Add endbr32/64 instruction at the start of asm functions

Only do so when compiling with -fcf-protection (=> __CET__ is defined). Inspired by: 004ff08738 Comment: Fixes SIGILL in tests on 11th gen intel (IBT) #23
Some cleanups in asm code
2024-04-13 15:36:12 +02:00 · 2024-04-13 09:59:07 +02:00
8 changed files with 2199 additions and 2225 deletions
--- a/src/lib/amd64_blend.S
+++ b/src/lib/amd64_blend.S
--- a/src/lib/amd64_blend_cmod.S
+++ b/src/lib/amd64_blend_cmod.S
--- a/src/lib/asm.h
+++ b/src/lib/asm.h
@ -17,4 +17,14 @@
    .size PR_(sym),.-PR_(sym); \
    .align 8;

+#ifdef __CET__
+#if defined(DO_MMX_ASM)
+#define ENDBR_ endbr32
+#elif defined(DO_AMD64_ASM)
+#define ENDBR_ endbr64
+#endif
+#else
+#define ENDBR_
+#endif
+
 #endif                          /* __ASM_H */
--- a/src/lib/asm_blend.S
+++ b/src/lib/asm_blend.S
@ -1,9 +1,7 @@
 #include <config.h>
 #include "asm.h"

-#ifdef DO_MMX_ASM
-
-/*\ 
+/*\
 |*| MMX assembly blending routines for Imlib2
 |*| Written by Willem Monsuwe <willem@stack.nl>
 |*|
@ -51,7 +49,7 @@ FN_(imlib_mmx_reshade_copy_rgb_to_rgba)

 #include "asm_loadimmq.S"

-	
+
 /*\ MMX register use:
 |*| %mm1 = Source value
 |*| %mm2 = Destination value
@ -63,6 +61,7 @@ FN_(imlib_mmx_reshade_copy_rgb_to_rgba)
 /*\ Common code \*/
 /*\ Set MMX mode, save registers, load common parameters \*/
 #define ENTER				\
+	ENDBR_				;\
 	pushl %ebp			;\
 	movl  %esp, %ebp		;\
 	pushl %ebx			;\
@ -95,7 +94,6 @@ FN_(imlib_mmx_reshade_copy_rgb_to_rgba)
 	decl %edx			;\
 	jns 8b

-
 /*\ Unset MMX mode, reset registers, return \*/
 #define LEAVE			\
 9:				;\
@ -281,7 +279,7 @@ PR_(imlib_mmx_copy_rgba_to_rgba):
 	/*\ Load source, save destination \*/
 	movq (%esi, %ecx, 4), %mm1
 	movq %mm1, (%edi, %ecx, 4)
-	
+
 	addl $2, %ecx
 	js 1b
 	jnz 3f
@ -358,7 +356,7 @@ PR_(imlib_mmx_add_blend_rgba_to_rgb):
 	/*\ Pack into lower 4 bytes and save \*/
 	packuswb %mm4, %mm2
 	movd %mm2, (%edi, %ecx, 4)
-	
+
 	incl %ecx
 	js 1b

@ -408,7 +406,7 @@ PR_(imlib_mmx_add_blend_rgba_to_rgba):
 	/*\ Pack into lower 4 bytes and save \*/
 	packuswb %mm4, %mm2
 	movd %mm2, (%edi, %ecx, 4)
-	
+
 	incl %ecx
 	js 1b

@ -440,7 +438,7 @@ PR_(imlib_mmx_add_copy_rgba_to_rgb):
 	/*\ d = d + s, unsigned saturation, and save \*/
 	paddusb %mm1, %mm2
 	movq %mm2, (%edi, %ecx, 4)
-	
+
 	addl $2, %ecx
 	js 1b
 	jnz 3f
@ -473,7 +471,7 @@ PR_(imlib_mmx_add_copy_rgba_to_rgba):
 	/*\ d = d + s, unsigned saturation, and save \*/
 	paddusb %mm1, %mm2
 	movq %mm2, (%edi, %ecx, 4)
-	
+
 	addl $2, %ecx
 	js 1b
 	jnz 3f
@ -511,7 +509,7 @@ PR_(imlib_mmx_add_copy_rgb_to_rgba):
 	/*\ Make result alpha 0xff \*/
 	por %mm5, %mm2
 	movq %mm2, (%edi, %ecx, 4)
-	
+
 	addl $2, %ecx
 	js 1b
 	jnz 3f
@ -559,7 +557,7 @@ PR_(imlib_mmx_subtract_blend_rgba_to_rgb):
 	/*\ Pack into lower 4 bytes and save \*/
 	packuswb %mm4, %mm2
 	movd %mm2, (%edi, %ecx, 4)
-	
+
 	incl %ecx
 	js 1b

