From 748e92bbb9ca4de9b3542338fc724689d090ddce Mon Sep 17 00:00:00 2001
From: James Zern <jzern@google.com>
Date: Sat, 30 Jul 2022 19:38:35 -0700
Subject: [PATCH] add WebPInt32ToMem

and use it in calls containing _mm_cvtsi32_si128; this calls
WebPUint32ToMem, but corrects the type to avoid runtime warnings with
clang -fsanitize=integer of the form:

implicit conversion from type 'int' of value -1904123502 (32-bit,
signed) to type 'uint32_t' (aka 'unsigned int') changed the value to
2390843794 (32-bit, unsigned)

Bug: b/229626362
Change-Id: I20545e822d8045fa44f688241879206055a0a148
---
 src/dsp/dec_sse2.c      | 52 ++++++++++++++++++++---------------------
 src/dsp/enc_sse2.c      | 42 ++++++++++++++++-----------------
 src/dsp/rescaler_sse2.c |  2 +-
 src/utils/utils.h       |  4 ++++
 4 files changed, 52 insertions(+), 48 deletions(-)

diff --git a/src/dsp/dec_sse2.c b/src/dsp/dec_sse2.c
index 0aa4cb16..8964b527 100644
--- a/src/dsp/dec_sse2.c
+++ b/src/dsp/dec_sse2.c
@@ -187,10 +187,10 @@ static void Transform_SSE2(const int16_t* in, uint8_t* dst, int do_two) {
       _mm_storel_epi64((__m128i*)(dst + 3 * BPS), dst3);
     } else {
       // Store four bytes/pixels per line.
-      WebPUint32ToMem(dst + 0 * BPS, _mm_cvtsi128_si32(dst0));
-      WebPUint32ToMem(dst + 1 * BPS, _mm_cvtsi128_si32(dst1));
-      WebPUint32ToMem(dst + 2 * BPS, _mm_cvtsi128_si32(dst2));
-      WebPUint32ToMem(dst + 3 * BPS, _mm_cvtsi128_si32(dst3));
+      WebPInt32ToMem(dst + 0 * BPS, _mm_cvtsi128_si32(dst0));
+      WebPInt32ToMem(dst + 1 * BPS, _mm_cvtsi128_si32(dst1));
+      WebPInt32ToMem(dst + 2 * BPS, _mm_cvtsi128_si32(dst2));
+      WebPInt32ToMem(dst + 3 * BPS, _mm_cvtsi128_si32(dst3));
     }
   }
 }
@@ -233,10 +233,10 @@ static void TransformAC3(const int16_t* in, uint8_t* dst) {
   dst2 = _mm_packus_epi16(dst2, dst2);
   dst3 = _mm_packus_epi16(dst3, dst3);
   // Store the results.
-  WebPUint32ToMem(dst + 0 * BPS, _mm_cvtsi128_si32(dst0));
-  WebPUint32ToMem(dst + 1 * BPS, _mm_cvtsi128_si32(dst1));
-  WebPUint32ToMem(dst + 2 * BPS, _mm_cvtsi128_si32(dst2));
-  WebPUint32ToMem(dst + 3 * BPS, _mm_cvtsi128_si32(dst3));
+  WebPInt32ToMem(dst + 0 * BPS, _mm_cvtsi128_si32(dst0));
+  WebPInt32ToMem(dst + 1 * BPS, _mm_cvtsi128_si32(dst1));
+  WebPInt32ToMem(dst + 2 * BPS, _mm_cvtsi128_si32(dst2));
+  WebPInt32ToMem(dst + 3 * BPS, _mm_cvtsi128_si32(dst3));
 }
 #undef MUL
 #endif   // USE_TRANSFORM_AC3
@@ -540,7 +540,7 @@ static WEBP_INLINE void Store4x4_SSE2(__m128i* const x,
                                       uint8_t* dst, int stride) {
   int i;
   for (i = 0; i < 4; ++i, dst += stride) {
-    WebPUint32ToMem(dst, _mm_cvtsi128_si32(*x));
+    WebPInt32ToMem(dst, _mm_cvtsi128_si32(*x));
     *x = _mm_srli_si128(*x, 4);
   }
 }
@@ -925,10 +925,10 @@ static void LD4_SSE2(uint8_t* dst) {   // Down-Left
   const __m128i lsb = _mm_and_si128(_mm_xor_si128(ABCDEFGH, CDEFGHH0), one);
   const __m128i avg2 = _mm_subs_epu8(avg1, lsb);
   const __m128i abcdefg = _mm_avg_epu8(avg2, BCDEFGH0);
-  WebPUint32ToMem(dst + 0 * BPS, _mm_cvtsi128_si32(               abcdefg    ));
-  WebPUint32ToMem(dst + 1 * BPS, _mm_cvtsi128_si32(_mm_srli_si128(abcdefg, 1)));
-  WebPUint32ToMem(dst + 2 * BPS, _mm_cvtsi128_si32(_mm_srli_si128(abcdefg, 2)));
-  WebPUint32ToMem(dst + 3 * BPS, _mm_cvtsi128_si32(_mm_srli_si128(abcdefg, 3)));
+  WebPInt32ToMem(dst + 0 * BPS, _mm_cvtsi128_si32(               abcdefg    ));
+  WebPInt32ToMem(dst + 1 * BPS, _mm_cvtsi128_si32(_mm_srli_si128(abcdefg, 1)));
+  WebPInt32ToMem(dst + 2 * BPS, _mm_cvtsi128_si32(_mm_srli_si128(abcdefg, 2)));
+  WebPInt32ToMem(dst + 3 * BPS, _mm_cvtsi128_si32(_mm_srli_si128(abcdefg, 3)));
 }
 
