vpx: hadamard: use ptrdiff_t instead of int for stride

Eliminates the following instruction for the x86 (64 bit)
intrinsic code:

movslq %esi,%rax

Change-Id: I8f5ebd40726f998708a668b0f52ea7a0576befae

test/hadamard_test.cc

@@ -22,7 +22,8 @@ namespace {
 
 using ::libvpx_test::ACMRandom;
 
-typedef void (*HadamardFunc)(const int16_t *a, int a_stride, tran_low_t *b);
+typedef void (*HadamardFunc)(const int16_t *a, ptrdiff_t a_stride,
+                             tran_low_t *b);
 
 void hadamard_loop(const int16_t *a, int a_stride, int16_t *out) {
   int16_t b[8];

vpx_dsp/arm/hadamard_neon.c

@@ -47,7 +47,7 @@ static void hadamard8x8_one_pass(int16x8_t *a0, int16x8_t *a1, int16x8_t *a2,
   *a7 = vaddq_s16(c1, c5);
 }
 
-void vpx_hadamard_8x8_neon(const int16_t *src_diff, int src_stride,
+void vpx_hadamard_8x8_neon(const int16_t *src_diff, ptrdiff_t src_stride,
                            tran_low_t *coeff) {
   int16x8_t a0 = vld1q_s16(src_diff);
   int16x8_t a1 = vld1q_s16(src_diff + src_stride);
@@ -76,7 +76,7 @@ void vpx_hadamard_8x8_neon(const int16_t *src_diff, int src_stride,
   store_s16q_to_tran_low(coeff + 56, a7);
 }
 
-void vpx_hadamard_16x16_neon(const int16_t *src_diff, int src_stride,
+void vpx_hadamard_16x16_neon(const int16_t *src_diff, ptrdiff_t src_stride,
                              tran_low_t *coeff) {
   int i;
 

vpx_dsp/avg.c

@@ -34,7 +34,7 @@ unsigned int vpx_avg_4x4_c(const uint8_t *s, int p) {
 
 // src_diff: first pass, 9 bit, dynamic range [-255, 255]
 //           second pass, 12 bit, dynamic range [-2040, 2040]
-static void hadamard_col8(const int16_t *src_diff, int src_stride,
+static void hadamard_col8(const int16_t *src_diff, ptrdiff_t src_stride,
                           int16_t *coeff) {
   int16_t b0 = src_diff[0 * src_stride] + src_diff[1 * src_stride];
   int16_t b1 = src_diff[0 * src_stride] - src_diff[1 * src_stride];
@@ -66,7 +66,7 @@ static void hadamard_col8(const int16_t *src_diff, int src_stride,
 
 // The order of the output coeff of the hadamard is not important. For
 // optimization purposes the final transpose may be skipped.
-void vpx_hadamard_8x8_c(const int16_t *src_diff, int src_stride,
+void vpx_hadamard_8x8_c(const int16_t *src_diff, ptrdiff_t src_stride,
                         tran_low_t *coeff) {
   int idx;
   int16_t buffer[64];
@@ -92,7 +92,7 @@ void vpx_hadamard_8x8_c(const int16_t *src_diff, int src_stride,
 }
 
 // In place 16x16 2D Hadamard transform
-void vpx_hadamard_16x16_c(const int16_t *src_diff, int src_stride,
+void vpx_hadamard_16x16_c(const int16_t *src_diff, ptrdiff_t src_stride,
                           tran_low_t *coeff) {
   int idx;
   for (idx = 0; idx < 4; ++idx) {

vpx_dsp/mips/avg_msa.c

@@ -56,7 +56,8 @@ uint32_t vpx_avg_4x4_msa(const uint8_t *src, int32_t src_stride) {
   return sum_out;
 }
 
-void vpx_hadamard_8x8_msa(const int16_t *src, int src_stride, int16_t *dst) {
+void vpx_hadamard_8x8_msa(const int16_t *src, ptrdiff_t src_stride,
+                          int16_t *dst) {
   v8i16 src0, src1, src2, src3, src4, src5, src6, src7;
   v8i16 tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
 
@@ -80,7 +81,8 @@ void vpx_hadamard_8x8_msa(const int16_t *src, int src_stride, int16_t *dst) {
   ST_SH8(src0, src1, src2, src3, src4, src5, src6, src7, dst, 8);
 }
 
-void vpx_hadamard_16x16_msa(const int16_t *src, int src_stride, int16_t *dst) {
+void vpx_hadamard_16x16_msa(const int16_t *src, ptrdiff_t src_stride,
+                            int16_t *dst) {
   v8i16 src0, src1, src2, src3, src4, src5, src6, src7, src8, src9, src10;
   v8i16 src11, src12, src13, src14, src15, tmp0, tmp1, tmp2, tmp3, tmp4, tmp5;
   v8i16 tmp6, tmp7, tmp8, tmp9, tmp10, tmp11, tmp12, tmp13, tmp14, tmp15;

vpx_dsp/ppc/hadamard_vsx.c

@@ -42,7 +42,7 @@ static void vpx_hadamard_s16_8x8_one_pass(int16x8_t v[8]) {
   v[7] = vec_add(c1, c5);
 }
 
