From b101c115102b83bb1fc4e28de6136dd4940796bc Mon Sep 17 00:00:00 2001
From: Siarhei Siamashka <siarhei.siamashka@nokia.com>
Date: Fri, 30 Oct 2009 17:02:14 +0000
Subject: ARM: initial 24bpp support
---
diff --git a/pixman/pixman-arm-neon-asm.S b/pixman/pixman-arm-neon-asm.S
index 35e6a7e..7f91ced 100644
--- a/pixman/pixman-arm-neon-asm.S
+++ b/pixman/pixman-arm-neon-asm.S
@@ -977,3 +977,32 @@ generate_composite_function \
pixman_composite_over_8888_n_8888_process_pixblock_head, \
pixman_composite_over_8888_n_8888_process_pixblock_tail, \
pixman_composite_over_8888_n_8888_process_pixblock_tail_head
+
+/******************************************************************************/
+
+.macro pixman_composite_src_0888_0888_process_pixblock_head
+.endm
+
+.macro pixman_composite_src_0888_0888_process_pixblock_tail
+.endm
+
+.macro pixman_composite_src_0888_0888_process_pixblock_tail_head
+ vst3.8 {d0, d1, d2}, [DST_W]!
+ vld3.8 {d0, d1, d2}, [SRC]!
+ cache_preload 8, 8
+.endm
+
+generate_composite_function \
+ pixman_composite_src_0888_0888_asm_neon, 24, 0, 24, \
+ FLAG_DST_WRITEONLY, \
+ 8, /* number of pixels, processed in a single block */ \
+ 10, /* prefetch distance */ \
+ default_init, \
+ default_cleanup, \
+ pixman_composite_src_0888_0888_process_pixblock_head, \
+ pixman_composite_src_0888_0888_process_pixblock_tail, \
+ pixman_composite_src_0888_0888_process_pixblock_tail_head, \
+ 0, /* dst_w_basereg */ \
+ 0, /* dst_r_basereg */ \
+ 0, /* src_basereg */ \
+ 0 /* mask_basereg */
diff --git a/pixman/pixman-arm-neon-asm.h b/pixman/pixman-arm-neon-asm.h
index a2941ae..1653ef4 100644
--- a/pixman/pixman-arm-neon-asm.h
+++ b/pixman/pixman-arm-neon-asm.h
@@ -95,6 +95,14 @@
op&.&elem_size {d®1[idx]}, [&mem_operand&]!
.endm
+.macro pixldst3 op, elem_size, reg1, reg2, reg3, mem_operand
+ op&.&elem_size {d®1, d®2, d®3}, [&mem_operand&]!
+.endm
+
+.macro pixldst30 op, elem_size, reg1, reg2, reg3, idx, mem_operand
+ op&.&elem_size {d®1[idx], d®2[idx], d®3[idx]}, [&mem_operand&]!
+.endm
+
.macro pixldst numbytes, op, elem_size, basereg, mem_operand, abits
.if numbytes == 32
pixldst4 op, elem_size, %(basereg+4), %(basereg+5), \
@@ -134,6 +142,18 @@
.if (bpp == 32) && (numpix == 8) && (DEINTERLEAVE_32BPP_ENABLED != 0)
pixldst4 vld4, 8, %(basereg+4), %(basereg+5), \
%(basereg+6), %(basereg+7), mem_operand, abits
+.elseif (bpp == 24) && (numpix == 8)
+ pixldst3 vld3, 8, %(basereg+3), %(basereg+4), %(basereg+5), mem_operand
+.elseif (bpp == 24) && (numpix == 4)
+ pixldst30 vld3, 8, %(basereg+0), %(basereg+1), %(basereg+2), 4, mem_operand
+ pixldst30 vld3, 8, %(basereg+0), %(basereg+1), %(basereg+2), 5, mem_operand
+ pixldst30 vld3, 8, %(basereg+0), %(basereg+1), %(basereg+2), 6, mem_operand
+ pixldst30 vld3, 8, %(basereg+0), %(basereg+1), %(basereg+2), 7, mem_operand
+.elseif (bpp == 24) && (numpix == 2)
