Initial commit of OpenSPARC T2 architecture model.

[OpenSPARC-T2-SAM] / sam-t2 / sam / cpus / vonk / ss / lib / cpu / src / SS_FastMemory.h

/*
* ========== Copyright Header Begin ==========================================
* 
* OpenSPARC T2 Processor File: SS_FastMemory.h
* Copyright (c) 2006 Sun Microsystems, Inc.  All Rights Reserved.
* DO NOT ALTER OR REMOVE COPYRIGHT NOTICES.
* 
* The above named program is free software; you can redistribute it and/or
* modify it under the terms of the GNU General Public
* License version 2 as published by the Free Software Foundation.
* 
* The above named program is distributed in the hope that it will be 
* useful, but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
* General Public License for more details.
* 
* You should have received a copy of the GNU General Public
* License along with this work; if not, write to the Free Software
* Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301, USA.
* 
* ========== Copyright Header End ============================================
*/
#ifndef __SS_FastMemory_h__
#define __SS_FastMemory_h__

#ifdef COMPILE_FOR_SAM

// See elsewhere

#else // Vonk's own memory

#include "SS_Types.h"

extern "C" uint8_t  ss_ldstub(              void* base, uint64_t ofs );
extern "C" uint32_t ss_swap  ( uint32_t rd, void* base, uint64_t ofs );
extern "C" uint32_t ss_cas   ( uint32_t rd, void* base, uint32_t rs2 );
extern "C" uint64_t ss_casx  ( uint64_t rd, void* base, uint64_t rs2 );
extern "C" void     ss_stp8  ( double   rd, void* base, uint64_t mask );

#include "utils.h"

#include "SS_Ld128Atomic.h"

#include <string.h>
#include <sys/mman.h>
#include "BL_Mutex.h"
#include "BL_Memory.h"

class SS_FastMemory : public BL_Memory
{
 public:
   SS_FastMemory();
   ~SS_FastMemory();

   void allocate( uint64_t _ram_size, uint64_t _rom_size, uint_t pa_bits );

   void load( const char* filename );
   void load( const char* filename, uint64_t addr );
   void save( const char* filename, uint64_t addr, uint64_t size );

   // Supported User Interface Operations
   
   void     poke8(   uint64_t addr, uint8_t  data )            { SS_FastMemory::st8(addr,data); }
   void     poke16(  uint64_t addr, uint16_t data )            { SS_FastMemory::st16(addr,data); }
   void     poke32(  uint64_t addr, uint32_t data )            { SS_FastMemory::st32(addr,data); }
   void     poke64(  uint64_t addr, uint64_t data )            { SS_FastMemory::st64(addr,data); }
   uint8_t  peek8u(  uint64_t addr )                           { return SS_FastMemory::ld8u(addr); }
   int8_t   peek8s(  uint64_t addr )                           { return SS_FastMemory::ld8s(addr); }
   uint16_t peek16u( uint64_t addr )                           { return SS_FastMemory::ld16u(addr); }
   int16_t  peek16s( uint64_t addr )                           { return SS_FastMemory::ld16s(addr); }
   uint32_t peek32u( uint64_t addr )                           { return SS_FastMemory::ld32u(addr); }
   int32_t  peek32s( uint64_t addr )                           { return SS_FastMemory::ld32s(addr); }
   uint64_t peek64(  uint64_t addr )                           { return SS_FastMemory::ld64(addr); }

   void peek128( uint64_t addr ,uint64_t data[2] )
   { 
     uint8_t* ptr = get_ptr(addr);
     data[0] = *(uint64_t*)(ptr +  0);
     data[1] = *(uint64_t*)(ptr +  8);
#if defined(ARCH_X64)
     data[0] = ss_byteswap64(data[0]);
     data[1] = ss_byteswap64(data[1]);
#endif
   }

   void peek256( uint64_t addr ,uint64_t data[4] )
   { 
     uint8_t* ptr = get_ptr(addr);
     data[0] = *(uint64_t*)(ptr +  0);
     data[1] = *(uint64_t*)(ptr +  8);
     data[2] = *(uint64_t*)(ptr +  16);
     data[3] = *(uint64_t*)(ptr +  24);
#if defined(ARCH_X64)
     data[0] = ss_byteswap64(data[0]);
     data[1] = ss_byteswap64(data[1]);
     data[2] = ss_byteswap64(data[2]);
     data[3] = ss_byteswap64(data[3]);
#endif
   }

