Initial-checkin for ON2 Software AVC/H264 decoder

o when neon is present, the performance gain of On2 AVC software decoder
  over PV software decoder is more than 30%.

o In addition, it fixes some known PV software decoder issues like missing
  output frames

o allow both pv and on2 software avc to be available for easy comparision

o change output frames from 8 to 16

Change-Id: I567ad1842025ead7092f0c47e3513d6d9ca232dd

5 files changed, 529 insertions, 0 deletions

diff --git a/media/libstagefright/codecs/on2/h264dec/source/arm_neon_asm/h264bsdClearMbLayer.s b/media/libstagefright/codecs/on2/h264dec/source/arm_neon_asm/h264bsdClearMbLayer.s
new file mode 100644
index 0000000..db11654
--- /dev/null
+++ b/media/libstagefright/codecs/on2/h264dec/source/arm_neon_asm/h264bsdClearMbLayer.s
@@ -0,0 +1,66 @@
+;
+; Copyright (C) 2009 The Android Open Source Project
+;
+; 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.
+;
+
+    REQUIRE8
+    PRESERVE8
+
+    AREA    |.text|, CODE
+
+    EXPORT h264bsdClearMbLayer
+
+; Input / output registers
+pMbLayer    RN  0
+size        RN  1
+pTmp        RN  2
+step        RN  3
+
+; -- NEON registers --
+
+qZero   QN  Q0.U8
+
+;/*------------------------------------------------------------------------------
+;
+;    Function: h264bsdClearMbLayer
+;
+;        Functional description:
+;
+;        Inputs:
+;
+;        Outputs:
+;
+;        Returns:
+;
+;------------------------------------------------------------------------------*/
+
+h264bsdClearMbLayer
+
+    VMOV    qZero, #0
+    ADD     pTmp, pMbLayer, #16
+    MOV     step, #32
+    SUBS    size, size, #64
+
+loop
+    VST1    qZero, [pMbLayer], step
+    SUBS    size, size, #64
+    VST1    qZero, [pTmp], step
+    VST1    qZero, [pMbLayer], step
+    VST1    qZero, [pTmp], step
+    BCS     loop
+
+    BX      lr
+    END
+
+
diff --git a/media/libstagefright/codecs/on2/h264dec/source/arm_neon_asm/h264bsdCountLeadingZeros.s b/media/libstagefright/codecs/on2/h264dec/source/arm_neon_asm/h264bsdCountLeadingZeros.s
new file mode 100644
index 0000000..c7bd73e
--- /dev/null
+++ b/media/libstagefright/codecs/on2/h264dec/source/arm_neon_asm/h264bsdCountLeadingZeros.s
@@ -0,0 +1,49 @@
+;
+; Copyright (C) 2009 The Android Open Source Project
+;
+; 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.
+;
+
+    REQUIRE8
+    PRESERVE8
+
+    AREA    |.text|, CODE
+
+    EXPORT h264bsdCountLeadingZeros
+
+; Input / output registers
+value    RN  0
+
+; -- NEON registers --
+
+;/*------------------------------------------------------------------------------
+;
+;    Function: h264bsdCountLeadingZeros
+;
+;        Functional description:
+;
+;        Inputs:
+;
+;        Outputs:
+;
+;        Returns:
+;
+;------------------------------------------------------------------------------*/
+
+h264bsdCountLeadingZeros
+
+    CLZ     value, value
+    BX      lr
+    END
+
+
diff --git a/media/libstagefright/codecs/on2/h264dec/source/arm_neon_asm/h264bsdFillRow7.s b/media/libstagefright/codecs/on2/h264dec/source/arm_neon_asm/h264bsdFillRow7.s
new file mode 100644
index 0000000..5bfac92
--- /dev/null
+++ b/media/libstagefright/codecs/on2/h264dec/source/arm_neon_asm/h264bsdFillRow7.s
@@ -0,0 +1,180 @@
+;
+; Copyright (C) 2009 The Android Open Source Project
+;
