Bug 979894. Add unwind information for pixman arm functions. r=jseward

This isn't technically correct because these functions will never call
functions that throw exceptions, however it lets the profiler unwind them.
The unwinding will also probably be wrong during prologue/epilogue. The right
solution is probably to use cfi.

gfx/cairo/libpixman/src/pixman-arm-neon-asm.S

@@ -954,6 +954,7 @@ generate_composite_function \
  */
 .macro pixman_composite_over_n_8_0565_init
     add         DUMMY, sp, #ARGS_STACK_OFFSET
+    .vsave      {d8-d15}
     vpush       {d8-d15}
     vld1.32     {d11[0]}, [DUMMY]
     vdup.8      d8, d11[0]
@@ -981,6 +982,7 @@ generate_composite_function \
 
 .macro pixman_composite_over_8888_n_0565_init
     add         DUMMY, sp, #(ARGS_STACK_OFFSET + 8)
+    .vsave      {d8-d15}
     vpush       {d8-d15}
     vld1.32     {d24[0]}, [DUMMY]
     vdup.8      d24, d24[3]
@@ -1447,6 +1449,7 @@ generate_composite_function \
 
 .macro pixman_composite_over_n_8_8888_init
     add         DUMMY, sp, #ARGS_STACK_OFFSET
+    .vsave      {d8-d15}
     vpush       {d8-d15}
     vld1.32     {d11[0]}, [DUMMY]
     vdup.8      d8, d11[0]
@@ -1518,6 +1521,7 @@ generate_composite_function \
 
 .macro pixman_composite_over_n_8_8_init
     add         DUMMY, sp, #ARGS_STACK_OFFSET
+    .vsave      {d8-d15}
     vpush       {d8-d15}
     vld1.32     {d8[0]}, [DUMMY]
     vdup.8      d8, d8[3]
@@ -1619,6 +1623,7 @@ generate_composite_function \
 
 .macro pixman_composite_over_n_8888_8888_ca_init
     add         DUMMY, sp, #ARGS_STACK_OFFSET
+    .vsave      {d8-d15}
     vpush       {d8-d15}
     vld1.32     {d11[0]}, [DUMMY]
     vdup.8      d8, d11[0]
@@ -1788,6 +1793,7 @@ generate_composite_function \
 
 .macro pixman_composite_over_n_8888_0565_ca_init
     add         DUMMY, sp, #ARGS_STACK_OFFSET
+    .vsave      {d8-d15}
     vpush       {d8-d15}
     vld1.32     {d11[0]}, [DUMMY]
     vdup.8      d8, d11[0]
@@ -1901,6 +1907,7 @@ generate_composite_function \
 
 .macro pixman_composite_add_n_8_8_init
     add         DUMMY, sp, #ARGS_STACK_OFFSET
+    .vsave      {d8-d15}
     vpush       {d8-d15}
     vld1.32     {d11[0]}, [DUMMY]
     vdup.8      d11, d11[3]
@@ -2207,6 +2214,7 @@ generate_composite_function_single_scanline \
 
 .macro pixman_composite_over_8888_n_8888_init
     add         DUMMY, sp, #48
+    .vsave      {d8-d15}
     vpush       {d8-d15}
     vld1.32     {d15[0]}, [DUMMY]
     vdup.8      d15, d15[3]
@@ -2579,6 +2587,7 @@ generate_composite_function \
 
 .macro pixman_composite_over_0565_n_0565_init
     add         DUMMY, sp, #(ARGS_STACK_OFFSET + 8)
+    .vsave      {d8-d15}
     vpush       {d8-d15}
     vld1.32     {d15[0]}, [DUMMY]
     vdup.8      d15, d15[3]
@@ -3132,13 +3141,16 @@ pixman_asm_function fname
     TMP4      .req      r9
     STRIDE    .req      r2
 
+    .fnstart
     mov       ip, sp
+    .save     {r4, r5, r6, r7, r8, r9}
     push      {r4, r5, r6, r7, r8, r9}
     mov       PF_OFFS, #prefetch_distance
     ldmia     ip, {WB, X, UX, WIDTH}
     mul       PF_OFFS, PF_OFFS, UX
 
 .if ((flags) & BILINEAR_FLAG_USE_ALL_NEON_REGS) != 0
+    .vsave    {d8-d15}
     vpush     {d8-d15}
 .endif
 
@@ -3232,6 +3244,7 @@ pixman_asm_function fname
 .endif
     pop       {r4, r5, r6, r7, r8, r9}
     bx        lr
+    .fnend
 
     .unreq    OUT
     .unreq    TOP

gfx/cairo/libpixman/src/pixman-arm-neon-asm.h

@@ -639,6 +639,8 @@ local skip1
     .type fname, %function
 #endif
 fname:
+    .fnstart
+    .save       {r4-r12, lr}
     push        {r4-r12, lr}        /* save all registers */
 
 /*
@@ -816,6 +818,7 @@ fname:
 
     init
 .if regs_shortage
+    .save       {r0, r1}
     push        {r0, r1}
 .endif
     subs        H, H, #1
@@ -901,6 +904,7 @@ fname:
 .endif
     cleanup
     pop         {r4-r12, pc}  /* exit */
+    .fnend
 
     .purgem     fetch_src_pixblock
     .purgem     pixld_src
@@ -953,6 +957,7 @@ fname:
     .type fname, %function
 #endif
 fname:
+    .fnstart
     .set PREFETCH_TYPE_CURRENT, PREFETCH_TYPE_NONE
 /*
  * Make some macro arguments globally visible and accessible
@@ -987,6 +992,7 @@ fname:
     .endm
 
     ldr         UNIT_X, [sp]
+    .save       {r4-r8, lr}
     push        {r4-r8, lr}
     ldr         SRC_WIDTH_FIXED, [sp, #(24 + 4)]
     .if mask_bpp != 0
@@ -1102,6 +1108,7 @@ fname:
     .purgem     fetch_src_pixblock
     .purgem     pixld_src
 
+    .fnend
     .endfunc
 .endm
 
@@ -1128,6 +1135,7 @@ fname:
  */
 
 .macro default_init_need_all_regs
+    .vsave      {d8-d15}
     vpush       {d8-d15}
 .endm
 

gfx/cairo/libpixman/src/pixman-arm-simd-asm.h

@@ -680,12 +680,15 @@
     ORIG_W      .req    r14 /* width (pixels) */
 
 fname:
+        .fnstart
+	.save   {r4-r11, lr}
         push    {r4-r11, lr}        /* save all registers */
 
         subs    Y, Y, #1
         blo     199f
 
 #ifdef DEBUG_PARAMS
+	.pad    #9*4
         sub     sp, sp, #9*4
 #endif
 
@@ -857,6 +860,7 @@ fname:
 #endif
 199:
         pop     {r4-r11, pc}  /* exit */
+	.fnend
 
  .ltorg