merge from open-source master

Change-Id: Ibc1c58aa757910f2b00acf6883127dd027f01d73

11 files changed, 415 insertions, 2 deletions

diff --git a/libpixelflinger/Android.mk b/libpixelflinger/Android.mk
index 0cc85d9..6491d24 100644
--- a/libpixelflinger/Android.mk
+++ b/libpixelflinger/Android.mk
@@ -40,7 +40,13 @@ PIXELFLINGER_SRC_FILES:= \
 	buffer.cpp
 
 ifeq ($(TARGET_ARCH),arm)
+ifeq ($(TARGET_ARCH_VERSION),armv7-a)
+PIXELFLINGER_SRC_FILES += col32cb16blend_neon.S
+PIXELFLINGER_SRC_FILES += col32cb16blend.S
+else
 PIXELFLINGER_SRC_FILES += t32cb16blend.S
+PIXELFLINGER_SRC_FILES += col32cb16blend.S
+endif
 endif
 
 ifeq ($(TARGET_ARCH),arm)
diff --git a/libpixelflinger/codeflinger/ARMAssembler.cpp b/libpixelflinger/codeflinger/ARMAssembler.cpp
index ff7b0b3..d3720c3 100644
--- a/libpixelflinger/codeflinger/ARMAssembler.cpp
+++ b/libpixelflinger/codeflinger/ARMAssembler.cpp
@@ -424,5 +424,15 @@ void ARMAssembler::SMLAW(int cc, int y,
     *mPC++ = (cc<<28) | 0x1200080 | (Rd<<16) | (Rn<<12) | (Rs<<8) | (y<<4) | Rm;
 }
 
+#if 0
+#pragma mark -
+#pragma mark Byte/half word extract and extend (ARMv6+ only)...
+#endif
+
+void ARMAssembler::UXTB16(int cc, int Rd, int Rm, int rotate)
+{
+    *mPC++ = (cc<<28) | 0x6CF0070 | (Rd<<12) | ((rotate >> 3) << 10) | Rm;
+}
+
 }; // namespace android
 
diff --git a/libpixelflinger/codeflinger/ARMAssembler.h b/libpixelflinger/codeflinger/ARMAssembler.h
index ef3b66a..a667cb5 100644
--- a/libpixelflinger/codeflinger/ARMAssembler.h
+++ b/libpixelflinger/codeflinger/ARMAssembler.h
@@ -123,6 +123,7 @@ public:
                 int RdHi, int RdLo, int Rs, int Rm);
     virtual void SMLAW(int cc, int y,
                 int Rd, int Rm, int Rs, int Rn);
+    virtual void UXTB16(int cc, int Rd, int Rm, int rotate);
 
 private:
                 ARMAssembler(const ARMAssembler& rhs);
diff --git a/libpixelflinger/codeflinger/ARMAssemblerInterface.h b/libpixelflinger/codeflinger/ARMAssemblerInterface.h
index 465b3bd..ff6af2a 100644
--- a/libpixelflinger/codeflinger/ARMAssemblerInterface.h
+++ b/libpixelflinger/codeflinger/ARMAssemblerInterface.h
@@ -203,6 +203,9 @@ public:
     virtual void SMLAW(int cc, int y,
                 int Rd, int Rm, int Rs, int Rn) = 0;
 
+    // byte/half word extract...
+    virtual void UXTB16(int cc, int Rd, int Rm, int rotate) = 0;
+
     // -----------------------------------------------------------------------
     // convenience...
     // -----------------------------------------------------------------------
diff --git a/libpixelflinger/codeflinger/ARMAssemblerProxy.cpp b/libpixelflinger/codeflinger/ARMAssemblerProxy.cpp
index 18c4618..7c422db 100644
--- a/libpixelflinger/codeflinger/ARMAssemblerProxy.cpp
+++ b/libpixelflinger/codeflinger/ARMAssemblerProxy.cpp
@@ -195,6 +195,9 @@ void ARMAssemblerProxy::SMLAW(int cc, int y, int Rd, int Rm, int Rs, int Rn) {
     mTarget->SMLAW(cc, y, Rd, Rm, Rs, Rn);
 }
 
