x06, 0x07,
+                            0x18, 0x19, 0x1A, 0x1B,
+                            0x1C, 0x1D, 0x1E, 0x1F);
     }
 
     vsrcAuc = vec_ld(0, src);
@@ -73,10 +73,8 @@ void PREFIX_h264_chroma_mc8_altivec(uint8_t * dst, uint8_t * src, int stride, in
     else
       vsrc1uc = vec_perm(vsrcAuc, vsrcBuc, vsrcperm1);
 
-    vsrc0ssH = (vector signed short)vec_mergeh((vector unsigned char)vzero,
-                                               (vector unsigned char)vsrc0uc);
-    vsrc1ssH = (vector signed short)vec_mergeh((vector unsigned char)vzero,
-                                               (vector unsigned char)vsrc1uc);
+    vsrc0ssH = (vec_s16_t)vec_mergeh(zero_u8v,(vec_u8_t)vsrc0uc);
+    vsrc1ssH = (vec_s16_t)vec_mergeh(zero_u8v,(vec_u8_t)vsrc1uc);
 
     if (!loadSecond) {// -> !reallyBadAlign
       for (i = 0 ; i < h ; i++) {
@@ -87,10 +85,8 @@ void PREFIX_h264_chroma_mc8_altivec(uint8_t * dst, uint8_t * src, int stride, in
         vsrc2uc = vec_perm(vsrcCuc, vsrcCuc, vsrcperm0);
         vsrc3uc = vec_perm(vsrcCuc, vsrcCuc, vsrcperm1);
 
-        vsrc2ssH = (vector signed short)vec_mergeh((vector unsigned char)vzero,
-                                                (vector unsigned char)vsrc2uc);
-        vsrc3ssH = (vector signed short)vec_mergeh((vector unsigned char)vzero,
-                                                (vector unsigned char)vsrc3uc);
+        vsrc2ssH = (vec_s16_t)vec_mergeh(zero_u8v,(vec_u8_t)vsrc2uc);
+        vsrc3ssH = (vec_s16_t)vec_mergeh(zero_u8v,(vec_u8_t)vsrc3uc);
 
         psum = vec_mladd(vA, vsrc0ssH, vec_splat_s16(0));
         psum = vec_mladd(vB, vsrc1ssH, psum);
@@ -100,7 +96,7 @@ void PREFIX_h264_chroma_mc8_altivec(uint8_t * dst, uint8_t * src, int stride, in
         psum = vec_sra(psum, v6us);
 
         vdst = vec_ld(0, dst);
-        ppsum = (vector unsigned char)vec_packsu(psum, psum);
+        ppsum = (vec_u8_t)vec_packsu(psum, psum);
         vfdst = vec_perm(vdst, ppsum, fperm);
 
         OP_U8_ALTIVEC(fsum, vfdst, vdst);
@@ -114,7 +110,7 @@ void PREFIX_h264_chroma_mc8_altivec(uint8_t * dst, uint8_t * src, int stride, in
         src += stride;
       }
     } else {
-        vector unsigned char vsrcDuc;
+        vec_u8_t vsrcDuc;
       for (i = 0 ; i < h ; i++) {
         vsrcCuc = vec_ld(stride + 0, src);
         vsrcDuc = vec_ld(stride + 16, src);
@@ -125,10 +121,8 @@ void PREFIX_h264_chroma_mc8_altivec(uint8_t * dst, uint8_t * src, int stride, in
         else
           vsrc3uc = vec_perm(vsrcCuc, vsrcDuc, vsrcperm1);
 
-        vsrc2ssH = (vector signed short)vec_mergeh((vector unsigned char)vzero,
-                                                (vector unsigned char)vsrc2uc);
-        vsrc3ssH = (vector signed short)vec_mergeh((vector unsigned char)vzero,
-                                                (vector unsigned char)vsrc3uc);
+        vsrc2ssH = (vec_s16_t)vec_mergeh(zero_u8v,(vec_u8_t)vsrc2uc);
+        vsrc3ssH = (vec_s16_t)vec_mergeh(zero_u8v,(vec_u8_t)vsrc3uc);
 
         psum = vec_mladd(vA, vsrc0ssH, vec_splat_s16(0));
         psum = vec_mladd(vB, vsrc1ssH, psum);
@@ -138,7 +132,7 @@ void PREFIX_h264_chroma_mc8_altivec(uint8_t * dst, uint8_t * src, int stride, in
         psum = vec_sr(psum, v6us);
 
         vdst = vec_ld(0, dst);
-        ppsum = (vector unsigned char)vec_pack(psum, psum);
+        ppsum = (vec_u8_t)vec_pack(psum, psum);
         vfdst = vec_perm(vdst, ppsum, fperm);
 