+; 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.
+;
+
+    REQUIRE8
+    PRESERVE8
+
+    AREA    |.text|, CODE
+
+    EXPORT h264bsdFillRow7
+
+; Input / output registers
+
+ref     RN 0
+fill    RN 1
+left    RN 2
+tmp2    RN 2
+center  RN 3
+right   RN 4
+tmp1    RN 5
+
+; -- NEON registers --
+
+qTmp0   QN  Q0.U8
+qTmp1   QN  Q1.U8
+dTmp0   DN  D0.U8
+dTmp1   DN  D1.U8
+dTmp2   DN  D2.U8
+dTmp3   DN  D3.U8
+
+
+;/*------------------------------------------------------------------------------
+;
+;    Function: h264bsdFillRow7
+;
+;        Functional description:
+;
+;        Inputs:
+;
+;        Outputs:
+;
+;        Returns:
+;
+;------------------------------------------------------------------------------*/
+
+h264bsdFillRow7
+        PUSH     {r4-r6,lr}
+        CMP      left, #0
+        LDR      right, [sp,#0x10]
+        BEQ      switch_center
+        LDRB     tmp1, [ref,#0]
+
+loop_left
+        SUBS     left, left, #1
+        STRB     tmp1, [fill], #1
+        BNE      loop_left
+
+switch_center
+        ASR      tmp2,center,#2
+        CMP      tmp2,#9
+        ADDCC    pc,pc,tmp2,LSL #2
+        B        loop_center
+        B        loop_center
+        B        case_1
+        B        case_2
+        B        case_3
+        B        case_4
+        B        case_5
+        B        case_6
+        B        case_7
+        B        case_8
+;case_8
+;        LDR      tmp2, [ref], #4
+;        SUB      center, center, #4
+;        STR      tmp2, [fill], #4
+;case_7
+;        LDR      tmp2, [ref], #4
+;        SUB      center, center, #4
+;        STR      tmp2, [fill], #4
+;case_6
+;        LDR      tmp2, [ref], #4
+;        SUB      center, center, #4
+;        STR      tmp2, [fill],#4
+;case_5
+;        LDR      tmp2, [ref], #4
+;        SUB      center, center, #4
+;        STR      tmp2, [fill],#4
+;case_4
+;        LDR      tmp2, [ref],#4
+;        SUB      center, center, #4
+;        STR      tmp2, [fill], #4
+;case_3
+;        LDR      tmp2, [ref],#4
+;        SUB      center, center, #4
+;        STR      tmp2, [fill], #4
+;case_2
+;        LDR      tmp2, [ref],#4
+;        SUB      center, center, #4
+;        STR      tmp2, [fill], #4
+;case_1
+;        LDR      tmp2, [ref],#4
+;        SUB      center, center, #4
+;        STR      tmp2, [fill], #4
+
+case_8
+        VLD1    {qTmp0, qTmp1}, [ref]!
+        SUB     center, center, #32
+        VST1    qTmp0, [fill]!
+        VST1    qTmp1, [fill]!
+        B       loop_center
+case_7
+        VLD1    {dTmp0,dTmp1,dTmp2}, [ref]!
+        SUB     center, center, #28
+        LDR     tmp2, [ref], #4
+        VST1    {dTmp0,dTmp1,dTmp2}, [fill]!
+        STR     tmp2, [fill],#4
+        B       loop_center
+case_6
+        VLD1    {dTmp0,dTmp1,dTmp2}, [ref]!
+        SUB     center, center, #24
+        VST1    {dTmp0,dTmp1,dTmp2}, [fill]!
+        B       loop_center
+case_5
+        VLD1    qTmp0, [ref]!
+        SUB     center, center, #20
+        LDR     tmp2, [ref], #4
+        VST1    qTmp0, [fill]!
+        STR     tmp2, [fill],#4
+        B       loop_center
+case_4
+        VLD1    qTmp0, [ref]!
+        SUB     center, center, #16
+        VST1    qTmp0, [fill]!
+        B       loop_center
+case_3
+        VLD1    dTmp0, [ref]!
+        SUB     center, center, #12