+void ARMAssemblerProxy::UXTB16(int cc, int Rd, int Rm, int rotate) {
+    mTarget->UXTB16(cc, Rd, Rm, rotate);
+}
 
 }; // namespace android
 
diff --git a/libpixelflinger/codeflinger/ARMAssemblerProxy.h b/libpixelflinger/codeflinger/ARMAssemblerProxy.h
index 4bdca9c..9134cce 100644
--- a/libpixelflinger/codeflinger/ARMAssemblerProxy.h
+++ b/libpixelflinger/codeflinger/ARMAssemblerProxy.h
@@ -114,6 +114,8 @@ public:
     virtual void SMLAW(int cc, int y,
                 int Rd, int Rm, int Rs, int Rn);
 
+    virtual void UXTB16(int cc, int Rd, int Rm, int rotate);
+
 private:
     ARMAssemblerInterface*  mTarget;
 };
diff --git a/libpixelflinger/codeflinger/disassem.c b/libpixelflinger/codeflinger/disassem.c
index 4676da0..c17f3ec 100644
--- a/libpixelflinger/codeflinger/disassem.c
+++ b/libpixelflinger/codeflinger/disassem.c
@@ -80,6 +80,7 @@
  * f - 1st fp operand (register) (bits 12-14)
  * g - 2nd fp operand (register) (bits 16-18)
  * h - 3rd fp operand (register/immediate) (bits 0-4)
+ * j - xtb rotate literal (bits 10-11)
  * b - branch address
  * t - thumb branch address (bits 24, 0-23)
  * k - breakpoint comment (bits 0-3, 8-19)
@@ -122,6 +123,7 @@ static const struct arm32_insn arm32_i[] = {
     { 0x0fe000f0, 0x00c00090, "smull",	"Sdnms" },
     { 0x0fe000f0, 0x00a00090, "umlal",	"Sdnms" },
     { 0x0fe000f0, 0x00e00090, "smlal",	"Sdnms" },
+    { 0x0fff03f0, 0x06cf0070, "uxtb16", "dmj" },
     { 0x0d700000, 0x04200000, "strt",	"daW" },
     { 0x0d700000, 0x04300000, "ldrt",	"daW" },
     { 0x0d700000, 0x04600000, "strbt",	"daW" },
@@ -276,7 +278,7 @@ static char const insn_fpaconstants[][8] = {
 
 #define insn_condition(x)	arm32_insn_conditions[(x >> 28) & 0x0f]
 #define insn_blktrans(x)	insn_block_transfers[(x >> 23) & 3]
-#define insn_stkblktrans(x)	insn_stack_block_transfers[(x >> 23) & 3]
+#define insn_stkblktrans(x)	insn_stack_block_transfers[(3*((x >> 20)&1))^((x >> 23)&3)]
 #define op2_shift(x)		op_shifts[(x >> 5) & 3]
 #define insn_fparnd(x)		insn_fpa_rounding[(x >> 5) & 0x03]
 #define insn_fpaprec(x)		insn_fpa_precision[(((x >> 18) & 2)|(x >> 7)) & 1]
@@ -406,6 +408,10 @@ disasm(const disasm_interface_t *di, u_int loc, int altfmt)
 			else
 				di->di_printf("f%d", insn & 7);
 			break;
+		/* j - xtb rotate literal (bits 10-11) */
+		case 'j':
+			di->di_printf("ror #%d", ((insn >> 10) & 3) << 3);
+			break;
 		/* b - branch address */
 		case 'b':
 			branch = ((insn << 2) & 0x03ffffff);
diff --git a/libpixelflinger/codeflinger/texturing.cpp b/libpixelflinger/codeflinger/texturing.cpp
index 90e6584..6baa28e 100644
--- a/libpixelflinger/codeflinger/texturing.cpp
+++ b/libpixelflinger/codeflinger/texturing.cpp
@@ -25,6 +25,9 @@
 
 #include "codeflinger/GGLAssembler.h"
 