         OP_U8_ALTIVEC(fsum, vfdst, vdst);
@@ -160,44 +154,39 @@ static void PREFIX_h264_qpel16_h_lowpass_altivec(uint8_t * dst, uint8_t * src, i
   POWERPC_PERF_DECLARE(PREFIX_h264_qpel16_h_lowpass_num, 1);
   register int i;
 
-  const vector signed int vzero = vec_splat_s32(0);
-  const vector unsigned char permM2 = vec_lvsl(-2, src);
-  const vector unsigned char permM1 = vec_lvsl(-1, src);
-  const vector unsigned char permP0 = vec_lvsl(+0, src);
-  const vector unsigned char permP1 = vec_lvsl(+1, src);
-  const vector unsigned char permP2 = vec_lvsl(+2, src);
-  const vector unsigned char permP3 = vec_lvsl(+3, src);
-  const vector signed short v5ss = vec_splat_s16(5);
-  const vector unsigned short v5us = vec_splat_u16(5);
-  const vector signed short v20ss = vec_sl(vec_splat_s16(5),vec_splat_u16(2));
-  const vector signed short v16ss = vec_sl(vec_splat_s16(1),vec_splat_u16(4));
-  const vector unsigned char dstperm = vec_lvsr(0, dst);
-  const vector unsigned char neg1 =
-                                (const vector unsigned char) vec_splat_s8(-1);
-
-  const vector unsigned char dstmask =
-                                vec_perm((const vector unsigned char)vzero,
-                                                               neg1, dstperm);
-
-  vector unsigned char srcM2, srcM1, srcP0, srcP1, srcP2, srcP3;
+  LOAD_ZERO;
+  const vec_u8_t permM2 = vec_lvsl(-2, src);
+  const vec_u8_t permM1 = vec_lvsl(-1, src);
+  const vec_u8_t permP0 = vec_lvsl(+0, src);
+  const vec_u8_t permP1 = vec_lvsl(+1, src);
+  const vec_u8_t permP2 = vec_lvsl(+2, src);
+  const vec_u8_t permP3 = vec_lvsl(+3, src);
+  const vec_s16_t v5ss = vec_splat_s16(5);
+  const vec_u16_t v5us = vec_splat_u16(5);
+  const vec_s16_t v20ss = vec_sl(vec_splat_s16(5),vec_splat_u16(2));
+  const vec_s16_t v16ss = vec_sl(vec_splat_s16(1),vec_splat_u16(4));
+  const vec_u8_t dstperm = vec_lvsr(0, dst);
+  const vec_u8_t neg1 = (const vec_u8_t) vec_splat_s8(-1);
+  const vec_u8_t dstmask = vec_perm(zero_u8v, neg1, dstperm);
+
+  vec_u8_t srcM2, srcM1, srcP0, srcP1, srcP2, srcP3;
 
   register int align = ((((unsigned long)src) - 2) % 16);
 
-  vector signed short srcP0A, srcP0B, srcP1A, srcP1B,
+  vec_s16_t srcP0A, srcP0B, srcP1A, srcP1B,
                       srcP2A, srcP2B, srcP3A, srcP3B,
                       srcM1A, srcM1B, srcM2A, srcM2B,
                       sum1A, sum1B, sum2A, sum2B, sum3A, sum3B,
                       pp1A, pp1B, pp2A, pp2B, pp3A, pp3B,
                       psumA, psumB, sumA, sumB;
 
-  vector unsigned char sum, dst1, dst2, vdst, fsum,
-                       rsum, fdst1, fdst2;
+  vec_u8_t sum, dst1, dst2, vdst, fsum, rsum, fdst1, fdst2;
 
   POWERPC_PERF_START_COUNT(PREFIX_h264_qpel16_h_lowpass_num, 1);
 
   for (i = 0 ; i < 16 ; i ++) {
-    vector unsigned char srcR1 = vec_ld(-2, src);
-    vector unsigned char srcR2 = vec_ld(14, src);
+    vec_u8_t srcR1 = vec_ld(-2, src);
+    vec_u8_t srcR2 = vec_ld(14, src);
 
