diff --git a/vpx_dsp/ppc/vpx_idct_altivec.c b/vpx_dsp/ppc/vpx_idct_altivec.c
index c1e633971..4f7a0fcee 100644
--- a/vpx_dsp/ppc/vpx_idct_altivec.c
+++ b/vpx_dsp/ppc/vpx_idct_altivec.c
@@ -77,7 +77,7 @@ vector unsigned char merge64_lvec ALIGN16 = {
 
 // DCT utility function to splat an arbitrary signed short. Can work
 // for other unit sizes.
-vector signed short ss_splat(signed short *what) {
+static vector signed short ss_splat(signed short *what) {
 	vector signed short result = vec_ld(0, what);
 	vector signed short temp = vec_ld(sizeof(*what) - 1, what);
 	vector unsigned char mover = vec_lvsl(0, what);
@@ -88,7 +88,7 @@ vector signed short ss_splat(signed short *what) {
 void vpx_idct4x4_1_add_altivec(const tran_low_t *input, uint8_t *dest, int stride) {
 	vector signed short dc_value, i0, i1, i2, i3;
 	vector unsigned char p0, p1, p2, p3;
-	vector unsigned char zero = vec_splat_s8(0);
+	vector unsigned char zero = vec_splat_u8(0);
 	vector unsigned int ltemp;
 	int a;
 	signed short a_s;
@@ -101,16 +101,16 @@ void vpx_idct4x4_1_add_altivec(const tran_low_t *input, uint8_t *dest, int strid
 	// I chose to interleave the vector calls to see if we can
 	// get better parallelism rather than a macro like SSE2's.	
 	
-	_unaligned_load32(p0, (uint32_t *)(dest + 0 * stride));
+	_unaligned_load32(p0, (unsigned char *)(dest + 0 * stride));
 	// SSE unpacklo(x,y) becomes vec_mergeh(y,x)
 	i0 = (vector signed short)vec_mergeh(zero, p0);
-	_unaligned_load32(p1, (uint32_t *)(dest + 1 * stride));
+	_unaligned_load32(p1, (unsigned char *)(dest + 1 * stride));
 	i0 = vec_add(i0, dc_value);
 	i1 = (vector signed short)vec_mergeh(zero, p1);
-	_unaligned_load32(p2, (uint32_t *)(dest + 2 * stride));
+	_unaligned_load32(p2, (unsigned char *)(dest + 2 * stride));
 	p0 = vec_packsu(i0, i0);
 	i1 = vec_add(i1, dc_value);
-	_unaligned_load32(p3, (uint32_t *)(dest + 3 * stride));
+	_unaligned_load32(p3, (unsigned char *)(dest + 3 * stride));
 	i2 = (vector signed short)vec_mergeh(zero, p2);
 	p1 = vec_packsu(i1, i1);
 	_unaligned_store32(p0, ltemp, (uint32_t *)(dest + 0 * stride));
@@ -125,9 +125,9 @@ void vpx_idct4x4_1_add_altivec(const tran_low_t *input, uint8_t *dest, int strid
 }
 
 void vpx_idct4x4_16_add_altivec(const tran_low_t *input, uint8_t *dest, int stride) {
-	vector unsigned char zero = vec_splat_s8(0);
-	vector unsigned char one = vec_splat_s8(1);
-	vector unsigned short four = vec_splat_s16(4);
+	vector unsigned char zero = vec_splat_u8(0);
+	vector unsigned char one = vec_splat_u8(1);
+	vector unsigned short four = vec_splat_u16(4);
 	vector signed short eight = vec_splat_s16(8);
 	vector signed short input0, input1, input2, input3;
 	vector signed int wnput0, wnput1, wnput2, wnput3;
@@ -142,8 +142,8 @@ void vpx_idct4x4_16_add_altivec(const tran_low_t *input, uint8_t *dest, int stri
 	};
 
 	// The Intel intrinsics used imply the input is aligned.
-	input0 = vec_ld(0, (uint32_t *)input);
-	input2 = vec_ld(16, (uint32_t *)input);
+	input0 = vec_ld(0, (const signed short *)input);
+	input2 = vec_ld(16, (const signed short *)input);
 
 	// ALL HAIL THE ALTIVEC PERMUTE UNIT!
 	// SUCK IT, INTEL!
@@ -242,12 +242,12 @@ void vpx_idct4x4_16_add_altivec(const tran_low_t *input, uint8_t *dest, int stri
 	// This is kind of a bass-ackwards merged unaligned store, really.
 	//
 	// Get the destination values we're going to merge with.
-	_unaligned_load32(wnput0, (uint32_t *)dest);
-	_unaligned_load32(wnput1, (uint32_t *)(dest + stride));
+	_unaligned_load32(wnput0, (signed int *)dest);
+	_unaligned_load32(wnput1, (signed int *)(dest + stride));
 	wnput0 = vec_mergeh(wnput0, wnput1);
-	_unaligned_load32(wnput2, (uint32_t *)(dest + stride + stride));
+	_unaligned_load32(wnput2, (signed int *)(dest + stride + stride));
 	wnput0 = (vector signed int)vec_mergeh(zero, (vector unsigned char)wnput0);
-	_unaligned_load32(wnput3, (uint32_t *)(dest + stride + stride + stride));
+	_unaligned_load32(wnput3, (signed int *)(dest + stride + stride + stride));
 	wnput2 = vec_mergeh(wnput3, wnput2);
 	wnput2 = (vector signed int)vec_mergeh(zero, (vector unsigned char)wnput2);
 
@@ -271,7 +271,7 @@ void vpx_idct4x4_16_add_altivec(const tran_low_t *input, uint8_t *dest, int stri
 void vpx_idct8x8_1_add_altivec(const tran_low_t *input, uint8_t *dest, int stride) {
 	vector signed short dc_value, i0, i1, i2, i3, i4, i5, i6, i7;
 	vector unsigned char p0, p1, p2, p3, p4, p5, p6, p7;
-	vector unsigned char zero = vec_splat_s8(0);
+	vector unsigned char zero = vec_splat_u8(0);
 	vector unsigned int ltemp;
 	int a;
 	signed short a_s;
@@ -290,15 +290,15 @@ void vpx_idct8x8_1_add_altivec(const tran_low_t *input, uint8_t *dest, int strid
 	// XXX: Would we spend more time constructing the optimized vectors
 	// to store than simply taking the hit for being unaligned?
 	
-	_unaligned_load64(p0, (uint32_t *)(dest + 0 * stride));
+	_unaligned_load64(p0, (unsigned char *)(dest + 0 * stride));
 	i0 = (vector signed short)vec_mergeh(zero, p0);
-	_unaligned_load64(p1, (uint32_t *)(dest + 1 * stride));
+	_unaligned_load64(p1, (unsigned char *)(dest + 1 * stride));
 	i4 = vec_add(i0, dc_value);
 	i1 = (vector signed short)vec_mergeh(zero, p1);
-	_unaligned_load64(p2, (uint32_t *)(dest + 2 * stride));
+	_unaligned_load64(p2, (unsigned char *)(dest + 2 * stride));
 	p0 = vec_packsu(i4, i4);
 	i5 = vec_add(i1, dc_value);
-	_unaligned_load64(p3, (uint32_t *)(dest + 3 * stride));
+	_unaligned_load64(p3, (unsigned char *)(dest + 3 * stride));
 	i2 = (vector signed short)vec_mergeh(zero, p2);
 	p1 = vec_packsu(i5, i5);
 	_unaligned_store64(p0, ltemp, (uint32_t *)(dest + 0 * stride));
@@ -311,16 +311,16 @@ void vpx_idct8x8_1_add_altivec(const tran_low_t *input, uint8_t *dest, int strid
 	p3 = vec_packsu(i7, i7);
 	_unaligned_store64(p3, ltemp, (uint32_t *)(dest + 3 * stride));
 	