+#ifdef __arm__
+#include <machine/cpu-features.h>
+#endif
 
 namespace android {
 
@@ -567,7 +570,7 @@ void GGLAssembler::build_textures(  fragment_parts_t& parts,
                     RSB(GE, 0, height, height, imm(0));
                     MUL(AL, 0, height, stride, height);
                 } else {
-                    // u has not been CLAMPed yet
+                    // v has not been CLAMPed yet
                     CMP(AL, height, reg_imm(v, ASR, FRAC_BITS));
                     MOV(LE, 0, v, reg_imm(height, LSL, FRAC_BITS));
                     MOV(LE, 0, height, imm(0));
@@ -868,6 +871,106 @@ void GGLAssembler::filter24(
     load(txPtr, texel, 0);
 }
 
+#if __ARM_ARCH__ >= 6
+// ARMv6 version, using UXTB16, and scheduled for Cortex-A8 pipeline
+void GGLAssembler::filter32(
+        const fragment_parts_t& parts,
+        pixel_t& texel, const texture_unit_t& tmu,
+        int U, int V, pointer_t& txPtr,
+        int FRAC_BITS)
+{
+    const int adjust = FRAC_BITS*2 - 8;
+    const int round  = 0;
+    const int prescale = 16 - adjust;
+
+    Scratch scratches(registerFile());
+    
+    int pixel= scratches.obtain();
+    int dh   = scratches.obtain();
+    int u    = scratches.obtain();
+    int k    = scratches.obtain();
+
+    int temp = scratches.obtain();
+    int dl   = scratches.obtain();
+
+    int offsetrt = scratches.obtain();
+    int offsetlb = scratches.obtain();
+
+    int pixellb = offsetlb;
+
+    // RB -> U * V
+    CONTEXT_LOAD(offsetrt, generated_vars.rt);
+    CONTEXT_LOAD(offsetlb, generated_vars.lb);
+    if(!round) {
+        MOV(AL, 0, U, reg_imm(U, LSL, prescale));
+    }
+    ADD(AL, 0, u, offsetrt, offsetlb);
+
+    LDR(AL, pixel, txPtr.reg, reg_scale_pre(u));
+    if (round) {
+        SMULBB(AL, u, U, V);
+        RSB(AL, 0, U, U, imm(1<<FRAC_BITS));
+    } else {
+        SMULWB(AL, u, U, V);
+        RSB(AL, 0, U, U, imm(1<<(FRAC_BITS+prescale)));
+    }
+    UXTB16(AL, temp, pixel, 0);
+    if (round) {
+        ADD(AL, 0, u, u, imm(1<<(adjust-1)));
+        MOV(AL, 0, u, reg_imm(u, LSR, adjust));
+    }
+    LDR(AL, pixellb, txPtr.reg, reg_scale_pre(offsetlb));
+    MUL(AL, 0, dh, temp, u);
+    UXTB16(AL, temp, pixel, 8);
+    MUL(AL, 0, dl, temp, u);
+    RSB(AL, 0, k, u, imm(0x100));
+
+    // LB -> (1-U) * V
+    if (round) {
+        SMULBB(AL, u, U, V);
+    } else {
+        SMULWB(AL, u, U, V);
+    }
+    UXTB16(AL, temp, pixellb, 0);
+    if (round) {
+        ADD(AL, 0, u, u, imm(1<<(adjust-1)));
+        MOV(AL, 0, u, reg_imm(u, LSR, adjust));
+    }
+    MLA(AL, 0, dh, temp, u, dh);    
+    UXTB16(AL, temp, pixellb, 8);
+    MLA(AL, 0, dl, temp, u, dl);
+    SUB(AL, 0, k, k, u);
+
+    // LT -> (1-U)*(1-V)
+    RSB(AL, 0, V, V, imm(1<<FRAC_BITS));
+    LDR(AL, pixel, txPtr.reg);
+    if (round) {
+        SMULBB(AL, u, U, V);
+    } else {
+        SMULWB(AL, u, U, V);
+    }
+    UXTB16(AL, temp, pixel, 0);
+    if (round) {
+        ADD(AL, 0, u, u, imm(1<<(adjust-1)));
+        MOV(AL, 0, u, reg_imm(u, LSR, adjust));
+    }
+    MLA(AL, 0, dh, temp, u, dh);    
+    UXTB16(AL, temp, pixel, 8);
+    MLA(AL, 0, dl, temp, u, dl);
+
+    // RT -> U*(1-V)            
+    LDR(AL, pixel, txPtr.reg, reg_scale_pre(offsetrt));
+    SUB(AL, 0, u, k, u);
+    UXTB16(AL, temp, pixel, 0);
+    MLA(AL, 0, dh, temp, u, dh);    
+    UXTB16(AL, temp, pixel, 8);
+    MLA(AL, 0, dl, temp, u, dl);
+
+    UXTB16(AL, dh, dh, 8);
+    UXTB16(AL, dl, dl, 8);
+    ORR(AL, 0, texel.reg, dh, reg_imm(dl, LSL, 8));
+}
+#else
 void GGLAssembler::filter32(
         const fragment_parts_t& parts,
         pixel_t& texel, const texture_unit_t& tmu,
@@ -955,6 +1058,7 @@ void GGLAssembler::filter32(
     AND(AL, 0, dl, dl, reg_imm(mask, LSL, 8));
     ORR(AL, 0, texel.reg, dh, dl);
 }
+#endif
 
