[qsd8k] Update cpu initialization from 'legacy' bootloader
diff --git a/platform/qsd8k/arch_init.S b/platform/qsd8k/arch_init.S
old mode 100644
new mode 100755
index 508330a..4080e28
--- a/platform/qsd8k/arch_init.S
+++ b/platform/qsd8k/arch_init.S
@@ -61,169 +61,210 @@
 
 //;; Number of TCXO cycles to count ring oscillations
 .equ TCXO_CNT_VAL,     0x100
-        
-SET_SA: 
-        //; clear CLK_TEST[19] (RING_OSC_DBG_SEL) to count ring osc
-        LDR     r5, =CLK_TEST
-        LDR     r4, [r5]
-        BIC     r4, r4, #0x80000
-        STR     r4, [r5]
-        
-        //; setup memory pointers
-        LDR     r5, =CLK_CTL_BASE
-        LDR     r6, =SPSS_CSR_BASE
-        
-        //; PRPH_WEB_NS_REG = 0x00000A00 - enable root and crc
-        MOV     r7, #0x0A00
-        STR     r7,[r5,#A_PRPH_WEB_NS_REG]
 
-        //; MISC_CLK_CTL = 0x0 to disable tcxo4cnt, ringosccnt
-        MOV     r0, #0
-        STR     r0,[r5,#A_MISC_CLK_CTL]
-        DSB
-        
-        //;; MSM_CLK_RINGOSC = 0x00000A0E
-        //;; the E selects Scorpion LVT RO(or DS) output, 
-        //;; the A enables RO CRC and CXC
-        LDR     r7, =0x0A0E
-        STR     r7,[r5,#A_MSM_CLK_RINGOSC]
+//; Raptor addresses
+.equ TCSR_SPARE2,      0xA8700060
 
-        MOV     r10, #0x1 //;  SCRINGOSC[1:0] = 01
-        MOV     r11, #TCXO_CNT_VAL
-        
-        //;;  initialize r9 - counts the 4 oscillators
-        MOV     r9,#0x0
-        LDR     r4,=RO_CNT
+/*
+;=======================================================================
+;  Set_SA code to set cache settings based on ring oscillator/sense amps
+;=======================================================================
+*/
+SET_SA:
+        //; no stack at this point and any registers we use will be 0'd
+        //; after we return
+        LDR     r0, =TCSR_SPARE2
+        LDR     r1, [r0]
+        LDR     r0, = 0x010F
+        AND     r2, r1, r0   //; concerned with bits [8, 3:0]
 
-ACQUIRE_FREQ:
-        //; SCRINGOSC[1:0] is set by r10, [13:11][10:8] = 0..7
-        //;MOV     r7, r10
-        BIC     r7, r10, #0x3F00
-        //;BFI     r7, r9, #8, #3
-        ORR     r7, r7, r9, LSL #8
-        //;BFI     r7, r9, #11, #3
-        ORR     r7, r7, r9, LSL #11
-        STR     r7,[r6,#A_SCRINGOSC]
+        //;--------------------------------------------------------------------
+        //; Fuse bits used to determine sense amp settings
+        //;--------------------------------------------------------------------
 
-        STR     r11,[r5,#A_TCXO_CNT]
-        DSB     //; Make sure the countdown value gets written before the ROs are enabled
+        LDR     r0, = 0x0105
+        AND     r4, r2, r0    //;  mask off all but L1 ACC2, L1 ACC1 and L1 ACC0
+        //;set to default of FC00
+        LDR     r5, =PVR0F0_6bits    //; point to PVR0F0
+        LDR     r3, =PVR2F0_6bits    //; point to PVR2F0
 
-        //; MISC_CLK_CTL = 0x3 to enable tcxo4cnt, ringosccnt
-        MOV     r7,#0x3
-        STR     r7,[r5,#A_MISC_CLK_CTL]
+ck_0:
+        //; if L1_[2:0] == 000 then ACC setting = FC00
+        LDR     r1, = 0x0
+        CMP     r4, r1
+        BNE     ck_1
+        B       WRITE_L1_SA_SETTINGS
 