   void peek512( uint64_t addr ,uint64_t data[8] )
   { 
     uint8_t* ptr = get_ptr(addr);
     data[0] = *(uint64_t*)(ptr +  0);
     data[1] = *(uint64_t*)(ptr +  8);
     data[2] = *(uint64_t*)(ptr +  16);
     data[3] = *(uint64_t*)(ptr +  24);
     data[4] = *(uint64_t*)(ptr +  32);
     data[5] = *(uint64_t*)(ptr +  40);
     data[6] = *(uint64_t*)(ptr +  48);
     data[7] = *(uint64_t*)(ptr +  56);
#if defined(ARCH_X64)
     data[0] = ss_byteswap64(data[0]);
     data[1] = ss_byteswap64(data[1]);
     data[2] = ss_byteswap64(data[2]);
     data[3] = ss_byteswap64(data[3]);
     data[4] = ss_byteswap64(data[4]);
     data[5] = ss_byteswap64(data[5]);
     data[6] = ss_byteswap64(data[6]);
     data[7] = ss_byteswap64(data[7]);
#endif
   }

   void poke128( uint64_t addr, uint64_t data[2] )
   { 
#if defined(ARCH_X64)
     data[0] = ss_byteswap64(data[0]);
     data[1] = ss_byteswap64(data[1]);
#endif
     uint8_t* ptr = get_ptr(addr);
     *(uint64_t*)(ptr +  0) = data[0]; 
     *(uint64_t*)(ptr +  8) = data[1]; 
   }

   void  poke512( uint64_t addr, uint64_t data[8] )
   { 
#if defined(ARCH_X64)
     data[0] = ss_byteswap64(data[0]);
     data[1] = ss_byteswap64(data[1]);
     data[2] = ss_byteswap64(data[2]);
     data[3] = ss_byteswap64(data[3]);
     data[4] = ss_byteswap64(data[4]);
     data[5] = ss_byteswap64(data[5]);
     data[6] = ss_byteswap64(data[6]);
     data[7] = ss_byteswap64(data[7]);
#endif
     uint8_t* ptr = get_ptr(addr);
     *(uint64_t*)(ptr +  0) = data[0]; 
     *(uint64_t*)(ptr +  8) = data[1]; 
     *(uint64_t*)(ptr + 16) = data[2]; 
     *(uint64_t*)(ptr + 24) = data[3]; 
     *(uint64_t*)(ptr + 32) = data[4]; 
     *(uint64_t*)(ptr + 40) = data[5]; 
     *(uint64_t*)(ptr + 48) = data[6]; 
     *(uint64_t*)(ptr + 56) = data[7]; 
   }


   // Supported Fetch Operation (instruction fetch) 
  
   virtual uint32_t fetch32( uint64_t addr )                           { return SS_FastMemory::ld32u(addr); }
   virtual void     fetch256( uint64_t addr, uint64_t data[4] )        { SS_FastMemory::ld256(addr,data); }
   virtual void     fetch512( uint64_t addr, uint64_t data[8] )        { SS_FastMemory::ld512(addr,data); }
   
   // Supported Store Operations. st8(), st16(), st32() and st64() are gueranteed to be atomic.
   // st128() and st512() are atomic per 64bit quantity.
  
   virtual void st8( uint64_t addr, uint8_t data )
   { 
     *(uint8_t*)(get_ptr(addr)) = data; 
   }
   virtual void st16( uint64_t addr, uint16_t data )
   { 
#if defined(ARCH_X64)
     data = ss_byteswap16(data);
#endif
     *(uint16_t*)(get_ptr(addr)) = data; 
   }
   virtual void st32( uint64_t addr, uint32_t data )
   { 
#if defined(ARCH_X64)
     data = ss_byteswap32(data);
#endif
     *(uint32_t*)(get_ptr(addr)) = data; 
   }
   virtual void st64( uint64_t addr, uint64_t data )
   { 
#if defined(ARCH_X64)
     data = ss_byteswap64(data);
#endif
     *(uint64_t*)(get_ptr(addr)) = data; 
   }
   virtual void st128( uint64_t addr, uint64_t data[2] )
   { 
#if defined(ARCH_X64)
     data[0] = ss_byteswap64(data[0]);
     data[1] = ss_byteswap64(data[1]);
#endif
     uint8_t* ptr = get_ptr(addr);
     *(uint64_t*)(ptr +  0) = data[0];
     *(uint64_t*)(ptr +  8) = data[1]; 
   }
   virtual void st512( uint64_t addr, uint64_t data[8] )
   { 
#if defined(ARCH_X64)
     data[0] = ss_byteswap64(data[0]);
     data[1] = ss_byteswap64(data[1]);
     data[2] = ss_byteswap64(data[2]);
     data[3] = ss_byteswap64(data[3]);
     data[4] = ss_byteswap64(data[4]);
     data[5] = ss_byteswap64(data[5]);
     data[6] = ss_byteswap64(data[6]);
     data[7] = ss_byteswap64(data[7]);
#endif
     uint8_t* ptr = get_ptr(addr);
     *(uint64_t*)(ptr +  0) = data[0]; 
     *(uint64_t*)(ptr +  8) = data[1]; 
     *(uint64_t*)(ptr + 16) = data[2]; 
     *(uint64_t*)(ptr + 24) = data[3]; 
     *(uint64_t*)(ptr + 32) = data[4]; 
     *(uint64_t*)(ptr + 40) = data[5]; 
     *(uint64_t*)(ptr + 48) = data[6]; 
     *(uint64_t*)(ptr + 56) = data[7]; 
   }