 static void VR4_SSE2(uint8_t* dst) {   // Vertical-Right
@@ -946,10 +946,10 @@ static void VR4_SSE2(uint8_t* dst) {   // Vertical-Right
   const __m128i lsb = _mm_and_si128(_mm_xor_si128(IXABCD, ABCD0), one);
   const __m128i avg2 = _mm_subs_epu8(avg1, lsb);
   const __m128i efgh = _mm_avg_epu8(avg2, XABCD);
-  WebPUint32ToMem(dst + 0 * BPS, _mm_cvtsi128_si32(               abcd    ));
-  WebPUint32ToMem(dst + 1 * BPS, _mm_cvtsi128_si32(               efgh    ));
-  WebPUint32ToMem(dst + 2 * BPS, _mm_cvtsi128_si32(_mm_slli_si128(abcd, 1)));
-  WebPUint32ToMem(dst + 3 * BPS, _mm_cvtsi128_si32(_mm_slli_si128(efgh, 1)));
+  WebPInt32ToMem(dst + 0 * BPS, _mm_cvtsi128_si32(               abcd    ));
+  WebPInt32ToMem(dst + 1 * BPS, _mm_cvtsi128_si32(               efgh    ));
+  WebPInt32ToMem(dst + 2 * BPS, _mm_cvtsi128_si32(_mm_slli_si128(abcd, 1)));
+  WebPInt32ToMem(dst + 3 * BPS, _mm_cvtsi128_si32(_mm_slli_si128(efgh, 1)));
 
   // these two are hard to implement in SSE2, so we keep the C-version:
   DST(0, 2) = AVG3(J, I, X);
@@ -971,10 +971,10 @@ static void VL4_SSE2(uint8_t* dst) {   // Vertical-Left
   const __m128i lsb2 = _mm_and_si128(abbc, lsb1);
   const __m128i avg4 = _mm_subs_epu8(avg3, lsb2);
   const uint32_t extra_out = _mm_cvtsi128_si32(_mm_srli_si128(avg4, 4));
-  WebPUint32ToMem(dst + 0 * BPS, _mm_cvtsi128_si32(               avg1    ));
-  WebPUint32ToMem(dst + 1 * BPS, _mm_cvtsi128_si32(               avg4    ));
-  WebPUint32ToMem(dst + 2 * BPS, _mm_cvtsi128_si32(_mm_srli_si128(avg1, 1)));
-  WebPUint32ToMem(dst + 3 * BPS, _mm_cvtsi128_si32(_mm_srli_si128(avg4, 1)));
+  WebPInt32ToMem(dst + 0 * BPS, _mm_cvtsi128_si32(               avg1    ));
+  WebPInt32ToMem(dst + 1 * BPS, _mm_cvtsi128_si32(               avg4    ));
+  WebPInt32ToMem(dst + 2 * BPS, _mm_cvtsi128_si32(_mm_srli_si128(avg1, 1)));
+  WebPInt32ToMem(dst + 3 * BPS, _mm_cvtsi128_si32(_mm_srli_si128(avg4, 1)));
 