-        //;; According to Raptor HDD, to read TCXO_CNT_DONE, use address for TCXO_CNT on pass1.
-        //;; Then read RINGOSC_CNT and store to debug_mem table
-WAIT_TCXO_CNT_DONE:
-        LDR     r7,[r5,#A_TCXO_CNT_DONE]
-        CMP     r7,#0x1
-        BNE     WAIT_TCXO_CNT_DONE
-        LDR     r7,[r5,#A_RINGOSC_CNT]
-        STR     r7,[r4],#0x04  //;  post-indexed update r4 to next empty table entry
+ck_1:
+        //; if L1_[2:0] == 001 then ACC setting = FC00
+        LDR     r1, = 0x01
+        CMP     r4, r1
+        BNE     ck_2
+        B       WRITE_L1_SA_SETTINGS
 
-        //; MISC_CLK_CTL = 0x0 to disable tcxo4cnt, ringosccnt
-        STR     r0,[r5,#A_MISC_CLK_CTL]
-        DSB
-         
-        //;; move to next oscillator - r9 counts oscillators 0..7
-        ADD     r9,r9,#1
-        CMP     r9,#0x4
-        BLT     ACQUIRE_FREQ
-        
-        //; average the 4 counts
-        LDR     r4,=RO_CNT
-        LDR     r5, [r4], #0x4
-        LDR     r6, [r4], #0x4
-        ADD     r5, r5, r6
-        LDR     r6, [r4], #0x4
-        ADD     r5, r5, r6
-        LDR     r6, [r4], #0x4
-        ADD     r5, r5, r6
-        MOV     r5, r5, LSR #2
-        
-.equ SLOW_THRESHOLD,    0x0E10
-.equ NOM_THRESHOLD,     0x123A
-        
-        //; compare average to slow part threshold
-        LDR     r4, =SLOW_THRESHOLD
-        CMP     r5, r4
-        LDRLT   r4, =SLOW_SA
-        BLT     WRITE_SA
-        
-        //; compare average to nominal part threshold
-        LDR     r4, =NOM_THRESHOLD
-        CMP     r5, r4
-        LDRLT   r4, =NOM_SA
-        BLT     WRITE_SA
-        
-        //; must be fast part
-        LDR     r4, =FAST_SA
-        B       WRITE_SA
-        
-        
-SLOW_SA:
-		.word 0x38001C00 //; PVR0F0
-		.word 0x1C000000 //; PVR2F0
-		.word 0x00212102 //; L2VR3F1
-        
-NOM_SA:
-		.word 0x38001C00 //; PVR0F0
-		.word 0x1C000000 //; PVR2F0
-		.word 0x00212102 //; L2VR3F1
-        
-FAST_SA:
-		.word 0x38003C00 //; PVR0F0
-		.word 0x1C000000 //; PVR2F0
-		.word 0x00212102 //; L2VR3F1
-        
-RO_CNT:
-		.word 0x0
-		.word 0x0
-		.word 0x0
-		.word 0x0
-        
-WRITE_SA:
-        //; write the sense amp settings
-        
-        //; PVR0F0[15:10] = I$
-        LDR     r5, [r4], #4
+ck_2:
+        //; if L1_[2:0] == 010 then ACC setting = 7C00
+        LDR     r1, = 0x04
+        CMP     r4, r1
+        BNE     ck_3
+        LDR     r5, =PVR0F0_5bits    //; point to PVR0F0
+        LDR     r3, =PVR2F0_5bits    //; point to PVR2F0
+        B       WRITE_L1_SA_SETTINGS
+
+ck_3:
+        //; if L1_[2:0] == 011 then ACC setting = FC00
+        LDR     r1, = 0x05
+        CMP     r4, r1
+        BNE     ck_4
+        LDR     r5, =PVR0F0_6bits    //; point to PVR0F0
+        LDR     r3, =PVR2F0_6bits    //; point to PVR2F0
+        B       WRITE_L1_SA_SETTINGS
+
+ck_4:
+        //; if L1_[2:0] == 0100 then ACC setting = 3C00
+        LDR     r1, = 0x0100
+        CMP     r4, r1
+        BNE     ck_5
+        LDR     r5, =PVR0F0_4bits    //; point to PVR0F0
+        LDR     r3, =PVR2F0_4bits    //; point to PVR2F0
+        B       WRITE_L1_SA_SETTINGS
+
+ck_5:
+        //; if L1_[2:0] == 0101 then ACC setting = 0400
+        LDR     r1, = 0x0101
+        CMP     r4, r1
+        BNE     ck_6
+        LDR     r5, =PVR0F0_1bits    //; point to PVR0F0
+        LDR     r3, =PVR2F0_1bits    //; point to PVR2F0