@ -609,7 +607,7 @@ PR_(imlib_mmx_subtract_blend_rgba_to_rgba):
 	/*\ Pack into lower 4 bytes and save \*/
 	packuswb %mm4, %mm2
 	movd %mm2, (%edi, %ecx, 4)
-	
+
 	incl %ecx
 	js 1b

@ -641,7 +639,7 @@ PR_(imlib_mmx_subtract_copy_rgba_to_rgb):
 	/*\ d = d - s, unsigned saturation, and save \*/
 	psubusb %mm1, %mm2
 	movq %mm2, (%edi, %ecx, 4)
-	
+
 	addl $2, %ecx
 	js 1b
 	jnz 3f
@ -683,7 +681,7 @@ PR_(imlib_mmx_subtract_copy_rgba_to_rgba):
 	/*\ Negate result alphas \*/
 	pxor %mm5, %mm2
 	movq %mm2, (%edi, %ecx, 4)
-	
+
 	addl $2, %ecx
 	js 1b
 	jnz 3f
@ -723,7 +721,7 @@ PR_(imlib_mmx_subtract_copy_rgb_to_rgba):
 	/*\ Make result alpha 0xff \*/
 	por %mm5, %mm2
 	movq %mm2, (%edi, %ecx, 4)
-	
+
 	addl $2, %ecx
 	js 1b
 	jnz 3f
@ -774,7 +772,7 @@ PR_(imlib_mmx_reshade_blend_rgba_to_rgb):
 	/*\ Pack into lower 4 bytes and save \*/
 	packuswb %mm4, %mm2
 	movd %mm2, (%edi, %ecx, 4)
-	
+
 	incl %ecx
 	js 1b

@ -827,7 +825,7 @@ PR_(imlib_mmx_reshade_blend_rgba_to_rgba):
 	/*\ Pack into lower 4 bytes and save \*/
 	packuswb %mm4, %mm2
 	movd %mm2, (%edi, %ecx, 4)
-	
+
 	incl %ecx
 	js 1b

@ -877,7 +875,7 @@ PR_(imlib_mmx_reshade_copy_rgba_to_rgb):
 	paddusb %mm1, %mm2
 	psubusb %mm3, %mm2
 	movq %mm2, (%edi, %ecx, 4)
-	
+
 	addl $2, %ecx
 	js 1b
 	jnz 3f
@ -942,7 +940,7 @@ PR_(imlib_mmx_reshade_copy_rgba_to_rgba):
 	paddusb %mm1, %mm2
 	psubusb %mm3, %mm2
 	movq %mm2, (%edi, %ecx, 4)
-	
+
 	addl $2, %ecx
 	js 1b
 	jnz 3f
@ -1001,7 +999,7 @@ PR_(imlib_mmx_reshade_copy_rgb_to_rgba):
 	/*\ Make result alpha 0xff \*/
 	por %mm7, %mm2
 	movq %mm2, (%edi, %ecx, 4)
-	
+
 	addl $2, %ecx
 	js 1b
 	jnz 3f
@ -1024,8 +1022,6 @@ PR_(imlib_mmx_reshade_copy_rgb_to_rgba):

 SIZE(imlib_mmx_reshade_copy_rgb_to_rgba)

-#endif
-
 #ifdef __ELF__
 .section .note.GNU-stack,"",@progbits
 #endif
--- a/src/lib/asm_blend_cmod.S
+++ b/src/lib/asm_blend_cmod.S
@ -1,9 +1,7 @@
 #include <config.h>
 #include "asm.h"