   // Supported Load Operations. ld8[su]() to ld64() are quaranteed to be atomic. ld128() and
   // above are atomic at the 64 bit granularity.

   virtual uint8_t ld8u ( uint64_t addr )
   { 
     return *(uint8_t *)(get_ptr(addr)); 
   }
   virtual int8_t ld8s( uint64_t addr )
   { 
     return *( int8_t *)(get_ptr(addr)); 
   }
   virtual uint16_t ld16u( uint64_t addr )
   { 
#if defined(ARCH_X64)
     uint16_t data = *(uint16_t*)(get_ptr(addr));
     return ss_byteswap16(data);
#else 
     return *(uint16_t*)(get_ptr(addr)); 
#endif
   }
   virtual int16_t ld16s( uint64_t addr )
   { 
#if defined(ARCH_X64)
     int16_t data = *( int16_t*)(get_ptr(addr));
     return ss_byteswap16(data);
#else
     return *( int16_t*)(get_ptr(addr)); 
#endif
   }
   virtual uint32_t ld32u( uint64_t addr )
   { 
#if defined(ARCH_X64)
     uint32_t data = *(uint32_t*)(get_ptr(addr)); 
     return ss_byteswap32(data);
#else
     return *(uint32_t*)(get_ptr(addr)); 
#endif
   }
   virtual int32_t ld32s( uint64_t addr )
   { 
#if defined(ARCH_X64)
     int32_t data = *( int32_t*)(get_ptr(addr)); 
     return ss_byteswap32(data);
#else
     return *( int32_t*)(get_ptr(addr)); 
#endif
   }
   virtual uint64_t ld64( uint64_t addr )
   { 
#if defined(ARCH_X64)
     uint64_t data = *(uint64_t*)(get_ptr(addr));
     return ss_byteswap64(data);
#else
     return *(uint64_t*)(get_ptr(addr)); 
#endif
   }
   virtual void ld128( uint64_t addr, uint64_t data[2] ) 
   { 
     uint8_t* ptr = get_ptr(addr);
     data[0] = *(uint64_t*)(ptr +  0);
     data[1] = *(uint64_t*)(ptr +  8); 
#if defined(ARCH_X64)
     data[0] = ss_byteswap64(data[0]);
     data[1] = ss_byteswap64(data[1]);
#endif 
   }
   virtual void ld512( uint64_t addr, uint64_t data[8] )
   { 
     uint8_t* ptr = get_ptr(addr);
     data[0] = *(uint64_t*)(ptr +  0);
     data[1] = *(uint64_t*)(ptr +  8);
     data[2] = *(uint64_t*)(ptr + 16);
     data[3] = *(uint64_t*)(ptr + 24);
     data[4] = *(uint64_t*)(ptr + 32);
     data[5] = *(uint64_t*)(ptr + 40);
     data[6] = *(uint64_t*)(ptr + 48);
     data[7] = *(uint64_t*)(ptr + 56); 
#if defined(ARCH_X64)
     data[0] = ss_byteswap64(data[0]);
     data[1] = ss_byteswap64(data[1]);
     data[2] = ss_byteswap64(data[2]);
     data[3] = ss_byteswap64(data[3]);
     data[4] = ss_byteswap64(data[4]);
     data[5] = ss_byteswap64(data[5]);
     data[6] = ss_byteswap64(data[6]);
     data[7] = ss_byteswap64(data[7]);
#endif 
   }
   virtual void ld256( uint64_t addr, uint64_t data[4] )
   { 
     uint8_t* ptr = get_ptr(addr);
     data[0] = *(uint64_t*)(ptr +  0);
     data[1] = *(uint64_t*)(ptr +  8);
     data[2] = *(uint64_t*)(ptr + 16);
     data[3] = *(uint64_t*)(ptr + 24); 
#if defined(ARCH_X64)
     data[0] = ss_byteswap64(data[0]);
     data[1] = ss_byteswap64(data[1]);
     data[2] = ss_byteswap64(data[2]);
     data[3] = ss_byteswap64(data[3]);
#endif 
   }
   