+        B       WRITE_L1_SA_SETTINGS
+
+ck_6:
+        //; if L1_[2:0] == 0110 then ACC setting = 0C00
+        LDR     r1, = 0x0104
+        CMP     r4, r1
+        BNE     ck_7
+        LDR     r5, =PVR0F0_2bits    //; point to PVR0F0
+        LDR     r3, =PVR2F0_2bits    //; point to PVR2F0
+        B       WRITE_L1_SA_SETTINGS
+
+ck_7:
+        //; if L1_[2:0] == 0111 then ACC setting = 1C00
+        LDR     r1, = 0x0105
+        CMP     r4, r1
+        LDREQ   r5, =PVR0F0_3bits    //; point to PVR0F0
+        LDREQ   r3, =PVR2F0_3bits    //; point to PVR2F0
+
+WRITE_L1_SA_SETTINGS:
+
+        LDR     r5, [r5]
+        LDR     r3, [r3]
+
         //;WCP15_PVR0F0   r5
-        MCR     p15,0x0,r5,c15,c15,0
-        
-        //; PVR2F0[31:26] = D$
-        LDR     r5, [r4], #4
-        //;WCP15_PVR2F0   r5
-        MCR     p15,0x2,r5,c15,c15,0
-        
-        //; L2VR3F1[23:0]
-        LDR     r5, [r4], #4
-        //;WCP15_L2VR3F1  r5
-        MCR     p15,0x3,r5,c15,c15,1
-        
-        ISB
-        
-        BX      lr
+        MCR     p15,0x0,r5,c15,c15,0   //; write R5 to PVR0F0
+
+        //;WCP15_PVR2F0   r3
+        MCR     p15,0x2,r3,c15,c15,0   //; write R3 to PVR2F0
+
+        AND     r4, r2, #0x000A    //;  mask off all but L2 array SA settings
+        LDR     r5, =HVT_010102    //; point to L2VR3F1 setting
+                       //;it gets ovewritten if its one of the other two cases
+        //; if L2_1 and L2_0 == 0     ACC setting = 010102
+        LDR     r1, = 0x0000
+        CMP     r4, r1
+        BEQ     WRITE_L2_SA_SETTINGS
+
+        //; if L2_1 = 0 & L2_0 = 1    ACC setting = 010102
+        LDR     R1, = 0x0002
+        CMP     r4, r1
+        BEQ     WRITE_L2_SA_SETTINGS
+
+        //; if L2_1 = 1 & L2_0 = 0    ACC setting = 010101
+        LDR     r5, =HVT_010101
+        LDR     R1, = 0x0008
+        CMP     r4, r1
+        BEQ     WRITE_L2_SA_SETTINGS
+
+        //; else L2_1 = 1 & L2_0 = 1  ACC setting = 212102
+        LDR     r5, =HVT_212102
+
+
+WRITE_L2_SA_SETTINGS:
+        //;WCP15_L2VR3F1  r4
+        LDR     r5, [r5]
+        MCR     p15,0x3,r5,c15,c15,1     //;write r4 to L2VR3F1
+
+        LDR     r0, =0                   //;make sure the registers we touched
+        LDR     r1, =0                   //;are cleared when we return
+        LDR     r2, =0
+        LDR     r3, =0
+        LDR     r4, =0
+        LDR     r5, =0
+
+        //; routine complete
+        BX      LR
+
+//; L1 SA settings according to LVT speed
+PVR0F0_0bits:
+.word 0x38000000 //; PVR0F0
+PVR2F0_0bits:
+.word 0x00000000 //; PVR2F0  0 bits set
+
+PVR0F0_1bits:
+.word 0x38000400 //; PVR0F0
+PVR2F0_1bits:
+.word 0x04000000 //; PVR2F0  1 bits set
+
+PVR0F0_2bits:
+.word 0x38000C00 //; PVR0F0
+PVR2F0_2bits:
+.word 0x0C000000 //; PVR2F0  2 bits set
+
+PVR0F0_3bits:
+.word 0x38001C00 //; PVR0F0
+PVR2F0_3bits:
+.word 0x1C000000 //; PVR2F0  3 bits set
+
+PVR0F0_4bits:
+.word 0x38003C00 //; PVR0F0
+PVR2F0_4bits:
+.word 0x3C000000 //; PVR2F0  4 bits set
+
+PVR0F0_5bits:
+.word 0x38007C00 //; PVR0F0
+PVR2F0_5bits:
+.word 0x7C000000 //; PVR2F0  5 bits set
+
+PVR0F0_6bits:
+.word 0x3800FC00 //; PVR0F0
+PVR2F0_6bits:
+.word 0xFC000000 //; PVR2F0  6 bits set
+
+//; L2 SA settings according to HVT speed
+HVT_212102:
+.word 0x00212102 //; L2VR3F1
+
+HVT_010102:
+.word 0x00010102 //; L2VR3F1
+
+HVT_010101:
+.word 0x00010101 //; L2VR3F1
+.ltorg
+
 
