Updating to current trunk

(bzr r8254.1.38)

3 files changed, 21 insertions, 453 deletions

diff --git a/src/libnr/Makefile_insert b/src/libnr/Makefile_insert
index 5cd2717be..4b19028f9 100644
--- a/src/libnr/Makefile_insert
+++ b/src/libnr/Makefile_insert
@@ -5,7 +5,6 @@ libnr_mmx_sources =	\
 	libnr/have_mmx.S	\
 	libnr/nr_mmx_R8G8B8A8_P_EMPTY_A8_RGBAP.S	\
 	libnr/nr_mmx_R8G8B8A8_P_R8G8B8A8_P_A8_RGBAP.S	\
-	libnr/nr_mmx_R8G8B8A8_P_R8G8B8A8_P_R8G8B8A8_N_TRANSFORM.S	\
 	libnr/nr_mmx_R8G8B8_R8G8B8_R8G8B8A8_P.S
 endif
 
diff --git a/src/libnr/nr-compose-transform.cpp b/src/libnr/nr-compose-transform.cpp
index afc8fd987..6e03faf2f 100644
--- a/src/libnr/nr-compose-transform.cpp
+++ b/src/libnr/nr-compose-transform.cpp
@@ -16,30 +16,25 @@
 #include "nr-pixops.h"
 #include "nr-matrix.h"
 
-
-#ifdef WITH_MMX
+/*#ifdef WITH_MMX
 #ifdef __cplusplus
 extern "C" {
-#endif /* __cplusplus */
-/* fixme: */
-int nr_have_mmx (void);
-void nr_mmx_R8G8B8A8_P_R8G8B8A8_P_R8G8B8A8_N_TRANSFORM_0 (unsigned char *px, int w, int h, int rs,
-							  const unsigned char *spx, int sw, int sh, int srs,
-							  const long *FFd2s, unsigned int alpha);
-void nr_mmx_R8G8B8A8_P_R8G8B8A8_P_R8G8B8A8_N_TRANSFORM_n (unsigned char *px, int w, int h, int rs,
-							  const unsigned char *spx, int sw, int sh, int srs,
-							  const long *FFd2s, const long *FF_S, unsigned int alpha, int dbits);
+#endif // __cplusplus
+/ * fixme: * /
+/ *int nr_have_mmx (void);
 #define NR_PIXOPS_MMX (1 && nr_have_mmx ())
 #ifdef __cplusplus
 }
 #endif //__cplusplus
 #endif
+*/
 
 /* fixme: Implement missing (Lauris) */
 /* fixme: PREMUL colors before calculating average (Lauris) */
 
 /* Fixed point precision */
 #define FBITS 12
+#define FBITS_HP 18 // In some places we need a higher precision
 
 void nr_R8G8B8A8_N_EMPTY_R8G8B8A8_N_TRANSFORM (unsigned char *px, int w, int h, int rs,
 					       const unsigned char *spx, int sw, int sh, int srs,
@@ -168,10 +163,10 @@ void nr_R8G8B8A8_N_R8G8B8A8_N_R8G8B8A8_P_TRANSFORM (unsigned char *px, int w, in
 static void
 nr_R8G8B8A8_P_R8G8B8A8_P_R8G8B8A8_N_TRANSFORM_0 (unsigned char *px, int w, int h, int rs,
 						 const unsigned char *spx, int sw, int sh, int srs,
-						 const long *FFd2s, unsigned int alpha)
+						 const long long *FFd2s, unsigned int alpha)
 {
-	unsigned char *d0;
-	int FFsx0, FFsy0;
+    unsigned char *d0;
+	long long FFsx0, FFsy0;
 	int x, y;
 
 	d0 = px;
@@ -180,15 +175,15 @@ nr_R8G8B8A8_P_R8G8B8A8_P_R8G8B8A8_N_TRANSFORM_0 (unsigned char *px, int w, int h
 
