Fix x86-64 ABI conformance issue in SIMD code

(descriptions cribbed by DRC from discussion in flutter#20)
In the x86-64 ABI, the high (unused) DWORD of a 32-bit argument's
register is undefined, so it was incorrect to use a 64-bit mov
instruction to transfer a JDIMENSION argument in the 64-bit SSE2 SIMD
functions.  The code worked thus far only because the existing compiler
optimizers weren't smart enough to do anything else with the register in
question, so the upper 32 bits happened to be all zeroes-- for the past
6 years, on every x86-64 compiler previously known to mankind.

The bleeding-edge Clang/LLVM compiler has a smarter optimizer, and
under certain circumstances, it will attempt to load-combine adjacent
32-bit integers from one of the libjpeg structures into a single 64-bit
integer and pass that 64-bit integer as a 32-bit argument to one of the
SIMD functions (which is allowed by the ABI, since the upper 32 bits of
the 32-bit argument's register are undefined.)  This caused the
libjpeg-turbo regression tests to crash.

Also enhance the documentation of JDIMENSION to explain that its size
is significant to the implementation of the SIMD code.

Closes flutter#20.  Refer also to http://crbug.com/532214.

Author: chandlerc

ChangeLog.txt

@@ -11,6 +11,16 @@ incorrectly encode certain JPEG images when quality=100 and the fast integer
 forward DCT were used.  This was known to cause 'make test' to fail when the
 library was built with '-march=haswell' on x86 systems.
 
+[3] Fixed an issue whereby libjpeg-turbo would crash when built with the latest
+& greatest development version of the Clang/LLVM compiler.  This was caused by
+an x86-64 ABI conformance issue in some of libjpeg-turbo's 64-bit SSE2 SIMD
+routines.  Those routines were incorrectly using a 64-bit mov instruction to
+transfer a 32-bit JDIMENSION argument, whereas the x86-64 ABI allows the upper
+(unused) 32 bits of a 32-bit argument's register to be undefined.  The new
+Clang/LLVM optimizer uses load combining to transfer multiple adjacent 32-bit
+structure members into a single 64-bit register, and this exposed the ABI
+conformance issue.
+
 
 1.4.1
 =====

jmorecfg.h

@@ -162,7 +162,9 @@ typedef long INT32;
  * images up to 64K*64K due to 16-bit fields in SOF markers.  Therefore
  * "unsigned int" is sufficient on all machines.  However, if you need to
  * handle larger images and you don't mind deviating from the spec, you
- * can change this datatype.
+ * can change this datatype.  (Note that changing this datatype will
+ * potentially require modifying the SIMD code.  The x86-64 SIMD extensions,
+ * in particular, assume a 32-bit JDIMENSION.)
  */
 
 typedef unsigned int JDIMENSION;

simd/jccolext-sse2-64.asm

@@ -50,14 +50,14 @@ EXTN(jsimd_rgb_ycc_convert_sse2):
         collect_args
         push    rbx
 
-        mov     rcx, r10
+        mov     ecx, r10d
         test    rcx,rcx
         jz      near .return
 
         push    rcx
 
         mov rsi, r12
-        mov rcx, r13
+        mov ecx, r13d
         mov     rdi, JSAMPARRAY [rsi+0*SIZEOF_JSAMPARRAY]
         mov     rbx, JSAMPARRAY [rsi+1*SIZEOF_JSAMPARRAY]
         mov     rdx, JSAMPARRAY [rsi+2*SIZEOF_JSAMPARRAY]

simd/jcgryext-sse2-64.asm

@@ -50,14 +50,14 @@ EXTN(jsimd_rgb_gray_convert_sse2):
         collect_args
         push    rbx
 
-        mov     rcx, r10
+        mov     ecx, r10d
         test    rcx,rcx
         jz      near .return
 
         push    rcx
 
         mov rsi, r12
-        mov rcx, r13
+        mov ecx, r13d
         mov     rdi, JSAMPARRAY [rsi+0*SIZEOF_JSAMPARRAY]
         lea     rdi, [rdi+rcx*SIZEOF_JSAMPROW]
 

simd/jcsample-sse2-64.asm

@@ -49,11 +49,11 @@ EXTN(jsimd_h2v1_downsample_sse2):
         mov     rbp,rsp
         collect_args
 
-        mov rcx, r13
+        mov ecx, r13d
         shl     rcx,3                   ; imul rcx,DCTSIZE (rcx = output_cols)
         jz      near .return
 
-        mov rdx, r10
+        mov edx, r10d
 
         ; -- expand_right_edge
 
@@ -90,7 +90,7 @@ EXTN(jsimd_h2v1_downsample_sse2):
 
         ; -- h2v1_downsample
 
-        mov     rax, r12        ; rowctr
+        mov     eax, r12d        ; rowctr
         test    eax,eax
         jle     near .return
 
@@ -193,11 +193,11 @@ EXTN(jsimd_h2v2_downsample_sse2):
         mov     rbp,rsp
         collect_args
 