     switch (align) {
     default: {
@@ -217,7 +206,7 @@ static void PREFIX_h264_qpel16_h_lowpass_altivec(uint8_t * dst, uint8_t * src, i
       srcP3 = srcR2;
     } break;
     case 12: {
-      vector unsigned char srcR3 = vec_ld(30, src);
+      vec_u8_t srcR3 = vec_ld(30, src);
       srcM2 = vec_perm(srcR1, srcR2, permM2);
       srcM1 = vec_perm(srcR1, srcR2, permM1);
       srcP0 = vec_perm(srcR1, srcR2, permP0);
@@ -226,7 +215,7 @@ static void PREFIX_h264_qpel16_h_lowpass_altivec(uint8_t * dst, uint8_t * src, i
       srcP3 = vec_perm(srcR2, srcR3, permP3);
     } break;
     case 13: {
-      vector unsigned char srcR3 = vec_ld(30, src);
+      vec_u8_t srcR3 = vec_ld(30, src);
       srcM2 = vec_perm(srcR1, srcR2, permM2);
       srcM1 = vec_perm(srcR1, srcR2, permM1);
       srcP0 = vec_perm(srcR1, srcR2, permP0);
@@ -235,7 +224,7 @@ static void PREFIX_h264_qpel16_h_lowpass_altivec(uint8_t * dst, uint8_t * src, i
       srcP3 = vec_perm(srcR2, srcR3, permP3);
     } break;
     case 14: {
-      vector unsigned char srcR3 = vec_ld(30, src);
+      vec_u8_t srcR3 = vec_ld(30, src);
       srcM2 = vec_perm(srcR1, srcR2, permM2);
       srcM1 = vec_perm(srcR1, srcR2, permM1);
       srcP0 = srcR2;
@@ -244,7 +233,7 @@ static void PREFIX_h264_qpel16_h_lowpass_altivec(uint8_t * dst, uint8_t * src, i
       srcP3 = vec_perm(srcR2, srcR3, permP3);
     } break;
     case 15: {
-      vector unsigned char srcR3 = vec_ld(30, src);
+      vec_u8_t srcR3 = vec_ld(30, src);
       srcM2 = vec_perm(srcR1, srcR2, permM2);
       srcM1 = srcR2;
       srcP0 = vec_perm(srcR2, srcR3, permP0);
@@ -254,32 +243,20 @@ static void PREFIX_h264_qpel16_h_lowpass_altivec(uint8_t * dst, uint8_t * src, i
     } break;
     }
 
-    srcP0A = (vector signed short)
-                vec_mergeh((vector unsigned char)vzero, srcP0);
-    srcP0B = (vector signed short)
-                vec_mergel((vector unsigned char)vzero, srcP0);
-    srcP1A = (vector signed short)
-                vec_mergeh((vector unsigned char)vzero, srcP1);
-    srcP1B = (vector signed short)
-                vec_mergel((vector unsigned char)vzero, srcP1);
-
-    srcP2A = (vector signed short)
-                vec_mergeh((vector unsigned char)vzero, srcP2);
-    srcP2B = (vector signed short)
-                vec_mergel((vector unsigned char)vzero, srcP2);
-    srcP3A = (vector signed short)
-                vec_mergeh((vector unsigned char)vzero, srcP3);
-    srcP3B = (vector signed short)
-                vec_mergel((vector unsigned char)vzero, srcP3);
-
-    srcM1A = (vector signed short)
-                vec_mergeh((vector unsigned char)vzero, srcM1);
-    srcM1B = (vector signed short)
-                vec_mergel((vector unsigned char)vzero, srcM1);
-    srcM2A = (vector signed short)
-                vec_mergeh((vector unsigned char)vzero, srcM2);
-    srcM2B = (vector signed short)
-                vec_mergel((vector unsigned char)vzero, srcM2);
+    srcP0A = (vec_s16_t) vec_mergeh(zero_u8v, srcP0);
+    srcP0B = (vec_s16_t) vec_mergel(zero_u8v, srcP0);
+    srcP1A = (vec_s16_t) vec_mergeh(zero_u8v, srcP1);
+    srcP1B = (vec_s16_t) vec_mergel(zero_u8v, srcP1);
+
+    srcP2A = (vec_s16_t) vec_mergeh(zero_u8v, srcP2);
+    srcP2B = (vec_s16_t) vec_mergel(zero_u8v, srcP2);
+    srcP3A = (vec_s16_t) vec_mergeh(zero_u8v, srcP3);
+    srcP3B = (vec_s16_t) vec_mergel(zero_u8v, srcP3);
+
+    srcM1A = (vec_s16_t) vec_mergeh(zero_u8v, srcM1);
+    srcM1B = (vec_s16_t) vec_mergel(zero_u8v, srcM1);
+    srcM2A = (vec_s16_t) vec_mergeh(zero_u8v, srcM2);
+    srcM2B = (vec_s16_t) vec_mergel(zero_u8v, srcM2);
 
     sum1A = vec_adds(srcP0A, srcP1A);
     sum1B = vec_adds(srcP0B, srcP1B);
@@ -291,8 +268,8 @@ static void PREFIX_h264_qpel16_h_lowpass_altivec(uint8_t * dst, uint8_t * src, i
     pp1A = vec_mladd(sum1A, v20ss, v16ss);
     pp1B = vec_mladd(sum1B, v20ss, v16ss);
 