-#ifdef DO_MMX_ASM
-
-/*\ 
+/*\
 |*| MMX assembly blending routines, with colour modding, for Imlib2
 |*| Written by Willem Monsuwe <willem@stack.nl>
 |*|
@ -69,7 +67,7 @@ FN_(imlib_mmx_reshade_copy_rgba_to_rgba_cmod)
 FN_(imlib_mmx_reshade_copy_rgb_to_rgba_cmod)

 #include "asm_loadimmq.S"
-	
+
 /*\ MMX register use:
 |*| %mm1 = Source value
 |*| %mm2 = Destination value
@ -81,6 +79,7 @@ FN_(imlib_mmx_reshade_copy_rgb_to_rgba_cmod)
 /*\ Common code \*/
 /*\ Set MMX mode, save registers, load common parameters \*/
 #define ENTER				\
+	ENDBR_				;\
 	pushl %ebp			;\
 	movl  %esp, %ebp		;\
 	pushl %ebx			;\
@ -113,7 +112,6 @@ FN_(imlib_mmx_reshade_copy_rgb_to_rgba_cmod)
 	decl %edx			;\
 	jns 8b

-
 /*\ Unset MMX mode, reset registers, return \*/
 #define LEAVE			\
 9:				;\
@ -622,7 +620,7 @@ PR_(imlib_mmx_add_blend_rgba_to_rgb_cmod):
 	/*\ Pack into lower 4 bytes and save \*/
 	packuswb %mm4, %mm2
 	movd %mm2, (%edi, %ecx, 4)
-	
+
 	incl %ecx
 	js 1b

@ -672,7 +670,7 @@ PR_(imlib_mmx_add_blend_rgba_to_rgba_cmod):
 	/*\ Pack into lower 4 bytes and save \*/
 	packuswb %mm4, %mm2
 	movd %mm2, (%edi, %ecx, 4)
-	
+
 	incl %ecx
 	js 1b

@ -713,7 +711,7 @@ PR_(imlib_mmx_add_blend_rgb_to_rgb_cmod):
 	/*\ Pack into lower 4 bytes and save \*/
 	packuswb %mm4, %mm2
 	movd %mm2, (%edi, %ecx, 4)
-	
+
 	incl %ecx
 	js 1b

@ -763,7 +761,7 @@ PR_(imlib_mmx_add_blend_rgb_to_rgba_cmod):
 	/*\ Pack into lower 4 bytes and save \*/
 	packuswb %mm4, %mm2
 	movd %mm2, (%edi, %ecx, 4)
-	
+
 	incl %ecx
 	js 1b

@ -795,7 +793,7 @@ PR_(imlib_mmx_add_copy_rgba_to_rgb_cmod):
 	/*\ d = d + s, unsigned saturation, and save \*/
 	paddusb %mm1, %mm2
 	movq %mm2, (%edi, %ecx, 4)
-	
+
 	addl $2, %ecx
 	js 1b
 	jnz 3f
@ -828,7 +826,7 @@ PR_(imlib_mmx_add_copy_rgba_to_rgba_cmod):
 	/*\ d = d + s, unsigned saturation, and save \*/
 	paddusb %mm1, %mm2
 	movq %mm2, (%edi, %ecx, 4)
-	
+
 	addl $2, %ecx
 	js 1b
 	jnz 3f
@ -860,7 +858,7 @@ PR_(imlib_mmx_add_copy_rgb_to_rgba_cmod):
 	/*\ d = d + s, unsigned saturation, and save \*/
 	paddusb %mm1, %mm2
 	movq %mm2, (%edi, %ecx, 4)
-	
+
 	addl $2, %ecx
 	js 1b
 	jnz 3f
@ -907,7 +905,7 @@ PR_(imlib_mmx_subtract_blend_rgba_to_rgb_cmod):
 	/*\ Pack into lower 4 bytes and save \*/
 	packuswb %mm4, %mm2
 	movd %mm2, (%edi, %ecx, 4)
-	
+
 	incl %ecx
 	js 1b