-	_unaligned_load64(p0, (uint32_t *)(dest + 4 * stride));
+	_unaligned_load64(p0, (unsigned char *)(dest + 4 * stride));
 	// SSE unpacklo(x,y) becomes vec_mergeh(y,x)
 	i0 = (vector signed short)vec_mergeh(zero, p0);
-	_unaligned_load64(p1, (uint32_t *)(dest + 5 * stride));
+	_unaligned_load64(p1, (unsigned char *)(dest + 5 * stride));
 	i4 = vec_add(i0, dc_value);
 	i1 = (vector signed short)vec_mergeh(zero, p1);
-	_unaligned_load64(p2, (uint32_t *)(dest + 6 * stride));
+	_unaligned_load64(p2, (unsigned char *)(dest + 6 * stride));
 	p0 = vec_packsu(i4, i4);
 	i5 = vec_add(i1, dc_value);
-	_unaligned_load64(p3, (uint32_t *)(dest + 7 * stride));
+	_unaligned_load64(p3, (unsigned char *)(dest + 7 * stride));
 	i2 = (vector signed short)vec_mergeh(zero, p2);
 	p1 = vec_packsu(i5, i5);
 	_unaligned_store64(p0, ltemp, (uint32_t *)(dest + 4 * stride));
@@ -337,7 +337,7 @@ void vpx_idct8x8_1_add_altivec(const tran_low_t *input, uint8_t *dest, int strid
 void vpx_idct16x16_1_add_altivec(const tran_low_t *input, uint8_t *dest, int stride) {
 	vector signed short dc_value, i0, i1, i2, i3, i4, i5, i6, i7;
 	vector unsigned char p0, p1, p2, p3, p4, p5, p6, p7;
-	vector unsigned char zero = vec_splat_s8(0);
+	vector unsigned char zero = vec_splat_u8(0);
 	vector unsigned int ltemp;
 	int a, i;
 	signed short a_s;
@@ -350,15 +350,15 @@ void vpx_idct16x16_1_add_altivec(const tran_low_t *input, uint8_t *dest, int str
 	// Massively unrolled. This is probably the limit for that.
 	
 	for(i=0; i<2; ++i) {
-		_unaligned_load64(p0, (uint32_t *)(dest + 0 * stride));
+		_unaligned_load64(p0, (unsigned char *)(dest + 0 * stride));
 		i0 = (vector signed short)vec_mergeh(zero, p0);
-		_unaligned_load64(p1, (uint32_t *)(dest + 1 * stride));
+		_unaligned_load64(p1, (unsigned char *)(dest + 1 * stride));
 		i4 = vec_add(i0, dc_value);
 		i1 = (vector signed short)vec_mergeh(zero, p1);
-		_unaligned_load64(p2, (uint32_t *)(dest + 2 * stride));
+		_unaligned_load64(p2, (unsigned char *)(dest + 2 * stride));
 		p0 = vec_packsu(i4, i4);
 		i5 = vec_add(i1, dc_value);
-		_unaligned_load64(p3, (uint32_t *)(dest + 3 * stride));
+		_unaligned_load64(p3, (unsigned char *)(dest + 3 * stride));
 		i2 = (vector signed short)vec_mergeh(zero, p2);
 		p1 = vec_packsu(i5, i5);
 		_unaligned_store64(p0, ltemp, (uint32_t *)(dest + 0 * stride));
@@ -371,16 +371,16 @@ void vpx_idct16x16_1_add_altivec(const tran_low_t *input, uint8_t *dest, int str
 		p3 = vec_packsu(i7, i7);
 		_unaligned_store64(p3, ltemp, (uint32_t *)(dest + 3 * stride));
 	
-		_unaligned_load64(p0, (uint32_t *)(dest + 4 * stride));
+		_unaligned_load64(p0, (unsigned char *)(dest + 4 * stride));
 		// SSE unpacklo(x,y) becomes vec_mergeh(y,x)
 		i0 = (vector signed short)vec_mergeh(zero, p0);
-		_unaligned_load64(p1, (uint32_t *)(dest + 5 * stride));
+		_unaligned_load64(p1, (unsigned char *)(dest + 5 * stride));
 		i4 = vec_add(i0, dc_value);
 		i1 = (vector signed short)vec_mergeh(zero, p1);
-		_unaligned_load64(p2, (uint32_t *)(dest + 6 * stride));
+		_unaligned_load64(p2, (unsigned char *)(dest + 6 * stride));
 		p0 = vec_packsu(i4, i4);
 		i5 = vec_add(i1, dc_value);
-		_unaligned_load64(p3, (uint32_t *)(dest + 7 * stride));
+		_unaligned_load64(p3, (unsigned char *)(dest + 7 * stride));
 		i2 = (vector signed short)vec_mergeh(zero, p2);
 		p1 = vec_packsu(i5, i5);
 		_unaligned_store64(p0, ltemp, (uint32_t *)(dest + 4 * stride));
@@ -393,15 +393,15 @@ void vpx_idct16x16_1_add_altivec(const tran_low_t *input, uint8_t *dest, int str
 		p3 = vec_packsu(i7, i7);
 		_unaligned_store64(p3, ltemp, (uint32_t *)(dest + 7 * stride));
 	
-		_unaligned_load64(p0, (uint32_t *)(dest + 8 * stride));
+		_unaligned_load64(p0, (unsigned char *)(dest + 8 * stride));
 		i0 = (vector signed short)vec_mergeh(zero, p0);
-		_unaligned_load64(p1, (uint32_t *)(dest + 9 * stride));
+		_unaligned_load64(p1, (unsigned char *)(dest + 9 * stride));
 		i4 = vec_add(i0, dc_value);
 		i1 = (vector signed short)vec_mergeh(zero, p1);
-		_unaligned_load64(p2, (uint32_t *)(dest + 10 * stride));
+		_unaligned_load64(p2, (unsigned char *)(dest + 10 * stride));
 		p0 = vec_packsu(i4, i4);
 		i5 = vec_add(i1, dc_value);
-		_unaligned_load64(p3, (uint32_t *)(dest + 11 * stride));
+		_unaligned_load64(p3, (unsigned char *)(dest + 11 * stride));
 		i2 = (vector signed short)vec_mergeh(zero, p2);
 		p1 = vec_packsu(i5, i5);
 		_unaligned_store64(p0, ltemp, (uint32_t *)(dest + 8 * stride));
@@ -414,15 +414,15 @@ void vpx_idct16x16_1_add_altivec(const tran_low_t *input, uint8_t *dest, int str
 		p3 = vec_packsu(i7, i7);
 		_unaligned_store64(p3, ltemp, (uint32_t *)(dest + 11 * stride));
 		