-    pp2A = vec_mladd(sum2A, v5ss, (vector signed short)vzero);
-    pp2B = vec_mladd(sum2B, v5ss, (vector signed short)vzero);
+    pp2A = vec_mladd(sum2A, v5ss, zero_s16v);
+    pp2B = vec_mladd(sum2B, v5ss, zero_s16v);
 
     pp3A = vec_add(sum3A, pp1A);
     pp3B = vec_add(sum3B, pp1B);
@@ -330,67 +307,56 @@ static void PREFIX_h264_qpel16_v_lowpass_altivec(uint8_t * dst, uint8_t * src, i
 
   register int i;
 
-  const vector signed int vzero = vec_splat_s32(0);
-  const vector unsigned char perm = vec_lvsl(0, src);
-  const vector signed short v20ss = vec_sl(vec_splat_s16(5),vec_splat_u16(2));
-  const vector unsigned short v5us = vec_splat_u16(5);
-  const vector signed short v5ss = vec_splat_s16(5);
-  const vector signed short v16ss = vec_sl(vec_splat_s16(1),vec_splat_u16(4));
-  const vector unsigned char dstperm = vec_lvsr(0, dst);
-  const vector unsigned char neg1 = (const vector unsigned char)vec_splat_s8(-1);
-  const vector unsigned char dstmask = vec_perm((const vector unsigned char)vzero, neg1, dstperm);
+  LOAD_ZERO;
+  const vec_u8_t perm = vec_lvsl(0, src);
+  const vec_s16_t v20ss = vec_sl(vec_splat_s16(5),vec_splat_u16(2));
+  const vec_u16_t v5us = vec_splat_u16(5);
+  const vec_s16_t v5ss = vec_splat_s16(5);
+  const vec_s16_t v16ss = vec_sl(vec_splat_s16(1),vec_splat_u16(4));
+  const vec_u8_t dstperm = vec_lvsr(0, dst);
+  const vec_u8_t neg1 = (const vec_u8_t)vec_splat_s8(-1);
+  const vec_u8_t dstmask = vec_perm(zero_u8v, neg1, dstperm);
 
   uint8_t *srcbis = src - (srcStride * 2);
 
-  const vector unsigned char srcM2a = vec_ld(0, srcbis);
-  const vector unsigned char srcM2b = vec_ld(16, srcbis);
-  const vector unsigned char srcM2 = vec_perm(srcM2a, srcM2b, perm);
+  const vec_u8_t srcM2a = vec_ld(0, srcbis);
+  const vec_u8_t srcM2b = vec_ld(16, srcbis);
+  const vec_u8_t srcM2 = vec_perm(srcM2a, srcM2b, perm);
 //  srcbis += srcStride;
-  const vector unsigned char srcM1a = vec_ld(0, srcbis += srcStride);
-  const vector unsigned char srcM1b = vec_ld(16, srcbis);
-  const vector unsigned char srcM1 = vec_perm(srcM1a, srcM1b, perm);
+  const vec_u8_t srcM1a = vec_ld(0, srcbis += srcStride);
+  const vec_u8_t srcM1b = vec_ld(16, srcbis);
+  const vec_u8_t srcM1 = vec_perm(srcM1a, srcM1b, perm);
 //  srcbis += srcStride;
-  const vector unsigned char srcP0a = vec_ld(0, srcbis += srcStride);
-  const vector unsigned char srcP0b = vec_ld(16, srcbis);
-  const vector unsigned char srcP0 = vec_perm(srcP0a, srcP0b, perm);
+  const vec_u8_t srcP0a = vec_ld(0, srcbis += srcStride);
+  const vec_u8_t srcP0b = vec_ld(16, srcbis);
+  const vec_u8_t srcP0 = vec_perm(srcP0a, srcP0b, perm);
 //  srcbis += srcStride;
-  const vector unsigned char srcP1a = vec_ld(0, srcbis += srcStride);
-  const vector unsigned char srcP1b = vec_ld(16, srcbis);
-  const vector unsigned char srcP1 = vec_perm(srcP1a, srcP1b, perm);
+  const vec_u8_t srcP1a = vec_ld(0, srcbis += srcStride);
+  const vec_u8_t srcP1b = vec_ld(16, srcbis);
+  const vec_u8_t srcP1 = vec_perm(srcP1a, srcP1b, perm);
 //  srcbis += srcStride;
-  const vector unsigned char srcP2a = vec_ld(0, srcbis += srcStride);
-  const vector unsigned char srcP2b = vec_ld(16, srcbis);
-  const vector unsigned char srcP2 = vec_perm(srcP2a, srcP2b, perm);
+  const vec_u8_t srcP2a = vec_ld(0, srcbis += srcStride);
+  const vec_u8_t srcP2b = vec_ld(16, srcbis);
+  const vec_u8_t srcP2 = vec_perm(srcP2a, srcP2b, perm);
 //  srcbis += srcStride;
 