 	for (y = 0; y < h; y++) {
 		unsigned char *d;
-		long FFsx, FFsy;
+		long long FFsx, FFsy;
 		d = d0;
 		FFsx = FFsx0;
 		FFsy = FFsy0;
 		for (x = 0; x < w; x++) {
 			long sx, sy;
-			sx = FFsx >> FBITS;
+			sx = long(FFsx >> FBITS_HP);
 			if ((sx >= 0) && (sx < sw)) {
-				sy = FFsy >> FBITS;
+				sy = long(FFsy >> FBITS_HP);
 				if ((sy >= 0) && (sy < sh)) {
 					const unsigned char *s;
 					unsigned int a;
@@ -224,11 +219,11 @@ nr_R8G8B8A8_P_R8G8B8A8_P_R8G8B8A8_N_TRANSFORM_0 (unsigned char *px, int w, int h
 static void
 nr_R8G8B8A8_P_R8G8B8A8_P_R8G8B8A8_N_TRANSFORM_n (unsigned char *px, int w, int h, int rs,
 						 const unsigned char *spx, int sw, int sh, int srs,
-						 const long *FFd2s, const long *FF_S, unsigned int alpha, int dbits)
+						 const long long *FFd2s, const long *FF_S, unsigned int alpha, int dbits)
 {
 	int size;
 	unsigned char *d0;
-	int FFsx0, FFsy0;
+	long long FFsx0, FFsy0;
 	int x, y;
 
 	size = (1 << dbits);
@@ -242,7 +237,7 @@ nr_R8G8B8A8_P_R8G8B8A8_P_R8G8B8A8_N_TRANSFORM_n (unsigned char *px, int w, int h
 
 	for (y = 0; y < h; y++) {
 		unsigned char *d;
-		long FFsx, FFsy;
+		long long FFsx, FFsy;
 		d = d0;
 		FFsx = FFsx0;
 		FFsy = FFsy0;
@@ -252,9 +247,9 @@ nr_R8G8B8A8_P_R8G8B8A8_P_R8G8B8A8_N_TRANSFORM_n (unsigned char *px, int w, int h
 			r = g = b = a = 0;
 			for (i = 0; i < size; i++) {
 				long sx, sy;
-				sx = (FFsx + FF_S[2 * i]) >> FBITS;
+				sx = (long (FFsx >> (FBITS_HP - FBITS)) + FF_S[2 * i]) >> FBITS;
 				if ((sx >= 0) && (sx < sw)) {
-					sy = (FFsy + FF_S[2 * i + 1]) >> FBITS;
+					sy = (long (FFsy >> (FBITS_HP - FBITS)) + FF_S[2 * i + 1]) >> FBITS;
 					if ((sy >= 0) && (sy < sh)) {
 						const unsigned char *s;
 						unsigned int ca;
@@ -302,6 +297,7 @@ void nr_R8G8B8A8_P_R8G8B8A8_P_R8G8B8A8_N_TRANSFORM (unsigned char *px, int w, in
 {
 	int dbits;
 	long FFd2s[6];
+	long long FFd2s_HP[6]; // with higher precision
 	int i;
 
 	if (alpha == 0) return;
@@ -310,17 +306,11 @@ void nr_R8G8B8A8_P_R8G8B8A8_P_R8G8B8A8_N_TRANSFORM (unsigned char *px, int w, in
 
 	for (i = 0; i < 6; i++) {
 		FFd2s[i] = (long) (d2s[i] * (1 << FBITS) + 0.5);
+		FFd2s_HP[i] = (long long) (d2s[i] * (1 << FBITS_HP) + 0.5);;
 	}
 
 	if (dbits == 0) {
-#ifdef WITH_MMX
-		if (NR_PIXOPS_MMX) {
-			/* WARNING: MMX composer REQUIRES w > 0 and h > 0 */
-			nr_mmx_R8G8B8A8_P_R8G8B8A8_P_R8G8B8A8_N_TRANSFORM_0 (px, w, h, rs, spx, sw, sh, srs, FFd2s, alpha);
-			return;
-		}
-#endif
-		nr_R8G8B8A8_P_R8G8B8A8_P_R8G8B8A8_N_TRANSFORM_0 (px, w, h, rs, spx, sw, sh, srs, FFd2s, alpha);
+		nr_R8G8B8A8_P_R8G8B8A8_P_R8G8B8A8_N_TRANSFORM_0 (px, w, h, rs, spx, sw, sh, srs, FFd2s_HP, alpha);
 	} else {
 		int xsize, ysize;
 		long FFs_x_x_S, FFs_x_y_S, FFs_y_x_S, FFs_y_y_S;
@@ -344,14 +334,7 @@ void nr_R8G8B8A8_P_R8G8B8A8_P_R8G8B8A8_N_TRANSFORM (unsigned char *px, int w, in
 			}
 		}
 
