diff --git a/README b/README
index 60da8a2..0592727 100644
--- a/README
+++ b/README
@@ -4,7 +4,7 @@
           \__\__/\____/\_____/__/ ____  ___
                 / _/ /    \    \ /  _ \/ _/
                /  \_/   / /   \ \   __/  \__
-               \____/____/\_____/_____/____/v1.0.3
+               \____/____/\_____/_____/____/v1.1.0
 
 Description:
 ============
diff --git a/README.android b/README.android
index 9c49f0f..94fd4d4 100644
--- a/README.android
+++ b/README.android
@@ -1,5 +1,5 @@
 URL: https://chromium.googlesource.com/webm/libwebp
-Version: v1.0.3
+Version: v1.1.0
 License: Google BSD like
 
 Local modifications:
@@ -8,7 +8,6 @@
   (e.g. bits.h) to leak into
 - Removed build files necessary for building via autoconf/automake tools
   These files are not required to build via Android.bp
-- cherry-pick 0e48d889 bugfix: last alpha rows were incorrectly decoded
 
 The Android.bp file creates WebP decoder and encoder static libraries which
 can be added to any application by adding libwebp-decode and libwebp-encode to
diff --git a/README.version b/README.version
index 6fe7274..f94ac05 100644
--- a/README.version
+++ b/README.version
@@ -1,3 +1,3 @@
-URL: https://storage.googleapis.com/downloads.webmproject.org/releases/webp/libwebp-1.0.3.tar.gz
-Version: 1.0.3
+URL: https://storage.googleapis.com/downloads.webmproject.org/releases/webp/libwebp-1.1.0.tar.gz
+Version: 1.1.0
 BugComponent: 20174
diff --git a/include/webp/decode.h b/include/webp/decode.h
index ae8bfe8..80dd0ef 100644
--- a/include/webp/decode.h
+++ b/include/webp/decode.h
@@ -20,7 +20,7 @@
 extern "C" {
 #endif
 
-#define WEBP_DECODER_ABI_VERSION 0x0208    // MAJOR(8b) + MINOR(8b)
+#define WEBP_DECODER_ABI_VERSION 0x0209    // MAJOR(8b) + MINOR(8b)
 
 // Note: forward declaring enumerations is not allowed in (strict) C and C++,
 // the types are left here for reference.
@@ -91,9 +91,6 @@
                                    uint8_t** u, uint8_t** v,
                                    int* stride, int* uv_stride);
 
-// Releases memory returned by the WebPDecode*() functions above.
-WEBP_EXTERN void WebPFree(void* ptr);
-
 // These five functions are variants of the above ones, that decode the image
 // directly into a pre-allocated buffer 'output_buffer'. The maximum storage
 // available in this buffer is indicated by 'output_buffer_size'. If this
diff --git a/include/webp/encode.h b/include/webp/encode.h
index 339f881..655166e 100644
--- a/include/webp/encode.h
+++ b/include/webp/encode.h
@@ -20,7 +20,7 @@
 extern "C" {
 #endif
 
-#define WEBP_ENCODER_ABI_VERSION 0x020e    // MAJOR(8b) + MINOR(8b)
+#define WEBP_ENCODER_ABI_VERSION 0x020f    // MAJOR(8b) + MINOR(8b)
 
 // Note: forward declaring enumerations is not allowed in (strict) C and C++,
 // the types are left here for reference.
@@ -79,9 +79,6 @@
                                           int width, int height, int stride,
                                           uint8_t** output);
 
-// Releases memory returned by the WebPEncode*() functions above.
-WEBP_EXTERN void WebPFree(void* ptr);
-
 //------------------------------------------------------------------------------
 // Coding parameters
 
@@ -306,7 +303,7 @@
   // YUV input (mostly used for input to lossy compression)
   WebPEncCSP colorspace;     // colorspace: should be YUV420 for now (=Y'CbCr).
   int width, height;         // dimensions (less or equal to WEBP_MAX_DIMENSION)
-  uint8_t *y, *u, *v;        // pointers to luma/chroma planes.
+  uint8_t* y, *u, *v;        // pointers to luma/chroma planes.
   int y_stride, uv_stride;   // luma/chroma strides.
   uint8_t* a;                // pointer to the alpha plane
   int a_stride;              // stride of the alpha plane
@@ -350,7 +347,7 @@
   uint32_t pad3[3];       // padding for later use
 
   // Unused for now
-  uint8_t *pad4, *pad5;
+  uint8_t* pad4, *pad5;
   uint32_t pad6[8];       // padding for later use
 
   // PRIVATE FIELDS
diff --git a/include/webp/mux.h b/include/webp/mux.h
index 66096a9..7d27489 100644
--- a/include/webp/mux.h
+++ b/include/webp/mux.h
@@ -57,7 +57,7 @@
   WebPMuxGetChunk(mux, "ICCP", &icc_profile);
   // ... (Consume icc_data).
   WebPMuxDelete(mux);
-  free(data);
+  WebPFree(data);
 */
 
 // Note: forward declaring enumerations is not allowed in (strict) C and C++,
@@ -245,7 +245,7 @@
     WebPMux* mux, const WebPMuxFrameInfo* frame, int copy_data);
 