@ -957,7 +955,7 @@ PR_(imlib_mmx_subtract_blend_rgba_to_rgba_cmod):
 	/*\ Pack into lower 4 bytes and save \*/
 	packuswb %mm4, %mm2
 	movd %mm2, (%edi, %ecx, 4)
-	
+
 	incl %ecx
 	js 1b

@ -998,7 +996,7 @@ PR_(imlib_mmx_subtract_blend_rgb_to_rgb_cmod):
 	/*\ Pack into lower 4 bytes and save \*/
 	packuswb %mm4, %mm2
 	movd %mm2, (%edi, %ecx, 4)
-	
+
 	incl %ecx
 	js 1b

@ -1048,7 +1046,7 @@ PR_(imlib_mmx_subtract_blend_rgb_to_rgba_cmod):
 	/*\ Pack into lower 4 bytes and save \*/
 	packuswb %mm4, %mm2
 	movd %mm2, (%edi, %ecx, 4)
-	
+
 	incl %ecx
 	js 1b

@ -1080,7 +1078,7 @@ PR_(imlib_mmx_subtract_copy_rgba_to_rgb_cmod):
 	/*\ d = d - s, unsigned saturation, and save \*/
 	psubusb %mm1, %mm2
 	movq %mm2, (%edi, %ecx, 4)
-	
+
 	addl $2, %ecx
 	js 1b
 	jnz 3f
@ -1122,7 +1120,7 @@ PR_(imlib_mmx_subtract_copy_rgba_to_rgba_cmod):
 	/*\ Negate result alphas \*/
 	pxor %mm5, %mm2
 	movq %mm2, (%edi, %ecx, 4)
-	
+
 	addl $2, %ecx
 	js 1b
 	jnz 3f
@ -1161,7 +1159,7 @@ PR_(imlib_mmx_subtract_copy_rgb_to_rgba_cmod):
 	psubusb %mm1, %mm2
 	pxor %mm5, %mm2
 	movq %mm2, (%edi, %ecx, 4)
-	
+
 	addl $2, %ecx
 	js 1b
 	jnz 3f
@ -1213,7 +1211,7 @@ PR_(imlib_mmx_reshade_blend_rgba_to_rgb_cmod):
 	/*\ Pack into lower 4 bytes and save \*/
 	packuswb %mm4, %mm2
 	movd %mm2, (%edi, %ecx, 4)
-	
+
 	incl %ecx
 	js 1b

@ -1266,7 +1264,7 @@ PR_(imlib_mmx_reshade_blend_rgba_to_rgba_cmod):
 	/*\ Pack into lower 4 bytes and save \*/
 	packuswb %mm4, %mm2
 	movd %mm2, (%edi, %ecx, 4)
-	
+
 	incl %ecx
 	js 1b

@ -1310,7 +1308,7 @@ PR_(imlib_mmx_reshade_blend_rgb_to_rgb_cmod):
 	/*\ Pack into lower 4 bytes and save \*/
 	packuswb %mm4, %mm2
 	movd %mm2, (%edi, %ecx, 4)
-	
+
 	incl %ecx
 	js 1b

@ -1363,7 +1361,7 @@ PR_(imlib_mmx_reshade_blend_rgb_to_rgba_cmod):
 	/*\ Pack into lower 4 bytes and save \*/
 	packuswb %mm4, %mm2
 	movd %mm2, (%edi, %ecx, 4)
-	
+
 	incl %ecx
 	js 1b

@ -1413,7 +1411,7 @@ PR_(imlib_mmx_reshade_copy_rgba_to_rgb_cmod):
 	paddusb %mm1, %mm2
 	psubusb %mm3, %mm2
 	movq %mm2, (%edi, %ecx, 4)
-	
+
 	addl $2, %ecx
 	js 1b
 	jnz 3f
@ -1478,7 +1476,7 @@ PR_(imlib_mmx_reshade_copy_rgba_to_rgba_cmod):
 	paddusb %mm1, %mm2
 	psubusb %mm3, %mm2
 	movq %mm2, (%edi, %ecx, 4)
-	
+
 	addl $2, %ecx
 	js 1b
 	jnz 3f
@ -1538,7 +1536,7 @@ PR_(imlib_mmx_reshade_copy_rgb_to_rgba_cmod):
 	paddusb %mm1, %mm2
 	psubusb %mm3, %mm2
 	movq %mm2, (%edi, %ecx, 4)
-	
+
 	addl $2, %ecx
 	js 1b
 	jnz 3f
@ -1563,8 +1561,6 @@ PR_(imlib_mmx_reshade_copy_rgb_to_rgba_cmod):