-  vector signed short srcM2ssA = (vector signed short)
-                                vec_mergeh((vector unsigned char)vzero, srcM2);
-  vector signed short srcM2ssB = (vector signed short)
-                                vec_mergel((vector unsigned char)vzero, srcM2);
-  vector signed short srcM1ssA = (vector signed short)
-                                vec_mergeh((vector unsigned char)vzero, srcM1);
-  vector signed short srcM1ssB = (vector signed short)
-                                vec_mergel((vector unsigned char)vzero, srcM1);
-  vector signed short srcP0ssA = (vector signed short)
-                                vec_mergeh((vector unsigned char)vzero, srcP0);
-  vector signed short srcP0ssB = (vector signed short)
-                                vec_mergel((vector unsigned char)vzero, srcP0);
-  vector signed short srcP1ssA = (vector signed short)
-                                vec_mergeh((vector unsigned char)vzero, srcP1);
-  vector signed short srcP1ssB = (vector signed short)
-                                vec_mergel((vector unsigned char)vzero, srcP1);
-  vector signed short srcP2ssA = (vector signed short)
-                                vec_mergeh((vector unsigned char)vzero, srcP2);
-  vector signed short srcP2ssB = (vector signed short)
-                                vec_mergel((vector unsigned char)vzero, srcP2);
-
-  vector signed short pp1A, pp1B, pp2A, pp2B, pp3A, pp3B,
+  vec_s16_t srcM2ssA = (vec_s16_t) vec_mergeh(zero_u8v, srcM2);
+  vec_s16_t srcM2ssB = (vec_s16_t) vec_mergel(zero_u8v, srcM2);
+  vec_s16_t srcM1ssA = (vec_s16_t) vec_mergeh(zero_u8v, srcM1);
+  vec_s16_t srcM1ssB = (vec_s16_t) vec_mergel(zero_u8v, srcM1);
+  vec_s16_t srcP0ssA = (vec_s16_t) vec_mergeh(zero_u8v, srcP0);
+  vec_s16_t srcP0ssB = (vec_s16_t) vec_mergel(zero_u8v, srcP0);
+  vec_s16_t srcP1ssA = (vec_s16_t) vec_mergeh(zero_u8v, srcP1);
+  vec_s16_t srcP1ssB = (vec_s16_t) vec_mergel(zero_u8v, srcP1);
+  vec_s16_t srcP2ssA = (vec_s16_t) vec_mergeh(zero_u8v, srcP2);
+  vec_s16_t srcP2ssB = (vec_s16_t) vec_mergel(zero_u8v, srcP2);
+
+  vec_s16_t pp1A, pp1B, pp2A, pp2B, pp3A, pp3B,
                       psumA, psumB, sumA, sumB,
                       srcP3ssA, srcP3ssB,
                       sum1A, sum1B, sum2A, sum2B, sum3A, sum3B;
 
-  vector unsigned char sum, dst1, dst2, vdst, fsum, rsum, fdst1, fdst2,
-                       srcP3a, srcP3b, srcP3;
+  vec_u8_t sum, dst1, dst2, vdst, fsum, rsum, fdst1, fdst2, srcP3a, srcP3b, srcP3;
 
   POWERPC_PERF_START_COUNT(PREFIX_h264_qpel16_v_lowpass_num, 1);
 
@@ -398,10 +364,8 @@ static void PREFIX_h264_qpel16_v_lowpass_altivec(uint8_t * dst, uint8_t * src, i
     srcP3a = vec_ld(0, srcbis += srcStride);
     srcP3b = vec_ld(16, srcbis);
     srcP3 = vec_perm(srcP3a, srcP3b, perm);
-    srcP3ssA = (vector signed short)
-                                vec_mergeh((vector unsigned char)vzero, srcP3);
-    srcP3ssB = (vector signed short)
-                                vec_mergel((vector unsigned char)vzero, srcP3);
+    srcP3ssA = (vec_s16_t) vec_mergeh(zero_u8v, srcP3);
+    srcP3ssB = (vec_s16_t) vec_mergel(zero_u8v, srcP3);
 //    srcbis += srcStride;
 
     sum1A = vec_adds(srcP0ssA, srcP1ssA);
@@ -425,8 +389,8 @@ static void PREFIX_h264_qpel16_v_lowpass_altivec(uint8_t * dst, uint8_t * src, i
     pp1A = vec_mladd(sum1A, v20ss, v16ss);
     pp1B = vec_mladd(sum1B, v20ss, v16ss);
 