 // Gets the nth frame from the mux object.
-// The content of 'frame->bitstream' is allocated using malloc(), and NOT
+// The content of 'frame->bitstream' is allocated using WebPMalloc(), and NOT
 // owned by the 'mux' object. It MUST be deallocated by the caller by calling
 // WebPDataClear().
 // nth=0 has a special meaning - last position.
@@ -376,10 +376,10 @@
 // Assembles all chunks in WebP RIFF format and returns in 'assembled_data'.
 // This function also validates the mux object.
 // Note: The content of 'assembled_data' will be ignored and overwritten.
-// Also, the content of 'assembled_data' is allocated using malloc(), and NOT
-// owned by the 'mux' object. It MUST be deallocated by the caller by calling
-// WebPDataClear(). It's always safe to call WebPDataClear() upon return,
-// even in case of error.
+// Also, the content of 'assembled_data' is allocated using WebPMalloc(), and
+// NOT owned by the 'mux' object. It MUST be deallocated by the caller by
+// calling WebPDataClear(). It's always safe to call WebPDataClear() upon
+// return, even in case of error.
 // Parameters:
 //   mux - (in/out) object whose chunks are to be assembled
 //   assembled_data - (out) assembled WebP data
diff --git a/include/webp/mux_types.h b/include/webp/mux_types.h
index ceea77d..2fe8195 100644
--- a/include/webp/mux_types.h
+++ b/include/webp/mux_types.h
@@ -14,7 +14,6 @@
 #ifndef WEBP_WEBP_MUX_TYPES_H_
 #define WEBP_WEBP_MUX_TYPES_H_
 
-#include <stdlib.h>  // free()
 #include <string.h>  // memset()
 #include "./types.h"
 
@@ -56,6 +55,7 @@
 
 // Data type used to describe 'raw' data, e.g., chunk data
 // (ICC profile, metadata) and WebP compressed image data.
+// 'bytes' memory must be allocated using WebPMalloc() and such.
 struct WebPData {
   const uint8_t* bytes;
   size_t size;
@@ -68,11 +68,11 @@
   }
 }
 
-// Clears the contents of the 'webp_data' object by calling free(). Does not
-// deallocate the object itself.
+// Clears the contents of the 'webp_data' object by calling WebPFree().
+// Does not deallocate the object itself.
 static WEBP_INLINE void WebPDataClear(WebPData* webp_data) {
   if (webp_data != NULL) {
-    free((void*)webp_data->bytes);
+    WebPFree((void*)webp_data->bytes);
     WebPDataInit(webp_data);
   }
 }
@@ -83,7 +83,7 @@
   if (src == NULL || dst == NULL) return 0;
   WebPDataInit(dst);
   if (src->bytes != NULL && src->size != 0) {
-    dst->bytes = (uint8_t*)malloc(src->size);
+    dst->bytes = (uint8_t*)WebPMalloc(src->size);
     if (dst->bytes == NULL) return 0;
     memcpy((void*)dst->bytes, src->bytes, src->size);
     dst->size = src->size;
diff --git a/include/webp/types.h b/include/webp/types.h
index 0ce2622..47f7f2b 100644
--- a/include/webp/types.h
+++ b/include/webp/types.h
@@ -7,7 +7,7 @@
 // be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
-//  Common types
+//  Common types + memory wrappers
 //
 // Author: Skal (pascal.massimino@gmail.com)
 
@@ -49,4 +49,20 @@
 // Macro to check ABI compatibility (same major revision number)
 #define WEBP_ABI_IS_INCOMPATIBLE(a, b) (((a) >> 8) != ((b) >> 8))
 
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+// Allocates 'size' bytes of memory. Returns NULL upon error. Memory
+// must be deallocated by calling WebPFree(). This function is made available
+// by the core 'libwebp' library.
+WEBP_EXTERN void* WebPMalloc(size_t size);
+
+// Releases memory returned by the WebPDecode*() functions (from decode.h).
+WEBP_EXTERN void WebPFree(void* ptr);
+
+#ifdef __cplusplus
+}    // extern "C"
+#endif
+
 #endif  // WEBP_WEBP_TYPES_H_
diff --git a/src/dec/frame_dec.c b/src/dec/frame_dec.c
index bda9e1a..04609a8 100644
--- a/src/dec/frame_dec.c
+++ b/src/dec/frame_dec.c
@@ -732,7 +732,7 @@
   mem += f_info_size;
   dec->thread_ctx_.id_ = 0;
   dec->thread_ctx_.f_info_ = dec->f_info_;
-  if (dec->mt_method_ > 0) {
+  if (dec->filter_type_ > 0 && dec->mt_method_ > 0) {
     // secondary cache line. The deblocking process need to make use of the
     // filtering strength from previous macroblock row, while the new ones
     // are being decoded in parallel. We'll just swap the pointers.
diff --git a/src/dec/idec_dec.c b/src/dec/idec_dec.c
index 9bc9166..9035df5 100644
--- a/src/dec/idec_dec.c
+++ b/src/dec/idec_dec.c
@@ -166,9 +166,11 @@
   VP8Decoder* const dec = (VP8Decoder*)idec->dec_;
   MemBuffer* const mem = &idec->mem_;
   const int need_compressed_alpha = NeedCompressedAlpha(idec);
-  const uint8_t* const old_start = mem->buf_ + mem->start_;
+  const uint8_t* const old_start =
+      (mem->buf_ == NULL) ? NULL : mem->buf_ + mem->start_;
   const uint8_t* const old_base =
       need_compressed_alpha ? dec->alpha_data_ : old_start;
+  assert(mem->buf_ != NULL || mem->start_ == 0);
   assert(mem->mode_ == MEM_MODE_APPEND);
   if (data_size > MAX_CHUNK_PAYLOAD) {
     // security safeguard: trying to allocate more than what the format
@@ -184,7 +186,7 @@
     uint8_t* const new_buf =
         (uint8_t*)WebPSafeMalloc(extra_size, sizeof(*new_buf));
     if (new_buf == NULL) return 0;
-    memcpy(new_buf, old_base, current_size);
+    if (old_base != NULL) memcpy(new_buf, old_base, current_size);
     WebPSafeFree(mem->buf_);
     mem->buf_ = new_buf;
     mem->buf_size_ = (size_t)extra_size;
@@ -192,6 +194,7 @@
     mem->end_ = current_size;
   }
 