-		_unaligned_load64(p0, (uint32_t *)(dest + 12 * stride));
+		_unaligned_load64(p0, (unsigned char *)(dest + 12 * stride));
 		i0 = (vector signed short)vec_mergeh(zero, p0);
-		_unaligned_load64(p1, (uint32_t *)(dest + 13 * stride));
+		_unaligned_load64(p1, (unsigned char *)(dest + 13 * stride));
 		i4 = vec_add(i0, dc_value);
 		i1 = (vector signed short)vec_mergeh(zero, p1);
-		_unaligned_load64(p2, (uint32_t *)(dest + 14 * stride));
+		_unaligned_load64(p2, (unsigned char *)(dest + 14 * stride));
 		p0 = vec_packsu(i4, i4);
 		i5 = vec_add(i1, dc_value);
-		_unaligned_load64(p3, (uint32_t *)(dest + 15 * stride));
+		_unaligned_load64(p3, (unsigned char *)(dest + 15 * stride));
 		i2 = (vector signed short)vec_mergeh(zero, p2);
 		p1 = vec_packsu(i5, i5);
 		_unaligned_store64(p0, ltemp, (uint32_t *)(dest + 12 * stride));
@@ -442,7 +442,7 @@ void vpx_idct16x16_1_add_altivec(const tran_low_t *input, uint8_t *dest, int str
 void vpx_idct32x32_1_add_altivec(const tran_low_t *input, uint8_t *dest, int stride) {
 	vector signed short dc_value, i0, i1;
 	vector unsigned char p0;
-	vector unsigned char zero = vec_splat_s8(0);
+	vector unsigned char zero = vec_splat_u8(0);
 	vector unsigned int ltemp;
 	int a, i, j;
 	signed short a_s;
@@ -457,7 +457,7 @@ void vpx_idct32x32_1_add_altivec(const tran_low_t *input, uint8_t *dest, int str
 	
 	for (i=0; i<4; ++i) {
 		for(j=0; j<32; ++j) {
-			_unaligned_load64(p0, (uint32_t *)(dest + j * stride));
+			_unaligned_load64(p0, (unsigned char *)(dest + j * stride));
 			i0 = (vector signed short)vec_mergeh(zero, p0);
 			i1 = vec_add(i0, dc_value);
 			p0 = vec_packsu(i1, i1);
@@ -515,7 +515,7 @@ static void iadst4_vmx(vector signed short *in) {
 	vector signed short k__sinpi_p03_m04 = short_pair_a(sinpi_3_9, -sinpi_4_9);
 	vector signed short k__sinpi_p03_p03 = short_pair_a(sinpi_3_9, sinpi_3_9);
 	vector signed short zero = vec_splat_s16(0);
-	vector unsigned int twoi = vec_splat_s32(2);
+	vector unsigned int twoi = vec_splat_u32(2);
 	vector signed short u[7], in7;
 	vector signed int v[7], w[7];
 	// For easier hand modeling I manually inlined the SSE2 transposition function.
@@ -574,13 +574,13 @@ void vp9_iht4x4_16_add_altivec(const tran_low_t *input, uint8_t *dest, int strid
 	vector unsigned int ltemp;
 	vector unsigned char result;
 	vector signed int wnput0, wnput1, wnput2, wnput3;
-	vector unsigned char zero = vec_splat_s8(0);
-	vector unsigned char one = vec_splat_s8(1);
-	vector unsigned short four = vec_splat_s16(4);
+	vector unsigned char zero = vec_splat_u8(0);
+	vector unsigned char one = vec_splat_u8(1);
+	vector unsigned short four = vec_splat_u16(4);
 	vector signed short eight = vec_splat_s16(8);
 
-	_unaligned_load128(in[0], (uint32_t *)(input));
-	_unaligned_load128(in[1], (uint32_t *)(input + 8));
+	_unaligned_load128(in[0], (const signed short *)(input));
+	_unaligned_load128(in[1], (const signed short *)(input + 8));
 
 	switch (tx_type) {
 		case 0: { // DCT_DCT
@@ -616,12 +616,12 @@ void vp9_iht4x4_16_add_altivec(const tran_low_t *input, uint8_t *dest, int strid
 	
 	// Reconstruct, pack and store (mostly the same as idct4x4_16).
 	// Get the destination values we're going to merge with.
-	_unaligned_load32(wnput0, (uint32_t *)dest);
-	_unaligned_load32(wnput1, (uint32_t *)(dest + stride));
+	_unaligned_load32(wnput0, (signed int *)dest);
+	_unaligned_load32(wnput1, (signed int *)(dest + stride));
 	wnput0 = vec_mergeh(wnput0, wnput1);
-	_unaligned_load32(wnput2, (uint32_t *)(dest + stride + stride));
+	_unaligned_load32(wnput2, (signed int *)(dest + stride + stride));
 	wnput0 = (vector signed int)vec_mergeh(zero, (vector unsigned char)wnput0);
-	_unaligned_load32(wnput3, (uint32_t *)(dest + stride + stride + stride));
+	_unaligned_load32(wnput3, (signed int *)(dest + stride + stride + stride));
 	wnput2 = vec_mergeh(wnput2, wnput3);
 	wnput2 = (vector signed int)vec_mergeh(zero, (vector unsigned char)wnput2);
 
@@ -779,7 +779,7 @@ void vpx_idct8x8_64_add_altivec(const tran_low_t *input, uint8_t *dest, int stri
 	vector signed short stp1_0, stp1_1, stp1_2, stp1_3, stp1_4, stp1_5, stp1_6, stp1_7;
 	vector signed int tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
 	vector signed short stp2_0, stp2_1, stp2_2, stp2_3, stp2_4, stp2_5, stp2_6, stp2_7;
-	vector unsigned char zero = vec_splat_s8(0);
+	vector unsigned char zero = vec_splat_u8(0);
 	vector unsigned int ltemp;
 	vector unsigned char p0, p1, p2, p3, p4, p5, p6, p7;
 	vector signed short i0, i1, i2, i3, i4, i5, i6, i7;
@@ -788,17 +788,17 @@ void vpx_idct8x8_64_add_altivec(const tran_low_t *input, uint8_t *dest, int stri
 	vector unsigned short four = vec_splat_u16(4);
 	vector unsigned short five = vec_splat_u16(5);
 	vector signed short final_rounding;
-	final_rounding = vec_rl(one, four);
+	final_rounding = (vector signed short)vec_rl(one, four);
 	
 	// The Intel mnemonic implies these loads are aligned.
-	in0 = vec_ld(0,   (uint32_t *)input);
-	in1 = vec_ld(16,  (uint32_t *)input);
-	in2 = vec_ld(32,  (uint32_t *)input);
-	in3 = vec_ld(48,  (uint32_t *)input);
-	in4 = vec_ld(64,  (uint32_t *)input);
-	in5 = vec_ld(80,  (uint32_t *)input);
-	in6 = vec_ld(96,  (uint32_t *)input);
-	in7 = vec_ld(112, (uint32_t *)input);
+	in0 = vec_ld(0,   (const signed short *)input);
+	in1 = vec_ld(16,  (const signed short *)input);
+	in2 = vec_ld(32,  (const signed short *)input);
+	in3 = vec_ld(48,  (const signed short *)input);
+	in4 = vec_ld(64,  (const signed short *)input);
+	in5 = vec_ld(80,  (const signed short *)input);
+	in6 = vec_ld(96,  (const signed short *)input);
+	in7 = vec_ld(112, (const signed short *)input);
 
 	TRANSPOSE_8X8(in0, in1, in2, in3, in4, in5, in6, in7,
 			in0, in1, in2, in3, in4, in5, in6, in7);
@@ -833,15 +833,15 @@ void vpx_idct8x8_64_add_altivec(const tran_low_t *input, uint8_t *dest, int stri
 