 SIZE(imlib_mmx_reshade_copy_rgb_to_rgba_cmod)

-#endif
-
 #ifdef __ELF__
 .section .note.GNU-stack,"",@progbits
 #endif
--- a/src/lib/asm_rgba.S
+++ b/src/lib/asm_rgba.S
@ -1,9 +1,7 @@
 #include <config.h>
 #include "asm.h"

-#ifdef DO_MMX_ASM
-
-/*\ 
+/*\
 |*| MMX assembly rgba rendering routines for Imlib2
 |*| Written by Willem Monsuwe <willem@stack.nl>
 |*|
@ -38,6 +36,7 @@ FN_(imlib_get_cpuid)
 /*\ Common code \*/
 /*\ Save registers, load common parameters \*/
 #define ENTER			\
+	ENDBR_;			\
 	pushl %ebp;		\
 	movl  %esp, %ebp;	\
 	pushl %ebx;		\
@ -81,7 +80,6 @@ FN_(imlib_get_cpuid)
 	ret


-
 PR_(imlib_mmx_bgr565_fast):
 	LOAD_IMMQ(mul_bgr565, %mm7)	/*\ This constant is the only difference \*/
 	CLEANUP_IMMQ_LOADS(1)
@ -273,8 +271,6 @@ PR_(imlib_get_cpuid):

 SIZE(imlib_get_cpuid)

-#endif
-
 #ifdef __ELF__
 .section .note.GNU-stack,"",@progbits
 #endif
--- a/src/lib/asm_rotate.S
+++ b/src/lib/asm_rotate.S
@ -1,9 +1,7 @@
 #include <config.h>
 #include "asm.h"

-#ifdef DO_MMX_ASM
-
-/*\ 
+/*\
 |*| MMX assembly rotation routine for Imlib2
 |*| Written by Willem Monsuwe <willem@stack.nl>
 \*/
@ -197,7 +195,7 @@ PR_(imlib_mmx_RotateAA):
 	paddw %mm3, %mm5
 	packuswb %mm5, %mm5
 	movd %mm5, (%edi, %ecx, 4)
-	
+
 	paddd dxh, %mm6

 	incl %ecx
@ -220,7 +218,7 @@ PR_(imlib_mmx_RotateAA):
 	decl %eax
 	sall $12, %eax
 	movl %eax, sht
-	
+
 	movl sow, %ebx
 	movl src, %edx
 .outside_loop_y:
@ -421,7 +419,7 @@ PR_(imlib_mmx_RotateAA):
 .outside_il_0:
 	movl $0, %eax
 	movl %eax, (%edi, %ecx, 4)
-	
+
 .outside_il_end:
 	paddd dxh, %mm6

@ -447,8 +445,6 @@ PR_(imlib_mmx_RotateAA):

 SIZE(imlib_mmx_RotateAA)

-#endif
-
 #ifdef __ELF__
 .section .note.GNU-stack,"",@progbits
 #endif
--- a/src/lib/asm_scale.S
+++ b/src/lib/asm_scale.S
@ -1,9 +1,7 @@
 #include <config.h>
 #include "asm.h"