 .globl __cpu_early_init
 __cpu_early_init:
-        //; Zero out r0 for use throughout this code. All other GPRs 
+		b SET_SA
+        //; Zero out r0 for use throughout this code. All other GPRs
         //; (r1-r3) are set throughout this code to help establish
         //; a consistent startup state for any code that follows.
         //; Users should add code at the end of this routine to establish
         //; their own stack address (r13), add translation page tables, enable
         //; the caches, etc.
-        MOV    r0,  #0x0                                      
+        MOV    r0,  #0x0
 
-        //; Write L2VR3F1 to make L2 array work properly at 998 on Raptor2
-        LDR    r1, =0x00212102 //; the l2 timing watch out on P2
-        MCR    p15, 3, r1, c15, c15, 1 //; WCP15_L2VR3F1
-        DSB
-        ISB
 
-        //; write the value for Icache to run at 998 on Raptor2
-        LDR    r1, =0x3800E000
-        MCR    p15, 0, r1, c15, c15, 0 //; WCP15_PVR0F0
-        ISB
+        //; Remove hardcoded cache settings. appsbl_handler.s calls Set_SA
+        //;   API to dynamically configure cache for slow/nominal/fast parts
 
         //; DCIALL to invalidate L2 cache bank (needs to be run 4 times, once per bank)
         //; This must be done early in code (prior to enabling the caches)
@@ -235,16 +276,18 @@
         MCR    p15, 0, r1, c9, c0, 6   //; DCIALL bank B ([15:14] == 2'b10)
         ADD    r1, r1, #0x00004000
         MCR    p15, 0, r1, c9, c0, 6   //; DCIALL bank A ([15:14] == 2'b11)
-       
+
         //; Initialize the BPCR - setup Global History Mask (GHRM) to all 1's
         //; and have all address bits (AM) participate.
         //; Different settings can be used to improve performance
-		.word 0xe30011ff // MOVW   r1, #0x01FF
-		.word 0xe34011ff // MOVT   r1, #0x01FF
+        // MOVW   r1, #0x01FF
+.word 0xe30011ff  // hardcoded MOVW instruction due to lack of compiler support
+        // MOVT   r1, #0x01FF
+.word 0xe34011ff  // hardcoded MOVT instruction due to lack of compiler support
         MCR    p15, 7, r1, c15, c0, 2   //; WCP15_BPCR
- 
-        
-        //; Initialize all I$ Victim Registers to 0 for startup 
+
+
+        //; Initialize all I$ Victim Registers to 0 for startup
         MCR    p15, 0, r0, c9, c1, 0    //; WCP15_ICVIC0    r0
         MCR    p15, 0, r0, c9, c1, 1    //; WCP15_ICVIC1    r0
         MCR    p15, 0, r0, c9, c1, 2    //; WCP15_ICVIC2    r0
@@ -283,21 +326,21 @@
         MCR    p15, 3, r0, c9, c1, 5    //; WCP15_DCFLOOR5  r0
         MCR    p15, 3, r0, c9, c1, 6    //; WCP15_DCFLOOR6  r0
         MCR    p15, 3, r0, c9, c1, 7    //; WCP15_DCFLOOR7  r0
-        
+
         //; Initialize ASID to zero
         MCR    p15, 0, r0, c13, c0, 1   //; WCP15_CONTEXTIDR r0
-       
+
         //; ICIALL to invalidate entire I-Cache
-        MCR    p15, 0, r0, c7, c5, 0    //; ICIALLU 
-               
+        MCR    p15, 0, r0, c7, c5, 0    //; ICIALLU