-#ifdef WITH_MMX
-		if (NR_PIXOPS_MMX) {
-			/* WARNING: MMX composer REQUIRES w > 0 and h > 0 */
-			nr_mmx_R8G8B8A8_P_R8G8B8A8_P_R8G8B8A8_N_TRANSFORM_n (px, w, h, rs, spx, sw, sh, srs, FFd2s, FF_S, alpha, dbits);
-			return;
-		}
-#endif
-		nr_R8G8B8A8_P_R8G8B8A8_P_R8G8B8A8_N_TRANSFORM_n (px, w, h, rs, spx, sw, sh, srs, FFd2s, FF_S, alpha, dbits);
+		nr_R8G8B8A8_P_R8G8B8A8_P_R8G8B8A8_N_TRANSFORM_n (px, w, h, rs, spx, sw, sh, srs, FFd2s_HP, FF_S, alpha, dbits);
 	}
 }
 
diff --git a/src/libnr/nr_mmx_R8G8B8A8_P_R8G8B8A8_P_R8G8B8A8_N_TRANSFORM.S b/src/libnr/nr_mmx_R8G8B8A8_P_R8G8B8A8_P_R8G8B8A8_N_TRANSFORM.S
deleted file mode 100644
index e30056af2..000000000
--- a/src/libnr/nr_mmx_R8G8B8A8_P_R8G8B8A8_P_R8G8B8A8_N_TRANSFORM.S
+++ /dev/null
@@ -1,414 +0,0 @@
-	.file	"nr-compose-transform.c"
-
-# Ensure Inkscape is execshield protected
-	.section .note.GNU-stack
-	.previous
-	
-	.text
-	.align 2
-.globl nr_mmx_R8G8B8A8_P_R8G8B8A8_P_R8G8B8A8_N_TRANSFORM_0
-	.type	nr_mmx_R8G8B8A8_P_R8G8B8A8_P_R8G8B8A8_N_TRANSFORM_0,@function
-
-/*
- * This code is in public domain
- *
- */
-
-nr_mmx_R8G8B8A8_P_R8G8B8A8_P_R8G8B8A8_N_TRANSFORM_0:
-	pushl	%ebp
-	movl	%esp, %ebp
-	pushl	%ebx
-	subl	$48, %esp
-	pushl	%edi
-	pushl	%esi
-
-/* Load %mm7 with [0 0 0 0] */
-	movl	$0, %eax
-	movd	%eax, %mm7
-
-/* Load %mm6 with [128 128 128 128] */
-	movl	$0x80808080, %eax
-	movd	%eax, %mm6
-	punpcklbw %mm7, %mm6
-
-/* Load %mm5 with [255 255 255 255] */
-	movl	$0xffffffff, %eax
-	movd	%eax, %mm5
-	punpcklbw %mm7, %mm5
-
-/* Load %mm0 with [a a a a] */
-	movzbl	44(%ebp), %eax
-	movd	%eax, %mm0
-	punpcklwd %mm0, %mm0
-	punpckldq %mm0, %mm0
-
-	movl	8(%ebp), %eax
-	movl	%eax, -8(%ebp)
-	movl	40(%ebp), %eax
-	addl	$16, %eax
-	movl	(%eax), %eax
-	movl	%eax, -12(%ebp)
-	movl	40(%ebp), %eax
-	addl	$20, %eax
-	movl	(%eax), %eax
-	movl	%eax, -16(%ebp)
-	movl	$0, -24(%ebp)
-.L29:
-	movl	-24(%ebp), %eax
-	cmpl	16(%ebp), %eax
-	jl	.L32
-	jmp	.L28
-.L32:
-	movl	-8(%ebp), %edi
-
-	movl	-12(%ebp), %eax
-	movl	%eax, %esi
-	movl	-16(%ebp), %eax