 void GGLAssembler::build_texture_environment(
         component_t& fragment,
diff --git a/libpixelflinger/col32cb16blend.S b/libpixelflinger/col32cb16blend.S
new file mode 100644
index 0000000..1450bde
--- /dev/null
+++ b/libpixelflinger/col32cb16blend.S
@@ -0,0 +1,78 @@
+/* libs/pixelflinger/col32cb16blend.S
+**
+** (C) COPYRIGHT 2009 ARM Limited.
+**
+** Licensed under the Apache License, Version 2.0 (the "License"); 
+** you may not use this file except in compliance with the License. 
+** You may obtain a copy of the License at 
+**
+**     http://www.apache.org/licenses/LICENSE-2.0 
+**
+** Unless required by applicable law or agreed to in writing, software 
+** distributed under the License is distributed on an "AS IS" BASIS, 
+** WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 
+** See the License for the specific language governing permissions and 
+** limitations under the License.
+**
+*/
+
+    .text
+    .align
+
+    .global scanline_col32cb16blend_arm
+
+//
+// This function alpha blends a fixed color into a destination scanline, using
+// the formula:
+//
+//     d = s + (((a + (a >> 7)) * d) >> 8)
+//
+// where d is the destination pixel,
+//       s is the source color,
+//       a is the alpha channel of the source color.
+//
+
+// r0 = destination buffer pointer
+// r1 = color value
+// r2 = count
+
+
+scanline_col32cb16blend_arm:
+    push        {r4-r10, lr}                    // stack ARM regs
+
+    mov         r5, r1, lsr #24                 // shift down alpha
+    mov         r9, #0xff                       // create mask
+    add         r5, r5, r5, lsr #7              // add in top bit
+    rsb         r5, r5, #256                    // invert alpha
+    and         r10, r1, #0xff                  // extract red
+    and         r12, r9, r1, lsr #8             // extract green
+    and         r4, r9, r1, lsr #16             // extract blue
+    mov         r10, r10, lsl #5                // prescale red
+    mov         r12, r12, lsl #6                // prescale green
+    mov         r4, r4, lsl #5                  // prescale blue
+    mov         r9, r9, lsr #2                  // create dest green mask
+
+1:
+    ldrh        r8, [r0]                        // load dest pixel
+    subs        r2, r2, #1                      // decrement loop counter
+    mov         r6, r8, lsr #11                 // extract dest red
+    and         r7, r9, r8, lsr #5              // extract dest green
+    and         r8, r8, #0x1f                   // extract dest blue
+
+    smlabb      r6, r6, r5, r10                 // dest red * alpha + src red
+    smlabb      r7, r7, r5, r12                 // dest green * alpha + src green
+    smlabb      r8, r8, r5, r4                  // dest blue * alpha + src blue
+
+    mov         r6, r6, lsr #8                  // shift down red
+    mov         r7, r7, lsr #8                  // shift down green
+    mov         r6, r6, lsl #11                 // shift red into 565
+    orr         r6, r7, lsl #5                  // shift green into 565
+    orr         r6, r8, lsr #8                  // shift blue into 565
+
+    strh        r6, [r0], #2                    // store pixel to dest, update ptr
+    bne         1b                              // if count != 0, loop
+
+    pop         {r4-r10, pc}                    // return
+
+
+
diff --git a/libpixelflinger/col32cb16blend_neon.S b/libpixelflinger/col32cb16blend_neon.S
new file mode 100644
index 0000000..17b0d01
--- /dev/null
+++ b/libpixelflinger/col32cb16blend_neon.S
@@ -0,0 +1,153 @@
+/* libs/pixelflinger/col32cb16blend_neon.S
+**
+** (C) COPYRIGHT 2009 ARM Limited.
+**
+** Licensed under the Apache License, Version 2.0 (the "License"); 
+** you may not use this file except in compliance with the License. 
+** You may obtain a copy of the License at 
+**
+**     http://www.apache.org/licenses/LICENSE-2.0 
+**
+** Unless required by applicable law or agreed to in writing, software 
+** distributed under the License is distributed on an "AS IS" BASIS, 
+** WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 
+** See the License for the specific language governing permissions and 
+** limitations under the License.
+**
+*/
+
+    .text
+    .align
+
+    .global scanline_col32cb16blend_neon
+
+//
+// This function alpha blends a fixed color into a destination scanline, using
+// the formula:
+//
+//     d = s + (((a + (a >> 7)) * d) >> 8)
+//
+// where d is the destination pixel,
+//       s is the source color,
+//       a is the alpha channel of the source color.
+//
+// The NEON implementation processes 16 pixels per iteration. The remaining 0 - 15
+// pixels are processed in ARM code.
+//
+
+// r0 = destination buffer pointer
+// r1 = color pointer
+// r2 = count
+
+
+scanline_col32cb16blend_neon:
+    push        {r4-r11, lr}                    // stack ARM regs
+
+    vmov.u16    q15, #256                       // create alpha constant
+    movs        r3, r2, lsr #4                  // calc. sixteens iterations
+    vmov.u16    q14, #0x1f                      // create blue mask
+
+    beq         2f                              // if r3 == 0, branch to singles
+
+    vld4.8      {d0[], d2[], d4[], d6[]}, [r1]  // load color into four registers
+                                                //  split and duplicate them, such that
+                                                //  d0 = 8 equal red values
+                                                //  d2 = 8 equal green values
+                                                //  d4 = 8 equal blue values
+                                                //  d6 = 8 equal alpha values
+    vshll.u8    q0, d0, #5                      // shift up red and widen
+    vshll.u8    q1, d2, #6                      // shift up green and widen
+    vshll.u8    q2, d4, #5                      // shift up blue and widen
+
+    vshr.u8     d7, d6, #7                      // extract top bit of alpha
+    vaddl.u8    q3, d6, d7                      // add top bit into alpha
+    vsub.u16    q3, q15, q3                     // invert alpha
+
+1:
+    // This loop processes 16 pixels per iteration. In the comments, references to
+    // the first eight pixels are suffixed with "0" (red0, green0, blue0), 
+    // the second eight are suffixed "1".
+                                                // q8  = dst red0
+                                                // q9  = dst green0
+                                                // q10 = dst blue0
+                                                // q13 = dst red1
+                                                // q12 = dst green1
+                                                // q11 = dst blue1
+
+    vld1.16     {d20, d21, d22, d23}, [r0]      // load 16 dest pixels
+    vshr.u16    q8, q10, #11                    // shift dst red0 to low 5 bits
+    pld         [r0, #63]                       // preload next dest pixels
+    vshl.u16    q9, q10, #5                     // shift dst green0 to top 6 bits
+    vand        q10, q10, q14                   // extract dst blue0
+    vshr.u16    q9, q9, #10                     // shift dst green0 to low 6 bits
+    vmul.u16    q8, q8, q3                      // multiply dst red0 by src alpha
+    vshl.u16    q12, q11, #5                    // shift dst green1 to top 6 bits
+    vmul.u16    q9, q9, q3                      // multiply dst green0 by src alpha
+    vshr.u16    q13, q11, #11                   // shift dst red1 to low 5 bits
+    vmul.u16    q10, q10, q3                    // multiply dst blue0 by src alpha
+    vshr.u16    q12, q12, #10                   // shift dst green1 to low 6 bits
+    vand        q11, q11, q14                   // extract dst blue1
+    vadd.u16    q8, q8, q0                      // add src red to dst red0
+    vmul.u16    q13, q13, q3                    // multiply dst red1 by src alpha
+    vadd.u16    q9, q9, q1                      // add src green to dst green0 
+    vmul.u16    q12, q12, q3                    // multiply dst green1 by src alpha
