bsnes/target-loki/debugger/debugger.cpp

480 lines
14 KiB
C++
Raw Normal View History

#include "../loki.hpp"
Debugger* debugger = nullptr;
Debugger::Debugger() {
debugger = this;
SFC::cpu.debugger.op_exec = {&Debugger::cpuExec, this};
SFC::cpu.debugger.op_read = {&Debugger::cpuRead, this};
SFC::cpu.debugger.op_write = {&Debugger::cpuWrite, this};
Update to v094r05 release. byuu says: Commands can be prefixed with: (cpu|smp|ppu|dsp|apu|vram|oam|cgram)/ to set their source. Eg "vram/hex 0800" or "smp/breakpoints.append execute ffc0"; default is cpu. These overlap a little bit in odd ways, but that's just the way the SNES works: it's not a very orthogonal system. CPU is both a processor and the main bus (ROM, RAM, WRAM, etc), APU is the shared memory by the SMP+DSP (eg use it to catch writes from either chip); PPU probably won't ever be used since it's broken down into three separate buses (VRAM, OAM, CGRAM), but DSP could be useful for tracking bugs like we found in Koushien 2 with the DSP echo buffer corrupting SMP opcodes. Technically the PPU memory pools are only ever tripped by the CPU poking at them, as the PPU doesn't ever write. I now have run.for, run.to, step.for, step.to. The difference is that run only prints the next instruction after running, whereas step prints all of the instructions along the way as well. run.to acts the same as "step over" here. Although it's not quite as nice, since you have to specify the address of the next instruction. Logging the Field/Vcounter/Hcounter on instruction listings now, good for timing information. Added in the tracer mask, as well as memory export, as well as VRAM/OAM/CGRAM/SMP read/write/execute breakpoints, as well as an APU usage map (it tracks DSP reads/writes separately, although I don't currently have debugger callbacks on DSP accesses just yet.) Have not hooked up actual SMP debugging just yet, but I plan to soon. Still thinking about how I want to allow / block interleaving of instructions (terminal output and tracing.) So ... remaining tasks at this point: - full SMP debugging - CPU+SMP interleave support - aliases - hotkeys - save states (will be kind of tricky ... will have to suppress breakpoints during synchronization, or abort a save in a break event.) - keep track of window geometry between runs
2014-02-05 11:30:08 +00:00
SFC::smp.debugger.op_exec = {&Debugger::smpExec, this};
SFC::smp.debugger.op_read = {&Debugger::smpRead, this};
SFC::smp.debugger.op_write = {&Debugger::smpWrite, this};
SFC::ppu.debugger.vram_read = {&Debugger::ppuVramRead, this};
SFC::ppu.debugger.vram_write = {&Debugger::ppuVramWrite, this};
SFC::ppu.debugger.oam_read = {&Debugger::ppuOamRead, this};
SFC::ppu.debugger.oam_write = {&Debugger::ppuOamWrite, this};
SFC::ppu.debugger.cgram_read = {&Debugger::ppuCgramRead, this};
SFC::ppu.debugger.cgram_write = {&Debugger::ppuCgramWrite, this};
}
void Debugger::load() {
Update to v094r05 release. byuu says: Commands can be prefixed with: (cpu|smp|ppu|dsp|apu|vram|oam|cgram)/ to set their source. Eg "vram/hex 0800" or "smp/breakpoints.append execute ffc0"; default is cpu. These overlap a little bit in odd ways, but that's just the way the SNES works: it's not a very orthogonal system. CPU is both a processor and the main bus (ROM, RAM, WRAM, etc), APU is the shared memory by the SMP+DSP (eg use it to catch writes from either chip); PPU probably won't ever be used since it's broken down into three separate buses (VRAM, OAM, CGRAM), but DSP could be useful for tracking bugs like we found in Koushien 2 with the DSP echo buffer corrupting SMP opcodes. Technically the PPU memory pools are only ever tripped by the CPU poking at them, as the PPU doesn't ever write. I now have run.for, run.to, step.for, step.to. The difference is that run only prints the next instruction after running, whereas step prints all of the instructions along the way