MIPS: DSPr2: Added bilinear over_8888_8_8888 fast path.

Performance numbers before/after on MIPS-74kc @ 1GHz

Referent (before):

cairo-perf-trace:
[ # ]  backend                         test   min(s) median(s) stddev. count
[ # ]    image: pixman 0.25.3
[  0]    image             firefox-fishtank 2289.180 2290.567   0.05%    5/6

Optimized:

cairo-perf-trace:
[ # ]  backend                         test   min(s) median(s) stddev. count
[ # ]    image: pixman 0.25.3
[  0]    image             firefox-fishtank 1700.925 1708.314   0.22%    5/6

4 files changed, 177 insertions, 0 deletions

diff --git a/pixman/pixman-mips-dspr2-asm.S b/pixman/pixman-mips-dspr2-asm.S
index f3908da9..87558f03 100644
--- a/pixman/pixman-mips-dspr2-asm.S
+++ b/pixman/pixman-mips-dspr2-asm.S
@@ -747,3 +747,66 @@ LEAF_MIPS_DSPR2(pixman_composite_over_n_8_0565_asm_mips)
      nop
 
 END(pixman_composite_over_n_8_0565_asm_mips)
+
+LEAF_MIPS_DSPR2(pixman_scaled_bilinear_scanline_8888_8_8888_OVER_asm_mips)
+/*
+ * a0     - dst        (a8r8g8b8)
+ * a1     - mask       (a8)
+ * a2     - src_top    (a8r8g8b8)
+ * a3     - src_bottom (a8r8g8b8)
+ * 16(sp) - wt
+ * 20(sp) - wb
+ * 24(sp) - vx
+ * 28(sp) - unit_x
+ * 32(sp) - w
+ */
+
+    SAVE_REGS_ON_STACK 28, v0, v1, s0, s1, s2, s3, s4, s5, s6, s7, s8
+
+    lw       v1, 60(sp)        /* v1 = w(sp + 32 + 28 save regs stack offset)*/
+    beqz     v1, 1f
+     nop
+
+    lw       s0, 44(sp)        /* s0 = wt */
+    lw       s1, 48(sp)        /* s1 = wb */
+    lw       s2, 52(sp)        /* s2 = vx */
+    lw       s3, 56(sp)        /* s3 = unit_x */
+    li       v0, 256
+    li       s8, 0x00ff00ff
+0:
+    andi     t4, s2, 0xffff    /* t4 = (short)vx */
+    srl      t4, t4, 8         /* t4 = vx >> 8 */
+    subu     t5, v0, t4        /* t5 = ( 256 - (vx>>8)) */
+
+    mul      s4, s0, t5        /* s4 = wt*(256-(vx>>8)) */
+    mul      s5, s0, t4        /* s5 = wt*(vx>>8) */
+    mul      s6, s1, t5        /* s6 = wb*(256-(vx>>8)) */
+    mul      s7, s1, t4        /* s7 = wb*(vx>>8) */
+
+    sra      t9, s2, 16
+    sll      t9, t9, 2
+    addiu    t8, t9, 4
+    lwx      t0, t9(a2)        /* t0 = tl */
+    lwx      t1, t8(a2)        /* t1 = tr */
+    addiu    v1, v1, -1
+    lwx      t2, t9(a3)        /* t2 = bl */
+    lwx      t3, t8(a3)        /* t3 = br */
+
+    BILINEAR_INTERPOLATE_SINGLE_PIXEL t0, t1, t2, t3, \
+                                      t4, t5, t6, t7, t8, t9, s4, s5, s6, s7
+    lbu      t1, 0(a1)         /* t1 = mask */
+    lw       t2, 0(a0)         /* t2 = dst */
+    addiu    a1, a1, 1
+    OVER_8888_8_8888 t0, t1, t2, t0, s8, t3, t4, t5, t6
+
+    addu     s2, s2, s3        /* vx += unit_x; */
+    sw       t0, 0(a0)
+    bnez     v1, 0b
+     addiu   a0, a0, 4
+
+1:
+    RESTORE_REGS_FROM_STACK 28, v0, v1, s0, s1, s2, s3, s4, s5, s6, s7, s8
+    j        ra
+     nop
+
+END(pixman_scaled_bilinear_scanline_8888_8_8888_OVER_asm_mips)
diff --git a/pixman/pixman-mips-dspr2-asm.h b/pixman/pixman-mips-dspr2-asm.h
index 83830604..7cf3281e 100644
--- a/pixman/pixman-mips-dspr2-asm.h
+++ b/pixman/pixman-mips-dspr2-asm.h
@@ -566,4 +566,60 @@ LEAF_MIPS32R2(symbol)                                   \
     addu_s.qb              \out2_8888, \d2_8888,  \scratch2
 .endm
 