+
         //; DCIALL to invalidate entire D-Cache
         MCR    p15, 0, r0, c9, c0, 6    //; DCIALL  r0
 
 
         //; The VBAR (Vector Base Address Register) should be initialized
-        //; early in your code. We are setting it to zero 
+        //; early in your code. We are setting it to zero
         MCR    p15, 0, r0, c12, c0, 0   //; WCP15_VBAR  r0
-        
+
         //; Ensure the MCR's above have completed their operation before continuing
         DSB
         ISB
@@ -305,16 +348,16 @@
         //;-------------------------------------------------------------------
         //; There are a number of registers that must be set prior to enabling
         //; the MMU. The DCAR is one of these registers. We are setting
-        //; it to zero (no access) to easily detect improper setup in subsequent 
+        //; it to zero (no access) to easily detect improper setup in subsequent
         //; code sequences
         //;-------------------------------------------------------------------
         //; Setup DACR (Domain Access Control Register) to zero
         MCR    p15, 0, r0, c3, c0, 0    //; WCP15_DACR  r0
-        
-        //; Setup DCLKCR to allow normal D-Cache line fills 
+
+        //; Setup DCLKCR to allow normal D-Cache line fills
         MCR    p15, 1, r0, c9, c0, 7    //; WCP15_DCLKCR r0
-    
-        //; Setup the TLBLKCR 
+
+        //; Setup the TLBLKCR
         //; Victim = 6'b000000; Floor = 6'b000000;
         //; IASIDCFG = 2'b00 (State-Machine); IALLCFG = 2'b01 (Flash); BNA = 1'b0;
         MOV    r1, #0x02
@@ -328,19 +371,38 @@
 
         //; Make sure UTLB request has been presented to macro before continuing
         ISB
-        
+
+        // Disable predecode repair cache on certain Scorpion revisions
+        // (Raptor V2 and earlier, or Halcyon V1)
+        MRC    p15, 0, r1, c0, c0, 0      //; MIDR
+        BIC    r2, r1, #0xf7              //; check for Raptor2 or below
+        LDR    r3, =0x510f0000
+        CMP    r2, r3
+        BEQ    DPRC
+        BIC    r2, r1, #0xf0              //; check for Halcyon V1
+        LDR    r3, =0x511f0000
+        CMP    r2, r3
+        BNE    SYSI2
+
+DPRC:
+        MRC    p15, 0, r1, c15, c15, 2    //; PVR0F2
+        ORR    r1, r1, #0x10              //; enable bit 4
+        MCR    p15, 0, r1, c15, c15, 2    //; disable predecode repair cache
+
+SYSI2:
         //; setup L2CR1 to some default Instruction and data prefetching values
-        //; Users may want specific settings for various performance enhancements 
-        MCR    p15, 3, r0, c15, c0, 3     //; WCP15_L2CR1  r0
+        //; Users may want specific settings for various performance enhancements
+        //; In Halcyon we do not have broadcasting barriers. So we need to turn
+        //  ; on bit 8 of L2CR1; which DBB:( Disable barrier broadcast )
+        MOV r2, #0x100
+        MCR    p15, 3, r2, c15, c0, 3     //; WCP15_L2CR1  r0
 
