This is to fix security issues found in Openmax SC16 code.
Verified with SNR testing code in Openmax folder.
R=aedla@chromium.org, rtoy@google.com
Review URL: https://webrtc-codereview.appspot.com/1715004
git-svn-id: http://webrtc.googlecode.com/svn/deps/third_party/openmax@4353 4adac7df-926f-26a2-2b94-8c16560cd09d
diff --git a/dl/sp/api/omxSP.h b/dl/sp/api/omxSP.h
index 40abe38..13c64e3 100644
--- a/dl/sp/api/omxSP.h
+++ b/dl/sp/api/omxSP.h
@@ -1887,8 +1887,7 @@
* must be aligned on a 32-byte boundary.
* pFFTSpec - pointer to the preallocated and initialized specification
* structure
- * scaleFactor - scale factor of the output. Valid value is 0
- * only.
+ * scaleFactor - scale factor of the output. Valid range is [0,16].
*
* Output Arguments:
* order
diff --git a/dl/sp/src/armSP_FFT_CToC_SC16_Radix4_fs_unsafe_s.S b/dl/sp/src/armSP_FFT_CToC_SC16_Radix4_fs_unsafe_s.S
index cdb42a9..1b0217e 100644
--- a/dl/sp/src/armSP_FFT_CToC_SC16_Radix4_fs_unsafe_s.S
+++ b/dl/sp/src/armSP_FFT_CToC_SC16_Radix4_fs_unsafe_s.S
@@ -142,7 +142,6 @@
RSB setStep,setStep,#16 @// setStep = - 3*pointStep+16
- VLD2 {dXr3,dXi3},[pSrc :128],setStep @// data[3]
MOV subFFTSize,#4 @// subFFTSize = 1 for the first stage
@@ -158,6 +157,7 @@
grpZeroSetLoop\name:
+ VLD2 {dXr3,dXi3},[pSrc :128],setStep @// data[3]
.ifeqs "\scaled", "TRUE"
@@ -178,9 +178,6 @@
VLD2 {dXr1,dXi1},[pSrc :128],step1 @// data[1]
VHADD qZ0,qY0,qY1 @// y0
- VLD2 {dXr3,dXi3},[pSrc :128],setStep
-
-
.ifeqs "\inverse", "TRUE"
VHSUB dZr3,dYr2,dYi3 @// y3
@@ -235,9 +232,6 @@
VLD2 {dXr1,dXi1},[pSrc :128],step1 @// data[1]
VADD qZ0,qY0,qY1 @// y0
- VLD2 {dXr3,dXi3},[pSrc :128],setStep
-
-
.ifeqs "\inverse", "TRUE"
VSUB dZr3,dYr2,dYi3 @// y3
diff --git a/dl/sp/src/armSP_FFT_CToC_SC16_Radix4_ls_unsafe_s.S b/dl/sp/src/armSP_FFT_CToC_SC16_Radix4_ls_unsafe_s.S
index 23e2c37..c5ef7a0 100644
--- a/dl/sp/src/armSP_FFT_CToC_SC16_Radix4_ls_unsafe_s.S
+++ b/dl/sp/src/armSP_FFT_CToC_SC16_Radix4_ls_unsafe_s.S
@@ -163,7 +163,6 @@
@// Define stack arguments
MOV pw2,pTwiddle
- VLD4 {dW2r,dW2i,dTmp0,dTmp1},[pw2 :256]!
MOV pw3,pTwiddle
MOV pw1,pTwiddle
@@ -171,43 +170,48 @@
@// pOut0+outPointStep == increment of 4*outPointStep bytes
MOV outPointStep,subFFTSize,LSL #2
- VLD3 {dW3rS32,dTmp1S32,dTmp2S32},[pw3 :64]!
MOV subFFTNum,#1 @//after the last stage
LSL grpCount,subFFTSize,#2
@// Update grpCount and grpSize rightaway
- VLD3 {dW3iS32,dTmp2S32,dTmp3S32},[pw3 :64]!
@// update subFFTSize for the next stage
MOV subFFTSize,grpCount
MOV dstStep,outPointStep,LSL #1
- VLD2 {dW1r,dW1i}, [pw1 :128]!