   // these two are hard to get and irregular
   DST(3, 2) = (extra_out >> 0) & 0xff;
@@ -998,10 +998,10 @@ static void RD4_SSE2(uint8_t* dst) {   // Down-right
   const __m128i lsb = _mm_and_si128(_mm_xor_si128(JIXABCD__, LKJIXABCD), one);
   const __m128i avg2 = _mm_subs_epu8(avg1, lsb);
   const __m128i abcdefg = _mm_avg_epu8(avg2, KJIXABCD_);
-  WebPUint32ToMem(dst + 3 * BPS, _mm_cvtsi128_si32(               abcdefg    ));
-  WebPUint32ToMem(dst + 2 * BPS, _mm_cvtsi128_si32(_mm_srli_si128(abcdefg, 1)));
-  WebPUint32ToMem(dst + 1 * BPS, _mm_cvtsi128_si32(_mm_srli_si128(abcdefg, 2)));
-  WebPUint32ToMem(dst + 0 * BPS, _mm_cvtsi128_si32(_mm_srli_si128(abcdefg, 3)));
+  WebPInt32ToMem(dst + 3 * BPS, _mm_cvtsi128_si32(               abcdefg    ));
+  WebPInt32ToMem(dst + 2 * BPS, _mm_cvtsi128_si32(_mm_srli_si128(abcdefg, 1)));
+  WebPInt32ToMem(dst + 1 * BPS, _mm_cvtsi128_si32(_mm_srli_si128(abcdefg, 2)));
+  WebPInt32ToMem(dst + 0 * BPS, _mm_cvtsi128_si32(_mm_srli_si128(abcdefg, 3)));
 }
 
 #undef DST
@@ -1021,7 +1021,7 @@ static WEBP_INLINE void TrueMotion_SSE2(uint8_t* dst, int size) {
       const int val = dst[-1] - top[-1];
       const __m128i base = _mm_set1_epi16(val);
       const __m128i out = _mm_packus_epi16(_mm_add_epi16(base, top_base), zero);
-      WebPUint32ToMem(dst, _mm_cvtsi128_si32(out));
+      WebPInt32ToMem(dst, _mm_cvtsi128_si32(out));
     }
   } else if (size == 8) {
     const __m128i top_values = _mm_loadl_epi64((const __m128i*)top);
diff --git a/src/dsp/enc_sse2.c b/src/dsp/enc_sse2.c
index f4babc62..600a019b 100644
--- a/src/dsp/enc_sse2.c
+++ b/src/dsp/enc_sse2.c
@@ -185,10 +185,10 @@ static void ITransform_SSE2(const uint8_t* ref, const int16_t* in, uint8_t* dst,
       _mm_storel_epi64((__m128i*)&dst[3 * BPS], ref3);
     } else {
       // Store four bytes/pixels per line.
-      WebPUint32ToMem(&dst[0 * BPS], _mm_cvtsi128_si32(ref0));
-      WebPUint32ToMem(&dst[1 * BPS], _mm_cvtsi128_si32(ref1));
-      WebPUint32ToMem(&dst[2 * BPS], _mm_cvtsi128_si32(ref2));
-      WebPUint32ToMem(&dst[3 * BPS], _mm_cvtsi128_si32(ref3));
+      WebPInt32ToMem(&dst[0 * BPS], _mm_cvtsi128_si32(ref0));
+      WebPInt32ToMem(&dst[1 * BPS], _mm_cvtsi128_si32(ref1));
+      WebPInt32ToMem(&dst[2 * BPS], _mm_cvtsi128_si32(ref2));
+      WebPInt32ToMem(&dst[3 * BPS], _mm_cvtsi128_si32(ref3));
     }
   }
 }
@@ -760,10 +760,10 @@ static WEBP_INLINE void LD4_SSE2(uint8_t* dst,
   const __m128i lsb = _mm_and_si128(_mm_xor_si128(ABCDEFGH, CDEFGHH0), one);
   const __m128i avg2 = _mm_subs_epu8(avg1, lsb);
   const __m128i abcdefg = _mm_avg_epu8(avg2, BCDEFGH0);
-  WebPUint32ToMem(dst + 0 * BPS, _mm_cvtsi128_si32(               abcdefg    ));
-  WebPUint32ToMem(dst + 1 * BPS, _mm_cvtsi128_si32(_mm_srli_si128(abcdefg, 1)));
-  WebPUint32ToMem(dst + 2 * BPS, _mm_cvtsi128_si32(_mm_srli_si128(abcdefg, 2)));
-  WebPUint32ToMem(dst + 3 * BPS, _mm_cvtsi128_si32(_mm_srli_si128(abcdefg, 3)));
+  WebPInt32ToMem(dst + 0 * BPS, _mm_cvtsi128_si32(               abcdefg    ));
+  WebPInt32ToMem(dst + 1 * BPS, _mm_cvtsi128_si32(_mm_srli_si128(abcdefg, 1)));
+  WebPInt32ToMem(dst + 2 * BPS, _mm_cvtsi128_si32(_mm_srli_si128(abcdefg, 2)));
+  WebPInt32ToMem(dst + 3 * BPS, _mm_cvtsi128_si32(_mm_srli_si128(abcdefg, 3)));
 }
 