 	// Unfortunately, these stores are NOT.
 	// Borrowed from idct8x8_1.
-	_unaligned_load64(p0, (uint32_t *)(dest + 0 * stride));
+	_unaligned_load64(p0, (unsigned char *)(dest + 0 * stride));
 	i0 = (vector signed short)vec_mergeh(zero, p0);
-	_unaligned_load64(p1, (uint32_t *)(dest + 1 * stride));
+	_unaligned_load64(p1, (unsigned char *)(dest + 1 * stride));
 	i4 = vec_add(i0, in0);
 	i1 = (vector signed short)vec_mergeh(zero, p1);
-	_unaligned_load64(p2, (uint32_t *)(dest + 2 * stride));
+	_unaligned_load64(p2, (unsigned char *)(dest + 2 * stride));
 	p0 = vec_packsu(i4, i4);
 	i5 = vec_add(i1, in1);
-	_unaligned_load64(p3, (uint32_t *)(dest + 3 * stride));
+	_unaligned_load64(p3, (unsigned char *)(dest + 3 * stride));
 	i2 = (vector signed short)vec_mergeh(zero, p2);
 	p1 = vec_packsu(i5, i5);
 	_unaligned_store64(p0, ltemp, (uint32_t *)(dest + 0 * stride));
@@ -854,15 +854,15 @@ void vpx_idct8x8_64_add_altivec(const tran_low_t *input, uint8_t *dest, int stri
 	p3 = vec_packsu(i7, i7);
 	_unaligned_store64(p3, ltemp, (uint32_t *)(dest + 3 * stride));
 	
-	_unaligned_load64(p0, (uint32_t *)(dest + 4 * stride));
+	_unaligned_load64(p0, (unsigned char *)(dest + 4 * stride));
 	i0 = (vector signed short)vec_mergeh(zero, p0);
-	_unaligned_load64(p1, (uint32_t *)(dest + 5 * stride));
+	_unaligned_load64(p1, (unsigned char *)(dest + 5 * stride));
 	i4 = vec_add(i0, in4);
 	i1 = (vector signed short)vec_mergeh(zero, p1);
-	_unaligned_load64(p2, (uint32_t *)(dest + 6 * stride));
+	_unaligned_load64(p2, (unsigned char *)(dest + 6 * stride));
 	p0 = vec_packsu(i4, i4);
 	i5 = vec_add(i1, in5);
-	_unaligned_load64(p3, (uint32_t *)(dest + 7 * stride));
+	_unaligned_load64(p3, (unsigned char *)(dest + 7 * stride));
 	i2 = (vector signed short)vec_mergeh(zero, p2);
 	p1 = vec_packsu(i5, i5);
 	_unaligned_store64(p0, ltemp, (uint32_t *)(dest + 4 * stride));
@@ -912,8 +912,8 @@ static void iadst8_vmx(vector signed short *in) {
 	vector signed short k__cospi_p16_m16 = short_pair_a(cospi_16_64, -cospi_16_64);
 	signed short a_s = (int16_t)cospi_16_64;
 	vector signed short k__cospi_p16_p16;
-	vector signed short k__const_0 = vec_splat_u16(0);
-	vector signed int zero_i = vec_splat_u16(0);
+	vector signed short k__const_0 = vec_splat_s16(0);
+	vector signed int zero_i = vec_splat_s32(0);
 	vector signed int u0, u1, u2, u3, u4, u5, u6, u7, u8, u9, u10, u11, u12, u13, u14, u15;
 	vector signed int v0, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15;
 	vector signed int w0, w1, w2, w3, w4, w5, w6, w7, w8, w9, w10, w11, w12, w13, w14, w15;
@@ -1130,7 +1130,7 @@ static void iadst8_vmx(vector signed short *in) {
 void vp9_iht8x8_64_add_altivec(const tran_low_t *input, uint8_t *dest, int stride,
                             int tx_type) {
 	vector signed short in[8];
-	vector unsigned char zero = vec_splat_s8(0);
+	vector unsigned char zero = vec_splat_u8(0);
 	vector unsigned int ltemp;
 	vector unsigned char p0, p1, p2, p3, p4, p5, p6, p7;
 	vector signed short i0, i1, i2, i3, i4, i5, i6, i7;
@@ -1139,17 +1139,17 @@ void vp9_iht8x8_64_add_altivec(const tran_low_t *input, uint8_t *dest, int strid
 	vector unsigned short four = vec_splat_u16(4);
 	vector unsigned short five = vec_splat_u16(5);
 	vector signed short final_rounding;
-	final_rounding = vec_rl(one, four);
+	final_rounding = (vector signed short)vec_rl(one, four);
 
 	// The Intel mnemonic implies these loads are aligned.
-	in[0] = vec_ld(0,   (uint32_t *)input);
-	in[1] = vec_ld(16,  (uint32_t *)input);
-	in[2] = vec_ld(32,  (uint32_t *)input);
-	in[3] = vec_ld(48,  (uint32_t *)input);
-	in[4] = vec_ld(64,  (uint32_t *)input);
-	in[5] = vec_ld(80,  (uint32_t *)input);
-	in[6] = vec_ld(96,  (uint32_t *)input);
-	in[7] = vec_ld(112, (uint32_t *)input);
+	in[0] = vec_ld(0,   (const signed short *)input);
+	in[1] = vec_ld(16,  (const signed short *)input);
+	in[2] = vec_ld(32,  (const signed short *)input);
+	in[3] = vec_ld(48,  (const signed short *)input);
+	in[4] = vec_ld(64,  (const signed short *)input);
+	in[5] = vec_ld(80,  (const signed short *)input);
+	in[6] = vec_ld(96,  (const signed short *)input);
+	in[7] = vec_ld(112, (const signed short *)input);
 
 	switch (tx_type) {
 		case 0:  // DCT_DCT
@@ -1193,15 +1193,15 @@ void vp9_iht8x8_64_add_altivec(const tran_low_t *input, uint8_t *dest, int strid
 	in[6] = vec_sra(in[6], five);
 	in[7] = vec_sra(in[7], five);
 
-	_unaligned_load64(p0, (uint32_t *)(dest + 0 * stride));
+	_unaligned_load64(p0, (unsigned char *)(dest + 0 * stride));
 	i0 = (vector signed short)vec_mergeh(zero, p0);
-	_unaligned_load64(p1, (uint32_t *)(dest + 1 * stride));
+	_unaligned_load64(p1, (unsigned char *)(dest + 1 * stride));
 	i4 = vec_add(i0, in[0]);
 	i1 = (vector signed short)vec_mergeh(zero, p1);
-	_unaligned_load64(p2, (uint32_t *)(dest + 2 * stride));
+	_unaligned_load64(p2, (unsigned char *)(dest + 2 * stride));
 	p0 = vec_packsu(i4, i4);
 	i5 = vec_add(i1, in[1]);
-	_unaligned_load64(p3, (uint32_t *)(dest + 3 * stride));
+	_unaligned_load64(p3, (unsigned char *)(dest + 3 * stride));
 	i2 = (vector signed short)vec_mergeh(zero, p2);
 	p1 = vec_packsu(i5, i5);
 	_unaligned_store64(p0, ltemp, (uint32_t *)(dest + 0 * stride));
@@ -1214,15 +1214,15 @@ void vp9_iht8x8_64_add_altivec(const tran_low_t *input, uint8_t *dest, int strid
 	p3 = vec_packsu(i7, i7);
 	_unaligned_store64(p3, ltemp, (uint32_t *)(dest + 3 * stride));
 	