+.macro BILINEAR_INTERPOLATE_SINGLE_PIXEL tl, tr, bl, br,         \
+                                         scratch1, scratch2,     \
+                                         alpha, red, green, blue \
+                                         wt1, wt2, wb1, wb2
+    andi            \scratch1, \tl,  0xff
+    andi            \scratch2, \tr,  0xff
+    andi            \alpha,    \bl,  0xff
+    andi            \red,      \br,  0xff
+
+    multu           $ac0,      \wt1, \scratch1
+    maddu           $ac0,      \wt2, \scratch2
+    maddu           $ac0,      \wb1, \alpha
+    maddu           $ac0,      \wb2, \red
+
+    ext             \scratch1, \tl,  8, 8
+    ext             \scratch2, \tr,  8, 8
+    ext             \alpha,    \bl,  8, 8
+    ext             \red,      \br,  8, 8
+
+    multu           $ac1,      \wt1, \scratch1
+    maddu           $ac1,      \wt2, \scratch2
+    maddu           $ac1,      \wb1, \alpha
+    maddu           $ac1,      \wb2, \red
+
+    ext             \scratch1, \tl,  16, 8
+    ext             \scratch2, \tr,  16, 8
+    ext             \alpha,    \bl,  16, 8
+    ext             \red,      \br,  16, 8
+
+    mflo            \blue,     $ac0
+
+    multu           $ac2,      \wt1, \scratch1
+    maddu           $ac2,      \wt2, \scratch2
+    maddu           $ac2,      \wb1, \alpha
+    maddu           $ac2,      \wb2, \red
+
+    ext             \scratch1, \tl,  24, 8
+    ext             \scratch2, \tr,  24, 8
+    ext             \alpha,    \bl,  24, 8
+    ext             \red,      \br,  24, 8
+
+    mflo            \green,    $ac1
+
+    multu           $ac3,      \wt1, \scratch1
+    maddu           $ac3,      \wt2, \scratch2
+    maddu           $ac3,      \wb1, \alpha
+    maddu           $ac3,      \wb2, \red
+
+    mflo            \red,      $ac2
+    mflo            \alpha,    $ac3
+
+    precr.qb.ph     \alpha,    \alpha, \red
+    precr.qb.ph     \scratch1, \green, \blue
+    precrq.qb.ph    \tl,       \alpha, \scratch1
+.endm
+
 #endif //PIXMAN_MIPS_DSPR2_ASM_H
diff --git a/pixman/pixman-mips-dspr2.c b/pixman/pixman-mips-dspr2.c
index 7081734c..06d4335d 100644
--- a/pixman/pixman-mips-dspr2.c
+++ b/pixman/pixman-mips-dspr2.c
@@ -58,6 +58,9 @@ PIXMAN_MIPS_BIND_FAST_PATH_N_MASK_DST (SKIP_ZERO_SRC, over_n_8_8888,
 PIXMAN_MIPS_BIND_FAST_PATH_N_MASK_DST (SKIP_ZERO_SRC, over_n_8_0565,
                                        uint8_t, 1, uint16_t, 1)
 