-    pp2A = vec_mladd(sum2A, v5ss, (vector signed short)vzero);
-    pp2B = vec_mladd(sum2B, v5ss, (vector signed short)vzero);
+    pp2A = vec_mladd(sum2A, v5ss, zero_s16v);
+    pp2B = vec_mladd(sum2B, v5ss, zero_s16v);
 
     pp3A = vec_add(sum3A, pp1A);
     pp3B = vec_add(sum3B, pp1B);
@@ -461,58 +425,56 @@ static void PREFIX_h264_qpel16_v_lowpass_altivec(uint8_t * dst, uint8_t * src, i
 static void PREFIX_h264_qpel16_hv_lowpass_altivec(uint8_t * dst, int16_t * tmp, uint8_t * src, int dstStride, int tmpStride, int srcStride) {
   POWERPC_PERF_DECLARE(PREFIX_h264_qpel16_hv_lowpass_num, 1);
   register int i;
-  const vector signed int vzero = vec_splat_s32(0);
-  const vector unsigned char permM2 = vec_lvsl(-2, src);
-  const vector unsigned char permM1 = vec_lvsl(-1, src);
-  const vector unsigned char permP0 = vec_lvsl(+0, src);
-  const vector unsigned char permP1 = vec_lvsl(+1, src);
-  const vector unsigned char permP2 = vec_lvsl(+2, src);
-  const vector unsigned char permP3 = vec_lvsl(+3, src);
-  const vector signed short v20ss = vec_sl(vec_splat_s16(5),vec_splat_u16(2));
-  const vector unsigned int v10ui = vec_splat_u32(10);
-  const vector signed short v5ss = vec_splat_s16(5);
-  const vector signed short v1ss = vec_splat_s16(1);
-  const vector signed int v512si = vec_sl(vec_splat_s32(1),vec_splat_u32(9));
-  const vector unsigned int v16ui = vec_sl(vec_splat_u32(1),vec_splat_u32(4));
+  LOAD_ZERO;
+  const vec_u8_t permM2 = vec_lvsl(-2, src);
+  const vec_u8_t permM1 = vec_lvsl(-1, src);
+  const vec_u8_t permP0 = vec_lvsl(+0, src);
+  const vec_u8_t permP1 = vec_lvsl(+1, src);
+  const vec_u8_t permP2 = vec_lvsl(+2, src);
+  const vec_u8_t permP3 = vec_lvsl(+3, src);
+  const vec_s16_t v20ss = vec_sl(vec_splat_s16(5),vec_splat_u16(2));
+  const vec_u32_t v10ui = vec_splat_u32(10);
+  const vec_s16_t v5ss = vec_splat_s16(5);
+  const vec_s16_t v1ss = vec_splat_s16(1);
+  const vec_s32_t v512si = vec_sl(vec_splat_s32(1),vec_splat_u32(9));
+  const vec_u32_t v16ui = vec_sl(vec_splat_u32(1),vec_splat_u32(4));
 
   register int align = ((((unsigned long)src) - 2) % 16);
 
-  const vector unsigned char neg1 = (const vector unsigned char)
-                                                        vec_splat_s8(-1);
+  const vec_u8_t neg1 = (const vec_u8_t) vec_splat_s8(-1);
 
-  vector signed short srcP0A, srcP0B, srcP1A, srcP1B,
+  vec_s16_t srcP0A, srcP0B, srcP1A, srcP1B,
                       srcP2A, srcP2B, srcP3A, srcP3B,
                       srcM1A, srcM1B, srcM2A, srcM2B,
                       sum1A, sum1B, sum2A, sum2B, sum3A, sum3B,
                       pp1A, pp1B, pp2A, pp2B, psumA, psumB;
 
-  const vector unsigned char dstperm = vec_lvsr(0, dst);
+  const vec_u8_t dstperm = vec_lvsr(0, dst);
 
-  const vector unsigned char dstmask = vec_perm((const vector unsigned char)vzero, neg1, dstperm);
+  const vec_u8_t dstmask = vec_perm(zero_u8v, neg1, dstperm);
 
-  const vector unsigned char mperm = (const vector unsigned char)
+  const vec_u8_t mperm = (const vec_u8_t)
     AVV(0x00, 0x08, 0x01, 0x09, 0x02, 0x0A, 0x03, 0x0B,
         0x04, 0x0C, 0x05, 0x0D, 0x06, 0x0E, 0x07, 0x0F);
   int16_t *tmpbis = tmp;
 
-  vector signed short tmpM1ssA, tmpM1ssB, tmpM2ssA, tmpM2ssB,
+  vec_s16_t tmpM1ssA, tmpM1ssB, tmpM2ssA, tmpM2ssB,
                       tmpP0ssA, tmpP0ssB, tmpP1ssA, tmpP1ssB,
                       tmpP2ssA, tmpP2ssB;
 
-  vector signed int pp1Ae, pp1Ao, pp1Be, pp1Bo, pp2Ae, pp2Ao, pp2Be, pp2Bo,
+  vec_s32_t pp1Ae, pp1Ao, pp1Be, pp1Bo, pp2Ae, pp2Ao, pp2Be, pp2Bo,
                     pp3Ae, pp3Ao, pp3Be, pp3Bo, pp1cAe, pp1cAo, pp1cBe, pp1cBo,
                     pp32Ae, pp32Ao, pp32Be, pp32Bo, sumAe, sumAo, sumBe, sumBo,
                     ssumAe, ssumAo, ssumBe, ssumBo;
-  vector unsigned char fsum, sumv, sum, dst1, dst2, vdst,
-                       rsum, fdst1, fdst2;
-  vector signed short ssume, ssumo;
+  vec_u8_t fsum, sumv, sum, dst1, dst2, vdst, rsum, fdst1, fdst2;
+  vec_s16_t ssume, ssumo;
 