-	_unaligned_load64(p0, (uint32_t *)(dest + 4 * stride));
+	_unaligned_load64(p0, (unsigned char *)(dest + 4 * stride));
 	i0 = (vector signed short)vec_mergeh(zero, p0);
-	_unaligned_load64(p1, (uint32_t *)(dest + 5 * stride));
+	_unaligned_load64(p1, (unsigned char *)(dest + 5 * stride));
 	i4 = vec_add(i0, in[4]);
 	i1 = (vector signed short)vec_mergeh(zero, p1);
-	_unaligned_load64(p2, (uint32_t *)(dest + 6 * stride));
+	_unaligned_load64(p2, (unsigned char *)(dest + 6 * stride));
 	p0 = vec_packsu(i4, i4);
 	i5 = vec_add(i1, in[5]);
-	_unaligned_load64(p3, (uint32_t *)(dest + 7 * stride));
+	_unaligned_load64(p3, (unsigned char *)(dest + 7 * stride));
 	i2 = (vector signed short)vec_mergeh(zero, p2);
 	p1 = vec_packsu(i5, i5);
 	_unaligned_store64(p0, ltemp, (uint32_t *)(dest + 4 * stride));
@@ -1259,12 +1259,12 @@ void vp9_iht8x8_64_add_altivec(const tran_low_t *input, uint8_t *dest, int strid
   {                                            \
 	vector signed short tr0_0 = vec_mergeh(in0, in1); \
 	vector signed short tr0_1 = vec_mergeh(in2, in3); \
-	out0 = vec_mergeh((vector signed int)tr0_0, (vector signed int)tr0_1); \
-	out1 = vec_mergel((vector signed int)tr0_0, (vector signed int)tr0_1); \
+	out0 = (vector signed short)vec_mergeh((vector signed int)tr0_0, (vector signed int)tr0_1); \
+	out1 = (vector signed short)vec_mergel((vector signed int)tr0_0, (vector signed int)tr0_1); \
   }
 
 void vpx_idct8x8_12_add_altivec(const tran_low_t *input, uint8_t *dest, int stride) {
-	vector signed short zero_s = vec_splat_u16(0);
+	vector signed short zero_s = vec_splat_s16(0);
 	vector unsigned char zero = vec_splat_u8(0);
 	vector signed short stg1_0 = short_pair_a(cospi_28_64, -cospi_4_64);
 	vector signed short stg1_1 = short_pair_a(cospi_4_64, cospi_28_64);
@@ -1289,14 +1289,14 @@ void vpx_idct8x8_12_add_altivec(const tran_low_t *input, uint8_t *dest, int stri
 	vector unsigned short four = vec_splat_u16(4);
 	vector unsigned short five = vec_splat_u16(5);
 	vector signed short final_rounding;
-	final_rounding = vec_rl(one, four);
+	final_rounding = (vector signed short)vec_rl(one, four);
 
 	// Rows. Load 4-row input data.
 	// Praise the Lord it's still aligned, yes?
-	in0 = vec_ld(0,  (uint32_t *)input);
-	in1 = vec_ld(16, (uint32_t *)input);
-	in2 = vec_ld(32, (uint32_t *)input);
-	in3 = vec_ld(48, (uint32_t *)input);
+	in0 = vec_ld(0,  (const signed short *)input);
+	in1 = vec_ld(16, (const signed short *)input);
+	in2 = vec_ld(32, (const signed short *)input);
+	in3 = vec_ld(48, (const signed short *)input);
 
 	// 8x4 transpose.
 	TRANSPOSE_8X8_10(in0, in1, in2, in3, in0, in1);
@@ -1399,15 +1399,15 @@ void vpx_idct8x8_12_add_altivec(const tran_low_t *input, uint8_t *dest, int stri
 	in6 = vec_sra(in6, five);
 	in7 = vec_sra(in7, five);
 
-	_unaligned_load64(p0, (uint32_t *)(dest + 0 * stride));
+	_unaligned_load64(p0, (unsigned char *)(dest + 0 * stride));
 	i0 = (vector signed short)vec_mergeh(zero, p0);
-	_unaligned_load64(p1, (uint32_t *)(dest + 1 * stride));
+	_unaligned_load64(p1, (unsigned char *)(dest + 1 * stride));
 	i4 = vec_add(i0, in0);
 	i1 = (vector signed short)vec_mergeh(zero, p1);
-	_unaligned_load64(p2, (uint32_t *)(dest + 2 * stride));
+	_unaligned_load64(p2, (unsigned char *)(dest + 2 * stride));
 	p0 = vec_packsu(i4, i4);
 	i5 = vec_add(i1, in1);
-	_unaligned_load64(p3, (uint32_t *)(dest + 3 * stride));
+	_unaligned_load64(p3, (unsigned char *)(dest + 3 * stride));
 	i2 = (vector signed short)vec_mergeh(zero, p2);
 	p1 = vec_packsu(i5, i5);
 	_unaligned_store64(p0, ltemp, (uint32_t *)(dest + 0 * stride));
@@ -1420,15 +1420,15 @@ void vpx_idct8x8_12_add_altivec(const tran_low_t *input, uint8_t *dest, int stri
 	p3 = vec_packsu(i7, i7);
 	_unaligned_store64(p3, ltemp, (uint32_t *)(dest + 3 * stride));
 	
-	_unaligned_load64(p0, (uint32_t *)(dest + 4 * stride));
+	_unaligned_load64(p0, (unsigned char *)(dest + 4 * stride));
 	i0 = (vector signed short)vec_mergeh(zero, p0);
-	_unaligned_load64(p1, (uint32_t *)(dest + 5 * stride));
+	_unaligned_load64(p1, (unsigned char *)(dest + 5 * stride));
 	i4 = vec_add(i0, in4);
 	i1 = (vector signed short)vec_mergeh(zero, p1);
-	_unaligned_load64(p2, (uint32_t *)(dest + 6 * stride));
+	_unaligned_load64(p2, (unsigned char *)(dest + 6 * stride));
 	p0 = vec_packsu(i4, i4);
 	i5 = vec_add(i1, in5);
-	_unaligned_load64(p3, (uint32_t *)(dest + 7 * stride));
+	_unaligned_load64(p3, (unsigned char *)(dest + 7 * stride));
 	i2 = (vector signed short)vec_mergeh(zero, p2);
 	p1 = vec_packsu(i5, i5);
 	_unaligned_store64(p0, ltemp, (uint32_t *)(dest + 4 * stride));
@@ -1639,13 +1639,13 @@ void vpx_idct16x16_256_add_altivec(const tran_low_t *input, uint8_t *dest,
 	vector unsigned char p0;
 	vector unsigned char zero = vec_splat_u8(0);
 	vector unsigned short one = vec_splat_u16(1);
-	vector unsigned short five = vec_splat_u8(5);
-	vector unsigned short six = vec_splat_u8(6);
+	vector unsigned short five = vec_splat_u16(5);
+	vector unsigned short six = vec_splat_u16(6);
 	vector signed short final_rounding;
 	vector unsigned int ltemp;
 	int i, j;
 
-	final_rounding = vec_rl(one, five);
+	final_rounding = (vector signed short)vec_rl(one, five);
 
 	// Part I.
 	// Little benefit to unrolling these loops since we alternate
@@ -1655,22 +1655,22 @@ void vpx_idct16x16_256_add_altivec(const tran_low_t *input, uint8_t *dest,
 		// 1-D IDCT.
 		// Load (aligned) input.
 		