+        LDR     tmp2, [ref], #4
+        VST1    dTmp0, [fill]!
+        STR     tmp2, [fill],#4
+        B       loop_center
+case_2
+        LDR      tmp2, [ref],#4
+        SUB      center, center, #4
+        STR      tmp2, [fill], #4
+case_1
+        LDR      tmp2, [ref],#4
+        SUB      center, center, #4
+        STR      tmp2, [fill], #4
+
+loop_center
+        CMP      center, #0
+        LDRBNE   tmp2, [ref], #1
+        SUBNE    center, center, #1
+        STRBNE   tmp2, [fill], #1
+        BNE      loop_center
+        CMP      right,#0
+        POPEQ    {r4-r6,pc}
+        LDRB     tmp2, [ref,#-1]
+
+loop_right
+        STRB     tmp2, [fill], #1
+        SUBS     right, right, #1
+        BNE      loop_right
+
+        POP      {r4-r6,pc}
+        END
+
diff --git a/media/libstagefright/codecs/on2/h264dec/source/arm_neon_asm/h264bsdFlushBits.s b/media/libstagefright/codecs/on2/h264dec/source/arm_neon_asm/h264bsdFlushBits.s
new file mode 100644
index 0000000..21335b8
--- /dev/null
+++ b/media/libstagefright/codecs/on2/h264dec/source/arm_neon_asm/h264bsdFlushBits.s
@@ -0,0 +1,82 @@
+;
+; Copyright (C) 2009 The Android Open Source Project
+;
+; 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.
+;
+
+    REQUIRE8
+    PRESERVE8
+
+    AREA    |.text|, CODE
+
+    EXPORT h264bsdFlushBits
+
+; Input / output registers
+pStrmData       RN  0
+numBits         RN  1
+readBits        RN  2
+strmBuffSize    RN  3
+pStrmBuffStart  RN  1
+pStrmCurrPos    RN  2
+bitPosInWord    RN  1
+
+; -- NEON registers --
+
+
+
+;/*------------------------------------------------------------------------------
+;
+;    Function: h264bsdFlushBits
+;
+;        Functional description:
+;
+;        Inputs:
+;
+;        Outputs:
+;
+;        Returns:
+;
+;------------------------------------------------------------------------------*/
+
+h264bsdFlushBits
+;//    PUSH     {r4-r6,lr}
+
+    LDR readBits, [pStrmData, #0x10]
+    LDR strmBuffSize, [pStrmData, #0xC]
+
+    ADD readBits, readBits, numBits
+    AND bitPosInWord, readBits, #7
+
+    STR readBits, [pStrmData, #0x10]
+    STR bitPosInWord, [pStrmData, #0x8]
+
+    LDR pStrmBuffStart, [pStrmData, #0x0]
+
+    CMP readBits, strmBuffSize, LSL #3
+
+    BHI end_of_stream
+
+    ADD pStrmCurrPos, pStrmBuffStart, readBits, LSR #3
+    STR pStrmCurrPos, [pStrmData, #0x4]
+    MOV r0, #0
+    BX  lr
+;//    POP      {r4-r6,pc}
+
+end_of_stream
+    MVN r0, #0
+    BX  lr
+;//    POP      {r4-r6,pc}
+
+    END
+
+
diff --git a/media/libstagefright/codecs/on2/h264dec/source/arm_neon_asm/h264bsdWriteMacroblock.s b/media/libstagefright/codecs/on2/h264dec/source/arm_neon_asm/h264bsdWriteMacroblock.s
new file mode 100644
index 0000000..38a0781
--- /dev/null
+++ b/media/libstagefright/codecs/on2/h264dec/source/arm_neon_asm/h264bsdWriteMacroblock.s
@@ -0,0 +1,152 @@
+;
+; Copyright (C) 2009 The Android Open Source Project
+;
+; 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.
+;
+
+    REQUIRE8
+    PRESERVE8
+
+    AREA    |.text|, CODE
+
+    EXPORT h264bsdWriteMacroblock
+
+; Input / output registers
+image   RN  0
+data    RN  1
+width   RN  2
+luma    RN  3
+cb      RN  4
+cr      RN  5
+cwidth  RN  6
+
+; -- NEON registers --
+
+qRow0   QN  Q0.U8
+qRow1   QN  Q1.U8
+qRow2   QN  Q2.U8
+qRow3   QN  Q3.U8
+qRow4   QN  Q4.U8