as well. run.to acts the same as "step over" here. Although it's not quite as nice, since you have to specify the address of the next instruction. Logging the Field/Vcounter/Hcounter on instruction listings now, good for timing information. Added in the tracer mask, as well as memory export, as well as VRAM/OAM/CGRAM/SMP read/write/execute breakpoints, as well as an APU usage map (it tracks DSP reads/writes separately, although I don't currently have debugger callbacks on DSP accesses just yet.) Have not hooked up actual SMP debugging just yet, but I plan to soon. Still thinking about how I want to allow / block interleaving of instructions (terminal output and tracing.) So ... remaining tasks at this point: - full SMP debugging - CPU+SMP interleave support - aliases - hotkeys - save states (will be kind of tricky ... will have to suppress breakpoints during synchronization, or abort a save in a break event.) - keep track of window geometry between runs
2014-02-05 11:30:08 +00:00
directory::create({interface->pathname, "loki/"});
cpuUsage = new uint8[0x1000000]();
apuUsage = new uint8[0x10000]();
file fp;
Update to v094r05 release. byuu says: Commands can be prefixed with: (cpu|smp|ppu|dsp|apu|vram|oam|cgram)/ to set their source. Eg "vram/hex 0800" or "smp/breakpoints.append execute ffc0"; default is cpu. These overlap a little bit in odd ways, but that's just the way the SNES works: it's not a very orthogonal system. CPU is both a processor and the main bus (ROM, RAM, WRAM, etc), APU is the shared memory by the SMP+DSP (eg use it to catch writes from either chip); PPU probably won't ever be used since it's broken down into three separate buses (VRAM, OAM, CGRAM), but DSP could be useful for tracking bugs like we found in Koushien 2 with the DSP echo buffer corrupting SMP opcodes. Technically the PPU memory pools are only ever tripped by the CPU poking at them, as the PPU doesn't ever write. I now have run.for, run.to, step.for, step.to. The difference is that run only prints the next instruction after running, whereas step prints all of the instructions along the way as well. run.to acts the same as "step over" here. Although it's not quite as nice, since you have to specify the address of the next instruction. Logging the Field/Vcounter/Hcounter on instruction listings now, good for timing information. Added in the tracer mask, as well as memory export, as well as VRAM/OAM/CGRAM/SMP read/write/execute breakpoints, as well as an APU usage map (it tracks DSP reads/writes separately, although I don't currently have debugger callbacks on DSP accesses just yet.) Have not hooked up actual SMP debugging just yet, but I plan to soon. Still thinking about how I want to allow / block interleaving of instructions (terminal output and tracing.) So ... remaining tasks at this point: - full SMP debugging - CPU+SMP interleave support - aliases - hotkeys - save states (will be kind of tricky ... will have to suppress breakpoints during synchronization, or abort a save in a break event.) - keep track of window geometry between runs
2014-02-05 11:30:08 +00:00
if(fp.open({interface->pathname, "loki/cpu.usage.map"}, file::mode::read)) {
if(fp.size() == 0x1000000) fp.read(cpuUsage, 0x1000000);
Update to v094r05 release. byuu says: Commands can be prefixed with: (cpu|smp|ppu|dsp|apu|vram|oam|cgram)/ to set their source. Eg "vram/hex 0800" or "smp/breakpoints.append execute ffc0"; default is cpu. These overlap a little bit in odd ways, but that's just the way the SNES works: it's not a very orthogonal system. CPU is both a processor and the main bus (ROM, RAM, WRAM, etc), APU is the shared memory by the SMP+DSP (eg use it to catch writes from either chip); PPU probably won't ever be used since it's broken down into three separate buses (VRAM, OAM, CGRAM), but DSP could be useful for tracking bugs like we found in Koushien 2 with the DSP