-
-
ADD dstStep,dstStep,outPointStep @// dstStep = 3*outPointStep
RSB dstStep,dstStep,#16 @// dstStep = - 3*outPointStep+16
- VLD4 {dButterfly1Real02,dButterfly1Imag02,dButterfly1Real13,dButterfly1Imag13},[pSrc :256]! @// AC.r AC.i BD.r BD.i
- VLD4 {dButterfly2Real02,dButterfly2Imag02,dButterfly2Real13,dButterfly2Imag13},[pSrc :256]! @// AC.r AC.i BD.r BD.i
-
@// Process 4 groups at a time
grpLoop\name:
+ VLD4 {dButterfly1Real02,dButterfly1Imag02,dButterfly1Real13,dButterfly1Imag13},[pSrc :256]! @// AC.r AC.i BD.r BD.i
+ VLD4 {dButterfly2Real02,dButterfly2Imag02,dButterfly2Real13,dButterfly2Imag13},[pSrc :256]! @// AC.r AC.i BD.r BD.i
+ @// Load the second twiddle for 4 groups : w^2
+ @// w^2 twiddle (2i+0,2i+2,2i+4,2i+6) for group 0,1,2,3
+ VLD4 {dW2r,dW2i,dTmp0,dTmp1},[pw2 :256]!
+
+ VUZP dButterfly1Real13, dButterfly2Real13 @// B.r D.r
+
+ @// Load the third twiddle for 4 groups : w^3
+ @// w^3 twiddle (3i+0,3i+3,3i+6,3i+9) for group 0,1,2,3
+ VLD3 {dW3rS32,dTmp1S32,dTmp2S32},[pw3 :64]!
+
+ VUZP dButterfly1Imag13, dButterfly2Imag13 @// B.i D.i
+ VUZP dButterfly1Real02, dButterfly2Real02 @// A.r C.r
+
+ VLD3 {dW3iS32,dTmp2S32,dTmp3S32},[pw3 :64]!
+
+ VUZP dButterfly1Imag02, dButterfly2Imag02 @// A.i C.i
+
+ VLD2 {dW1r,dW1i}, [pw1 :128]!
@// Rearrange the third twiddle
VUZP dW3r,dW3i
SUBS grpCount,grpCount,#16 @// grpCount is multiplied by 4
-
- VUZP dButterfly1Real13, dButterfly2Real13 @// B.r D.r
- VUZP dButterfly1Imag13, dButterfly2Imag13 @// B.i D.i
- VUZP dButterfly1Real02, dButterfly2Real02 @// A.r C.r
- VUZP dButterfly1Imag02, dButterfly2Imag02 @// A.i C.i
-
-
.ifeqs "\inverse", "TRUE"
VMULL qT0,dXr1,dW1r
VMLAL qT0,dXi1,dW1i @// real part
@@ -225,8 +229,6 @@
@// Load the first twiddle for 4 groups : w^1
@// w^1 twiddle (i+0,i+1,i+2,i+3) for group 0,1,2,3
- VLD2 {dW1r,dW1i}, [pw1 :128]!
-
.ifeqs "\inverse", "TRUE"
VMULL qT2,dXr2,dW2r
VMLAL qT2,dXi2,dW2i @// real part
@@ -260,24 +262,12 @@
.ENDIF
- @// Load the second twiddle for 4 groups : w^2
- @// w^2 twiddle (2i+0,2i+2,2i+4,2i+6) for group 0,1,2,3
- VLD4 {dW2r,dW2i,dTmp0,dTmp1},[pw2 :256]!
-
-
VRSHRN dZr2,qT2,#15
VRSHRN dZi2,qT3,#15
- @// Load the third twiddle for 4 groups : w^3
- @// w^3 twiddle (3i+0,3i+3,3i+6,3i+9) for group 0,1,2,3
-
- VLD3 {dW3rS32,dTmp1S32,dTmp2S32},[pw3 :64]!
-
VRSHRN dZr3,qT0,#15
VRSHRN dZi3,qT1,#15
- VLD3 {dW3iS32,dTmp2S32,dTmp3S32},[pw3 :64]!