   POWERPC_PERF_START_COUNT(PREFIX_h264_qpel16_hv_lowpass_num, 1);
   src -= (2 * srcStride);
   for (i = 0 ; i < 21 ; i ++) {
-    vector unsigned char srcM2, srcM1, srcP0, srcP1, srcP2, srcP3;
-    vector unsigned char srcR1 = vec_ld(-2, src);
-    vector unsigned char srcR2 = vec_ld(14, src);
+    vec_u8_t srcM2, srcM1, srcP0, srcP1, srcP2, srcP3;
+    vec_u8_t srcR1 = vec_ld(-2, src);
+    vec_u8_t srcR2 = vec_ld(14, src);
 
     switch (align) {
     default: {
@@ -532,7 +494,7 @@ static void PREFIX_h264_qpel16_hv_lowpass_altivec(uint8_t * dst, int16_t * tmp,
       srcP3 = srcR2;
     } break;
     case 12: {
-      vector unsigned char srcR3 = vec_ld(30, src);
+      vec_u8_t srcR3 = vec_ld(30, src);
       srcM2 = vec_perm(srcR1, srcR2, permM2);
       srcM1 = vec_perm(srcR1, srcR2, permM1);
       srcP0 = vec_perm(srcR1, srcR2, permP0);
@@ -541,7 +503,7 @@ static void PREFIX_h264_qpel16_hv_lowpass_altivec(uint8_t * dst, int16_t * tmp,
       srcP3 = vec_perm(srcR2, srcR3, permP3);
     } break;
     case 13: {
-      vector unsigned char srcR3 = vec_ld(30, src);
+      vec_u8_t srcR3 = vec_ld(30, src);
       srcM2 = vec_perm(srcR1, srcR2, permM2);
       srcM1 = vec_perm(srcR1, srcR2, permM1);
       srcP0 = vec_perm(srcR1, srcR2, permP0);
@@ -550,7 +512,7 @@ static void PREFIX_h264_qpel16_hv_lowpass_altivec(uint8_t * dst, int16_t * tmp,
       srcP3 = vec_perm(srcR2, srcR3, permP3);
     } break;
     case 14: {
-      vector unsigned char srcR3 = vec_ld(30, src);
+      vec_u8_t srcR3 = vec_ld(30, src);
       srcM2 = vec_perm(srcR1, srcR2, permM2);
       srcM1 = vec_perm(srcR1, srcR2, permM1);
       srcP0 = srcR2;
@@ -559,7 +521,7 @@ static void PREFIX_h264_qpel16_hv_lowpass_altivec(uint8_t * dst, int16_t * tmp,
       srcP3 = vec_perm(srcR2, srcR3, permP3);
     } break;
     case 15: {
-      vector unsigned char srcR3 = vec_ld(30, src);
+      vec_u8_t srcR3 = vec_ld(30, src);
       srcM2 = vec_perm(srcR1, srcR2, permM2);
       srcM1 = srcR2;
       srcP0 = vec_perm(srcR2, srcR3, permP0);
@@ -569,32 +531,20 @@ static void PREFIX_h264_qpel16_hv_lowpass_altivec(uint8_t * dst, int16_t * tmp,
     } break;
     }
 
