+#if 1
+
//
// GPU Core
//
// Originally by David Raingeard (Cal2)
// GCC/SDL port by Niels Wagenaar (Linux/WIN32) and Caz (BeOS)
-// Cleanups, endian wrongness, and bad ASM amelioration by James L. Hammons
+// Cleanups, endian wrongness, and bad ASM amelioration by James Hammons
+// (C) 2010 Underground Software
+//
+// JLH = James Hammons <jlhamm@acm.org>
+//
+// Who When What
+// --- ---------- -------------------------------------------------------------
+// JLH 01/16/2010 Created this log ;-)
+// JLH 11/26/2011 Added fixes for LOAD/STORE alignment issues
+
+//
// Note: Endian wrongness probably stems from the MAME origins of this emu and
// the braindead way in which MAME handles memory. :-)
//
#include "gpu.h"
-#include <string.h> // For memset
#include <stdlib.h>
-#include "log.h"
+#include <string.h> // For memset
+#include "dsp.h"
+#include "jagdasm.h"
#include "jaguar.h"
-#include "m68k.h"
+#include "log.h"
+#include "m68000/m68kinterface.h"
+//#include "memory.h"
#include "tom.h"
-#include "memory.h"
-#include "jagdasm.h"
-#include "dsp.h"
+
+// Seems alignment in loads & stores was off...
+#define GPU_CORRECT_ALIGNMENT
//#define GPU_DEBUG
// For GPU dissasembly...
+#if 0
#define GPU_DIS_ABS
#define GPU_DIS_ADD
#define GPU_DIS_ADDC
#define GPU_DIS_SUBQT
#define GPU_DIS_XOR
-bool doGPUDis = false;
-//bool doGPUDis = true;
-//*/
+//bool doGPUDis = false;
+bool doGPUDis = true;
+#endif
+
/*
GPU opcodes use (BIOS flying ATARI logo):
+ add 357416
// Private function prototypes
void GPUUpdateRegisterBanks(void);
-
void GPUDumpDisassembly(void);
void GPUDumpRegisters(void);
void GPUDumpMemory(void);
static void gpu_opcode_pack(void);
// This is wrong, since it doesn't take pipeline effects into account. !!! FIX !!!
-/*uint8 gpu_opcode_cycles[64] =
+/*uint8 gpu_opcode_cycles[64] =
{
3, 3, 3, 3, 3, 3, 3, 3,
3, 3, 3, 3, 3, 3, 3, 3,
//This is wrong, wrong, WRONG, but it seems to work for the time being...
//(That is, it fixes Flip Out which relies on GPU timing rather than semaphores. Bad developers! Bad!)
//What's needed here is a way to take pipeline effects into account (including pipeline stalls!)...
-/*uint8 gpu_opcode_cycles[64] =
+/*uint8 gpu_opcode_cycles[64] =
{
1, 1, 1, 1, 1, 1, 1, 1,
1, 1, 1, 1, 1, 1, 1, 1,
1, 1, 1, 1, 1, 1, 4, 1,
1, 1, 3, 3, 1, 1, 1, 1
};//*/
-uint8 gpu_opcode_cycles[64] =
+uint8 gpu_opcode_cycles[64] =
{
1, 1, 1, 1, 1, 1, 1, 1,
1, 1, 1, 1, 1, 1, 1, 1,
1, 1, 1, 1, 1, 1, 1, 1
};//*/
-void (*gpu_opcode[64])()=
-{
+void (*gpu_opcode[64])()=
+{
gpu_opcode_add, gpu_opcode_addc, gpu_opcode_addq, gpu_opcode_addqt,
gpu_opcode_sub, gpu_opcode_subc, gpu_opcode_subq, gpu_opcode_subqt,
gpu_opcode_neg, gpu_opcode_and, gpu_opcode_or, gpu_opcode_xor,
#define RESET_FLAG_Z() gpu_flag_z = 0;
#define RESET_FLAG_N() gpu_flag_n = 0;
-#define RESET_FLAG_C() gpu_flag_c = 0;
+#define RESET_FLAG_C() gpu_flag_c = 0;
#define CLR_Z (gpu_flag_z = 0)
#define CLR_ZN (gpu_flag_z = gpu_flag_n = 0)
uint32 gpu_opcode_use[64];
-const char * gpu_opcode_str[64]=
-{
+const char * gpu_opcode_str[64]=
+{