-
.ifeqs "\scaled", "TRUE"
@// finish first stage of 4 point FFT
@@ -285,7 +275,6 @@
VHADD qY0,qX0,qZ2
VHSUB qY2,qX0,qZ2
VHADD qY1,qZ1,qZ3
- VLD4 {dButterfly1Real02,dButterfly1Imag02,dButterfly1Real13,dButterfly1Imag13},[pSrc :256]! @// AC.r AC.i BD.r BD.i
VHSUB qY3,qZ1,qZ3
@@ -293,7 +282,6 @@
VHSUB qZ0,qY2,qY1
VHADD qZ2,qY2,qY1
- VLD4 {dButterfly2Real02,dButterfly2Imag02,dButterfly2Real13,dButterfly2Imag13},[pSrc :256]! @// AC.r AC.i BD.r BD.i
.ifeqs "\inverse", "TRUE"
@@ -329,7 +317,6 @@
VADD qY0,qX0,qZ2
VSUB qY2,qX0,qZ2
VADD qY1,qZ1,qZ3
- VLD4 {dButterfly1Real02,dButterfly1Imag02,dButterfly1Real13,dButterfly1Imag13},[pSrc :256]! @// AC.r AC.i BD.r BD.i
VSUB qY3,qZ1,qZ3
@@ -337,7 +324,6 @@
VSUB qZ0,qY2,qY1
VADD qZ2,qY2,qY1
- VLD4 {dButterfly2Real02,dButterfly2Imag02,dButterfly2Real13,dButterfly2Imag13},[pSrc :256]! @// AC.r AC.i BD.r BD.i
.ifeqs "\inverse", "TRUE"
@@ -376,7 +362,6 @@
@// Reset and Swap pSrc and pDst for the next stage
MOV pTmp,pDst
- SUB pSrc,pSrc,#64 @// Extra increment currently done in the loop
SUB pDst,pSrc,outPointStep,LSL #2 @// pDst -= size; pSrc -= 4*size bytes
SUB pSrc,pTmp,outPointStep
diff --git a/dl/sp/src/armSP_FFT_CToC_SC16_Radix4_unsafe_s.S b/dl/sp/src/armSP_FFT_CToC_SC16_Radix4_unsafe_s.S
index 0eba385..0cb4701 100644
--- a/dl/sp/src/armSP_FFT_CToC_SC16_Radix4_unsafe_s.S
+++ b/dl/sp/src/armSP_FFT_CToC_SC16_Radix4_unsafe_s.S
@@ -154,7 +154,6 @@
MOV srcStep,pointStep,LSL #1 @// srcStep = 2*pointStep
VLD1 dW2,[pTwiddle :64] @//[wi | wr]
ADD setStep,srcStep,pointStep @// setStep = 3*pointStep
- SUB srcStep,srcStep,#16 @// srcStep = 2*pointStep-16
VLD1 dW3,[pTwiddle :64]
@//RSB setStep,setStep,#16 @// setStep = - 3*pointStep+16
RSB setStep,setStep,#0 @// setStep = - 3*pointStep
@@ -167,26 +166,23 @@
grpLoop\name:
- VLD2 {dXr0,dXi0},[pSrc :128],pointStep @// data[0]
ADD stepTwiddle,stepTwiddle,pointStep
- VLD2 {dXr1,dXi1},[pSrc :128],pointStep @// data[1]
ADD pTwiddle,pTwiddle,stepTwiddle @// set pTwiddle to the first point
- VLD2 {dXr2,dXi2},[pSrc :128],pointStep @// data[2]
MOV twStep,stepTwiddle,LSL #2
- VLD2 {dXr3,dXi3},[pSrc :128],setStep @// data[3] & reset pSrc
SUB twStep,stepTwiddle,twStep @// twStep = -3*stepTwiddle
MOV setCount,pointStep,LSR #2
- ADD pSrc,pSrc,#16 @// set pSrc to data[0] of the next set
- ADD pSrc,pSrc,pointStep @// increment to data[1] of the next set
+ ADD pSrc,pSrc,pointStep @// increment to data[1] of the next set
@// Loop on the sets : 4 at a time
setLoop\name:
+ VLD2 {dXr1,dXi1},[pSrc :128],pointStep @// data[1]
+ VLD2 {dXr2,dXi2},[pSrc :128],pointStep @// data[2]
- SUBS setCount,setCount,#4 @// decrement the loop counter
+ SUBS setCount,setCount,#4 @// decrement the loop counter
.ifeqs "\inverse", "TRUE"
VMULL qT0,dXr1,dW1[0]