-		in[0]  = vec_ld(0,   (uint32_t *)input);
-		in[8]  = vec_ld(16,  (uint32_t *)input);
-		in[1]  = vec_ld(32,  (uint32_t *)input);
-		in[9]  = vec_ld(48,  (uint32_t *)input);
-		in[2]  = vec_ld(64,  (uint32_t *)input);
-		in[10] = vec_ld(80,  (uint32_t *)input);
-		in[3]  = vec_ld(96,  (uint32_t *)input);
-		in[11] = vec_ld(112, (uint32_t *)input);
-		in[4]  = vec_ld(128, (uint32_t *)input);
-		in[12] = vec_ld(144, (uint32_t *)input);
-		in[5]  = vec_ld(160, (uint32_t *)input);
-		in[13] = vec_ld(176, (uint32_t *)input);
-		in[6]  = vec_ld(192, (uint32_t *)input);
-		in[14] = vec_ld(208, (uint32_t *)input);
-		in[7]  = vec_ld(224, (uint32_t *)input);
-		in[15] = vec_ld(240, (uint32_t *)input);
+		in[0]  = vec_ld(0,   (const signed short *)input);
+		in[8]  = vec_ld(16,  (const signed short *)input);
+		in[1]  = vec_ld(32,  (const signed short *)input);
+		in[9]  = vec_ld(48,  (const signed short *)input);
+		in[2]  = vec_ld(64,  (const signed short *)input);
+		in[10] = vec_ld(80,  (const signed short *)input);
+		in[3]  = vec_ld(96,  (const signed short *)input);
+		in[11] = vec_ld(112, (const signed short *)input);
+		in[4]  = vec_ld(128, (const signed short *)input);
+		in[12] = vec_ld(144, (const signed short *)input);
+		in[5]  = vec_ld(160, (const signed short *)input);
+		in[13] = vec_ld(176, (const signed short *)input);
+		in[6]  = vec_ld(192, (const signed short *)input);
+		in[14] = vec_ld(208, (const signed short *)input);
+		in[7]  = vec_ld(224, (const signed short *)input);
+		in[15] = vec_ld(240, (const signed short *)input);
 		array_transpose_8x8(in, in);
 		array_transpose_8x8(in + 8, in + 8);
 
@@ -1730,7 +1730,7 @@ void vpx_idct16x16_256_add_altivec(const tran_low_t *input, uint8_t *dest,
 			in[j] = vec_sra(in[j], six);
 			
 			// Reconstitute and store.
-			_unaligned_load64(p0, (uint32_t *)(dest + j * stride));
+			_unaligned_load64(p0, (unsigned char *)(dest + j * stride));
 			i0 = (vector signed short)vec_mergeh(zero, p0);
 			i0 = vec_add(in[j], i0);
 			p0 = vec_packsu(i0, i0);
@@ -1750,8 +1750,8 @@ void vpx_idct16x16_256_add_altivec(const tran_low_t *input, uint8_t *dest,
 	vector signed short tr0_0 = vec_mergeh(in0, in1); \
 	vector signed short tr0_1 = vec_mergeh(in2, in3); \
                                                         \
-	out0 = vec_mergeh((vector signed int)tr0_0, (vector signed int)tr0_1);  /* i1 i0 */  \
-	out1 = vec_mergel((vector signed int)tr0_0, (vector signed int)tr0_1);  /* i3 i2 */  \
+	out0 = (vector signed short)vec_mergeh((vector signed int)tr0_0, (vector signed int)tr0_1);  /* i1 i0 */  \
+	out1 = (vector signed short)vec_mergel((vector signed int)tr0_0, (vector signed int)tr0_1);  /* i3 i2 */  \
   }
 
 // However, *this* is correct.
@@ -1919,7 +1919,7 @@ void vpx_idct16x16_10_add_altivec(const tran_low_t *input, uint8_t *dest,
 	vector signed short smp0, smp1, smp2, smp3;
 	vector signed int tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
 	vector unsigned char p0;
-	vector signed short zero = vec_splat_u16(0);
+	vector signed short zero = vec_splat_s16(0);
 	vector unsigned char zero_c = vec_splat_u8(0);
 	vector unsigned short one = vec_splat_u16(1);
 	vector unsigned short five = vec_splat_u16(5);
@@ -1928,14 +1928,14 @@ void vpx_idct16x16_10_add_altivec(const tran_low_t *input, uint8_t *dest,
 	vector unsigned int ltemp;
 	int i, j;
 
-	final_rounding = vec_rl(one, five);
+	final_rounding = (vector signed short)vec_rl(one, five);
 
 	// First 1-D inverse DCT.
 	// Stage 1: load input data (aligned).
-	in[0] = vec_ld(0,  (uint32_t *)input);
-	in[1] = vec_ld(32, (uint32_t *)input); // Yes, really. We're skipping every other.
-	in[2] = vec_ld(64, (uint32_t *)input);
-	in[3] = vec_ld(96, (uint32_t *)input);
+	in[0] = vec_ld(0,  (const signed short *)input);
+	in[1] = vec_ld(32, (const signed short *)input); // Yes, really. We're skipping every other.
+	in[2] = vec_ld(64, (const signed short *)input);
+	in[3] = vec_ld(96, (const signed short *)input);
 
 	TRANSPOSE_8X4(in[0], in[1], in[2], in[3], in[0], in[1]);
 
@@ -2002,10 +2002,10 @@ void vpx_idct16x16_10_add_altivec(const tran_low_t *input, uint8_t *dest,
 
 	// Stages 5 and 6.
 	{
-		tmp0 = vec_add(stp2_8, stp2_11);
-		tmp1 = vec_sub(stp2_8, stp2_11);
-		tmp2 = vec_add(stp2_9, stp2_10);
-		tmp3 = vec_sub(stp2_9, stp2_10);
+		tmp0 = (vector signed int)vec_add(stp2_8, stp2_11);
+		tmp1 = (vector signed int)vec_sub(stp2_8, stp2_11);
+		tmp2 = (vector signed int)vec_add(stp2_9, stp2_10);
+		tmp3 = (vector signed int)vec_sub(stp2_9, stp2_10);
 
 		stp1_9  = (vector signed short)vec_perm(tmp2, (vector signed int)zero, merge64_hvec);
 		stp1_10 = (vector signed short)vec_perm(tmp3, (vector signed int)zero, merge64_hvec);
@@ -2105,7 +2105,7 @@ void vpx_idct16x16_10_add_altivec(const tran_low_t *input, uint8_t *dest,
 			in[j] = vec_sra(in[j], six);
 			
 			// Reconstitute and store.
-			_unaligned_load64(p0, (uint32_t *)(dest + j * stride));
+			_unaligned_load64(p0, (unsigned char *)(dest + j * stride));
 			i0 = (vector signed short)vec_mergeh(zero_c, p0);
 			i0 = vec_add(in[j], i0);
 			p0 = vec_packsu(i0, i0);
@@ -2438,7 +2438,7 @@ void vpx_idct32x32_34_add_altivec(const tran_low_t *input, uint8_t *dest,
 	vector signed short stg4_6 = short_pair_a(-cospi_24_64, -cospi_8_64);
 
 	vector signed short stg6_0 = short_pair_a(-cospi_16_64, cospi_16_64);
-	vector signed short zero = vec_splat_u16(0);
+	vector signed short zero = vec_splat_s16(0);
 	vector unsigned char zero_c = vec_splat_u8(0);
 	vector unsigned short one = vec_splat_u16(1);
 	vector unsigned short five = vec_splat_u16(5);
@@ -2461,17 +2461,17 @@ void vpx_idct32x32_34_add_altivec(const tran_low_t *input, uint8_t *dest,
 	vector signed int tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
 	int i, j;
 
-	final_rounding = vec_rl(one, five);
+	final_rounding = (vector signed short)vec_rl(one, five);
 	
 	// Load vectors from top left 8x8 block.
-	in[0] = vec_ld(0,   (uint32_t *)input);
-	in[1] = vec_ld(64,  (uint32_t *)input);
-	in[2] = vec_ld(128, (uint32_t *)input);
-	in[3] = vec_ld(192, (uint32_t *)input);
-	in[4] = vec_ld(256, (uint32_t *)input);
-	in[5] = vec_ld(320, (uint32_t *)input);
-	in[6] = vec_ld(384, (uint32_t *)input);
-	in[7] = vec_ld(448, (uint32_t *)input);
+	in[0] = vec_ld(0,   (const signed short *)input);
+	in[1] = vec_ld(64,  (const signed short *)input);
+	in[2] = vec_ld(128, (const signed short *)input);
+	in[3] = vec_ld(192, (const signed short *)input);
+	in[4] = vec_ld(256, (const signed short *)input);
+	in[5] = vec_ld(320, (const signed short *)input);
+	in[6] = vec_ld(384, (const signed short *)input);
+	in[7] = vec_ld(448, (const signed short *)input);
 