 static WEBP_INLINE void VR4_SSE2(uint8_t* dst,
@@ -782,10 +782,10 @@ static WEBP_INLINE void VR4_SSE2(uint8_t* dst,
   const __m128i lsb = _mm_and_si128(_mm_xor_si128(IXABCD, ABCD0), one);
   const __m128i avg2 = _mm_subs_epu8(avg1, lsb);
   const __m128i efgh = _mm_avg_epu8(avg2, XABCD);
-  WebPUint32ToMem(dst + 0 * BPS, _mm_cvtsi128_si32(               abcd    ));
-  WebPUint32ToMem(dst + 1 * BPS, _mm_cvtsi128_si32(               efgh    ));
-  WebPUint32ToMem(dst + 2 * BPS, _mm_cvtsi128_si32(_mm_slli_si128(abcd, 1)));
-  WebPUint32ToMem(dst + 3 * BPS, _mm_cvtsi128_si32(_mm_slli_si128(efgh, 1)));
+  WebPInt32ToMem(dst + 0 * BPS, _mm_cvtsi128_si32(               abcd    ));
+  WebPInt32ToMem(dst + 1 * BPS, _mm_cvtsi128_si32(               efgh    ));
+  WebPInt32ToMem(dst + 2 * BPS, _mm_cvtsi128_si32(_mm_slli_si128(abcd, 1)));
+  WebPInt32ToMem(dst + 3 * BPS, _mm_cvtsi128_si32(_mm_slli_si128(efgh, 1)));
 
   // these two are hard to implement in SSE2, so we keep the C-version:
   DST(0, 2) = AVG3(J, I, X);
@@ -808,10 +808,10 @@ static WEBP_INLINE void VL4_SSE2(uint8_t* dst,
   const __m128i lsb2 = _mm_and_si128(abbc, lsb1);
   const __m128i avg4 = _mm_subs_epu8(avg3, lsb2);
   const uint32_t extra_out = _mm_cvtsi128_si32(_mm_srli_si128(avg4, 4));
-  WebPUint32ToMem(dst + 0 * BPS, _mm_cvtsi128_si32(               avg1    ));
-  WebPUint32ToMem(dst + 1 * BPS, _mm_cvtsi128_si32(               avg4    ));
-  WebPUint32ToMem(dst + 2 * BPS, _mm_cvtsi128_si32(_mm_srli_si128(avg1, 1)));
-  WebPUint32ToMem(dst + 3 * BPS, _mm_cvtsi128_si32(_mm_srli_si128(avg4, 1)));
+  WebPInt32ToMem(dst + 0 * BPS, _mm_cvtsi128_si32(               avg1    ));
+  WebPInt32ToMem(dst + 1 * BPS, _mm_cvtsi128_si32(               avg4    ));
+  WebPInt32ToMem(dst + 2 * BPS, _mm_cvtsi128_si32(_mm_srli_si128(avg1, 1)));
+  WebPInt32ToMem(dst + 3 * BPS, _mm_cvtsi128_si32(_mm_srli_si128(avg4, 1)));
 
