Upgrade libjpeg to 2.0.2

Test: build
Change-Id: If51cbf56f0b48fde26fac4f36d4f0219068659ac

8 files changed, 67 insertions, 18 deletions

diff --git a/simd/CMakeLists.txt b/simd/CMakeLists.txt
index 346994c..8dbd7f1 100755
--- a/simd/CMakeLists.txt
+++ b/simd/CMakeLists.txt
@@ -262,7 +262,7 @@ endif()
 # MIPS (GAS)
 ###############################################################################
 
-elseif(CPU_TYPE STREQUAL "mips")
+elseif(CPU_TYPE STREQUAL "mips" OR CPU_TYPE STREQUAL "mipsel")
 
 enable_language(ASM)
 
@@ -293,7 +293,7 @@ if(NOT HAVE_DSPR2)
   return()
 endif()
 
-add_library(simd OBJECT ${CPU_TYPE}/jsimd_dspr2.S ${CPU_TYPE}/jsimd.c)
+add_library(simd OBJECT mips/jsimd_dspr2.S mips/jsimd.c)
 
 if(CMAKE_POSITION_INDEPENDENT_CODE OR ENABLE_SHARED)
   set_target_properties(simd PROPERTIES POSITION_INDEPENDENT_CODE 1)
diff --git a/simd/arm64/jsimd_neon.S b/simd/arm64/jsimd_neon.S
index 447dedf..d30715a 100644
--- a/simd/arm64/jsimd_neon.S
+++ b/simd/arm64/jsimd_neon.S
@@ -309,7 +309,7 @@ _\fname:
     trn2            \x1\literal, \xi\literal, \x1\literal
 .endm
 
-/* Transpose elements of 2 differnet registers */
+/* Transpose elements of 2 different registers */
 .macro transpose x0, x1, xi, xilen, literal
     mov             \xi\xilen, \x0\xilen
     trn1            \x0\literal, \x0\literal, \x1\literal
diff --git a/simd/i386/jsimdcpu.asm b/simd/i386/jsimdcpu.asm
index 50a0d51..faddd38 100644
--- a/simd/i386/jsimdcpu.asm
+++ b/simd/i386/jsimdcpu.asm
@@ -94,9 +94,10 @@ EXTN(jpeg_simd_cpu_support):
 
     xor         ecx, ecx
     xgetbv
-    test        eax, 6                  ; O/S does not manage XMM/YMM state
+    and         eax, 6
+    cmp         eax, 6                  ; O/S does not manage XMM/YMM state
                                         ; using XSAVE
-    jz          short .no_avx2
+    jnz         short .no_avx2
 