"add", "addc", "addq", "addqt",
"sub", "subc", "subq", "subqt",
"neg", "and", "or", "xor",
static uint32 gpu_in_exec = 0;
static uint32 gpu_releaseTimeSlice_flag = 0;
-void gpu_releaseTimeslice(void)
+void GPUReleaseTimeslice(void)
{
gpu_releaseTimeSlice_flag = 1;
}
-uint32 gpu_get_pc(void)
+uint32 GPUGetPC(void)
{
return gpu_pc;
}
gpu_flag_n = (gpu_flag_n ? 1 : 0);
gpu_flags = (gpu_flags & 0xFFFFFFF8) | (gpu_flag_n << 2) | (gpu_flag_c << 1) | gpu_flag_z;
-
+
return gpu_flags & 0xFFFFC1FF;
case 0x04:
return gpu_matrix_control;
else
gpu_div_control = (gpu_div_control & 0x0000FFFF) | ((data & 0xFFFF) << 16);
}
- else
+ else
{
//WriteLog("[GPU W16:%08X,%04X]", offset, data);
uint32 old_data = GPUReadLong(offset & 0xFFFFFFC, who);
+
if (offset & 0x02)
old_data = (old_data & 0xFFFF0000) | (data & 0xFFFF);
else
old_data = (old_data & 0x0000FFFF) | ((data & 0xFFFF) << 16);
+
GPUWriteLong(offset & 0xFFFFFFC, old_data, who);
}
+
return;
}
else if ((offset == GPU_WORK_RAM_BASE + 0x0FFF) || (GPU_CONTROL_RAM_BASE + 0x1F))
case 0x00:
{
bool IMASKCleared = (gpu_flags & IMASK) && !(data & IMASK);
- gpu_flags = data;
+ // NOTE: According to the JTRM, writing a 1 to IMASK has no effect; only the
+ // IRQ logic can set it. So we mask it out here to prevent problems...
+ gpu_flags = data & (~IMASK);
gpu_flag_z = gpu_flags & ZERO_FLAG;
gpu_flag_c = (gpu_flags & CARRY_FLAG) >> 1;
gpu_flag_n = (gpu_flags & NEGA_FLAG) >> 2;
#endif // GPU_DEBUG
break;
case 0x14:
- {
+ {
// uint32 gpu_was_running = GPU_RUNNING;
data &= ~0xF7C0; // Disable writes to INT_LAT0-4 & TOM version number
if (data & 0x02)
{
//WriteLog("GPU->CPU interrupt\n");
- if (tom_irq_enabled(IRQ_GPU))
+ if (TOMIRQEnabled(IRQ_GPU))
{
- if ((tom_irq_enabled(IRQ_GPU)) && (jaguar_interrupt_handler_is_valid(64)))
+//This is the programmer's responsibility, to make sure the handler is valid, not ours!
+// if ((TOMIRQEnabled(IRQ_GPU))// && (JaguarInterruptHandlerIsValid(64)))
{
- tom_set_pending_gpu_int();
- m68k_set_irq(7); // Set 68000 NMI
- gpu_releaseTimeslice();
+ TOMSetPendingGPUInt();
+ m68k_set_irq(2); // Set 68000 IPL 2
+ GPUReleaseTimeslice();
}
}
data &= ~0x02;
//WriteLog("CPU->GPU interrupt\n");
GPUSetIRQLine(0, ASSERT_LINE);
m68k_end_timeslice();
- dsp_releaseTimeslice();
+ DSPReleaseTimeslice();
data &= ~0x04;
}
{
WriteLog("GPU: Write32--About to do stupid braindead GPU execution for 200 cycles.\n");
#endif // GPU_DEBUG
- gpu_exec(200);
+ GPUExec(200);
#ifdef GPU_DEBUG
}
#endif // GPU_DEBUG//*/
#else
if (gpu_control & 0x18)
- gpu_exec(1);
+ GPUExec(1);
#endif // #ifndef GPU_SINGLE_STEPPING
#ifdef GPU_DEBUG
WriteLog("Write to GPU CTRL by %s: %08X ", whoName[who], data);
for(int x=0; x<2; x++)
{
JaguarWriteLong(dst, JaguarReadLong(src));
-
+
src += 4;
dst += 4;
}
// Get the interrupt latch & enable bits
uint32 bits = (gpu_control >> 6) & 0x1F, mask = (gpu_flags >> 4) & 0x1F;
-
+
// Bail out if latched interrupts aren't enabled
bits &= mask;
if (!bits)
return;
-
+
// Determine which interrupt to service
uint32 which = 0; //Isn't there a #pragma to disable this warning???