+ pixldst30 vld3, 8, %(basereg+0), %(basereg+1), %(basereg+2), 2, mem_operand
+ pixldst30 vld3, 8, %(basereg+0), %(basereg+1), %(basereg+2), 3, mem_operand
+.elseif (bpp == 24) && (numpix == 1)
+ pixldst30 vld3, 8, %(basereg+0), %(basereg+1), %(basereg+2), 1, mem_operand
.else
pixldst %(numpix * bpp / 8), vld1, %(bpp), basereg, mem_operand, abits
.endif
@@ -145,6 +165,18 @@
.if (bpp == 32) && (numpix == 8) && (DEINTERLEAVE_32BPP_ENABLED != 0)
pixldst4 vst4, 8, %(basereg+4), %(basereg+5), \
%(basereg+6), %(basereg+7), mem_operand, abits
+.elseif (bpp == 24) && (numpix == 8)
+ pixldst3 vst3, 8, %(basereg+3), %(basereg+4), %(basereg+5), mem_operand
+.elseif (bpp == 24) && (numpix == 4)
+ pixldst30 vst3, 8, %(basereg+0), %(basereg+1), %(basereg+2), 4, mem_operand
+ pixldst30 vst3, 8, %(basereg+0), %(basereg+1), %(basereg+2), 5, mem_operand
+ pixldst30 vst3, 8, %(basereg+0), %(basereg+1), %(basereg+2), 6, mem_operand
+ pixldst30 vst3, 8, %(basereg+0), %(basereg+1), %(basereg+2), 7, mem_operand
+.elseif (bpp == 24) && (numpix == 2)
+ pixldst30 vst3, 8, %(basereg+0), %(basereg+1), %(basereg+2), 2, mem_operand
+ pixldst30 vst3, 8, %(basereg+0), %(basereg+1), %(basereg+2), 3, mem_operand
+.elseif (bpp == 24) && (numpix == 1)
+ pixldst30 vst3, 8, %(basereg+0), %(basereg+1), %(basereg+2), 1, mem_operand
.else
pixldst %(numpix * bpp / 8), vst1, %(bpp), basereg, mem_operand, abits
.endif
@@ -334,6 +366,8 @@ fname:
.if src_bpp == 32
.set src_bpp_shift, 2
+.elseif src_bpp == 24
+ .set src_bpp_shift, 0
.elseif src_bpp == 16
.set src_bpp_shift, 1
.elseif src_bpp == 8
@@ -345,6 +379,8 @@ fname:
.endif
.if mask_bpp == 32
.set mask_bpp_shift, 2
+.elseif mask_bpp == 24
+ .set mask_bpp_shift, 0
.elseif mask_bpp == 8
.set mask_bpp_shift, 0
.elseif mask_bpp == 0
@@ -354,6 +390,8 @@ fname:
.endif
.if dst_w_bpp == 32
.set dst_bpp_shift, 2
+.elseif dst_w_bpp == 24
+ .set dst_bpp_shift, 0
.elseif dst_w_bpp == 16
.set dst_bpp_shift, 1
.elseif dst_w_bpp == 8
@@ -398,6 +436,19 @@ fname:
PF mov PF_CTL, H, lsl #4
PF add PF_CTL, #(prefetch_distance - 0x10)
+.if src_bpp == 24
+ sub SRC_STRIDE, SRC_STRIDE, W
+ sub SRC_STRIDE, SRC_STRIDE, W, lsl #1
+.endif
+.if mask_bpp == 24
+ sub MASK_STRIDE, MASK_STRIDE, W
+ sub MASK_STRIDE, MASK_STRIDE, W, lsl #1
+.endif
+.if dst_w_bpp == 24
+ sub DST_STRIDE, DST_STRIDE, W
+ sub DST_STRIDE, DST_STRIDE, W, lsl #1
+.endif
+
init
.if regs_shortage
push {r0, r1}
@@ -412,7 +463,8 @@ fname:
cmp W, #(pixblock_size * 2)
blt 8f
0:
- /* ensure 16 byte alignment of the destination buffer */
+ /* ensure 16 byte alignment of the destination buffer, except for 24bpp */
+.if dst_w_bpp != 24
tst DST_R, #0xF
beq 2f
@@ -454,6 +506,7 @@ fname:
.endif
.endr
2:
+.endif
pixld_a pixblock_size, dst_r_bpp, \
(dst_r_basereg - pixblock_size * dst_r_bpp / 64), DST_R
@@ -520,11 +573,13 @@ fname:
.if mask_bpp != 0
add MASK, MASK, MASK_STRIDE, lsl #mask_bpp_shift
.endif
+.if (dst_w_bpp != 24)