 
-        //; Enable Z bit to enable branch prediction (default is off)   
- 
+        //; Enable Z bit to enable branch prediction (default is off)
         MRC    p15, 0, r2, c1, c0, 0      //; RCP15_SCTLR  r2
         ORR    r2, r2, #0x00000800
         MCR    p15, 0, r2, c1, c0, 0      //; WCP15_SCTLR  r2
 
-		mov r1, lr
         //; Make sure Link stack is initialized with branch and links to sequential addresses
         //; This aids in creating a predictable startup environment
        BL      SEQ1
@@ -351,57 +413,87 @@
 SEQ5:  BL      SEQ6
 SEQ6:  BL      SEQ7
 SEQ7:  BL      SEQ8
-SEQ8: 
-
-		mov lr, r1
+SEQ8:
 
         //; REMOVE FOLLOWING THREE INSTRUCTIONS WHEN POWER COLLAPSE IS ENA
         //;Make sure the DBGOSLSR[LOCK] bit is cleared to allow access to the debug registers
         //; Writing anything but the "secret code" to the DBGOSLAR clears the DBGOSLSR[LOCK] bit
         MCR    p14, 0, r0, c1, c0, 4       //; WCP14_DBGOSLAR r0
-        
-        
+
+
         //; Read the DBGPRSR to clear the DBGPRSR[STICKYPD]
         //; Any read to DBGPRSR clear the STICKYPD bit
-        //; ISB guarantees the read completes before attempting to 
+        //; ISB guarantees the read completes before attempting to
         //; execute a CP14 instruction.
         MRC    p14, 0, r3, c1, c5, 4       //; RCP14_DBGPRSR r3
         ISB
-        
+
         //; Initialize the Watchpoint Control Registers to zero (optional)
-//;;;        MCR    p14, 0, r0, c0, c0, 7       ; WCP14_DBGWCR0  r0
-//;;;        MCR    p14, 0, r0, c0, c1, 7       ; WCP14_DBGWCR1  r0
+        //;;; MCR    p14, 0, r0, c0, c0, 7       ; WCP14_DBGWCR0  r0
+        //;;; MCR    p14, 0, r0, c0, c1, 7       ; WCP14_DBGWCR1  r0
+
+
+        //;----------------------------------------------------------------------
+        //; The saved Program Status Registers (SPSRs) should be setup
+        //; prior to any automatic mode switches. The following
+        //; code sets these registers up to a known state. Users will need to
+        //; customize these settings to meet their needs.
+        //;----------------------------------------------------------------------
+        MOV    r2,  #0x1f
+        MOV    r1,  #0x17                 //;ABT mode
+        msr    cpsr_c, r1                 //;ABT mode
+        msr    spsr_cxfs, r2              //;clear the spsr
+        MOV    r1,  #0x1b                 //;UND mode
+        msr    cpsr_c, r1                 //;UND mode
+        msr    spsr_cxfs, r2              //;clear the spsr
+        MOV    r1,  #0x11                 //;FIQ mode
+        msr    cpsr_c, r1                 //;FIQ mode
+        msr    spsr_cxfs, r2              //;clear the spsr
+        MOV    r1,  #0x12                 //;IRQ mode
+        msr    cpsr_c, r1                 //;IRQ mode
+        msr    spsr_cxfs, r2              //;clear the spsr
+        MOV    r1,  #0x16                 //;Monitor mode
+        msr    cpsr_c, r1                 //;Monitor mode
+        msr    spsr_cxfs, r2              //;clear the spsr
+        MOV    r1,  #0x13                 //;SVC mode
+        msr    cpsr_c, r1                 //;SVC mode
+        msr    spsr_cxfs, r2              //;clear the spsr
+
 