-	movl	%eax, -36(%ebp)
-
-	movl	12(%ebp), %ebx
-.for_x_0:
-
-	movl	%esi, %ecx
-	cmpl	$0, %ecx
-	js	.clip_0
-	sarl	$12, %ecx
-	cmpl	28(%ebp), %ecx
-	jge	.clip_0
-	shll	$2, %ecx
-
-	movl	-36(%ebp), %eax
-	cmpl	$0, %eax
-	js	.clip_0
-	sarl	$12, %eax
-	cmpl	32(%ebp), %eax
-	jge	.clip_0
-	imull	36(%ebp), %eax
-
-	addl	%ecx, %eax
-	addl	24(%ebp), %eax
-
-/* Fg -> %mm1 */
-	movl	(%eax), %eax
-	testl	$0xff000000, %eax
-	jz	.clip_0
-	movd	%eax, %mm1
-	punpcklbw %mm7, %mm1
-
-/* [a a a 255] -> %mm3 */
-	shrl	$24, %eax
-	movl	$0x10101, %edx
-	mull	%edx
-	orl	$0xff000000, %eax
-	movd	%eax, %mm3
-	punpcklbw %mm7, %mm3
-
-/* [Fg * a] -> mm1 */
-	pmullw	%mm3, %mm1
-	paddw	%mm6, %mm1
-	movq	%mm1, %mm4
-	psrlw	$8, %mm4
-	paddw	%mm4, %mm1
-	psrlw	$8, %mm1
-
-/* Multiply by alpha */
-	pmullw	%mm0, %mm1
-	paddw	%mm6, %mm1
-	movq	%mm1, %mm4
-	psrlw	$8, %mm4
-	paddw	%mm4, %mm1
-	psrlw	$8, %mm1
-
-/* [255 - FgA] -> mm2 */
-	movq	%mm1, %mm2
-	punpckhwd %mm2, %mm2
-	punpckhdq %mm2, %mm2
-	pxor	%mm5, %mm2
-
-/* Bg -> mm3 */
-	movd	(%edi), %mm3
-	punpcklbw %mm7, %mm3
-
-/* Fg + ((255 - FgA) * Bg) / 255 */
-
-	pmullw	%mm2, %mm3
-	paddw	%mm6, %mm3
-	movq	%mm3, %mm4
-	psrlw	$8, %mm4
-	paddw	%mm4, %mm3
-	psrlw	$8, %mm3
-	paddw	%mm1, %mm3
-
-/* Store pixel */
-	packuswb %mm3, %mm3
-	movd	%mm3, (%edi)
-
-.clip_0:
-.L37:
-	movl	40(%ebp), %ecx
-	movl	(%ecx), %edx
-	addl	%edx, %esi
-	movl	4(%ecx), %edx
-	addl	%edx, -36(%ebp)
-
-	addl	$4, %edi
-
-	decl	%ebx
-	jnz	.for_x_0
-
-.L34:
-	movl	8(%ecx), %edx
-	addl	%edx, -12(%ebp)
-	movl	12(%ecx), %edx
-	addl	%edx, -16(%ebp)
-
-	movl	20(%ebp), %edx
-	leal	-8(%ebp), %eax
-	addl	%edx, (%eax)
-	leal	-24(%ebp), %eax
-	incl	(%eax)
-	jmp	.L29
-.L28:
-	emms
-	popl	%esi
-	popl	%edi
-	addl	$48, %esp
-	popl	%ebx
-	popl	%ebp
-	ret
-.Lfe2:
-	.size	nr_mmx_R8G8B8A8_P_R8G8B8A8_P_R8G8B8A8_N_TRANSFORM_0,.Lfe2-nr_mmx_R8G8B8A8_P_R8G8B8A8_P_R8G8B8A8_N_TRANSFORM_0
-
-/*
- *
- *	dbits	52(%ebp)
- *	alpha	48(%ebp)
- *	FF_S	44(%ebp)
- *
- *	d	-32(%ebp) -> %edi
- *	i	-60(%ebp) -> %esi
- *	sx	-64(%ebp) -> %ebx
- *	sy	-68(%ebp)