+PIXMAN_MIPS_BIND_SCALED_BILINEAR_SRC_A8_DST (SKIP_ZERO_SRC, 8888_8_8888, OVER,
+                                             uint32_t, uint32_t)
+
 static pixman_bool_t
 pixman_fill_mips (uint32_t *bits,
                   int       stride,
@@ -206,6 +209,9 @@ static const pixman_fast_path_t mips_dspr2_fast_paths[] =
     PIXMAN_STD_FAST_PATH (OVER, solid,    a8,       r5g6b5,   mips_composite_over_n_8_0565),
     PIXMAN_STD_FAST_PATH (OVER, solid,    a8,       b5g6r5,   mips_composite_over_n_8_0565),
 
+    SIMPLE_BILINEAR_A8_MASK_FAST_PATH (OVER, a8r8g8b8, a8r8g8b8, mips_8888_8_8888),
+    SIMPLE_BILINEAR_A8_MASK_FAST_PATH (OVER, a8r8g8b8, x8r8g8b8, mips_8888_8_8888),
+
     { PIXMAN_OP_NONE },
 };
 
diff --git a/pixman/pixman-mips-dspr2.h b/pixman/pixman-mips-dspr2.h
index cc35d02e..5036938c 100644
--- a/pixman/pixman-mips-dspr2.h
+++ b/pixman/pixman-mips-dspr2.h
@@ -127,4 +127,56 @@ mips_composite_##name (pixman_implementation_t *imp,                \
     }                                                               \
 }
 
+/*****************************************************************************/
+
+#define PIXMAN_MIPS_BIND_SCALED_BILINEAR_SRC_A8_DST(flags, name, op,          \
+                                                src_type, dst_type)           \
+void                                                                          \
+pixman_scaled_bilinear_scanline_##name##_##op##_asm_mips (                    \
+                                             dst_type *       dst,            \
+                                             const uint8_t *  mask,           \
+                                             const src_type * top,            \
+                                             const src_type * bottom,         \
+                                             int              wt,             \
+                                             int              wb,             \
+                                             pixman_fixed_t   x,              \
+                                             pixman_fixed_t   ux,             \
+                                             int              width);         \
+                                                                              \
+static force_inline void                                                      \
+scaled_bilinear_scanline_mips_##name##_##op (dst_type *       dst,            \
+                                             const uint8_t *  mask,           \
+                                             const src_type * src_top,        \
+                                             const src_type * src_bottom,     \
+                                             int32_t          w,              \
+                                             int              wt,             \
+                                             int              wb,             \
+                                             pixman_fixed_t   vx,             \
+                                             pixman_fixed_t   unit_x,         \
+                                             pixman_fixed_t   max_vx,         \
+                                             pixman_bool_t    zero_src)       \
+{                                                                             \
+    if ((flags & SKIP_ZERO_SRC) && zero_src)                                  \
+        return;                                                               \
+    pixman_scaled_bilinear_scanline_##name##_##op##_asm_mips (                \
+                      dst, mask, src_top, src_bottom, wt, wb, vx, unit_x, w); \
+}                                                                             \
+                                                                              \
+FAST_BILINEAR_MAINLOOP_COMMON (mips_##name##_cover_##op,                      \
+                       scaled_bilinear_scanline_mips_##name##_##op,           \
+                       src_type, uint8_t, dst_type, COVER,                    \
+                       FLAG_HAVE_NON_SOLID_MASK)                              \
+FAST_BILINEAR_MAINLOOP_COMMON (mips_##name##_none_##op,                       \
+                       scaled_bilinear_scanline_mips_##name##_##op,           \
+                       src_type, uint8_t, dst_type, NONE,                     \
+                       FLAG_HAVE_NON_SOLID_MASK)                              \
+FAST_BILINEAR_MAINLOOP_COMMON (mips_##name##_pad_##op,                        \
+                       scaled_bilinear_scanline_mips_##name##_##op,           \
+                       src_type, uint8_t, dst_type, PAD,                      \
+                       FLAG_HAVE_NON_SOLID_MASK)                              \
+FAST_BILINEAR_MAINLOOP_COMMON (mips_##name##_normal_##op,                     \
+                       scaled_bilinear_scanline_mips_##name##_##op,           \
+                       src_type, uint8_t, dst_type, NORMAL,                   \
+                       FLAG_HAVE_NON_SOLID_MASK)
+
 #endif //PIXMAN_MIPS_DSPR2_H