-void vpx_hadamard_8x8_vsx(const int16_t *src_diff, int src_stride,
+void vpx_hadamard_8x8_vsx(const int16_t *src_diff, ptrdiff_t src_stride,
                           tran_low_t *coeff) {
   int16x8_t v[8];
 
@@ -71,7 +71,7 @@ void vpx_hadamard_8x8_vsx(const int16_t *src_diff, int src_stride,
   store_tran_low(v[7], 0, coeff + 56);
 }
 
-void vpx_hadamard_16x16_vsx(const int16_t *src_diff, int src_stride,
+void vpx_hadamard_16x16_vsx(const int16_t *src_diff, ptrdiff_t src_stride,
                             tran_low_t *coeff) {
   int i;
   const uint16x8_t ones = vec_splat_u16(1);

vpx_dsp/vpx_dsp_rtcd_defs.pl

@@ -765,19 +765,19 @@ if (vpx_config("CONFIG_VP9_ENCODER") eq "yes") {
   specialize qw/vpx_minmax_8x8 sse2 neon msa/;
 
   if (vpx_config("CONFIG_VP9_HIGHBITDEPTH") eq "yes") {
-    add_proto qw/void vpx_hadamard_8x8/, "const int16_t *src_diff, int src_stride, tran_low_t *coeff";
+    add_proto qw/void vpx_hadamard_8x8/, "const int16_t *src_diff, ptrdiff_t src_stride, tran_low_t *coeff";
     specialize qw/vpx_hadamard_8x8 sse2 neon vsx/, "$ssse3_x86_64";
 
-    add_proto qw/void vpx_hadamard_16x16/, "const int16_t *src_diff, int src_stride, tran_low_t *coeff";
+    add_proto qw/void vpx_hadamard_16x16/, "const int16_t *src_diff, ptrdiff_t src_stride, tran_low_t *coeff";
     specialize qw/vpx_hadamard_16x16 avx2 sse2 neon vsx/;
 
     add_proto qw/int vpx_satd/, "const tran_low_t *coeff, int length";
     specialize qw/vpx_satd sse2 neon/;
   } else {
-    add_proto qw/void vpx_hadamard_8x8/, "const int16_t *src_diff, int src_stride, int16_t *coeff";
+    add_proto qw/void vpx_hadamard_8x8/, "const int16_t *src_diff, ptrdiff_t src_stride, int16_t *coeff";
     specialize qw/vpx_hadamard_8x8 sse2 neon msa vsx/, "$ssse3_x86_64";
 
-    add_proto qw/void vpx_hadamard_16x16/, "const int16_t *src_diff, int src_stride, int16_t *coeff";
+    add_proto qw/void vpx_hadamard_16x16/, "const int16_t *src_diff, ptrdiff_t src_stride, int16_t *coeff";
     specialize qw/vpx_hadamard_16x16 avx2 sse2 neon msa vsx/;
 
     add_proto qw/int vpx_satd/, "const int16_t *coeff, int length";

vpx_dsp/x86/avg_intrin_avx2.c

@@ -91,7 +91,7 @@ static void hadamard_col8x2_avx2(__m256i *in, int iter) {
   }
 }
 
-static void hadamard_8x8x2_avx2(int16_t const *src_diff, int src_stride,
+static void hadamard_8x8x2_avx2(int16_t const *src_diff, ptrdiff_t src_stride,
                                 int16_t *coeff) {
   __m256i src[8];
   src[0] = _mm256_loadu_si256((const __m256i *)src_diff);
@@ -131,7 +131,7 @@ static void hadamard_8x8x2_avx2(int16_t const *src_diff, int src_stride,
                       _mm256_permute2x128_si256(src[6], src[7], 0x31));
 }
 
-void vpx_hadamard_16x16_avx2(int16_t const *src_diff, int src_stride,
+void vpx_hadamard_16x16_avx2(int16_t const *src_diff, ptrdiff_t src_stride,
                              tran_low_t *coeff) {
   int idx;
 #if CONFIG_VP9_HIGHBITDEPTH

vpx_dsp/x86/avg_intrin_sse2.c

@@ -214,7 +214,7 @@ static void hadamard_col8_sse2(__m128i *in, int iter) {
   }
 }
 
-void vpx_hadamard_8x8_sse2(int16_t const *src_diff, int src_stride,
+void vpx_hadamard_8x8_sse2(int16_t const *src_diff, ptrdiff_t src_stride,
                            tran_low_t *coeff) {
   __m128i src[8];
   src[0] = _mm_load_si128((const __m128i *)src_diff);
@@ -246,7 +246,7 @@ void vpx_hadamard_8x8_sse2(int16_t const *src_diff, int src_stride,
   store_tran_low(src[7], coeff);
 }
 
-void vpx_hadamard_16x16_sse2(int16_t const *src_diff, int src_stride,
+void vpx_hadamard_16x16_sse2(int16_t const *src_diff, ptrdiff_t src_stride,
                              tran_low_t *coeff) {
   int idx;
   for (idx = 0; idx < 4; ++idx) {