+  assert(mem->buf_ != NULL);
   memcpy(mem->buf_ + mem->end_, data, data_size);
   mem->end_ += data_size;
   assert(mem->end_ <= mem->buf_size_);
@@ -204,7 +207,9 @@
                           const uint8_t* const data, size_t data_size) {
   MemBuffer* const mem = &idec->mem_;
   const uint8_t* const old_buf = mem->buf_;
-  const uint8_t* const old_start = old_buf + mem->start_;
+  const uint8_t* const old_start =
+      (old_buf == NULL) ? NULL : old_buf + mem->start_;
+  assert(old_buf != NULL || mem->start_ == 0);
   assert(mem->mode_ == MEM_MODE_MAP);
 
   if (data_size < mem->buf_size_) return 0;  // can't remap to a shorter buffer!
diff --git a/src/dec/vp8i_dec.h b/src/dec/vp8i_dec.h
index 3de8d86..600a684 100644
--- a/src/dec/vp8i_dec.h
+++ b/src/dec/vp8i_dec.h
@@ -31,8 +31,8 @@
 
 // version numbers
 #define DEC_MAJ_VERSION 1
-#define DEC_MIN_VERSION 0
-#define DEC_REV_VERSION 3
+#define DEC_MIN_VERSION 1
+#define DEC_REV_VERSION 0
 
 // YUV-cache parameters. Cache is 32-bytes wide (= one cacheline).
 // Constraints are: We need to store one 16x16 block of luma samples (y),
diff --git a/src/dec/vp8li_dec.h b/src/dec/vp8li_dec.h
index 0a4d613..72b2e86 100644
--- a/src/dec/vp8li_dec.h
+++ b/src/dec/vp8li_dec.h
@@ -37,7 +37,7 @@
   int                    bits_;   // subsampling bits defining transform window.
   int                    xsize_;  // transform window X index.
   int                    ysize_;  // transform window Y index.
-  uint32_t              *data_;   // transform data.
+  uint32_t*              data_;   // transform data.
 };
 
 typedef struct {
@@ -48,23 +48,23 @@
   int             huffman_mask_;
   int             huffman_subsample_bits_;
   int             huffman_xsize_;
-  uint32_t       *huffman_image_;
+  uint32_t*       huffman_image_;
   int             num_htree_groups_;
-  HTreeGroup     *htree_groups_;
-  HuffmanCode    *huffman_tables_;
+  HTreeGroup*     htree_groups_;
+  HuffmanCode*    huffman_tables_;
 } VP8LMetadata;
 
 typedef struct VP8LDecoder VP8LDecoder;
 struct VP8LDecoder {
   VP8StatusCode    status_;
   VP8LDecodeState  state_;
-  VP8Io           *io_;
+  VP8Io*           io_;
 
-  const WebPDecBuffer *output_;    // shortcut to io->opaque->output
+  const WebPDecBuffer* output_;    // shortcut to io->opaque->output
 
-  uint32_t        *pixels_;        // Internal data: either uint8_t* for alpha
+  uint32_t*        pixels_;        // Internal data: either uint8_t* for alpha
                                    // or uint32_t* for BGRA.
-  uint32_t        *argb_cache_;    // Scratch buffer for temporary BGRA storage.
+  uint32_t*        argb_cache_;    // Scratch buffer for temporary BGRA storage.
 