     or          edi, JSIMD_AVX2
 .no_avx2:
diff --git a/simd/loongson/jccolext-mmi.c b/simd/loongson/jccolext-mmi.c
index e1c4e69..6cdeb5e 100644
--- a/simd/loongson/jccolext-mmi.c
+++ b/simd/loongson/jccolext-mmi.c
@@ -2,12 +2,13 @@
  * Loongson MMI optimizations for libjpeg-turbo
  *
  * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
- * Copyright (C) 2014-2015, D. R. Commander.  All Rights Reserved.
- * Copyright (C) 2016-2017, Loongson Technology Corporation Limited, BeiJing.
+ * Copyright (C) 2014-2015, 2019, D. R. Commander.  All Rights Reserved.
+ * Copyright (C) 2016-2018, Loongson Technology Corporation Limited, BeiJing.
  *                          All Rights Reserved.
  * Authors:  ZhuChen     <zhuchen@loongson.cn>
  *           SunZhangzhi <sunzhangzhi-cq@loongson.cn>
  *           CaiWanwei   <caiwanwei@loongson.cn>
+ *           ZhangLixia  <zhanglixia-hf@loongson.cn>
  *
  * Based on the x86 SIMD extension for IJG JPEG library
  * Copyright (C) 1999-2006, MIYASAKA Masaru.
@@ -184,9 +185,15 @@ void jsimd_rgb_ycc_convert_mmi(JDIMENSION image_width, JSAMPARRAY input_buf,
               "$14", "memory"
            );
       } else {
-        mmA = _mm_load_si64((__m64 *)&inptr[0]);
-        mmG = _mm_load_si64((__m64 *)&inptr[8]);
-        mmF = _mm_load_si64((__m64 *)&inptr[16]);
+        if (!(((long)inptr) & 7)) {
+          mmA = _mm_load_si64((__m64 *)&inptr[0]);
+          mmG = _mm_load_si64((__m64 *)&inptr[8]);
+          mmF = _mm_load_si64((__m64 *)&inptr[16]);
+        } else {
+          mmA = _mm_loadu_si64((__m64 *)&inptr[0]);
+          mmG = _mm_loadu_si64((__m64 *)&inptr[8]);
+          mmF = _mm_loadu_si64((__m64 *)&inptr[16]);
+        }
         inptr += RGB_PIXELSIZE * 8;
       }
       mmD = mmA;
@@ -268,10 +275,17 @@ void jsimd_rgb_ycc_convert_mmi(JDIMENSION image_width, JSAMPARRAY input_buf,
             : "$f0", "$f2", "$8", "$9", "$10", "$11", "$13", "memory"
            );
       } else {
-        mmA = _mm_load_si64((__m64 *)&inptr[0]);
-        mmF = _mm_load_si64((__m64 *)&inptr[8]);
-        mmD = _mm_load_si64((__m64 *)&inptr[16]);
-        mmC = _mm_load_si64((__m64 *)&inptr[24]);
+        if (!(((long)inptr) & 7)) {
+          mmA = _mm_load_si64((__m64 *)&inptr[0]);
+          mmF = _mm_load_si64((__m64 *)&inptr[8]);
+          mmD = _mm_load_si64((__m64 *)&inptr[16]);
+          mmC = _mm_load_si64((__m64 *)&inptr[24]);
+        } else {
+          mmA = _mm_loadu_si64((__m64 *)&inptr[0]);
+          mmF = _mm_loadu_si64((__m64 *)&inptr[8]);
+          mmD = _mm_loadu_si64((__m64 *)&inptr[16]);
+          mmC = _mm_loadu_si64((__m64 *)&inptr[24]);
+        }
         inptr += RGB_PIXELSIZE * 8;
       }
       mmB = mmA;
diff --git a/simd/loongson/loongson-mmintrin.h b/simd/loongson/loongson-mmintrin.h
index 4aea763..50d166b 100644
--- a/simd/loongson/loongson-mmintrin.h
+++ b/simd/loongson/loongson-mmintrin.h
@@ -1,8 +1,9 @@
 /*
  * Loongson MMI optimizations for libjpeg-turbo
  *
- * Copyright (C) 2016-2017, Loongson Technology Corporation Limited, BeiJing.
+ * Copyright (C) 2016-2018, Loongson Technology Corporation Limited, BeiJing.
  *                          All Rights Reserved.
+ * Copyright (C) 2019, D. R. Commander.  All Rights Reserved.
  *
  * This software is provided 'as-is', without any express or implied
  * warranty.  In no event will the authors be held liable for any damages
@@ -41,7 +42,7 @@ typedef float __m32;
 
 /********** Set Operations **********/
 
-extern __inline __m64
+extern __inline __m64 FUNCTION_ATTRIBS
 _mm_setzero_si64(void)
 {
   return 0.0;
@@ -1245,6 +1246,22 @@ _mm_load_si64(const __m64 *src)
   asm("ldc1 %0, %1\n\t"
       : "=f" (ret)
       : "m" (*src)
+      : "memory"
+     );
+
+  return ret;
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_loadu_si64(const __m64 *src)
+{
+  __m64 ret;
+
+  asm("gsldlc1 %0,  7(%1)\n\t"
+      "gsldrc1 %0,  0(%1)\n\t"
+      : "=f" (ret)
+      : "r" (src)
+      : "memory"
      );
 
   return ret;
diff --git a/simd/mips/jsimd.c b/simd/mips/jsimd.c
index af886f6..454cc99 100644
--- a/simd/mips/jsimd.c
+++ b/simd/mips/jsimd.c
@@ -692,8 +692,10 @@ jsimd_can_convsamp_float(void)
   if (sizeof(ISLOW_MULT_TYPE) != 2)
     return 0;
 