-        mov     rcx, r13
+        mov     ecx, r13d
         shl     rcx,3                   ; imul rcx,DCTSIZE (rcx = output_cols)
         jz      near .return
 
-        mov     rdx, r10
+        mov     edx, r10d
 
         ; -- expand_right_edge
 
@@ -234,7 +234,7 @@ EXTN(jsimd_h2v2_downsample_sse2):
 
         ; -- h2v2_downsample
 
-        mov     rax, r12        ; rowctr
+        mov     eax, r12d        ; rowctr
         test    rax,rax
         jle     near .return
 

simd/jdcolext-sse2-64.asm

@@ -52,14 +52,14 @@ EXTN(jsimd_ycc_rgb_convert_sse2):
         collect_args
         push    rbx
 
-        mov     rcx, r10        ; num_cols
+        mov     ecx, r10d        ; num_cols
         test    rcx,rcx
         jz      near .return
 
         push    rcx
 
         mov     rdi, r11
-        mov     rcx, r12
+        mov     ecx, r12d
         mov     rsi, JSAMPARRAY [rdi+0*SIZEOF_JSAMPARRAY]
         mov     rbx, JSAMPARRAY [rdi+1*SIZEOF_JSAMPARRAY]
         mov     rdx, JSAMPARRAY [rdi+2*SIZEOF_JSAMPARRAY]

simd/jdmrgext-sse2-64.asm

@@ -52,14 +52,14 @@ EXTN(jsimd_h2v1_merged_upsample_sse2):
         collect_args
         push    rbx
 
-        mov     rcx, r10        ; col
+        mov     ecx, r10d        ; col
         test    rcx,rcx
         jz      near .return
 
         push    rcx
 
         mov     rdi, r11
-        mov     rcx, r12
+        mov     ecx, r12d
         mov     rsi, JSAMPARRAY [rdi+0*SIZEOF_JSAMPARRAY]
         mov     rbx, JSAMPARRAY [rdi+1*SIZEOF_JSAMPARRAY]
         mov     rdx, JSAMPARRAY [rdi+2*SIZEOF_JSAMPARRAY]
@@ -455,10 +455,10 @@ EXTN(jsimd_h2v2_merged_upsample_sse2):
         collect_args
         push    rbx
 
-        mov     rax, r10
+        mov     eax, r10d
 
         mov     rdi, r11
-        mov     rcx, r12
+        mov     ecx, r12d
         mov     rsi, JSAMPARRAY [rdi+0*SIZEOF_JSAMPARRAY]
         mov     rbx, JSAMPARRAY [rdi+1*SIZEOF_JSAMPARRAY]
         mov     rdx, JSAMPARRAY [rdi+2*SIZEOF_JSAMPARRAY]

simd/jdsample-sse2-64.asm

@@ -67,7 +67,7 @@ EXTN(jsimd_h2v1_fancy_upsample_sse2):
         mov     rbp,rsp
         collect_args
 
-        mov     rax, r11  ; colctr
+        mov     eax, r11d  ; colctr
         test    rax,rax
         jz      near .return
 
@@ -214,7 +214,7 @@ EXTN(jsimd_h2v2_fancy_upsample_sse2):
         collect_args
         push    rbx
 
-        mov     rax, r11  ; colctr
+        mov     eax, r11d  ; colctr
         test    rax,rax
         jz      near .return
 
@@ -506,7 +506,7 @@ EXTN(jsimd_h2v1_upsample_sse2):
         mov     rbp,rsp
         collect_args
 
-        mov     rdx, r11
+        mov     edx, r11d
         add     rdx, byte (2*SIZEOF_XMMWORD)-1
         and     rdx, byte -(2*SIZEOF_XMMWORD)
         jz      near .return
@@ -596,7 +596,7 @@ EXTN(jsimd_h2v2_upsample_sse2):
         collect_args
         push    rbx
 
-        mov     rdx, r11
+        mov     edx, r11d
         add     rdx, byte (2*SIZEOF_XMMWORD)-1
         and     rdx, byte -(2*SIZEOF_XMMWORD)
         jz      near .return

simd/jidctflt-sse2-64.asm

@@ -326,7 +326,7 @@ EXTN(jsimd_idct_float_sse2):
         mov     rax, [original_rbp]
         lea     rsi, [workspace]                        ; FAST_FLOAT * wsptr
         mov     rdi, r12        ; (JSAMPROW *)
-        mov     rax, r13
+        mov     eax, r13d
         mov     rcx, DCTSIZE/4                          ; ctr
 .rowloop:
 

simd/jidctfst-sse2-64.asm

@@ -323,7 +323,7 @@ EXTN(jsimd_idct_ifast_sse2):
 
         mov     rax, [original_rbp]
         mov     rdi, r12        ; (JSAMPROW *)
-        mov     rax, r13
+        mov     eax, r13d
 
         ; -- Even part