+    vadd.u16    q10, q10, q2                    // add src blue to dst blue0
+    vmul.u16    q11, q11, q3                    // multiply dst blue1 by src alpha
+    vshr.u16    q8, q8, #8                      // shift down red0
+    vadd.u16    q13, q13, q0                    // add src red to dst red1
+    vshr.u16    q9, q9, #8                      // shift down green0
+    vadd.u16    q12, q12, q1                    // add src green to dst green1
+    vshr.u16    q10, q10, #8                    // shift down blue0
+    vadd.u16    q11, q11, q2                    // add src blue to dst blue1
+    vsli.u16    q10, q9, #5                     // shift & insert green0 into blue0
+    vshr.u16    q13, q13, #8                    // shift down red1
+    vsli.u16    q10, q8, #11                    // shift & insert red0 into blue0    
+    vshr.u16    q12, q12, #8                    // shift down green1
+    vshr.u16    q11, q11, #8                    // shift down blue1
+    subs        r3, r3, #1                      // decrement loop counter
+    vsli.u16    q11, q12, #5                    // shift & insert green1 into blue1
+    vsli.u16    q11, q13, #11                   // shift & insert red1 into blue1
+
+    vst1.16     {d20, d21, d22, d23}, [r0]!     // write 16 pixels back to dst
+    bne         1b                              // if count != 0, loop
+
+2:
+    ands        r3, r2, #15                     // calc. single iterations 
+    beq         4f                              // if r3 == 0, exit
+
+    ldr         r4, [r1]                        // load source color
+    mov         r5, r4, lsr #24                 // shift down alpha
+    add         r5, r5, r5, lsr #7              // add in top bit
+    rsb         r5, r5, #256                    // invert alpha
+    and         r11, r4, #0xff                  // extract red
+    ubfx        r12, r4, #8, #8                 // extract green
+    ubfx        r4, r4, #16, #8                 // extract blue
+    mov         r11, r11, lsl #5                // prescale red
+    mov         r12, r12, lsl #6                // prescale green
+    mov         r4, r4, lsl #5                  // prescale blue
+
+3:
+    ldrh        r8, [r0]                        // load dest pixel
+    subs        r3, r3, #1                      // decrement loop counter
+    mov         r6, r8, lsr #11                 // extract dest red
+    ubfx        r7, r8, #5, #6                  // extract dest green
+    and         r8, r8, #0x1f                   // extract dest blue
+
+    smlabb      r6, r6, r5, r11                 // dest red * alpha + src red
+    smlabb      r7, r7, r5, r12                 // dest green * alpha + src green
+    smlabb      r8, r8, r5, r4                  // dest blue * alpha + src blue
+
+    mov         r6, r6, lsr #8                  // shift down red
+    mov         r7, r7, lsr #8                  // shift down green
+    mov         r6, r6, lsl #11                 // shift red into 565
+    orr         r6, r7, lsl #5                  // shift green into 565
+    orr         r6, r8, lsr #8                  // shift blue into 565
+
+    strh        r6, [r0], #2                    // store pixel to dest, update ptr
+    bne         3b                              // if count != 0, loop
+4:
+
+    pop         {r4-r11, pc}                    // return
+
+
+
diff --git a/libpixelflinger/scanline.cpp b/libpixelflinger/scanline.cpp
index f700306..a2f43eb 100644
--- a/libpixelflinger/scanline.cpp
+++ b/libpixelflinger/scanline.cpp
@@ -80,6 +80,7 @@ static void scanline_perspective(context_t* c);
 static void scanline_perspective_single(context_t* c);
 static void scanline_t32cb16blend(context_t* c);
 static void scanline_t32cb16(context_t* c);
+static void scanline_col32cb16blend(context_t* c);
 static void scanline_memcpy(context_t* c);
 static void scanline_memset8(context_t* c);
 static void scanline_memset16(context_t* c);
@@ -93,6 +94,8 @@ static void rect_memcpy(context_t* c, size_t yc);
 