   VP8LBitReader    br_;
   int              incremental_;   // if true, incremental decoding is expected
@@ -86,8 +86,8 @@
   // or'd bitset storing the transforms types.
   uint32_t         transforms_seen_;
 
-  uint8_t         *rescaler_memory;  // Working memory for rescaling work.
-  WebPRescaler    *rescaler;         // Common rescaler for all channels.
+  uint8_t*         rescaler_memory;  // Working memory for rescaling work.
+  WebPRescaler*    rescaler;         // Common rescaler for all channels.
 };
 
 //------------------------------------------------------------------------------
diff --git a/src/demux/demux.c b/src/demux/demux.c
index ab6433e..1b3cc2e 100644
--- a/src/demux/demux.c
+++ b/src/demux/demux.c
@@ -24,8 +24,8 @@
 #include "src/webp/format_constants.h"
 
 #define DMUX_MAJ_VERSION 1
-#define DMUX_MIN_VERSION 0
-#define DMUX_REV_VERSION 3
+#define DMUX_MIN_VERSION 1
+#define DMUX_REV_VERSION 0
 
 typedef struct {
   size_t start_;        // start location of the data
diff --git a/src/dsp/dec_neon.c b/src/dsp/dec_neon.c
index ffa697f..239ec41 100644
--- a/src/dsp/dec_neon.c
+++ b/src/dsp/dec_neon.c
@@ -1361,7 +1361,8 @@
   const uint32_t J = dst[-1 + 1 * BPS];
   const uint32_t K = dst[-1 + 2 * BPS];
   const uint32_t L = dst[-1 + 3 * BPS];
-  const uint64x1_t LKJI____ = vcreate_u64(L | (K << 8) | (J << 16) | (I << 24));
+  const uint64x1_t LKJI____ =
+      vcreate_u64((uint64_t)L | (K << 8) | (J << 16) | (I << 24));
   const uint64x1_t LKJIXABC = vorr_u64(LKJI____, ____XABC);
   const uint8x8_t KJIXABC_ = vreinterpret_u8_u64(vshr_n_u64(LKJIXABC, 8));
   const uint8x8_t JIXABC__ = vreinterpret_u8_u64(vshr_n_u64(LKJIXABC, 16));
@@ -1427,10 +1428,16 @@
 
   if (do_top) {
     const uint8x8_t A = vld1_u8(dst - BPS);  // top row
+#if defined(__aarch64__)
+    const uint16x8_t B = vmovl_u8(A);
+    const uint16_t p2 = vaddvq_u16(B);
+    sum_top = vdupq_n_u16(p2);
+#else
     const uint16x4_t p0 = vpaddl_u8(A);  // cascading summation of the top
     const uint16x4_t p1 = vpadd_u16(p0, p0);
     const uint16x4_t p2 = vpadd_u16(p1, p1);
     sum_top = vcombine_u16(p2, p2);
+#endif
   }
 
   if (do_left) {
diff --git a/src/dsp/dsp.h b/src/dsp/dsp.h
index fafc2d0..a784de3 100644
--- a/src/dsp/dsp.h
+++ b/src/dsp/dsp.h
@@ -246,9 +246,9 @@
 extern VP8WHT VP8FTransformWHT;
 // Predictions
 // *dst is the destination block. *top and *left can be NULL.
-typedef void (*VP8IntraPreds)(uint8_t *dst, const uint8_t* left,
+typedef void (*VP8IntraPreds)(uint8_t* dst, const uint8_t* left,
                               const uint8_t* top);
-typedef void (*VP8Intra4Preds)(uint8_t *dst, const uint8_t* top);
+typedef void (*VP8Intra4Preds)(uint8_t* dst, const uint8_t* top);
 extern VP8Intra4Preds VP8EncPredLuma4;
 extern VP8IntraPreds VP8EncPredLuma16;
 extern VP8IntraPreds VP8EncPredChroma8;
diff --git a/src/dsp/lossless.c b/src/dsp/lossless.c
index d05af84..aad5f43 100644
--- a/src/dsp/lossless.c
+++ b/src/dsp/lossless.c
@@ -81,7 +81,7 @@
 
 // gcc <= 4.9 on ARM generates incorrect code in Select() when Sub3() is
 // inlined.
-#if defined(__arm__) && LOCAL_GCC_VERSION <= 0x409
+#if defined(__arm__) && defined(__GNUC__) && LOCAL_GCC_VERSION <= 0x409
 # define LOCAL_INLINE __attribute__ ((noinline))
 #else
 # define LOCAL_INLINE WEBP_INLINE
@@ -167,15 +167,20 @@
   return pred;
 }
 