if (bits & 0x01)
if (start_logging)
WriteLog("GPU: Generating IRQ #%i\n", which);
- // set the interrupt flag
+ // set the interrupt flag
gpu_flags |= IMASK;
GPUUpdateRegisterBanks();
- // subqt #4,r31 ; pre-decrement stack pointer
- // move pc,r30 ; address of interrupted code
+ // subqt #4,r31 ; pre-decrement stack pointer
+ // move pc,r30 ; address of interrupted code
// store r30,(r31) ; store return address
gpu_reg[31] -= 4;
GPUWriteLong(gpu_reg[31], gpu_pc - 2, GPU);
-
- // movei #service_address,r30 ; pointer to ISR entry
- // jump (r30) ; jump to ISR
+
+ // movei #service_address,r30 ; pointer to ISR entry
+ // jump (r30) ; jump to ISR
// nop
gpu_pc = gpu_reg[30] = GPU_WORK_RAM_BASE + (which * 0x10);
}
//#include "gpu2.h"
//#include "gpu3.h"
-void gpu_init(void)
+void GPUInit(void)
{
// memory_malloc_secure((void **)&gpu_ram_8, 0x1000, "GPU work RAM");
// memory_malloc_secure((void **)&gpu_reg_bank_0, 32 * sizeof(int32), "GPU bank 0 regs");
build_branch_condition_table();
- gpu_reset();
+ GPUReset();
//TEMPORARY: Testing only!
// gpu2_init();
// gpu3_init();
}
-void gpu_reset(void)
+void GPUReset(void)
{
// GPU registers (directly visible)
gpu_flags = 0x00000000;
memset(gpu_ram_8, 0xFF, 0x1000);
gpu_in_exec = 0;
//not needed GPUInterruptPending = false;
- gpu_reset_stats();
+ GPUResetStats();
}
-uint32 gpu_read_pc(void)
+uint32 GPUReadPC(void)
{
return gpu_pc;
}
-void gpu_reset_stats(void)
+void GPUResetStats(void)
{
for(uint32 i=0; i<64; i++)
gpu_opcode_use[i] = 0;
gpu_ram_8[i+1], gpu_ram_8[i+2], gpu_ram_8[i+3]);
}
-void gpu_done(void)
-{
+void GPUDone(void)
+{
WriteLog("GPU: Stopped at PC=%08X (GPU %s running)\n", (unsigned int)gpu_pc, GPU_RUNNING ? "was" : "wasn't");
- // Get the interrupt latch & enable bits
+ // Get the interrupt latch & enable bits
uint8 bits = (gpu_control >> 6) & 0x1F, mask = (gpu_flags >> 4) & 0x1F;
WriteLog("GPU: Latch bits = %02X, enable bits = %02X\n", bits, mask);
static int testCount = 1;
static int len = 0;
static bool tripwire = false;
-void gpu_exec(int32 cycles)
+void GPUExec(int32 cycles)
{
if (!GPU_RUNNING)
return;
/* gpu_flag_c = (gpu_flag_c ? 1 : 0);
gpu_flag_z = (gpu_flag_z ? 1 : 0);
gpu_flag_n = (gpu_flag_n ? 1 : 0);*/
-
+
uint16 opcode = GPUReadWord(gpu_pc, GPU);
uint32 index = opcode >> 10;
gpu_instruction = opcode; // Added for GPU #3...
// gpu3_opcode[index]();
// BIOS hacking
-//GPU: [00F03548] jr nz,00F03560 (0xd561) (RM=00F03114, RN=00000004) -> --> JR: Branch taken.
+//GPU: [00F03548] jr nz,00F03560 (0xd561) (RM=00F03114, RN=00000004) -> --> JR: Branch taken.