- *	s	-72(%ebp)
- *
- *	%mm0	a a a a
- *	%mm1	FgA
- *	%mm2	SumFgA
- *	%mm3	a a a 255
- *	%mm4
-*/
-
-	.align 2
-.globl nr_mmx_R8G8B8A8_P_R8G8B8A8_P_R8G8B8A8_N_TRANSFORM_n
-	.type	nr_mmx_R8G8B8A8_P_R8G8B8A8_P_R8G8B8A8_N_TRANSFORM_n,@function
-nr_mmx_R8G8B8A8_P_R8G8B8A8_P_R8G8B8A8_N_TRANSFORM_n:
-	pushl	%ebp
-	movl	%esp, %ebp
-	pushl	%ebx
-	subl	$72, %esp
-	pushl	%edi
-	pushl	%esi
-
-/* Load %mm7 with [0 0 0 0] */
-	movl	$0, %eax
-	movd	%eax, %mm7
-
-/* Load %mm6 with [128 128 128 128] */
-	movl	$0x80808080, %eax
-	movd	%eax, %mm6
-	punpcklbw %mm7, %mm6
-
-/* Load %mm5 with [255 255 255 255] */
-	movl	$0xffffffff, %eax
-	movd	%eax, %mm5
-	punpcklbw %mm7, %mm5
-
-/* Load %mm0 with [a a a a] */
-	movzbl	48(%ebp), %eax
-	movd	%eax, %mm0
-	punpcklwd %mm0, %mm0
-	punpckldq %mm0, %mm0
-
-	movl	$1, %eax
-	movzbl	52(%ebp), %ecx
-	sall	%cl, %eax
-	movl	%eax, -8(%ebp)
-	movl	8(%ebp), %eax
-	movl	%eax, -12(%ebp)
-	movl	40(%ebp), %eax
-	addl	$16, %eax
-	movl	(%eax), %eax
-	movl	%eax, -16(%ebp)
-	movl	40(%ebp), %eax
-	addl	$20, %eax
-	movl	(%eax), %eax
-	movl	%eax, -20(%ebp)
-	movl	$0, -28(%ebp)
-.L44:
-	movl	-28(%ebp), %eax
-	cmpl	16(%ebp), %eax
-	jl	.L47
-	jmp	.exit_n
-.L47:
-	movl	-12(%ebp), %eax
-	movl	%eax, -32(%ebp)
-	movl	-16(%ebp), %eax
-	movl	%eax, -36(%ebp)
-	movl	-20(%ebp), %eax
-	movl	%eax, -40(%ebp)
-	movl	$0, -24(%ebp)
-.L48:
-	movl	-24(%ebp), %eax
-	cmpl	12(%ebp), %eax
-	jl	.L51
-	jmp	.L49
-.L51:
-
-/* Zero accumulator */
-	movq	%mm7, %mm2
-
-/* Set i to dptr (size - 1) */
-	movl	-8(%ebp), %esi
-	sub	$1, %esi
-	shll	$3, %esi
-
-	movl	44(%ebp), %edi
-	movl	-36(%ebp), %ecx
-
-.for_i_n:
-	movl	(%edi,%esi), %ebx
-	addl	%ecx, %ebx
-/* Test negative before shift */
-	cmpl	$0, %ebx
-	js	.next_i_n
-	sarl	$12, %ebx
-	cmpl	28(%ebp), %ebx
-	jge	.next_i_n
-/* We multiply sx by 4 here */
-	shll	$2, %ebx
-
-	movl	4(%edi,%esi), %eax
-	addl	-40(%ebp), %eax
-/* Test negative before shift */
-	cmpl	$0, %eax
-	js	.next_i_n
-	sarl	$12, %eax
-	cmpl	32(%ebp), %eax
-	jge	.next_i_n
-/* We multiply sy by srs here */
-	imull	36(%ebp), %eax