-GENERATE_PREDICTOR_ADD(Predictor0_C, PredictorAdd0_C)
+static void PredictorAdd0_C(const uint32_t* in, const uint32_t* upper,
+                            int num_pixels, uint32_t* out) {
+  int x;
+  (void)upper;
+  for (x = 0; x < num_pixels; ++x) out[x] = VP8LAddPixels(in[x], ARGB_BLACK);
+}
 static void PredictorAdd1_C(const uint32_t* in, const uint32_t* upper,
                             int num_pixels, uint32_t* out) {
   int i;
   uint32_t left = out[-1];
+  (void)upper;
   for (i = 0; i < num_pixels; ++i) {
     out[i] = left = VP8LAddPixels(in[i], left);
   }
-  (void)upper;
 }
 GENERATE_PREDICTOR_ADD(Predictor2_C, PredictorAdd2_C)
 GENERATE_PREDICTOR_ADD(Predictor3_C, PredictorAdd3_C)
diff --git a/src/dsp/lossless_common.h b/src/dsp/lossless_common.h
index a2648d1..9c2ebe6 100644
--- a/src/dsp/lossless_common.h
+++ b/src/dsp/lossless_common.h
@@ -177,6 +177,7 @@
 static void PREDICTOR_ADD(const uint32_t* in, const uint32_t* upper, \
                           int num_pixels, uint32_t* out) {           \
   int x;                                                             \
+  assert(upper != NULL);                                             \
   for (x = 0; x < num_pixels; ++x) {                                 \
     const uint32_t pred = (PREDICTOR)(out[x - 1], upper + x);        \
     out[x] = VP8LAddPixels(in[x], pred);                             \
@@ -189,6 +190,7 @@
 static void PREDICTOR_SUB(const uint32_t* in, const uint32_t* upper, \
                           int num_pixels, uint32_t* out) {           \
   int x;                                                             \
+  assert(upper != NULL);                                             \
   for (x = 0; x < num_pixels; ++x) {                                 \
     const uint32_t pred = (PREDICTOR)(in[x - 1], upper + x);         \
     out[x] = VP8LSubPixels(in[x], pred);                             \
diff --git a/src/dsp/lossless_enc_sse2.c b/src/dsp/lossless_enc_sse2.c
index 8adc521..e676f6f 100644
--- a/src/dsp/lossless_enc_sse2.c
+++ b/src/dsp/lossless_enc_sse2.c
@@ -455,8 +455,9 @@
     _mm_storeu_si128((__m128i*)&out[i], res);
   }
   if (i != num_pixels) {
-    VP8LPredictorsSub_C[0](in + i, upper + i, num_pixels - i, out + i);
+    VP8LPredictorsSub_C[0](in + i, NULL, num_pixels - i, out + i);
   }
+  (void)upper;
 }
 
 #define GENERATE_PREDICTOR_1(X, IN)                                           \
diff --git a/src/dsp/lossless_sse2.c b/src/dsp/lossless_sse2.c
index 17d7576..aef0cee 100644
--- a/src/dsp/lossless_sse2.c
+++ b/src/dsp/lossless_sse2.c
@@ -191,8 +191,9 @@
     _mm_storeu_si128((__m128i*)&out[i], res);
   }
   if (i != num_pixels) {
-    VP8LPredictorsAdd_C[0](in + i, upper + i, num_pixels - i, out + i);
+    VP8LPredictorsAdd_C[0](in + i, NULL, num_pixels - i, out + i);
   }
+  (void)upper;
 }
 
 // Predictor1: left.
diff --git a/src/dsp/upsampling_msa.c b/src/dsp/upsampling_msa.c
index 99eea70..f2e03e8 100644
--- a/src/dsp/upsampling_msa.c
+++ b/src/dsp/upsampling_msa.c
@@ -576,9 +576,9 @@
   const uint32_t l_uv = ((cur_u[0]) | ((cur_v[0]) << 16));               \
   const uint32_t uv0 = (3 * tl_uv + l_uv + 0x00020002u) >> 2;            \
   const uint8_t* ptop_y = &top_y[1];                                     \
-  uint8_t *ptop_dst = top_dst + XSTEP;                                   \
+  uint8_t* ptop_dst = top_dst + XSTEP;                                   \
   const uint8_t* pbot_y = &bot_y[1];                                     \
-  uint8_t *pbot_dst = bot_dst + XSTEP;                                   \
+  uint8_t* pbot_dst = bot_dst + XSTEP;                                   \
                                                                          \
   FUNC(top_y[0], uv0 & 0xff, (uv0 >> 16), top_dst);                      \
   if (bot_y != NULL) {                                                   \
diff --git a/src/dsp/upsampling_neon.c b/src/dsp/upsampling_neon.c
index 17cbc9f..6ba71a7 100644
--- a/src/dsp/upsampling_neon.c
+++ b/src/dsp/upsampling_neon.c
@@ -58,8 +58,8 @@
 } while (0)
 
 // Turn the macro into a function for reducing code-size when non-critical
-static void Upsample16Pixels_NEON(const uint8_t *r1, const uint8_t *r2,
-                                  uint8_t *out) {
+static void Upsample16Pixels_NEON(const uint8_t* r1, const uint8_t* r2,
+                                  uint8_t* out) {
   UPSAMPLE_16PIXELS(r1, r2, out);
 }
 
