MIPS: DSPr2: Added more fast-paths for REVERSE operation:

 - out_reverse_8_0565
 - out_reverse_8_8888

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

lowlevel-blt-bench results

Referent (before):
        out_reverse_8_0565 =  L1:  14.29  L2:  13.58  M: 12.14 ( 24.16%)  HT:  9.23  VT:  9.12  R:  8.84  RT:  4.75 (  36Kops/s)
        out_reverse_8_8888 =  L1:  27.46  L2:  23.24  M: 17.41 ( 57.73%)  HT: 12.61  VT: 12.47  R: 11.79  RT:  5.86 (  41Kops/s)

Optimized:
        out_reverse_8_0565 =  L1:  28.24  L2:  25.64  M: 20.63 ( 41.05%)  HT: 16.69  VT: 16.14  R: 15.50  RT:  8.69 (  52Kops/s)
        out_reverse_8_8888 =  L1:  52.78  L2:  41.44  M: 23.50 ( 77.94%)  HT: 18.79  VT: 18.16  R: 16.90  RT:  9.11 (  53Kops/s)

diff --git a/pixman/pixman-mips-dspr2-asm.S b/pixman/pixman-mips-dspr2-asm.S
index ba22e62..64ef660 100644 (file)
--- a/pixman/pixman-mips-dspr2-asm.S
+++ b/pixman/pixman-mips-dspr2-asm.S
@@ -2100,6 +2100,116 @@ LEAF_MIPS_DSPR2(pixman_composite_add_8888_8888_asm_mips)
 
 END(pixman_composite_add_8888_8888_asm_mips)
 
+LEAF_MIPS_DSPR2(pixman_composite_out_reverse_8_0565_asm_mips)
+/*
+ * a0 - dst  (r5g6b5)
+ * a1 - src  (a8)
+ * a2 - w
+ */
+
+    beqz     a2, 4f
+     nop
+
+    SAVE_REGS_ON_STACK 0, s0, s1, s2, s3
+    li       t2, 0xf800f800
+    li       t3, 0x07e007e0
+    li       t4, 0x001F001F
+    li       t5, 0x00ff00ff
+
+    addiu    t1, a2, -1
+    beqz     t1, 2f
+     nop
+1:
+    lbu      t0, 0(a1) /* t0 = source      (a8) */
+    lbu      t1, 1(a1) /* t1 = source      (a8) */
+    lhu      t6, 0(a0) /* t6 = destination (r5g6b5) */
+    lhu      t7, 2(a0) /* t7 = destination (r5g6b5) */
+    addiu    a1, a1, 2
+
+    not      t0, t0
+    not      t1, t1
+    andi     t0, 0xff  /* t0 = neg source1 */
+    andi     t1, 0xff  /* t1 = neg source2 */
+    CONVERT_2x0565_TO_2x8888 t6, t7, t8, t9, t3, t4, s0, s1, s2, s3
+    MIPS_2xUN8x4_MUL_2xUN8   t8, t9, t0, t1, t6, t7, t5, s0, s1, s2, s3, t8, t9
+    CONVERT_2x8888_TO_2x0565 t6, t7, t8, t9, t2, t3, t4, s0, s1
+
+    sh       t8, 0(a0)
+    sh       t9, 2(a0)
+    addiu    a2, a2, -2
+    addiu    t1, a2, -1
+    bgtz     t1, 1b
+     addiu   a0, a0, 4
+2:
+    beqz     a2, 3f
+     nop
+    lbu      t0, 0(a1) /* t0 = source      (a8) */
+    lhu      t1, 0(a0) /* t1 = destination (r5g6b5) */
+
+    not      t0, t0
+    andi     t0, 0xff  /* t0 = neg source */
+    CONVERT_1x0565_TO_1x8888 t1, t2, t3, t4
+    MIPS_UN8x4_MUL_UN8        t2, t0, t1, t5, t3, t4, t6
+    CONVERT_1x8888_TO_1x0565 t1, t2, t3, t4
+
+    sh       t2, 0(a0)
+3:
+    RESTORE_REGS_FROM_STACK 0, s0, s1, s2, s3
+4:
+    j        ra
+     nop
+
+END(pixman_composite_out_reverse_8_0565_asm_mips)
+
+LEAF_MIPS_DSPR2(pixman_composite_out_reverse_8_8888_asm_mips)
+/*
+ * a0 - dst  (a8r8g8b8)
+ * a1 - src  (a8)
+ * a2 - w
+ */
+
+    beqz     a2, 3f
+     nop
+    li       t4, 0x00ff00ff
+    addiu    t1, a2, -1
+    beqz     t1, 2f
+     nop
+1:
+    lbu      t0, 0(a1) /* t0 = source      (a8) */
+    lbu      t1, 1(a1) /* t1 = source      (a8) */
+    lw       t2, 0(a0) /* t2 = destination (a8r8g8b8) */
+    lw       t3, 4(a0) /* t3 = destination (a8r8g8b8) */
+    addiu    a1, a1, 2
+    not      t0, t0
+    not      t1, t1
+    andi     t0, 0xff  /* t0 = neg source */
+    andi     t1, 0xff  /* t1 = neg source */
+
+    MIPS_2xUN8x4_MUL_2xUN8 t2, t3, t0, t1, t5, t6, t4, t7, t8, t9, t2, t3, t0
+
+    sw       t5, 0(a0)
+    sw       t6, 4(a0)
+    addiu    a2, a2, -2
+    addiu    t1, a2, -1
+    bgtz     t1, 1b
+     addiu   a0, a0, 8
+2:
+    beqz     a2, 3f
+     nop
+    lbu      t0, 0(a1) /* t0 = source      (a8) */
+    lw       t1, 0(a0) /* t1 = destination (a8r8g8b8) */
+    not      t0, t0
+    andi     t0, 0xff  /* t0 = neg source */
+
+    MIPS_UN8x4_MUL_UN8 t1, t0, t2, t4, t3, t5, t6
+
+    sw       t2, 0(a0)
+3:
+    j        ra
+     nop
+
+END(pixman_composite_out_reverse_8_8888_asm_mips)
+
 LEAF_MIPS_DSPR2(pixman_scaled_nearest_scanline_8888_8_0565_OVER_asm_mips)
 /*
  * a0     - dst  (r5g6b5)