+qRow5   QN  Q5.U8
+qRow6   QN  Q6.U8
+qRow7   QN  Q7.U8
+qRow8   QN  Q8.U8
+qRow9   QN  Q9.U8
+qRow10  QN  Q10.U8
+qRow11  QN  Q11.U8
+qRow12  QN  Q12.U8
+qRow13  QN  Q13.U8
+qRow14  QN  Q14.U8
+qRow15  QN  Q15.U8
+
+dRow0   DN  D0.U8
+dRow1   DN  D1.U8
+dRow2   DN  D2.U8
+dRow3   DN  D3.U8
+dRow4   DN  D4.U8
+dRow5   DN  D5.U8
+dRow6   DN  D6.U8
+dRow7   DN  D7.U8
+dRow8   DN  D8.U8
+dRow9   DN  D9.U8
+dRow10  DN  D10.U8
+dRow11  DN  D11.U8
+dRow12  DN  D12.U8
+dRow13  DN  D13.U8
+dRow14  DN  D14.U8
+dRow15  DN  D15.U8
+
+;/*------------------------------------------------------------------------------
+;
+;    Function: h264bsdWriteMacroblock
+;
+;        Functional description:
+;            Write one macroblock into the image. Both luma and chroma
+;            components will be written at the same time.
+;
+;        Inputs:
+;            data    pointer to macroblock data to be written, 256 values for
+;                    luma followed by 64 values for both chroma components
+;
+;        Outputs:
+;            image   pointer to the image where the macroblock will be written
+;
+;        Returns:
+;            none
+;
+;------------------------------------------------------------------------------*/
+
+h264bsdWriteMacroblock
+    PUSH    {r4-r6,lr}
+    VPUSH   {q4-q7}
+
+    LDR     width, [image, #4]
+    LDR     luma, [image, #0xC]
+    LDR     cb, [image, #0x10]
+    LDR     cr, [image, #0x14]
+
+
+;   Write luma
+    VLD1    {qRow0, qRow1}, [data]!
+    LSL     width, width, #4
+    VLD1    {qRow2, qRow3}, [data]!
+    LSR     cwidth, width, #1
+    VST1    {qRow0}, [luma@128], width
+    VLD1    {qRow4, qRow5}, [data]!
+    VST1    {qRow1}, [luma@128], width
+    VLD1    {qRow6, qRow7}, [data]!
+    VST1    {qRow2}, [luma@128], width
+    VLD1    {qRow8, qRow9}, [data]!
+    VST1    {qRow3}, [luma@128], width
+    VLD1    {qRow10, qRow11}, [data]!
+    VST1    {qRow4}, [luma@128], width
+    VLD1    {qRow12, qRow13}, [data]!
+    VST1    {qRow5}, [luma@128], width
+    VLD1    {qRow14, qRow15}, [data]!
+    VST1    {qRow6}, [luma@128], width
+
+    VLD1    {qRow0, qRow1}, [data]! ;cb rows 0,1,2,3
+    VST1    {qRow7}, [luma@128], width
+    VLD1    {qRow2, qRow3}, [data]! ;cb rows 4,5,6,7
+    VST1    {qRow8}, [luma@128], width
+    VLD1    {qRow4, qRow5}, [data]! ;cr rows 0,1,2,3
+    VST1    {qRow9}, [luma@128], width
+    VLD1    {qRow6, qRow7}, [data]! ;cr rows 4,5,6,7
+    VST1    {qRow10}, [luma@128], width
+    VST1    {dRow0}, [cb@64], cwidth
+    VST1    {dRow8}, [cr@64], cwidth
+    VST1    {qRow11}, [luma@128], width
+    VST1    {dRow1}, [cb@64], cwidth
+    VST1    {dRow9}, [cr@64], cwidth
+    VST1    {qRow12}, [luma@128], width
+    VST1    {dRow2}, [cb@64], cwidth
+    VST1    {dRow10}, [cr@64], cwidth
+    VST1    {qRow13}, [luma@128], width
+    VST1    {dRow3}, [cb@64], cwidth
+    VST1    {dRow11}, [cr@64], cwidth
+    VST1    {qRow14}, [luma@128], width
+    VST1    {dRow4}, [cb@64], cwidth
+    VST1    {dRow12}, [cr@64], cwidth
+    VST1    {qRow15}, [luma]
+    VST1    {dRow5}, [cb@64], cwidth
+    VST1    {dRow13}, [cr@64], cwidth
+    VST1    {dRow6}, [cb@64], cwidth
+    VST1    {dRow14}, [cr@64], cwidth
+    VST1    {dRow7}, [cb@64]
+    VST1    {dRow15}, [cr@64]
+
+    VPOP    {q4-q7}
+    POP     {r4-r6,pc}
+    END
+
+