@@ -202,8 +198,6 @@
.ENDIF
- VLD2 {dXr1,dXi1},[pSrc :128],pointStep @// data[1]
-
.ifeqs "\inverse", "TRUE"
VMULL qT2,dXr2,dW2[0]
VMLAL qT2,dXi2,dW2[1] @// real part
@@ -218,11 +212,13 @@
.ENDIF
+ VLD2 {dXr3,dXi3},[pSrc :128],setStep @// data[3] & update pSrc for the next set
+
VRSHRN dZr1,qT0,#15
VRSHRN dZi1,qT1,#15
-
- VLD2 {dXr2,dXi2},[pSrc :128],pointStep @// data[2]
+ VLD2 {dXr0,dXi0},[pSrc :128],pointStep @// data[0]
+ ADD pSrc,pSrc,#16 @// set pSrc to data[1] of the next set
.ifeqs "\inverse", "TRUE"
VMULL qT0,dXr3,dW3[0]
@@ -244,7 +240,6 @@
VRSHRN dZr3,qT0,#15
VRSHRN dZi3,qT1,#15
- VLD2 {dXr3,dXi3},[pSrc :128],setStep @// data[3] & update pSrc for the next set
.ifeqs "\scaled", "TRUE"
@@ -253,7 +248,6 @@
VHADD qY0,qX0,qZ2
VHSUB qY2,qX0,qZ2
- VLD2 {dXr0,dXi0},[pSrc :128]! @// data[0]
VHADD qY1,qZ1,qZ3
VHSUB qY3,qZ1,qZ3
@@ -303,7 +297,6 @@
VADD qY0,qX0,qZ2
VSUB qY2,qX0,qZ2
- VLD2 {dXr0,dXi0},[pSrc]! @// data[0]
VADD qY1,qZ1,qZ3
VSUB qY3,qZ1,qZ3
@@ -351,7 +344,6 @@
.ENDIF
- ADD pSrc,pSrc,pointStep @// increment to data[1] of the next set
BGT setLoop\name
VLD1 dW1,[pTwiddle :64],stepTwiddle @//[wi | wr]
diff --git a/dl/sp/src/armSP_FFT_CToC_SC16_Radix8_fs_unsafe_s.S b/dl/sp/src/armSP_FFT_CToC_SC16_Radix8_fs_unsafe_s.S
index 588c319..ba7c344 100644
--- a/dl/sp/src/armSP_FFT_CToC_SC16_Radix8_fs_unsafe_s.S
+++ b/dl/sp/src/armSP_FFT_CToC_SC16_Radix8_fs_unsafe_s.S
@@ -233,12 +233,12 @@
VLD2 {dXr4,dXi4},[pSrc :128],pointStep @// data[4]
VLD2 {dXr5,dXi5},[pSrc :128],pointStep @// data[5]
VLD2 {dXr6,dXi6},[pSrc :128],pointStep @// data[6]
- VLD2 {dXr7,dXi7},[pSrc :128],setStep @// data[7] & update pSrc for the next set
- @// setStep = -7*pointStep + 16
@// grp = 0 a special case since all the twiddle factors are 1
@// Loop on the sets : 4 sets at a time
grpZeroSetLoop\name:
+ VLD2 {dXr7,dXi7},[pSrc :128],setStep @// data[7] & update pSrc for the next set
+ @// setStep = -7*pointStep + 16
@// Decrement setcount
SUBS setCount,setCount,#4 @// decrement the set loop counter
@@ -348,9 +348,6 @@
VSUB dVi7,dVi7,dT1
SUB pDst, pDst, step2 @// set pDst to y1
- VLD2 {dXr7,dXi7},[pSrc :128],setStep @// data[7]
-
-
VHSUB dYr3,dVr3,dVr7
VHSUB dYi3,dVi3,dVi7
VST2 {dYr1,dYi1},[pDst :128],step1 @// store y1
@@ -388,7 +385,6 @@
VSUB dVr5,dT1,dVi5 @// a * V5
VADD dVi5,dT1,dVi5
- VLD2 {dXr7,dXi7},[pSrc :128],setStep @// data[7]
VHSUB qY5,qV1,qV5
@@ -514,9 +510,6 @@
VSUB dVi7,dVi7,dT1
SUB pDst, pDst, step2 @// set pDst to y1
- VLD2 {dXr7,dXi7},[pSrc :128],setStep @// data[7]
-
-
VSUB dYr3,dVr3,dVr7
VSUB dYi3,dVi3,dVi7
VST2 {dYr1,dYi1},[pDst :128],step1 @// store y1
@@ -554,7 +547,6 @@
VSUB dVr5,dT1,dVi5 @// a * V5
VADD dVi5,dT1,dVi5
- VLD2 {dXr7,dXi7},[pSrc :128],setStep @// data[7]
VSUB qY5,qV1,qV5