/*static bool firstTime = true;
if (gpu_pc == 0xF03548 && firstTime)
{
if (gpu_start_log)
WriteLog(" --> JUMP: Branch taken.\n");
uint32 delayed_pc = RM;
- gpu_exec(1);
+ GPUExec(1);
gpu_pc = delayed_pc;
/* uint16 opcode = GPUReadWord(gpu_pc, GPU);
gpu_opcode_first_parameter = (opcode >> 5) & 0x1F;
WriteLog(" --> JR: Branch taken.\n");
int32 offset = (IMM_1 & 0x10 ? 0xFFFFFFF0 | IMM_1 : IMM_1); // Sign extend IMM_1
int32 delayed_pc = gpu_pc + (offset * 2);
- gpu_exec(1);
+ GPUExec(1);
gpu_pc = delayed_pc;
/* uint16 opcode = GPUReadWord(gpu_pc, GPU);
gpu_opcode_first_parameter = (opcode >> 5) & 0x1F;
if (doGPUDis)
WriteLog("%06X: STORE R%02u, (R14+$%02X) [NCZ:%u%u%u, R%02u=%08X, R14+$%02X=%08X]\n", gpu_pc-2, IMM_2, gpu_convert_zero[IMM_1] << 2, gpu_flag_n, gpu_flag_c, gpu_flag_z, IMM_2, RN, gpu_convert_zero[IMM_1] << 2, gpu_reg[14]+(gpu_convert_zero[IMM_1] << 2));
#endif
+#ifdef GPU_CORRECT_ALIGNMENT
+ uint32 address = gpu_reg[14] + (gpu_convert_zero[IMM_1] << 2);
+
+ if (address >= 0xF03000 && address <= 0xF03FFF)
+ GPUWriteLong(address & 0xFFFFFFFC, RN, GPU);
+ else
+ GPUWriteLong(address, RN, GPU);
+#else
GPUWriteLong(gpu_reg[14] + (gpu_convert_zero[IMM_1] << 2), RN, GPU);
+#endif
}
static void gpu_opcode_store_r15_indexed(void)
if (doGPUDis)
WriteLog("%06X: STORE R%02u, (R15+$%02X) [NCZ:%u%u%u, R%02u=%08X, R15+$%02X=%08X]\n", gpu_pc-2, IMM_2, gpu_convert_zero[IMM_1] << 2, gpu_flag_n, gpu_flag_c, gpu_flag_z, IMM_2, RN, gpu_convert_zero[IMM_1] << 2, gpu_reg[15]+(gpu_convert_zero[IMM_1] << 2));
#endif
+#ifdef GPU_CORRECT_ALIGNMENT
+ uint32 address = gpu_reg[15] + (gpu_convert_zero[IMM_1] << 2);
+
+ if (address >= 0xF03000 && address <= 0xF03FFF)
+ GPUWriteLong(address & 0xFFFFFFFC, RN, GPU);
+ else
+ GPUWriteLong(address, RN, GPU);
+#else
GPUWriteLong(gpu_reg[15] + (gpu_convert_zero[IMM_1] << 2), RN, GPU);
+#endif
}
static void gpu_opcode_load_r14_ri(void)
if (doGPUDis)
WriteLog("%06X: LOAD (R14+R%02u), R%02u [NCZ:%u%u%u, R14+R%02u=%08X, R%02u=%08X] -> ", gpu_pc-2, IMM_1, IMM_2, gpu_flag_n, gpu_flag_c, gpu_flag_z, IMM_1, RM+gpu_reg[14], IMM_2, RN);
#endif
+#ifdef GPU_CORRECT_ALIGNMENT
+ uint32 address = gpu_reg[14] + RM;
+
+ if (address >= 0xF03000 && address <= 0xF03FFF)
+ RN = GPUReadLong(address & 0xFFFFFFFC, GPU);
+ else
+ RN = GPUReadLong(address, GPU);
+#else
RN = GPUReadLong(gpu_reg[14] + RM, GPU);
+#endif
#ifdef GPU_DIS_LOAD14R
if (doGPUDis)
WriteLog("[NCZ:%u%u%u, R%02u=%08X]\n", gpu_flag_n, gpu_flag_c, gpu_flag_z, IMM_2, RN);
if (doGPUDis)
WriteLog("%06X: LOAD (R15+R%02u), R%02u [NCZ:%u%u%u, R15+R%02u=%08X, R%02u=%08X] -> ", gpu_pc-2, IMM_1, IMM_2, gpu_flag_n, gpu_flag_c, gpu_flag_z, IMM_1, RM+gpu_reg[15], IMM_2, RN);