@@ -190,14 +190,14 @@
 }
 
 #define NEON_UPSAMPLE_FUNC(FUNC_NAME, FMT, XSTEP)                       \
-static void FUNC_NAME(const uint8_t *top_y, const uint8_t *bottom_y,    \
-                      const uint8_t *top_u, const uint8_t *top_v,       \
-                      const uint8_t *cur_u, const uint8_t *cur_v,       \
-                      uint8_t *top_dst, uint8_t *bottom_dst, int len) { \
+static void FUNC_NAME(const uint8_t* top_y, const uint8_t* bottom_y,    \
+                      const uint8_t* top_u, const uint8_t* top_v,       \
+                      const uint8_t* cur_u, const uint8_t* cur_v,       \
+                      uint8_t* top_dst, uint8_t* bottom_dst, int len) { \
   int block;                                                            \
   /* 16 byte aligned array to cache reconstructed u and v */            \
   uint8_t uv_buf[2 * 32 + 15];                                          \
-  uint8_t *const r_uv = (uint8_t*)((uintptr_t)(uv_buf + 15) & ~15);     \
+  uint8_t* const r_uv = (uint8_t*)((uintptr_t)(uv_buf + 15) & ~15);     \
   const int uv_len = (len + 1) >> 1;                                    \
   /* 9 pixels must be read-able for each block */                       \
   const int num_blocks = (uv_len - 1) >> 3;                             \
diff --git a/src/enc/histogram_enc.c b/src/enc/histogram_enc.c
index d89b985..a4e6bf3 100644
--- a/src/enc/histogram_enc.c
+++ b/src/enc/histogram_enc.c
@@ -641,7 +641,7 @@
 // Merges some histograms with same bin_id together if it's advantageous.
 // Sets the remaining histograms to NULL.
 static void HistogramCombineEntropyBin(VP8LHistogramSet* const image_histo,
-                                       int *num_used,
+                                       int* num_used,
                                        const uint16_t* const clusters,
                                        uint16_t* const cluster_mappings,
                                        VP8LHistogram* cur_combo,
diff --git a/src/enc/picture_csp_enc.c b/src/enc/picture_csp_enc.c
index 02d9df7..718e014 100644
--- a/src/enc/picture_csp_enc.c
+++ b/src/enc/picture_csp_enc.c
@@ -29,11 +29,15 @@
 #define USE_INVERSE_ALPHA_TABLE
 
 #ifdef WORDS_BIGENDIAN
-#define ALPHA_OFFSET 0   // uint32_t 0xff000000 is 0xff,00,00,00 in memory
+// uint32_t 0xff000000 is 0xff,00,00,00 in memory
+#define CHANNEL_OFFSET(i) (i)
 #else
-#define ALPHA_OFFSET 3   // uint32_t 0xff000000 is 0x00,00,00,ff in memory
+// uint32_t 0xff000000 is 0x00,00,00,ff in memory
+#define CHANNEL_OFFSET(i) (3-(i))
 #endif
 
+#define ALPHA_OFFSET CHANNEL_OFFSET(0)
+
 //------------------------------------------------------------------------------
 // Detection of non-trivial transparency
 
@@ -997,10 +1001,10 @@
     return WebPEncodingSetError(picture, VP8_ENC_ERROR_INVALID_CONFIGURATION);
   } else {
     const uint8_t* const argb = (const uint8_t*)picture->argb;
-    const uint8_t* const a = argb + (0 ^ ALPHA_OFFSET);
-    const uint8_t* const r = argb + (1 ^ ALPHA_OFFSET);
-    const uint8_t* const g = argb + (2 ^ ALPHA_OFFSET);
-    const uint8_t* const b = argb + (3 ^ ALPHA_OFFSET);
+    const uint8_t* const a = argb + CHANNEL_OFFSET(0);
+    const uint8_t* const r = argb + CHANNEL_OFFSET(1);
+    const uint8_t* const g = argb + CHANNEL_OFFSET(2);
+    const uint8_t* const b = argb + CHANNEL_OFFSET(3);
 
     picture->colorspace = WEBP_YUV420;
     return ImportYUVAFromRGBA(r, g, b, a, 4, 4 * picture->argb_stride,
@@ -1050,7 +1054,7 @@
     const int height = picture->height;
     const int argb_stride = 4 * picture->argb_stride;
     uint8_t* dst = (uint8_t*)picture->argb;
-    const uint8_t *cur_u = picture->u, *cur_v = picture->v, *cur_y = picture->y;
+    const uint8_t* cur_u = picture->u, *cur_v = picture->v, *cur_y = picture->y;
     WebPUpsampleLinePairFunc upsample =
         WebPGetLinePairConverter(ALPHA_OFFSET > 0);
 
diff --git a/src/enc/vp8i_enc.h b/src/enc/vp8i_enc.h
index 24e1944..fedcaee 100644
--- a/src/enc/vp8i_enc.h
+++ b/src/enc/vp8i_enc.h
@@ -31,8 +31,8 @@
 
 // version numbers
 #define ENC_MAJ_VERSION 1
-#define ENC_MIN_VERSION 0
-#define ENC_REV_VERSION 3
+#define ENC_MIN_VERSION 1
+#define ENC_REV_VERSION 0
 
 enum { MAX_LF_LEVELS = 64,       // Maximum loop filter level
        MAX_VARIABLE_LEVEL = 67,  // last (inclusive) level with variable cost
@@ -249,7 +249,7 @@
   int           percent0_;         // saved initial progress percent
 
   DError        left_derr_;        // left error diffusion (u/v)
-  DError       *top_derr_;         // top diffusion error - NULL if disabled
+  DError*       top_derr_;         // top diffusion error - NULL if disabled
 
   uint8_t* y_left_;    // left luma samples (addressable from index -1 to 15).
   uint8_t* u_left_;    // left u samples (addressable from index -1 to 7)
diff --git a/src/mux/muxi.h b/src/mux/muxi.h
index 7bc0b07..ad3e1bd 100644
--- a/src/mux/muxi.h
+++ b/src/mux/muxi.h
@@ -28,8 +28,8 @@
 // Defines and constants.
 