sub DST_W, DST_W, W, lsl #dst_bpp_shift
-.if src_bpp != 0
+.endif
+.if (src_bpp != 24) && (src_bpp != 0)
sub SRC, SRC, W, lsl #src_bpp_shift
.endif
-.if mask_bpp != 0
+.if (mask_bpp != 24) && (mask_bpp != 0)
sub MASK, MASK, W, lsl #mask_bpp_shift
.endif
subs H, H, #1
@@ -539,7 +594,7 @@ fname:
cleanup
pop {r4-r12, pc} /* exit */
-8: /* handle small rectangle, width up to 15 pixels */
+8: /* handle small rectangle, width up to (pixblock_size * 2 - 1) pixels */
tst W, #pixblock_size
beq 1f
pixld pixblock_size, dst_r_bpp, \
@@ -592,11 +647,13 @@ fname:
.if mask_bpp != 0
add MASK, MASK, MASK_STRIDE, lsl #mask_bpp_shift
.endif
+.if (dst_w_bpp != 24)
sub DST_W, DST_W, W, lsl #dst_bpp_shift
-.if src_bpp != 0
+.endif
+.if (src_bpp != 24) && (src_bpp != 0)
sub SRC, SRC, W, lsl #src_bpp_shift
.endif
-.if mask_bpp != 0
+.if (mask_bpp != 24) && (mask_bpp != 0)
sub MASK, MASK, W, lsl #mask_bpp_shift
.endif
subs H, H, #1
diff --git a/pixman/pixman-arm-neon.c b/pixman/pixman-arm-neon.c
index 2811099..f3f38a9 100644
--- a/pixman/pixman-arm-neon.c
+++ b/pixman/pixman-arm-neon.c
@@ -2065,6 +2065,43 @@ neon_composite_src_8888_8888 (pixman_implementation_t *imp,
}
void
+pixman_composite_src_0888_0888_asm_neon (int32_t w,
+ int32_t h,
+ uint8_t *dst,
+ int32_t dst_stride,
+ uint8_t *src,
+ int32_t src_stride);
+
+static void
+neon_composite_src_0888_0888 (pixman_implementation_t *imp,
+ pixman_op_t op,
+ pixman_image_t * src_image,
+ pixman_image_t * mask_image,
+ pixman_image_t * dst_image,
+ int32_t src_x,
+ int32_t src_y,
+ int32_t mask_x,
+ int32_t mask_y,
+ int32_t dest_x,
+ int32_t dest_y,
+ int32_t width,
+ int32_t height)
+{
+ uint8_t *dst_line;
+ uint8_t *src_line;
+ int32_t dst_stride, src_stride;
+
+ PIXMAN_IMAGE_GET_LINE (src_image, src_x, src_y, uint8_t,
+ src_stride, src_line, 3);
+ PIXMAN_IMAGE_GET_LINE (dst_image, dest_x, dest_y, uint8_t,
+ dst_stride, dst_line, 3);
+
+ pixman_composite_src_0888_0888_asm_neon (width, height,
+ dst_line, dst_stride,
+ src_line, src_stride);
+}
+
+void
pixman_composite_over_8888_8888_asm_neon (int32_t w,
int32_t h,
uint32_t *dst,
@@ -2449,6 +2486,7 @@ static const pixman_fast_path_t arm_neon_fast_path_array[] =
{ PIXMAN_OP_SRC, PIXMAN_a8b8g8r8, PIXMAN_null, PIXMAN_x8b8g8r8, neon_composite_src_8888_8888, 0 },
{ PIXMAN_OP_SRC, PIXMAN_x8b8g8r8, PIXMAN_null, PIXMAN_x8b8g8r8, neon_composite_src_8888_8888, 0 },
{ PIXMAN_OP_SRC, PIXMAN_b5g6r5, PIXMAN_null, PIXMAN_b5g6r5, neon_composite_src_0565_0565, 0 },
+ { PIXMAN_OP_SRC, PIXMAN_r8g8b8, PIXMAN_null, PIXMAN_r8g8b8, neon_composite_src_0888_0888, 0 },
{ PIXMAN_OP_OVER, PIXMAN_a8r8g8b8, PIXMAN_null, PIXMAN_r5g6b5, neon_composite_over_8888_0565, 0 },
{ PIXMAN_OP_OVER, PIXMAN_a8b8g8r8, PIXMAN_null, PIXMAN_b5g6r5, neon_composite_over_8888_0565, 0 },
{ PIXMAN_OP_OVER, PIXMAN_a8r8g8b8, PIXMAN_null, PIXMAN_a8r8g8b8, neon_composite_over_8888_8888, 0 },
--
cgit v0.8.2