#endif
+#ifdef GPU_CORRECT_ALIGNMENT
+ uint32 address = gpu_reg[15] + RM;
+
+ if (address >= 0xF03000 && address <= 0xF03FFF)
+ RN = GPUReadLong(address & 0xFFFFFFFC, GPU);
+ else
+ RN = GPUReadLong(address, GPU);
+#else
RN = GPUReadLong(gpu_reg[15] + RM, GPU);
+#endif
#ifdef GPU_DIS_LOAD15R
if (doGPUDis)
WriteLog("[NCZ:%u%u%u, R%02u=%08X]\n", gpu_flag_n, gpu_flag_c, gpu_flag_z, IMM_2, RN);
if (doGPUDis)
WriteLog("%06X: STORE R%02u, (R14+R%02u) [NCZ:%u%u%u, R%02u=%08X, R14+R%02u=%08X]\n", gpu_pc-2, IMM_2, IMM_1, gpu_flag_n, gpu_flag_c, gpu_flag_z, IMM_2, RN, IMM_1, RM+gpu_reg[14]);
#endif
+#ifdef GPU_CORRECT_ALIGNMENT
+ uint32 address = gpu_reg[14] + RM;
+
+ if (address >= 0xF03000 && address <= 0xF03FFF)
+ GPUWriteLong(address & 0xFFFFFFFC, RN, GPU);
+ else
+ GPUWriteLong(address, RN, GPU);
+#else
GPUWriteLong(gpu_reg[14] + RM, RN, GPU);
+#endif
}
static void gpu_opcode_store_r15_ri(void)
if (doGPUDis)
WriteLog("%06X: STORE R%02u, (R15+R%02u) [NCZ:%u%u%u, R%02u=%08X, R15+R%02u=%08X]\n", gpu_pc-2, IMM_2, IMM_1, gpu_flag_n, gpu_flag_c, gpu_flag_z, IMM_2, RN, IMM_1, RM+gpu_reg[15]);
#endif
+#ifdef GPU_CORRECT_ALIGNMENT_STORE
+ uint32 address = gpu_reg[15] + RM;
+
+ if (address >= 0xF03000 && address <= 0xF03FFF)
+ GPUWriteLong(address & 0xFFFFFFFC, RN, GPU);
+ else
+ GPUWriteLong(address, RN, GPU);
+#else
GPUWriteLong(gpu_reg[15] + RM, RN, GPU);
+#endif
}
static void gpu_opcode_nop(void)
if (doGPUDis)
WriteLog("%06X: STOREW R%02u, (R%02u) [NCZ:%u%u%u, R%02u=%08X, R%02u=%08X]\n", gpu_pc-2, IMM_2, IMM_1, gpu_flag_n, gpu_flag_c, gpu_flag_z, IMM_2, RN, IMM_1, RM);
#endif
+#ifdef GPU_CORRECT_ALIGNMENT
+ if ((RM >= 0xF03000) && (RM <= 0xF03FFF))
+ GPUWriteLong(RM & 0xFFFFFFFE, RN & 0xFFFF, GPU);
+ else
+ JaguarWriteWord(RM, RN, GPU);
+#else
if ((RM >= 0xF03000) && (RM <= 0xF03FFF))
GPUWriteLong(RM, RN & 0xFFFF, GPU);
else
JaguarWriteWord(RM, RN, GPU);
+#endif
}
static void gpu_opcode_store(void)
if (doGPUDis)
WriteLog("%06X: STORE R%02u, (R%02u) [NCZ:%u%u%u, R%02u=%08X, R%02u=%08X]\n", gpu_pc-2, IMM_2, IMM_1, gpu_flag_n, gpu_flag_c, gpu_flag_z, IMM_2, RN, IMM_1, RM);
#endif
+#ifdef GPU_CORRECT_ALIGNMENT
+ if ((RM >= 0xF03000) && (RM <= 0xF03FFF))
+ GPUWriteLong(RM & 0xFFFFFFFC, RN, GPU);
+ else
+ GPUWriteLong(RM, RN, GPU);
+#else
GPUWriteLong(RM, RN, GPU);
+#endif
}
static void gpu_opcode_storep(void)
{
+#ifdef GPU_CORRECT_ALIGNMENT
+ if ((RM >= 0xF03000) && (RM <= 0xF03FFF))
+ {
+ GPUWriteLong((RM & 0xFFFFFFF8) + 0, gpu_hidata, GPU);
+ GPUWriteLong((RM & 0xFFFFFFF8) + 4, RN, GPU);