         //;----------------------------------------------------------------------
         //; Enabling Error reporting is something users may want to do at
-        //; some other point in time. We have chosen some default settings  
+        //; some other point in time. We have chosen some default settings
         //; that should be reviewed. Most of these registers come up in an
-        //; unpredictable state after reset.   
+        //; unpredictable state after reset.
         //;----------------------------------------------------------------------
 //;Start of error and control setting
 
         //; setup L2CR0 with various L2/TCM control settings
         //; enable out of order bus attributes and error reporting
-        //; this register comes up unpredictable after reset 
-		.word 0xe3001f0f // MOVW   r1, #0x0F0F
-		.word 0xe34c1005 // MOVT   r1, #0xC005
-
-        MCR    p15, 3, r1, c15, c0, 1      //; WCP15_L2CR0  r1 
+        //; this register comes up unpredictable after reset
+        // MOVW   r1, #0x0F0F
+.word 0xe3001f0f  // hardcoded MOVW instruction due to lack of compiler support
+        // MOVT   r1, #0xC005
+.word 0xe34c1005  // hardcoded MOVW instruction due to lack of compiler support
+        MCR    p15, 3, r1, c15, c0, 1    //; WCP15_L2CR0  r1
 
         //; setup L2CPUCR
-        MOV    r2, #0
-        MCR    p15, 3, r2, c15, c0, 2       //; WCP15_L2CPUCR  r2
+        //; MOV    r2, #0xFF
+        //; Enable I and D cache parity
+        //;L2CPUCR[7:5] = 3~Rh7 ~V enable parity error reporting for modified,
+        //;tag, and data parity errors
+        MOV    r2, #0xe0
+        MCR    p15, 3, r2, c15, c0, 2    //; WCP15_L2CPUCR  r2
 
         //; setup SPCR
         //; enable all error reporting (reset value is unpredicatble for most bits)
         MOV    r3, #0x0F
-        MCR    p15, 0, r3, c9, c7, 0        //; WCP15_SPCR  r3
+        MCR    p15, 0, r3, c9, c7, 0     //; WCP15_SPCR  r3
 
-        
         //; setup DMACHCRs (reset value unpredictable)
         //; control setting and enable all error reporting
         MOV   r1, #0x0F
-        
+
         //; DMACHCR0 = 0000000F
         MOV   r2, #0x00                  //; channel 0
         MCR   p15, 0, r2, c11, c0, 0     //; WCP15_DMASELR  r2
@@ -424,22 +516,36 @@
 
         //; Set ACTLR (reset unpredictable)
         //; Set AVIVT control, error reporting, etc.
-        MOV   r3, #0x04
-
-        //; keep d-cache disabled on Raptor Rev 1.0 silicon
-        //; orr   r3, r3, #0x4000 //;disable dcache
+        //; MOV   r3, #0x07
+        //; Enable I and D cache parity
+        //;ACTLR[2:0] = 3'h7 - enable parity error reporting from L2/I$/D$)
+        //;ACTLR[5:4] = 2'h3 - enable parity
+        //;ACTLR[19:18] =2'h3 - always generate and check parity(when MMU disabled).
+        //;Value to be written #0xC0037
+        // MOVW   r3, #0x0037
+.word 0xe3003037  // hardcoded MOVW instruction due to lack of compiler support
+        // MOVT   r3, #0x000C
+.word 0xe340300c  // hardcoded MOVW instruction due to lack of compiler support
+            //; read the version_id to determine if d-cache should be disabled
+            LDR r2, = 0xa8e00270  //;Read HW_REVISION_NUMBER, HWIO_HW_REVISION_NUMBER_ADDR
+            LDR r2,[r2]
+            AND r2,r2,#0xf0000000 //;hw_revision mask off bits 28-31
+                //;if HW_revision is 1.0 or older, (revision==0)
+                CMP r2,#0
+        //; Disable d-cache on older QSD8650 (Rev 1.0) silicon
+        orreq   r3, r3, #0x4000          //;disable dcache
         MCR   p15, 0, r3, c1, c0, 1      //; WCP15_ACTLR  r3
 