-    srcP0A = (vector signed short)
-                            vec_mergeh((vector unsigned char)vzero, srcP0);
-    srcP0B = (vector signed short)
-                            vec_mergel((vector unsigned char)vzero, srcP0);
-    srcP1A = (vector signed short)
-                            vec_mergeh((vector unsigned char)vzero, srcP1);
-    srcP1B = (vector signed short)
-                            vec_mergel((vector unsigned char)vzero, srcP1);
-
-    srcP2A = (vector signed short)
-                            vec_mergeh((vector unsigned char)vzero, srcP2);
-    srcP2B = (vector signed short)
-                            vec_mergel((vector unsigned char)vzero, srcP2);
-    srcP3A = (vector signed short)
-                            vec_mergeh((vector unsigned char)vzero, srcP3);
-    srcP3B = (vector signed short)
-                            vec_mergel((vector unsigned char)vzero, srcP3);
-
-    srcM1A = (vector signed short)
-                            vec_mergeh((vector unsigned char)vzero, srcM1);
-    srcM1B = (vector signed short)
-                            vec_mergel((vector unsigned char)vzero, srcM1);
-    srcM2A = (vector signed short)
-                            vec_mergeh((vector unsigned char)vzero, srcM2);
-    srcM2B = (vector signed short)
-                            vec_mergel((vector unsigned char)vzero, srcM2);
+    srcP0A = (vec_s16_t) vec_mergeh(zero_u8v, srcP0);
+    srcP0B = (vec_s16_t) vec_mergel(zero_u8v, srcP0);
+    srcP1A = (vec_s16_t) vec_mergeh(zero_u8v, srcP1);
+    srcP1B = (vec_s16_t) vec_mergel(zero_u8v, srcP1);
+
+    srcP2A = (vec_s16_t) vec_mergeh(zero_u8v, srcP2);
+    srcP2B = (vec_s16_t) vec_mergel(zero_u8v, srcP2);
+    srcP3A = (vec_s16_t) vec_mergeh(zero_u8v, srcP3);
+    srcP3B = (vec_s16_t) vec_mergel(zero_u8v, srcP3);
+
+    srcM1A = (vec_s16_t) vec_mergeh(zero_u8v, srcM1);
+    srcM1B = (vec_s16_t) vec_mergel(zero_u8v, srcM1);
+    srcM2A = (vec_s16_t) vec_mergeh(zero_u8v, srcM2);
+    srcM2B = (vec_s16_t) vec_mergel(zero_u8v, srcM2);
 
     sum1A = vec_adds(srcP0A, srcP1A);
     sum1B = vec_adds(srcP0B, srcP1B);
@@ -606,8 +556,8 @@ static void PREFIX_h264_qpel16_hv_lowpass_altivec(uint8_t * dst, int16_t * tmp,
     pp1A = vec_mladd(sum1A, v20ss, sum3A);
     pp1B = vec_mladd(sum1B, v20ss, sum3B);
 
-    pp2A = vec_mladd(sum2A, v5ss, (vector signed short)vzero);
-    pp2B = vec_mladd(sum2B, v5ss, (vector signed short)vzero);
+    pp2A = vec_mladd(sum2A, v5ss, zero_s16v);
+    pp2B = vec_mladd(sum2B, v5ss, zero_s16v);
 
     psumA = vec_sub(pp1A, pp2A);
     psumB = vec_sub(pp1B, pp2B);
@@ -636,15 +586,15 @@ static void PREFIX_h264_qpel16_hv_lowpass_altivec(uint8_t * dst, int16_t * tmp,
   tmpbis += tmpStride;
 
   for (i = 0 ; i < 16 ; i++) {
-    const vector signed short tmpP3ssA = vec_ld(0, tmpbis);
-    const vector signed short tmpP3ssB = vec_ld(16, tmpbis);
+    const vec_s16_t tmpP3ssA = vec_ld(0, tmpbis);
+    const vec_s16_t tmpP3ssB = vec_ld(16, tmpbis);
 
-    const vector signed short sum1A = vec_adds(tmpP0ssA, tmpP1ssA);
-    const vector signed short sum1B = vec_adds(tmpP0ssB, tmpP1ssB);
-    const vector signed short sum2A = vec_adds(tmpM1ssA, tmpP2ssA);
-    const vector signed short sum2B = vec_adds(tmpM1ssB, tmpP2ssB);
-    const vector signed short sum3A = vec_adds(tmpM2ssA, tmpP3ssA);
-    const vector signed short sum3B = vec_adds(tmpM2ssB, tmpP3ssB);
+    const vec_s16_t sum1A = vec_adds(tmpP0ssA, tmpP1ssA);
+    const vec_s16_t sum1B = vec_adds(tmpP0ssB, tmpP1ssB);
+    const vec_s16_t sum2A = vec_adds(tmpM1ssA, tmpP2ssA);
+    const vec_s16_t sum2B = vec_adds(tmpM1ssB, tmpP2ssB);
+    const vec_s16_t sum3A = vec_adds(tmpM2ssA, tmpP3ssA);
+    const vec_s16_t sum3B = vec_adds(tmpM2ssB, tmpP3ssB);
 
     tmpbis += tmpStride;
 
@@ -669,9 +619,9 @@ static void PREFIX_h264_qpel16_hv_lowpass_altivec(uint8_t * dst, int16_t * tmp,
     pp2Be = vec_mule(sum2B, v5ss);
     pp2Bo = vec_mulo(sum2B, v5ss);
 
-    pp3Ae = vec_sra((vector signed int)sum3A, v16ui);
+    pp3Ae = vec_sra((vec_s32_t)sum3A, v16ui);
     pp3Ao = vec_mulo(sum3A, v1ss);
-    pp3Be = vec_sra((vector signed int)sum3B, v16ui);
+    pp3Be = vec_sra((vec_s32_t)sum3B, v16ui);
     pp3Bo = vec_mulo(sum3B, v1ss);
 
     pp1cAe = vec_add(pp1Ae, v512si);