diff --git a/pixman/pixman-mips-dspr2.c b/pixman/pixman-mips-dspr2.c
index 11f1254..a7e6f8a 100644 (file)
--- a/pixman/pixman-mips-dspr2.c
+++ b/pixman/pixman-mips-dspr2.c
@@ -54,6 +54,10 @@ PIXMAN_MIPS_BIND_FAST_PATH_SRC_DST (0, add_8_8,
                                     uint8_t, 1, uint8_t, 1)
 PIXMAN_MIPS_BIND_FAST_PATH_SRC_DST (0, add_8888_8888,
                                     uint32_t, 1, uint32_t, 1)
+PIXMAN_MIPS_BIND_FAST_PATH_SRC_DST (0, out_reverse_8_0565,
+                                    uint8_t, 1, uint16_t, 1)
+PIXMAN_MIPS_BIND_FAST_PATH_SRC_DST (0, out_reverse_8_8888,
+                                    uint8_t,  1, uint32_t, 1)
 
 PIXMAN_MIPS_BIND_FAST_PATH_N_MASK_DST (0, src_n_8_8888,
                                        uint8_t, 1, uint32_t, 1)
@@ -324,6 +328,10 @@ static const pixman_fast_path_t mips_dspr2_fast_paths[] =
     PIXMAN_STD_FAST_PATH (ADD,  a8,       null,     a8,       mips_composite_add_8_8),
     PIXMAN_STD_FAST_PATH (ADD,  a8r8g8b8, null,     a8r8g8b8, mips_composite_add_8888_8888),
     PIXMAN_STD_FAST_PATH (ADD,  a8b8g8r8, null,     a8b8g8r8, mips_composite_add_8888_8888),
+    PIXMAN_STD_FAST_PATH (OUT_REVERSE, a8,    null, r5g6b5,   mips_composite_out_reverse_8_0565),
+    PIXMAN_STD_FAST_PATH (OUT_REVERSE, a8,    null, b5g6r5,   mips_composite_out_reverse_8_0565),
+    PIXMAN_STD_FAST_PATH (OUT_REVERSE, a8,    null, a8r8g8b8, mips_composite_out_reverse_8_8888),
+    PIXMAN_STD_FAST_PATH (OUT_REVERSE, a8,    null, a8b8g8r8, mips_composite_out_reverse_8_8888),
 
     PIXMAN_MIPS_SIMPLE_NEAREST_A8_MASK_FAST_PATH (OVER, a8r8g8b8, r5g6b5, mips_8888_8_0565),
     PIXMAN_MIPS_SIMPLE_NEAREST_A8_MASK_FAST_PATH (OVER, a8b8g8r8, b5g6r5, mips_8888_8_0565),