-
-	addl	%ebx, %eax
-	addl	24(%ebp), %eax
-
-/* Fg -> %mm1 */
-	movl	(%eax), %eax
-	testl	$0xff000000, %eax
-	jz	.next_i_n
-	movd	%eax, %mm1
-	punpcklbw %mm7, %mm1
-
-/* [a a a 255] -> %mm3 */
-	shrl	$24, %eax
-	movl	$0x10101, %edx
-	mull	%edx
-	orl	$0xff000000, %eax
-	movd	%eax, %mm3
-	punpcklbw %mm7, %mm3
-
-/* [Fg * a] -> mm1 */
-	pmullw	%mm3, %mm1
-	paddw	%mm6, %mm1
-	movq	%mm1, %mm4
-	psrlw	$8, %mm4
-	paddw	%mm4, %mm1
-	psrlw	$8, %mm1
-
-/* Add to accumulator */
-	paddw	%mm1, %mm2
-
-.next_i_n:
-	subl	$8, %esi
-	jnb	.for_i_n
-
-/* Divide components by sample size */
-	movd	52(%ebp), %mm3
-	psrlw	%mm3, %mm2
-
-/* Multiply by alpha */
-	pmullw	%mm0, %mm2
-	paddw	%mm6, %mm2
-	movq	%mm2, %mm4
-	psrlw	$8, %mm4
-	paddw	%mm4, %mm2
-	psrlw	$8, %mm2
-
-/* [255 - FgA] -> mm1 */
-	movq	%mm2, %mm1
-	punpckhwd %mm1, %mm1
-	punpckhdq %mm1, %mm1
-	pxor	%mm5, %mm1
-
-	movl	-32(%ebp), %edi
-/* Bg -> mm3 */
-	movd	(%edi), %mm3
-	punpcklbw %mm7, %mm3
-
-/* Fg + ((255 - FgA) * Bg) / 255 */
-
-	pmullw	%mm1, %mm3
-	paddw	%mm6, %mm3
-	movq	%mm3, %mm4
-	psrlw	$8, %mm4
-	paddw	%mm4, %mm3
-	psrlw	$8, %mm3
-	paddw	%mm2, %mm3
-
-/* Store pixel */
-	packuswb %mm3, %mm3
-	movd	%mm3, (%edi)
-
-.L58:
-	movl	40(%ebp), %eax
-	movl	(%eax), %edx
-	leal	-36(%ebp), %eax
-	addl	%edx, (%eax)
-	movl	40(%ebp), %eax
-	addl	$4, %eax
-	movl	(%eax), %edx
-	leal	-40(%ebp), %eax
-	addl	%edx, (%eax)
-	leal	-32(%ebp), %eax
-	addl	$4, (%eax)
-	leal	-24(%ebp), %eax
-	incl	(%eax)
-	jmp	.L48
-.L49:
-	movl	40(%ebp), %eax
-	addl	$8, %eax
-	movl	(%eax), %edx
-	leal	-16(%ebp), %eax
-	addl	%edx, (%eax)
-	movl	40(%ebp), %eax
-	addl	$12, %eax
-	movl	(%eax), %edx
-	leal	-20(%ebp), %eax
-	addl	%edx, (%eax)
-	movl	20(%ebp), %edx
-	leal	-12(%ebp), %eax
-	addl	%edx, (%eax)
-	leal	-28(%ebp), %eax
-	incl	(%eax)
-	jmp	.L44
-
-.exit_n:
-	emms
-	popl	%esi
-	popl	%edi
-	addl	$72, %esp
-	popl	%ebx
-	popl	%ebp
-	ret
-.Lfe3:
-	.size	nr_mmx_R8G8B8A8_P_R8G8B8A8_P_R8G8B8A8_N_TRANSFORM_n,.Lfe3-nr_mmx_R8G8B8A8_P_R8G8B8A8_P_R8G8B8A8_N_TRANSFORM_n
-	.ident	"GCC: (GNU) 3.2"