+ }
+ else
+ {
+ GPUWriteLong(RM + 0, gpu_hidata, GPU);
+ GPUWriteLong(RM + 4, RN, GPU);
+ }
+#else
GPUWriteLong(RM + 0, gpu_hidata, GPU);
GPUWriteLong(RM + 4, RN, GPU);
+#endif
}
static void gpu_opcode_loadb(void)
if (doGPUDis)
WriteLog("%06X: LOADW (R%02u), R%02u [NCZ:%u%u%u, R%02u=%08X, R%02u=%08X] -> ", gpu_pc-2, IMM_1, IMM_2, gpu_flag_n, gpu_flag_c, gpu_flag_z, IMM_1, RM, IMM_2, RN);
#endif
+#ifdef GPU_CORRECT_ALIGNMENT
+ if ((RM >= 0xF03000) && (RM <= 0xF03FFF))
+ RN = GPUReadLong(RM & 0xFFFFFFFE, GPU) & 0xFFFF;
+ else
+ RN = JaguarReadWord(RM, GPU);
+#else
if ((RM >= 0xF03000) && (RM <= 0xF03FFF))
RN = GPUReadLong(RM, GPU) & 0xFFFF;
else
RN = JaguarReadWord(RM, GPU);
+#endif
#ifdef GPU_DIS_LOADW
if (doGPUDis)
WriteLog("[NCZ:%u%u%u, R%02u=%08X]\n", gpu_flag_n, gpu_flag_c, gpu_flag_z, IMM_2, RN);
#endif
}
+// According to the docs, & "Do The Same", this address is long aligned...
+// So let's try it:
+// And it works!!! Need to fix all instances...
+// Also, Power Drive Rally seems to contradict the idea that only LOADs in
+// the $F03000-$F03FFF range are aligned...
+#warning "!!! Alignment issues, need to find definitive final word on this !!!"
+/*
+Preliminary testing on real hardware seems to confirm that something strange goes on
+with unaligned reads in main memory. When the address is off by 1, the result is the
+same as the long address with the top byte replaced by something. So if the read is
+from $401, and $400 has 12 34 56 78, the value read will be $nn345678, where nn is a currently unknown vlaue.
+When the address is off by 2, the result would be $nnnn5678, where nnnn is unknown.
+When the address is off by 3, the result would be $nnnnnn78, where nnnnnn is unknown.
+It may be that the "unknown" values come from the prefetch queue, but not sure how
+to test that. They seem to be stable, though, which would indicate such a mechanism.
+Sometimes, however, the off by 2 case returns $12345678!
+*/
static void gpu_opcode_load(void)
{
#ifdef GPU_DIS_LOAD
if (doGPUDis)
WriteLog("%06X: LOAD (R%02u), R%02u [NCZ:%u%u%u, R%02u=%08X, R%02u=%08X] -> ", gpu_pc-2, IMM_1, IMM_2, gpu_flag_n, gpu_flag_c, gpu_flag_z, IMM_1, RM, IMM_2, RN);
#endif
+#ifdef GPU_CORRECT_ALIGNMENT
+ uint32 mask[4] = { 0x00000000, 0xFF000000, 0xFFFF0000, 0xFFFFFF00 };
+// if ((RM >= 0xF03000) && (RM <= 0xF03FFF))
+ RN = GPUReadLong(RM & 0xFFFFFFFC, GPU);
+// RN = GPUReadLong(RM & 0x00FFFFFC, GPU);
+// else
+// RN = GPUReadLong(RM, GPU);
+ // Simulate garbage in unaligned reads...
+//seems that this behavior is different in GPU mem vs. main mem...
