Port the more accurate (and slightly faster) floating point IDCT implementation from jpeg-8a and later. New research revealed that the SSE/SSE2 floating point IDCT implementation was actually more accurate than the jpeg-6b implementation, not less, which is why its mathematical results have always differed from those of the jpeg-6b implementation. This patch brings the accuracy of the C code in line with that of the SSE/SSE2 code.
git-svn-id: svn+ssh://svn.code.sf.net/p/libjpeg-turbo/code/trunk@1288 632fc199-4ca6-4c93-a231-07263d6284db
diff --git a/Makefile.am b/Makefile.am
index f14ddbf..1844672 100644
--- a/Makefile.am
+++ b/Makefile.am
@@ -175,15 +175,13 @@
MD5_PPM_GRAY_ISLOW = 8d3596c56eace32f205deccc229aa5ed
MD5_PPM_GRAY_RGB_ISLOW = 116424ac07b79e5e801f00508eab48ec
MD5_JPEG_420S_IFAST_OPT = 388708217ac46273ca33086b22827ed8
-# The SSE/SSE2 DCT/IDCT implementation has always produced a slight round-off
-# error relative to the C code, so in this case, we just test for regression
-# rather than verifying that the output matches libjpeg.
+# See README-turbo.txt for more details on why this next bit is necessary.
if WITH_SSE_FLOAT_DCT
MD5_JPEG_3x2_FLOAT_PROG = 343e3f8caf8af5986ebaf0bdc13b5c71
MD5_PPM_3x2_FLOAT = 1a75f36e5904d6fc3a85a43da9ad89bb
else
MD5_JPEG_3x2_FLOAT_PROG = 9bca803d2042bd1eb03819e2bf92b3e5
-MD5_PPM_3x2_FLOAT = ef6a420e369440edd6b918a0cf54ba3f
+MD5_PPM_3x2_FLOAT = f6bfab038438ed8f5522fbd33595dcdc
endif
MD5_JPEG_420_ISLOW_ARI = e986fb0a637a8d833d96e8a6d6d84ea1
MD5_JPEG_444_ISLOW_PROGARI = 0a8f1c8f66e113c3cf635df0a475a617