   // st64partial() performs 8 byte partial store. The bytes to store are specified by mask. A 1 in bit N of 
   // mask denotes that byte (data >> (8*N)) & 0xff should be written to memory

   virtual void st64partial( uint64_t addr, uint64_t data, uint64_t mask )  
   {
     ss_stp8(*(double*)&data,get_ptr(addr),mask);
   }
   
   // ld128atomic() (aka load twin double, load quad atomic) atomically loads two
   // 64bit values from memory at addr into rd. rd[0] is the value at addr, rd[1]
   // is the value at addr + 8. Note ld128 does() not guarantee atomicity.
   
   virtual void ld128atomic( uint64_t addr, uint64_t data[2] )
   {
     ss_ld128atomic(get_ptr(addr),data);
   }

   // ldstub() return a byte from memory at addr, and set the byte at addr
   // to 0xff. The ldstub() operation is atomic.
  
   virtual uint8_t ldstub( uint64_t addr )
   {
     return ss_ldstub(get_ptr(addr),0);
   }

   // swap() stores the 32bit value rd with the 32bit value at addr.
   // The old 32bit value at addr is returned.  The operation is atomic.
   
   virtual uint32_t swap( uint64_t addr, uint32_t rd )
   {
     return ss_swap(rd,get_ptr(addr),0);
   }

   // casx() compares the 64bit value rs2 with the 64bit value at addr.
   // If the two values are equal, the value rd is stored in the
   // 64bit value at addr. In both cases the old 64bit value at addr is
   // returned, that is the value at addr before the storei happened.
   // The casx() operation is atomic.
   
   virtual uint64_t casx( uint64_t addr, uint64_t rd, uint64_t rs2 )
   {
     return ss_casx(rd,get_ptr(addr),rs2);
   }

   // cas() is as casx, but for 32bit.

   virtual uint32_t cas( uint64_t addr, uint32_t rd, uint32_t rs2 )
   {
     return ss_cas(rd,get_ptr(addr),rs2);
   }

   // prefetch() prefetches data from memory into the cache hierarchy.
   
   void prefetch( uint64_t addr, uint_t size ) {}

   // flush() writes dirty data in the cache back to memory.

   void flush( uint64_t addr, uint_t size ) {}   // process does not provide data.
 
   static SS_FastMemory memory;  
 private:
   enum 
   {
     // Configure the 55 bits of physical address space which is the 
     // current SunSparc limit. Note, there is no processor has 
     // implemented this many address lines yet. Currently the max 
     // is 48. To avoid complains about memory size from people we 
     // should provide 55 bits always. To make a choose, a toplevel 
     // of 2^23 entries of 4GB pages is nice. In v8plus mode however
     // we use much less ... just enough to make it work.
  
#if defined(ARCH_V9) || defined(ARCH_X64)
     L1BITS = 23,                      // 2^L1BITS of 
     L2BITS = 32,                      // 4GB of NORESERVED mmapped space
#elif defined(ARCH_V8)
     L1BITS = 29,                      // In v8 mode we have less va space ... 
     L2BITS = 19,                      // 1MB of NORESERVED mmapped chunks
#else
#error "Oops"
#endif
     L1SIZE = 1 << L1BITS,
     L1MASK = L1SIZE - 1
   };
   
   uint8_t* get_ptr( uint64_t addr )
   {
     uint8_t** o1 = l1 + ((addr >> L2BITS) & L1MASK);
     uint8_t*  l2 = *o1;
     if (l2 == 0)
     {
       l2lock.lock();
       if (*o1 == 0)
       {
         l2 = *o1 = (uint8_t*)mmap((char*)0,1ull << L2BITS,PROT_READ|PROT_WRITE,
                                   MAP_NORESERVE|MAP_PRIVATE|MAP_ANON|MAP_ALIGN,-1,0);
       }
       l2lock.unlock();
     }
     return l2 + (addr & uint64_t((1ull << L2BITS) - 1));
   }
   
   uint64_t ram_size;
   uint64_t rom_size;
   uint64_t page_size;
   uint64_t page_mask;

   uint8_t** l1;
   BL_Mutex l2lock;
};

#endif /* COMPILE_FOR_SAM */
#endif /* __SS_FastMemory_h__ */