 //;End of error and control setting
-         
+
         //;----------------------------------------------------------------------
         //; Unlock ETM and read StickyPD to halt the ETM clocks from running.
         //; This is required for power saving whether the ETM is used or not.
         //;----------------------------------------------------------------------
-        
+
         //;Clear ETMOSLSR[LOCK] bit
         MOV   r1, #0x00000000
-        MCR   p14, 1, r1, c1, c0, 4        //; WCP14_ETMOSLAR       r1
+        MCR   p14, 1, r1, c1, c0, 4        //; WCP14_ETMOSLAR      r1
 
         //;Clear ETMPDSR[STICKYPD] bit
         MRC   p14, 1, r2, c1, c5, 4        //; RCP14_ETMPDSR       r2
@@ -448,11 +554,11 @@
 #ifdef APPSBL_ETM_ENABLE
         ;----------------------------------------------------------------------
         ; Optionally Enable the ETM (Embedded Trace Macro) which is used for debug
-        ;---------------------------------------------------------------------- 
-         
+        ;----------------------------------------------------------------------
+
         ; enable ETM clock if disabled
         MRC   p15, 7, r1, c15, c0, 5       ; RCP15_CPMR           r1
-        ORR   r1, r1, #0x00000008 
+        ORR   r1, r1, #0x00000008
         MCR   p15, 7, r1, c15, c0, 5       ; WCP15_CPMR           r1
         ISB
 
@@ -486,7 +592,7 @@
         ; clear view data address range comparator usage and exclude nothing
         MOV   r2, #0x00010000
         MCR   p14, 1, r2, c0, c15, 0       ;  WCP14_ETMVDCR3       r2
-        
+
         ; set counter1 to 194
         MOV   r2, #0x000000C2
         MCR   p14, 1, r2, c0, c0, 5        ;  WCP14_ETMCNTRLDVR1   r2
@@ -508,18 +614,18 @@
         ;  - Set the ETM to perform data address tracing
         MOV   r2, #0x00002008
         MCR   p14, 1, r2, c0, c0, 0        ; WCP14_ETMCR          r2
-        ISB        
+        ISB
 #endif *//* APPSBL_ETM_ENABLE */
 
 /*
 #ifdef APPSBL_VFP_ENABLE
        ;----------------------------------------------------------------------
-       ; Perform the following operations if you intend to make use of 
-       ; the VFP/Neon unit. Note that the FMXR instruction requires a CPU ID 
+       ; Perform the following operations if you intend to make use of
+       ; the VFP/Neon unit. Note that the FMXR instruction requires a CPU ID
        ; indicating the VFP unit is present (i.e.Cortex-A8). .
        ; Some tools will require full double precision floating point support
        ; which will become available in Scorpion pass 2
-       ;---------------------------------------------------------------------- 
+       ;----------------------------------------------------------------------
        ; allow full access to CP 10 and 11 space for VFP/NEON use
         MRC   p15, 0, r1, c1, c0, 2        ; Read CP Access Control Register
         ORR   r1, r1, #0x00F00000          ; enable full access for p10,11
@@ -527,14 +633,16 @@
 
         ;make sure the CPACR is complete before continuing
         ISB
-        
+
        ; Enable VFP itself (certain OSes may want to dynamically set/clear
-       ; the enable bit based on the application being executed 
+       ; the enable bit based on the application being executed
         MOV   r1, #0x40000000
         FMXR  FPEXC, r1
 #endif *//* APPSBL_VFP_ENABLE */
 
-		/* we have no stack, so just tail-call into the SET_SA routine... */
-		b SET_SA
+/*
+   MEND
+
+#endif *//* _ARM_ASM_ */
 
 .ltorg