 	// 1-D IDCT.
 	// Store intermediate results for each 8x32 block.
@@ -2554,7 +2554,7 @@ void vpx_idct32x32_34_add_altivec(const tran_low_t *input, uint8_t *dest,
 			in[j] = vec_adds(in[j], final_rounding);
 			in[j] = vec_sra(in[j], six);
 			
-			_unaligned_load64(p0, (uint32_t *)(dest + j * stride));
+			_unaligned_load64(p0, (unsigned char *)(dest + j * stride));
 			i0 = (vector signed short)vec_mergeh(zero_c, p0);
 			i0 = vec_add(in[j], i0);
 			p0 = vec_packsu(i0, i0);
@@ -2953,7 +2953,7 @@ void vpx_idct32x32_1024_add_altivec(const tran_low_t *input, uint8_t *dest,
 		stp2_30, stp2_31;
 	vector signed int tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
 	int i, j, i32;
-	vector signed short zero = vec_splat_u16(0);
+	vector signed short zero = vec_splat_s16(0);
 	vector unsigned char zero_c = vec_splat_u8(0);
 	vector unsigned short one = vec_splat_u16(1);
 	vector unsigned short five = vec_splat_u16(5);
@@ -2962,45 +2962,45 @@ void vpx_idct32x32_1024_add_altivec(const tran_low_t *input, uint8_t *dest,
 	vector unsigned int ltemp;
 	vector unsigned char p0;
 
-	final_rounding = vec_rl(one, five);
+	final_rounding = (vector signed short)vec_rl(one, five);
 	
 	for(i=0;i<4;i++) {
 		// 1-D IDCT.
 		// Load next block.
 		i32 = (i << 5);
 		
-		in[0]  = vec_ld(0,   (uint32_t *)input);
-		in[8]  = vec_ld(16,  (uint32_t *)input);
-		in[16] = vec_ld(32,  (uint32_t *)input);
-		in[24] = vec_ld(48,  (uint32_t *)input);
-		in[1]  = vec_ld(64,  (uint32_t *)input);
-		in[9]  = vec_ld(80,  (uint32_t *)input);
-		in[17] = vec_ld(96,  (uint32_t *)input);
-		in[25] = vec_ld(112, (uint32_t *)input);
-		in[2]  = vec_ld(128, (uint32_t *)input);
-		in[10] = vec_ld(144, (uint32_t *)input);
-		in[18] = vec_ld(160, (uint32_t *)input);
-		in[26] = vec_ld(176, (uint32_t *)input);
-		in[3]  = vec_ld(192, (uint32_t *)input);
-		in[11] = vec_ld(208, (uint32_t *)input);
-		in[19] = vec_ld(224, (uint32_t *)input);
-		in[27] = vec_ld(240, (uint32_t *)input);
-		in[4]  = vec_ld(256, (uint32_t *)input);
-		in[12] = vec_ld(272, (uint32_t *)input);
-		in[20] = vec_ld(288, (uint32_t *)input);
-		in[28] = vec_ld(304, (uint32_t *)input);
-		in[5]  = vec_ld(320, (uint32_t *)input);
-		in[13] = vec_ld(336, (uint32_t *)input);
-		in[21] = vec_ld(352, (uint32_t *)input);
-		in[29] = vec_ld(368, (uint32_t *)input);
-		in[6]  = vec_ld(384, (uint32_t *)input);
-		in[14] = vec_ld(400, (uint32_t *)input);
-		in[22] = vec_ld(416, (uint32_t *)input);
-		in[30] = vec_ld(432, (uint32_t *)input);
-		in[7]  = vec_ld(448, (uint32_t *)input);
-		in[15] = vec_ld(464, (uint32_t *)input);
-		in[23] = vec_ld(480, (uint32_t *)input);
-		in[31] = vec_ld(496, (uint32_t *)input);
+		in[0]  = vec_ld(0,   (const signed short *)input);
+		in[8]  = vec_ld(16,  (const signed short *)input);
+		in[16] = vec_ld(32,  (const signed short *)input);
+		in[24] = vec_ld(48,  (const signed short *)input);
+		in[1]  = vec_ld(64,  (const signed short *)input);
+		in[9]  = vec_ld(80,  (const signed short *)input);
+		in[17] = vec_ld(96,  (const signed short *)input);
+		in[25] = vec_ld(112, (const signed short *)input);
+		in[2]  = vec_ld(128, (const signed short *)input);
+		in[10] = vec_ld(144, (const signed short *)input);
+		in[18] = vec_ld(160, (const signed short *)input);
+		in[26] = vec_ld(176, (const signed short *)input);
+		in[3]  = vec_ld(192, (const signed short *)input);
+		in[11] = vec_ld(208, (const signed short *)input);
+		in[19] = vec_ld(224, (const signed short *)input);
+		in[27] = vec_ld(240, (const signed short *)input);
+		in[4]  = vec_ld(256, (const signed short *)input);
+		in[12] = vec_ld(272, (const signed short *)input);
+		in[20] = vec_ld(288, (const signed short *)input);
+		in[28] = vec_ld(304, (const signed short *)input);
+		in[5]  = vec_ld(320, (const signed short *)input);
+		in[13] = vec_ld(336, (const signed short *)input);
+		in[21] = vec_ld(352, (const signed short *)input);
+		in[29] = vec_ld(368, (const signed short *)input);
+		in[6]  = vec_ld(384, (const signed short *)input);
+		in[14] = vec_ld(400, (const signed short *)input);
+		in[22] = vec_ld(416, (const signed short *)input);
+		in[30] = vec_ld(432, (const signed short *)input);
+		in[7]  = vec_ld(448, (const signed short *)input);
+		in[15] = vec_ld(464, (const signed short *)input);
+		in[23] = vec_ld(480, (const signed short *)input);
+		in[31] = vec_ld(496, (const signed short *)input);
 		input += 256;
 		
 		// Are all entries zero? If so, the result is zero.
@@ -3041,38 +3041,38 @@ void vpx_idct32x32_1024_add_altivec(const tran_low_t *input, uint8_t *dest,
 		zero_idx[14] = vec_or(zero_idx[12], zero_idx[13]);
 