 #define MUX_MAJ_VERSION 1
-#define MUX_MIN_VERSION 0
-#define MUX_REV_VERSION 3
+#define MUX_MIN_VERSION 1
+#define MUX_REV_VERSION 0
 
 // Chunk object.
 typedef struct WebPChunk WebPChunk;
diff --git a/src/mux/muxread.c b/src/mux/muxread.c
index 268f6ac..ae3b876 100644
--- a/src/mux/muxread.c
+++ b/src/mux/muxread.c
@@ -100,7 +100,7 @@
                          WebPMuxImage* const wpi) {
   const uint8_t* bytes = chunk->data_.bytes;
   size_t size = chunk->data_.size;
-  const uint8_t* const last = bytes + size;
+  const uint8_t* const last = (bytes == NULL) ? NULL : bytes + size;
   WebPChunk subchunk;
   size_t subchunk_size;
   WebPChunk** unknown_chunk_list = &wpi->unknown_;
diff --git a/src/utils/color_cache_utils.h b/src/utils/color_cache_utils.h
index ec21d51..b45d47c 100644
--- a/src/utils/color_cache_utils.h
+++ b/src/utils/color_cache_utils.h
@@ -26,7 +26,7 @@
 
 // Main color cache struct.
 typedef struct {
-  uint32_t *colors_;  // color entries
+  uint32_t* colors_;  // color entries
   int hash_shift_;    // Hash shift: 32 - hash_bits_.
   int hash_bits_;
 } VP8LColorCache;
diff --git a/src/utils/thread_utils.c b/src/utils/thread_utils.c
index 438296b..4e470e1 100644
--- a/src/utils/thread_utils.c
+++ b/src/utils/thread_utils.c
@@ -73,7 +73,7 @@
 #endif
 
 static int pthread_create(pthread_t* const thread, const void* attr,
-                          unsigned int (__stdcall *start)(void*), void* arg) {
+                          unsigned int (__stdcall* start)(void*), void* arg) {
   (void)attr;
 #ifdef USE_CREATE_THREAD
   *thread = CreateThread(NULL,   /* lpThreadAttributes */
diff --git a/src/utils/utils.c b/src/utils/utils.c
index 44d5c14..764f752 100644
--- a/src/utils/utils.c
+++ b/src/utils/utils.c
@@ -216,9 +216,14 @@
   free(ptr);
 }
 
-// Public API function.
+// Public API functions.
+
+void* WebPMalloc(size_t size) {
+  return WebPSafeMalloc(1, size);
+}
+
 void WebPFree(void* ptr) {
-  free(ptr);
+  WebPSafeFree(ptr);
 }
 
 //------------------------------------------------------------------------------
diff --git a/src/webp/decode.h b/src/webp/decode.h
index ae8bfe8..80dd0ef 100644
--- a/src/webp/decode.h
+++ b/src/webp/decode.h
@@ -20,7 +20,7 @@
 extern "C" {
 #endif
 
-#define WEBP_DECODER_ABI_VERSION 0x0208    // MAJOR(8b) + MINOR(8b)
+#define WEBP_DECODER_ABI_VERSION 0x0209    // MAJOR(8b) + MINOR(8b)
 
 // Note: forward declaring enumerations is not allowed in (strict) C and C++,
 // the types are left here for reference.
@@ -91,9 +91,6 @@
                                    uint8_t** u, uint8_t** v,
                                    int* stride, int* uv_stride);
 
-// Releases memory returned by the WebPDecode*() functions above.
-WEBP_EXTERN void WebPFree(void* ptr);
-
 // These five functions are variants of the above ones, that decode the image
 // directly into a pre-allocated buffer 'output_buffer'. The maximum storage
 // available in this buffer is indicated by 'output_buffer_size'. If this
diff --git a/src/webp/encode.h b/src/webp/encode.h
index 339f881..655166e 100644
--- a/src/webp/encode.h
+++ b/src/webp/encode.h
@@ -20,7 +20,7 @@
 extern "C" {
 #endif
 
-#define WEBP_ENCODER_ABI_VERSION 0x020e    // MAJOR(8b) + MINOR(8b)
+#define WEBP_ENCODER_ABI_VERSION 0x020f    // MAJOR(8b) + MINOR(8b)
 
 // Note: forward declaring enumerations is not allowed in (strict) C and C++,
 // the types are left here for reference.
@@ -79,9 +79,6 @@
                                           int width, int height, int stride,
                                           uint8_t** output);
 
-// Releases memory returned by the WebPEncode*() functions above.
-WEBP_EXTERN void WebPFree(void* ptr);
-
 //------------------------------------------------------------------------------
 // Coding parameters
 
@@ -306,7 +303,7 @@
   // YUV input (mostly used for input to lossy compression)
   WebPEncCSP colorspace;     // colorspace: should be YUV420 for now (=Y'CbCr).
   int width, height;         // dimensions (less or equal to WEBP_MAX_DIMENSION)
-  uint8_t *y, *u, *v;        // pointers to luma/chroma planes.
+  uint8_t* y, *u, *v;        // pointers to luma/chroma planes.
   int y_stride, uv_stride;   // luma/chroma strides.
   uint8_t* a;                // pointer to the alpha plane
   int a_stride;              // stride of the alpha plane
@@ -350,7 +347,7 @@
   uint32_t pad3[3];       // padding for later use
 