   // these two are hard to get and irregular
   DST(3, 2) = (extra_out >> 0) & 0xff;
@@ -829,10 +829,10 @@ static WEBP_INLINE void RD4_SSE2(uint8_t* dst,
   const __m128i lsb = _mm_and_si128(_mm_xor_si128(JIXABCD__, LKJIXABCD), one);
   const __m128i avg2 = _mm_subs_epu8(avg1, lsb);
   const __m128i abcdefg = _mm_avg_epu8(avg2, KJIXABCD_);
-  WebPUint32ToMem(dst + 3 * BPS, _mm_cvtsi128_si32(               abcdefg    ));
-  WebPUint32ToMem(dst + 2 * BPS, _mm_cvtsi128_si32(_mm_srli_si128(abcdefg, 1)));
-  WebPUint32ToMem(dst + 1 * BPS, _mm_cvtsi128_si32(_mm_srli_si128(abcdefg, 2)));
-  WebPUint32ToMem(dst + 0 * BPS, _mm_cvtsi128_si32(_mm_srli_si128(abcdefg, 3)));
+  WebPInt32ToMem(dst + 3 * BPS, _mm_cvtsi128_si32(               abcdefg    ));
+  WebPInt32ToMem(dst + 2 * BPS, _mm_cvtsi128_si32(_mm_srli_si128(abcdefg, 1)));
+  WebPInt32ToMem(dst + 1 * BPS, _mm_cvtsi128_si32(_mm_srli_si128(abcdefg, 2)));
+  WebPInt32ToMem(dst + 0 * BPS, _mm_cvtsi128_si32(_mm_srli_si128(abcdefg, 3)));
 }
 
 static WEBP_INLINE void HU4_SSE2(uint8_t* dst, const uint8_t* top) {
@@ -882,7 +882,7 @@ static WEBP_INLINE void TM4_SSE2(uint8_t* dst, const uint8_t* top) {
     const int val = top[-2 - y] - top[-1];
     const __m128i base = _mm_set1_epi16(val);
     const __m128i out = _mm_packus_epi16(_mm_add_epi16(base, top_base), zero);
-    WebPUint32ToMem(dst, _mm_cvtsi128_si32(out));
+    WebPInt32ToMem(dst, _mm_cvtsi128_si32(out));
   }
 }
 
diff --git a/src/dsp/rescaler_sse2.c b/src/dsp/rescaler_sse2.c
index 8b231d85..bcd765dc 100644
--- a/src/dsp/rescaler_sse2.c
+++ b/src/dsp/rescaler_sse2.c
@@ -85,7 +85,7 @@ static void RescalerImportRowExpand_SSE2(WebPRescaler* const wrk,
       const __m128i mult = _mm_cvtsi32_si128(((x_add - accum) << 16) | accum);
       const __m128i out = _mm_madd_epi16(cur_pixels, mult);
       assert(sizeof(*frow) == sizeof(uint32_t));
-      WebPUint32ToMem((uint8_t*)frow, _mm_cvtsi128_si32(out));
+      WebPInt32ToMem((uint8_t*)frow, _mm_cvtsi128_si32(out));
       frow += 1;
       if (frow >= frow_end) break;
       accum -= wrk->x_sub;
diff --git a/src/utils/utils.h b/src/utils/utils.h
index 5211b81b..c7cafafc 100644
--- a/src/utils/utils.h
+++ b/src/utils/utils.h
@@ -82,6 +82,10 @@ static WEBP_INLINE void WebPUint32ToMem(uint8_t* const ptr, uint32_t val) {
   memcpy(ptr, &val, sizeof(val));
 }
 
+static WEBP_INLINE void WebPInt32ToMem(uint8_t* const ptr, int val) {
+  WebPUint32ToMem(ptr, (uint32_t)val);
+}
+
 //------------------------------------------------------------------------------
 // Reading/writing data.