 extern "C" void scanline_t32cb16blend_arm(uint16_t*, uint32_t*, size_t);
 extern "C" void scanline_t32cb16_arm(uint16_t *dst, uint32_t *src, size_t ct);
+extern "C" void scanline_col32cb16blend_neon(uint16_t *dst, uint32_t *col, size_t ct);
+extern "C" void scanline_col32cb16blend_arm(uint16_t *dst, uint32_t col, size_t ct);
 
 // ----------------------------------------------------------------------------
 
@@ -111,6 +114,9 @@ static shortcut_t shortcuts[] = {
     { { { 0x03010104, 0x00000077, { 0x00000A01, 0x00000000 } },
         { 0xFFFFFFFF, 0xFFFFFFFF, { 0xFFFFFFFF, 0x0000003F } } },
         "565 fb, 8888 tx", scanline_t32cb16, init_y_noop  },  
+    { { { 0x03515104, 0x00000077, { 0x00000000, 0x00000000 } },
+        { 0xFFFFFFFF, 0xFFFFFFFF, { 0xFFFFFFFF, 0xFFFFFFFF } } },
+        "565 fb, 8888 fixed color", scanline_col32cb16blend, init_y_packed  },  
     { { { 0x00000000, 0x00000000, { 0x00000000, 0x00000000 } },
         { 0x00000000, 0x00000007, { 0x00000000, 0x00000000 } } },
         "(nop) alpha test", scanline_noop, init_y_noop },
@@ -943,6 +949,8 @@ void init_y_packed(context_t* c, int32_t y0)
     uint8_t f = c->state.buffers.color.format;
     c->packed = ggl_pack_color(c, f,
             c->shade.r0, c->shade.g0, c->shade.b0, c->shade.a0);
+    c->packed8888 = ggl_pack_color(c, GGL_PIXEL_FORMAT_RGBA_8888,
+            c->shade.r0, c->shade.g0, c->shade.b0, c->shade.a0);
     c->iterators.y = y0;
     c->step_y = step_y__nop;
     // choose the rectangle blitter
@@ -1253,6 +1261,45 @@ finish:
 
 // ----------------------------------------------------------------------------
 
+void scanline_col32cb16blend(context_t* c)
+{
+    int32_t x = c->iterators.xl;
+    size_t ct = c->iterators.xr - x;
+    int32_t y = c->iterators.y;
+    surface_t* cb = &(c->state.buffers.color);
+    union {
+        uint16_t* dst;
+        uint32_t* dst32;
+    };
+    dst = reinterpret_cast<uint16_t*>(cb->data) + (x+(cb->stride*y));
+
+#if ((ANDROID_CODEGEN >= ANDROID_CODEGEN_ASM) && defined(__arm__))
+#if defined(__ARM_HAVE_NEON) && BYTE_ORDER == LITTLE_ENDIAN
+    scanline_col32cb16blend_neon(dst, &(c->packed8888), ct);
+#else  // defined(__ARM_HAVE_NEON) && BYTE_ORDER == LITTLE_ENDIAN
+    scanline_col32cb16blend_arm(dst, GGL_RGBA_TO_HOST(c->packed8888), ct);
+#endif // defined(__ARM_HAVE_NEON) && BYTE_ORDER == LITTLE_ENDIAN
+#else
+    uint32_t s = GGL_RGBA_TO_HOST(c->packed8888);
+    int sA = (s>>24);
+    int f = 0x100 - (sA + (sA>>7));
+    while (ct--) {
+        uint16_t d = *dst;
+        int dR = (d>>11)&0x1f;
+        int dG = (d>>5)&0x3f;
+        int dB = (d)&0x1f;
+        int sR = (s >> (   3))&0x1F;
+        int sG = (s >> ( 8+2))&0x3F;
+        int sB = (s >> (16+3))&0x1F;
+        sR += (f*dR)>>8;
+        sG += (f*dG)>>8;
+        sB += (f*dB)>>8;
+        *dst++ = uint16_t((sR<<11)|(sG<<5)|sB);
+    }
+#endif
+
+}
+
 void scanline_t32cb16(context_t* c)
 {
     int32_t x = c->iterators.xl;