   // Unused for now
-  uint8_t *pad4, *pad5;
+  uint8_t* pad4, *pad5;
   uint32_t pad6[8];       // padding for later use
 
   // PRIVATE FIELDS
diff --git a/src/webp/mux.h b/src/webp/mux.h
index 66096a9..7d27489 100644
--- a/src/webp/mux.h
+++ b/src/webp/mux.h
@@ -57,7 +57,7 @@
   WebPMuxGetChunk(mux, "ICCP", &icc_profile);
   // ... (Consume icc_data).
   WebPMuxDelete(mux);
-  free(data);
+  WebPFree(data);
 */
 
 // Note: forward declaring enumerations is not allowed in (strict) C and C++,
@@ -245,7 +245,7 @@
     WebPMux* mux, const WebPMuxFrameInfo* frame, int copy_data);
 
 // Gets the nth frame from the mux object.
-// The content of 'frame->bitstream' is allocated using malloc(), and NOT
+// The content of 'frame->bitstream' is allocated using WebPMalloc(), and NOT
 // owned by the 'mux' object. It MUST be deallocated by the caller by calling
 // WebPDataClear().
 // nth=0 has a special meaning - last position.
@@ -376,10 +376,10 @@
 // Assembles all chunks in WebP RIFF format and returns in 'assembled_data'.
 // This function also validates the mux object.
 // Note: The content of 'assembled_data' will be ignored and overwritten.
-// Also, the content of 'assembled_data' is allocated using malloc(), and NOT
-// owned by the 'mux' object. It MUST be deallocated by the caller by calling
-// WebPDataClear(). It's always safe to call WebPDataClear() upon return,
-// even in case of error.
+// Also, the content of 'assembled_data' is allocated using WebPMalloc(), and
+// NOT owned by the 'mux' object. It MUST be deallocated by the caller by
+// calling WebPDataClear(). It's always safe to call WebPDataClear() upon
+// return, even in case of error.
 // Parameters:
 //   mux - (in/out) object whose chunks are to be assembled
 //   assembled_data - (out) assembled WebP data
diff --git a/src/webp/mux_types.h b/src/webp/mux_types.h
index ceea77d..2fe8195 100644
--- a/src/webp/mux_types.h
+++ b/src/webp/mux_types.h
@@ -14,7 +14,6 @@
 #ifndef WEBP_WEBP_MUX_TYPES_H_
 #define WEBP_WEBP_MUX_TYPES_H_
 
-#include <stdlib.h>  // free()
 #include <string.h>  // memset()
 #include "./types.h"
 
@@ -56,6 +55,7 @@
 
 // Data type used to describe 'raw' data, e.g., chunk data
 // (ICC profile, metadata) and WebP compressed image data.
+// 'bytes' memory must be allocated using WebPMalloc() and such.
 struct WebPData {
   const uint8_t* bytes;
   size_t size;
@@ -68,11 +68,11 @@
   }
 }
 
-// Clears the contents of the 'webp_data' object by calling free(). Does not
-// deallocate the object itself.
+// Clears the contents of the 'webp_data' object by calling WebPFree().
+// Does not deallocate the object itself.
 static WEBP_INLINE void WebPDataClear(WebPData* webp_data) {
   if (webp_data != NULL) {
-    free((void*)webp_data->bytes);
+    WebPFree((void*)webp_data->bytes);
     WebPDataInit(webp_data);
   }
 }
@@ -83,7 +83,7 @@
   if (src == NULL || dst == NULL) return 0;
   WebPDataInit(dst);
   if (src->bytes != NULL && src->size != 0) {
-    dst->bytes = (uint8_t*)malloc(src->size);
+    dst->bytes = (uint8_t*)WebPMalloc(src->size);
     if (dst->bytes == NULL) return 0;
     memcpy((void*)dst->bytes, src->bytes, src->size);
     dst->size = src->size;
diff --git a/src/webp/types.h b/src/webp/types.h
index 0ce2622..47f7f2b 100644
--- a/src/webp/types.h
+++ b/src/webp/types.h
@@ -7,7 +7,7 @@
 // be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
-//  Common types
+//  Common types + memory wrappers
 //
 // Author: Skal (pascal.massimino@gmail.com)
 
@@ -49,4 +49,20 @@
 // Macro to check ABI compatibility (same major revision number)
 #define WEBP_ABI_IS_INCOMPATIBLE(a, b) (((a) >> 8) != ((b) >> 8))
 
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+// Allocates 'size' bytes of memory. Returns NULL upon error. Memory
+// must be deallocated by calling WebPFree(). This function is made available
+// by the core 'libwebp' library.
+WEBP_EXTERN void* WebPMalloc(size_t size);
+
+// Releases memory returned by the WebPDecode*() functions (from decode.h).
+WEBP_EXTERN void WebPFree(void* ptr);
+
+#ifdef __cplusplus
+}    // extern "C"
+#endif
+
 #endif  // WEBP_WEBP_TYPES_H_