 		if(vec_all_eq(zero_idx[14], zero)) {
-			col[i32 +  0] = vec_splat_u16(0);
-			col[i32 +  1] = vec_splat_u16(0);
-			col[i32 +  2] = vec_splat_u16(0);
-			col[i32 +  3] = vec_splat_u16(0);
-			col[i32 +  4] = vec_splat_u16(0);
-			col[i32 +  5] = vec_splat_u16(0);
-			col[i32 +  6] = vec_splat_u16(0);
-			col[i32 +  7] = vec_splat_u16(0);
-			col[i32 +  8] = vec_splat_u16(0);
-			col[i32 +  9] = vec_splat_u16(0);
-			col[i32 + 10] = vec_splat_u16(0);
-			col[i32 + 11] = vec_splat_u16(0);
-			col[i32 + 12] = vec_splat_u16(0);
-			col[i32 + 13] = vec_splat_u16(0);
-			col[i32 + 14] = vec_splat_u16(0);
-			col[i32 + 15] = vec_splat_u16(0);
-			col[i32 + 16] = vec_splat_u16(0);
-			col[i32 + 17] = vec_splat_u16(0);
-			col[i32 + 18] = vec_splat_u16(0);
-			col[i32 + 19] = vec_splat_u16(0);
-			col[i32 + 20] = vec_splat_u16(0);
-			col[i32 + 21] = vec_splat_u16(0);
-			col[i32 + 22] = vec_splat_u16(0);
-			col[i32 + 23] = vec_splat_u16(0);
-			col[i32 + 24] = vec_splat_u16(0);
-			col[i32 + 25] = vec_splat_u16(0);
-			col[i32 + 26] = vec_splat_u16(0);
-			col[i32 + 27] = vec_splat_u16(0);
-			col[i32 + 28] = vec_splat_u16(0);
-			col[i32 + 29] = vec_splat_u16(0);
-			col[i32 + 30] = vec_splat_u16(0);
-			col[i32 + 31] = vec_splat_u16(0);
+			col[i32 +  0] = vec_splat_s16(0);
+			col[i32 +  1] = vec_splat_s16(0);
+			col[i32 +  2] = vec_splat_s16(0);
+			col[i32 +  3] = vec_splat_s16(0);
+			col[i32 +  4] = vec_splat_s16(0);
+			col[i32 +  5] = vec_splat_s16(0);
+			col[i32 +  6] = vec_splat_s16(0);
+			col[i32 +  7] = vec_splat_s16(0);
+			col[i32 +  8] = vec_splat_s16(0);
+			col[i32 +  9] = vec_splat_s16(0);
+			col[i32 + 10] = vec_splat_s16(0);
+			col[i32 + 11] = vec_splat_s16(0);
+			col[i32 + 12] = vec_splat_s16(0);
+			col[i32 + 13] = vec_splat_s16(0);
+			col[i32 + 14] = vec_splat_s16(0);
+			col[i32 + 15] = vec_splat_s16(0);
+			col[i32 + 16] = vec_splat_s16(0);
+			col[i32 + 17] = vec_splat_s16(0);
+			col[i32 + 18] = vec_splat_s16(0);
+			col[i32 + 19] = vec_splat_s16(0);
+			col[i32 + 20] = vec_splat_s16(0);
+			col[i32 + 21] = vec_splat_s16(0);
+			col[i32 + 22] = vec_splat_s16(0);
+			col[i32 + 23] = vec_splat_s16(0);
+			col[i32 + 24] = vec_splat_s16(0);
+			col[i32 + 25] = vec_splat_s16(0);
+			col[i32 + 26] = vec_splat_s16(0);
+			col[i32 + 27] = vec_splat_s16(0);
+			col[i32 + 28] = vec_splat_s16(0);
+			col[i32 + 29] = vec_splat_s16(0);
+			col[i32 + 30] = vec_splat_s16(0);
+			col[i32 + 31] = vec_splat_s16(0);
 			continue;
 		}
 		
@@ -3167,7 +3167,7 @@ void vpx_idct32x32_1024_add_altivec(const tran_low_t *input, uint8_t *dest,
 			in[j] = vec_adds(in[j], final_rounding);
 			in[j] = vec_sra(in[j], six);
 			
-			_unaligned_load64(p0, (uint32_t *)(dest + j * stride));
+			_unaligned_load64(p0, (unsigned char *)(dest + j * stride));
 			i0 = (vector signed short)vec_mergeh(zero_c, p0);
 			i0 = vec_add(in[j], i0);
 			p0 = vec_packsu(i0, i0);
@@ -3195,22 +3195,22 @@ static inline void array_transpose_16x16(vector signed short *res0, vector signe
 }
 
 static inline void load_buffer_8x16(const tran_low_t *input, vector signed short *in) {
-  	in[0]  = vec_ld(0,   (uint32_t *)input);
-  	in[1]  = vec_ld(32,  (uint32_t *)input);
-  	in[2]  = vec_ld(64,  (uint32_t *)input);
-  	in[3]  = vec_ld(96,  (uint32_t *)input);
-  	in[4]  = vec_ld(128, (uint32_t *)input);
-  	in[5]  = vec_ld(160, (uint32_t *)input);
-  	in[6]  = vec_ld(192, (uint32_t *)input);
-  	in[7]  = vec_ld(224, (uint32_t *)input);
-  	in[8]  = vec_ld(256, (uint32_t *)input);
-  	in[9]  = vec_ld(288, (uint32_t *)input);
-  	in[10] = vec_ld(320, (uint32_t *)input);
-  	in[11] = vec_ld(352, (uint32_t *)input);
-  	in[12] = vec_ld(384, (uint32_t *)input);
-  	in[13] = vec_ld(416, (uint32_t *)input);
-  	in[14] = vec_ld(448, (uint32_t *)input);
-  	in[15] = vec_ld(480, (uint32_t *)input);
+  	in[0]  = vec_ld(0,   (const signed short *)input);
+  	in[1]  = vec_ld(32,  (const signed short *)input);
+  	in[2]  = vec_ld(64,  (const signed short *)input);
+  	in[3]  = vec_ld(96,  (const signed short *)input);
+  	in[4]  = vec_ld(128, (const signed short *)input);
+  	in[5]  = vec_ld(160, (const signed short *)input);
+  	in[6]  = vec_ld(192, (const signed short *)input);
+  	in[7]  = vec_ld(224, (const signed short *)input);
+  	in[8]  = vec_ld(256, (const signed short *)input);
+  	in[9]  = vec_ld(288, (const signed short *)input);
+  	in[10] = vec_ld(320, (const signed short *)input);
+  	in[11] = vec_ld(352, (const signed short *)input);
+  	in[12] = vec_ld(384, (const signed short *)input);
+  	in[13] = vec_ld(416, (const signed short *)input);
+  	in[14] = vec_ld(448, (const signed short *)input);
+  	in[15] = vec_ld(480, (const signed short *)input);
 }
 
 static inline void write_buffer_8x16(uint8_t *dest, vector signed short *in, int stride) {
@@ -3223,14 +3223,14 @@ static inline void write_buffer_8x16(uint8_t *dest, vector signed short *in, int
 	vector unsigned char p0;
 	int i;
 	
-	final_rounding = vec_rl(one, five);
+	final_rounding = (vector signed short)vec_rl(one, five);
 	
 	// Round, shift, reconstitute and store, again.
 	for(i=0;i<16;i++) {
 		in[i] = vec_adds(in[i], final_rounding);
 		in[i] = vec_sra(in[i], six);
 
-		_unaligned_load64(p0, (uint32_t *)(dest + i * stride));
+		_unaligned_load64(p0, (unsigned char *)(dest + i * stride));
 		i0 = (vector signed short)vec_mergeh(zero_c, p0);
 		i0 = vec_add(in[i], i0);
 		p0 = vec_packsu(i0, i0);
@@ -3272,8 +3272,8 @@ static void iadst16_8col(vector signed short *in) {
 	vector signed short k__cospi_p16_p16;
 	vector signed short k__cospi_p16_m16 = short_pair_a(cospi_16_64, -cospi_16_64);
 	vector signed short k__cospi_m16_p16 = short_pair_a(-cospi_16_64, cospi_16_64);
-	vector signed short kZero = vec_splat_u16(0);
-	vector signed int mzero = vec_splat_u16(0);
+	vector signed short kZero = vec_splat_s16(0);
+	vector signed int mzero = vec_splat_s32(0);
 	
 	signed short a_s = ((int16_t)-cospi_16_64);
 	signed short b_s = ((int16_t)cospi_16_64);
@@ -3748,7 +3748,7 @@ static void idct16_8col(vector signed short *in) {
 	vector signed short s[16], t[16], u[16];
 	vector signed int v[16], w[16];
 	
-	signed short a_s = ((int16_t)cospi_16_64); int j;
+	signed short a_s = ((int16_t)cospi_16_64);
 	k__cospi_p16_p16 = ss_splat(&a_s);
 
 	// Stage 1.