+// if ((RM < 0xF03000) || (RM > 0xF0BFFF))
+// RN |= mask[RM & 0x03];
+#else
RN = GPUReadLong(RM, GPU);
+#endif
#ifdef GPU_DIS_LOAD
if (doGPUDis)
WriteLog("[NCZ:%u%u%u, R%02u=%08X]\n", gpu_flag_n, gpu_flag_c, gpu_flag_z, IMM_2, RN);
static void gpu_opcode_loadp(void)
{
+#ifdef GPU_CORRECT_ALIGNMENT
+ if ((RM >= 0xF03000) && (RM <= 0xF03FFF))
+ {
+ gpu_hidata = GPUReadLong((RM & 0xFFFFFFF8) + 0, GPU);
+ RN = GPUReadLong((RM & 0xFFFFFFF8) + 4, GPU);
+ }
+ else
+ {
+ gpu_hidata = GPUReadLong(RM + 0, GPU);
+ RN = GPUReadLong(RM + 4, GPU);
+ }
+#else
gpu_hidata = GPUReadLong(RM + 0, GPU);
RN = GPUReadLong(RM + 4, GPU);
+#endif
}
static void gpu_opcode_load_r14_indexed(void)
if (doGPUDis)
WriteLog("%06X: LOAD (R14+$%02X), R%02u [NCZ:%u%u%u, R14+$%02X=%08X, R%02u=%08X] -> ", gpu_pc-2, gpu_convert_zero[IMM_1] << 2, IMM_2, gpu_flag_n, gpu_flag_c, gpu_flag_z, gpu_convert_zero[IMM_1] << 2, gpu_reg[14]+(gpu_convert_zero[IMM_1] << 2), IMM_2, RN);
#endif
+#ifdef GPU_CORRECT_ALIGNMENT
+ uint32 address = gpu_reg[14] + (gpu_convert_zero[IMM_1] << 2);
+
+ if ((RM >= 0xF03000) && (RM <= 0xF03FFF))
+ RN = GPUReadLong(address & 0xFFFFFFFC, GPU);
+ else
+ RN = GPUReadLong(address, GPU);
+#else
RN = GPUReadLong(gpu_reg[14] + (gpu_convert_zero[IMM_1] << 2), GPU);
+#endif
#ifdef GPU_DIS_LOAD14I
if (doGPUDis)
WriteLog("[NCZ:%u%u%u, R%02u=%08X]\n", gpu_flag_n, gpu_flag_c, gpu_flag_z, IMM_2, RN);
if (doGPUDis)
WriteLog("%06X: LOAD (R15+$%02X), R%02u [NCZ:%u%u%u, R15+$%02X=%08X, R%02u=%08X] -> ", gpu_pc-2, gpu_convert_zero[IMM_1] << 2, IMM_2, gpu_flag_n, gpu_flag_c, gpu_flag_z, gpu_convert_zero[IMM_1] << 2, gpu_reg[15]+(gpu_convert_zero[IMM_1] << 2), IMM_2, RN);
#endif
+#ifdef GPU_CORRECT_ALIGNMENT
+ uint32 address = gpu_reg[15] + (gpu_convert_zero[IMM_1] << 2);
+
+ if ((RM >= 0xF03000) && (RM <= 0xF03FFF))
+ RN = GPUReadLong(address & 0xFFFFFFFC, GPU);
+ else
+ RN = GPUReadLong(address, GPU);
+#else
RN = GPUReadLong(gpu_reg[15] + (gpu_convert_zero[IMM_1] << 2), GPU);
+#endif
#ifdef GPU_DIS_LOAD15I
if (doGPUDis)
WriteLog("[NCZ:%u%u%u, R%02u=%08X]\n", gpu_flag_n, gpu_flag_c, gpu_flag_z, IMM_2, RN);
if (gpu_matrix_control & 0x10) // Column stepping
{
for(int i=0; i<count; i++)
- {
+ {
int16 a;
if (i & 0x01)
a = (int16)((gpu_alternate_reg[IMM_1 + (i >> 1)] >> 16) & 0xFFFF);
//Temporary: Testing only!
//#include "gpu2.cpp"
//#include "gpu3.cpp"
+
+#else
+
+// New thread-safe GPU core
+
+int GPUCore(void * data)
+{
+}
+
+#endif