-#ifdef DO_MMX_ASM
-
-/*\ 
+/*\
 |*| MMX assembly scaling routine for Imlib2
 |*| Written by Willem Monsuwe <willem@stack.nl>
 \*/
@ -293,7 +291,7 @@ PR_(imlib_Scale_mmx_AARGBA):
 	punpcklbw %mm7, %mm0
 	psllw $6, %mm0
 	pmulhw %mm5, %mm0
-	
+
 	/*\ i = 0x4000 - My \*/
 	movl $0x4000, %ebx
 	subl My, %ebx
@ -307,18 +305,18 @@ PR_(imlib_Scale_mmx_AARGBA):
 	psllw $6, %mm1
 	pmulhw %mm4, %mm1
 	paddw %mm1, %mm0
-	
+
 	/*\ i -= Cy; while (i > Cy) \*/
 	subl Cy, %ebx
 2:
 	cmpl Cy, %ebx
 	jg 1b
-	
+
 	/*\ mm6 = i \*/
 	movd %ebx, %mm6
 	punpcklwd %mm6, %mm6
 	punpckldq %mm6, %mm6
-	
+
 	/*\ p += sow; v += (*p * i) >> 10 \*/
 	addl sow_4, %eax
 	movd (%eax), %mm1
@ -336,7 +334,7 @@ PR_(imlib_Scale_mmx_AARGBA):
 	movd %eax, %mm3
 	punpcklwd %mm3, %mm3
 	punpckldq %mm3, %mm3
-	
+
 	/*\ p + 1 \*/
 	movl %esi, %eax
 	addl $4, %eax
@ -345,7 +343,7 @@ PR_(imlib_Scale_mmx_AARGBA):
 	punpcklbw %mm7, %mm2
 	psllw $6, %mm2
 	pmulhw %mm5, %mm2
-	
+
 	/*\ i = 0x4000 - My \*/
 	movl $0x4000, %ebx
 	subl My, %ebx
@ -359,13 +357,13 @@ PR_(imlib_Scale_mmx_AARGBA):
 	psllw $6, %mm1
 	pmulhw %mm4, %mm1
 	paddw %mm1, %mm2
-	
+
 	/*\ i -= Cy; while (i > Cy) \*/
 	subl Cy, %ebx
 2:
 	cmpl Cy, %ebx
 	jg 1b
-	
+
 	/*\ p += sow; v += (*p * i) >> 10 \*/
 	addl sow_4, %eax
 	movd (%eax), %mm1
@ -425,7 +423,7 @@ PR_(imlib_Scale_mmx_AARGBA):
 	movd %eax, %mm3
 	punpcklwd %mm3, %mm3
 	punpckldq %mm3, %mm3
-	
+
 	/*\ x = -dw \*/
 	movl dw, %ecx
 	negl %ecx
@ -459,7 +457,7 @@ PR_(imlib_Scale_mmx_AARGBA):
 	punpcklbw %mm7, %mm0
 	psllw $6, %mm0
 	pmulhw %mm5, %mm0
-	
+
 	/*\ i = 0x4000 - Mx \*/
 	movl $0x4000, %ebx
 	subl Mx, %ebx
@ -473,18 +471,18 @@ PR_(imlib_Scale_mmx_AARGBA):
 	psllw $6, %mm1
 	pmulhw %mm4, %mm1
 	paddw %mm1, %mm0
-	
+
 	/*\ i -= Cx; while (i > Cx) \*/
 	subl Cx, %ebx
 2:
 	cmpl Cx, %ebx
 	jg 1b
-	
+
 	/*\ mm6 = i \*/
 	movd %ebx, %mm6
 	punpcklwd %mm6, %mm6
 	punpckldq %mm6, %mm6
-	
+
 	/*\ p += sow; v += (*p * i) >> 10 \*/
 	addl $4, %eax
 	movd (%eax), %mm1
@ -504,7 +502,7 @@ PR_(imlib_Scale_mmx_AARGBA):
 	punpcklbw %mm7, %mm2
 	psllw $6, %mm2
 	pmulhw %mm5, %mm2
-	
+
 	/*\ i = 0x4000 - Mx \*/
 	movl $0x4000, %ebx
 	subl Mx, %ebx
@ -518,13 +516,13 @@ PR_(imlib_Scale_mmx_AARGBA):
 	psllw $6, %mm1
 	pmulhw %mm4, %mm1
 	paddw %mm1, %mm2
-	
+