+#ifndef __mips_soft_float
   if (simd_support & JSIMD_DSPR2)
     return 1;
+#endif
 
   return 0;
 }
@@ -709,7 +711,9 @@ GLOBAL(void)
 jsimd_convsamp_float(JSAMPARRAY sample_data, JDIMENSION start_col,
                      FAST_FLOAT *workspace)
 {
+#ifndef __mips_soft_float
   jsimd_convsamp_float_dspr2(sample_data, start_col, workspace);
+#endif
 }
 
 GLOBAL(int)
@@ -805,8 +809,10 @@ jsimd_can_quantize_float(void)
   if (sizeof(ISLOW_MULT_TYPE) != 2)
     return 0;
 
+#ifndef __mips_soft_float
   if (simd_support & JSIMD_DSPR2)
     return 1;
+#endif
 
   return 0;
 }
@@ -821,7 +827,9 @@ GLOBAL(void)
 jsimd_quantize_float(JCOEFPTR coef_block, FAST_FLOAT *divisors,
                      FAST_FLOAT *workspace)
 {
+#ifndef __mips_soft_float
   jsimd_quantize_float_dspr2(coef_block, divisors, workspace);
+#endif
 }
 
 GLOBAL(int)
diff --git a/simd/mips/jsimd_dspr2.S b/simd/mips/jsimd_dspr2.S
index 2ec543e..a28c116 100644
--- a/simd/mips/jsimd_dspr2.S
+++ b/simd/mips/jsimd_dspr2.S
@@ -2810,6 +2810,8 @@ LEAF_DSPR2(jsimd_quantize_dspr2)
 END(jsimd_quantize_dspr2)
 
 
+#ifndef __mips_soft_float
+
 /*****************************************************************************/
 LEAF_DSPR2(jsimd_quantize_float_dspr2)
 /*
@@ -2890,6 +2892,8 @@ LEAF_DSPR2(jsimd_quantize_float_dspr2)
 
 END(jsimd_quantize_float_dspr2)
 
+#endif
+
 
 /*****************************************************************************/
 LEAF_DSPR2(jsimd_idct_2x2_dspr2)
@@ -4110,6 +4114,8 @@ LEAF_DSPR2(jsimd_convsamp_dspr2)
 END(jsimd_convsamp_dspr2)
 
 
+#ifndef __mips_soft_float
+
 /*****************************************************************************/
 LEAF_DSPR2(jsimd_convsamp_float_dspr2)
 /*
@@ -4468,4 +4474,6 @@ LEAF_DSPR2(jsimd_convsamp_float_dspr2)
 
 END(jsimd_convsamp_float_dspr2)
 
+#endif
+
 /*****************************************************************************/
diff --git a/simd/x86_64/jsimdcpu.asm b/simd/x86_64/jsimdcpu.asm
index 42979be..38e1a7b 100644
--- a/simd/x86_64/jsimdcpu.asm
+++ b/simd/x86_64/jsimdcpu.asm
@@ -60,9 +60,10 @@ EXTN(jpeg_simd_cpu_support):
 
     xor         rcx, rcx
     xgetbv
-    test        rax, 6                  ; O/S does not manage XMM/YMM state
+    and         rax, 6
+    cmp         rax, 6                  ; O/S does not manage XMM/YMM state
                                         ; using XSAVE
-    jz          short .return
+    jnz         short .return
 
     or          rdi, JSIMD_AVX2