 	/*\ i -= Cx; while (i > Cx) \*/
 	subl Cx, %ebx
 2:
 	cmpl Cx, %ebx
 	jg 1b
-	
+
 	/*\ p += sow; v += (*p * i) >> 10 \*/
 	addl $4, %eax
 	movd (%eax), %mm1
@ -604,14 +602,14 @@ PR_(imlib_Scale_mmx_AARGBA):
 	movd %ebx, %mm5
 	punpcklwd %mm5, %mm5
 	punpckldq %mm5, %mm5
-	
+
 	/*\ p = sptr; v = (*p * Mx) >> 9 \*/
 	movl %esi, %eax
 	movd (%eax), %mm0
 	punpcklbw %mm7, %mm0
 	psllw $7, %mm0
 	pmulhw %mm5, %mm0
-	
+
 	/*\ i = 0x4000 - Mx \*/
 	movl $0x4000, %ebx
 	subl Mx, %ebx
@ -625,18 +623,18 @@ PR_(imlib_Scale_mmx_AARGBA):
 	psllw $7, %mm1
 	pmulhw %mm3, %mm1
 	paddw %mm1, %mm0
-	
+
 	/*\ i -= Cx; while (i > Cx) \*/
 	subl Cx, %ebx
 2:
 	cmpl Cx, %ebx
 	jg 1b
-	
+
 	/*\ mm6 = i \*/
 	movd %ebx, %mm6
 	punpcklwd %mm6, %mm6
 	punpckldq %mm6, %mm6
-	
+
 	/*\ v += (*++p * i) >> 9 \*/
 	addl $4, %eax
 	movd (%eax), %mm1
@ -651,7 +649,7 @@ PR_(imlib_Scale_mmx_AARGBA):
 	punpckldq %mm4, %mm4
 	psllw $2, %mm0
 	pmulhw %mm4, %mm0
-	
+
 	/*\ j = 0x4000 - My \*/
 	movl $0x4000, %edx
 	subl My, %edx
@ -666,7 +664,7 @@ PR_(imlib_Scale_mmx_AARGBA):
 	punpcklbw %mm7, %mm1
 	psllw $7, %mm1
 	pmulhw %mm5, %mm1
-	
+
 	/*\ i = 0x4000 - Mx \*/
 	movl $0x4000, %ebx
 	subl Mx, %ebx
@ -680,13 +678,13 @@ PR_(imlib_Scale_mmx_AARGBA):
 	psllw $7, %mm2
 	pmulhw %mm3, %mm2
 	paddw %mm2, %mm1
-	
+
 	/*\ i -= Cx; while (i > Cx) \*/
 	subl Cx, %ebx
 2:
 	cmpl Cx, %ebx
 	jg 1b
-	
+
 	/*\ vx += (*++p * i) >> 9 \*/
 	addl $4, %eax
 	movd (%eax), %mm2
@ -702,13 +700,13 @@ PR_(imlib_Scale_mmx_AARGBA):
 	psllw $2, %mm1
 	pmulhw %mm4, %mm1
 	paddw %mm1, %mm0
-	
+
 	/*\ j -= Cy; while (j > Cy) \*/
 	subl Cy, %edx
 4:
 	cmpl Cy, %edx
 	jg 3b
-	
+
 	/*\ sptr += sow; p = sptr \*/
 	addl sow_4, %esi
 	movl %esi, %eax
@ -717,7 +715,7 @@ PR_(imlib_Scale_mmx_AARGBA):
 	punpcklbw %mm7, %mm1
 	psllw $7, %mm1
 	pmulhw %mm5, %mm1
-	
+
 	/*\ i = 0x4000 - Mx \*/
 	movl $0x4000, %ebx
 	subl Mx, %ebx
@ -731,13 +729,13 @@ PR_(imlib_Scale_mmx_AARGBA):
 	psllw $7, %mm2
 	pmulhw %mm3, %mm2
 	paddw %mm2, %mm1
-	
+
 	/*\ i -= Cx; while (i > Cx) \*/
 	subl Cx, %ebx
 2:
 	cmpl Cx, %ebx
 	jg 1b
-	
+
 	/*\ vx += (*++p * i) >> 9 \*/
 	addl $4, %eax
 	movd (%eax), %mm2
@ -788,8 +786,6 @@ PR_(imlib_Scale_mmx_AARGBA):

 SIZE(imlib_Scale_mmx_AARGBA)

-#endif
-
 #ifdef __ELF__
 .section .note.GNU-stack,"",@progbits
 #endif