echo buffer corrupting SMP opcodes. Technically the PPU memory pools are only ever tripped by the CPU poking at them, as the PPU doesn't ever write. I now have run.for, run.to, step.for, step.to. The difference is that run only prints the next instruction after running, whereas step prints all of the instructions along the way as well. run.to acts the same as "step over" here. Although it's not quite as nice, since you have to specify the address of the next instruction. Logging the Field/Vcounter/Hcounter on instruction listings now, good for timing information. Added in the tracer mask, as well as memory export, as well as VRAM/OAM/CGRAM/SMP read/write/execute breakpoints, as well as an APU usage map (it tracks DSP reads/writes separately, although I don't currently have debugger callbacks on DSP accesses just yet.) Have not hooked up actual SMP debugging just yet, but I plan to soon. Still thinking about how I want to allow / block interleaving of instructions (terminal output and tracing.) So ... remaining tasks at this point: - full SMP debugging - CPU+SMP interleave support - aliases - hotkeys - save states (will be kind of tricky ... will have to suppress breakpoints during synchronization, or abort a save in a break event.) - keep track of window geometry between runs
2014-02-05 11:30:08 +00:00
fp.close();
}
if(fp.open({interface->pathname, "loki/apu.usage.map"}, file::mode::read)) {
if(fp.size() == 0x10000) fp.read(apuUsage, 0x10000);
fp.close();
}
}
void Debugger::unload() {
if(cpuTracerFile.open()) cpuTracerFile.close();
if(smpTracerFile.open()) smpTracerFile.close();
file::write({interface->pathname, "loki/cpu.usage.map"}, cpuUsage, 0x1000000);
file::write({interface->pathname, "loki/apu.usage.map"}, apuUsage, 0x10000);
delete[] cpuUsage;
delete[] apuUsage;
cpuUsage = nullptr;
apuUsage = nullptr;
}
Update to v094r05 release. byuu says: Commands can be prefixed with: (cpu|smp|ppu|dsp|apu|vram|oam|cgram)/ to set their source. Eg "vram/hex 0800" or "smp/breakpoints.append execute ffc0"; default is cpu. These overlap a little bit in odd ways, but that's just the way the SNES works: it's not a very orthogonal system. CPU is both a processor and the main bus (ROM, RAM, WRAM, etc), APU is the shared memory by the SMP+DSP (eg use it to catch writes from either chip); PPU probably won't ever be used since it's broken down into three separate buses (VRAM, OAM, CGRAM), but DSP could be useful for tracking bugs like we found in Koushien 2 with the DSP echo buffer corrupting SMP opcodes. Technically the PPU memory pools are only ever tripped by the CPU poking at them, as the PPU doesn't ever write. I now have run.for, run.to, step.for, step.to. The difference is that run only prints the next instruction after running, whereas step prints all of the instructions along the way as well. run.to acts the same as "step over" here. Although it's not quite as nice, since you have to specify the address of the next instruction. Logging the Field/Vcounter/Hcounter on instruction listings now, good for timing information. Added in the tracer mask, as well as memory export, as well as VRAM/OAM/CGRAM/SMP read/write/execute breakpoints, as well as an APU usage map (it tracks DSP reads/writes separately, although I don't currently have debugger callbacks on DSP accesses just yet.) Have not hooked up actual SMP debugging just yet, but I plan to soon. Still thinking about how I want to allow / block interleaving of instructions (terminal output and tracing.) So ... remaining tasks at this point: - full SMP debugging - CPU+SMP interleave support - aliases - hotkeys - save states (will be kind of tricky ... will have to suppress breakpoints during synchronization, or abort a save in a break event.) - keep track of window geometry between runs
2014-02-05 11:30:08 +00:00
void Debugger::main() {
if(running == false) {
usleep(20 * 1000);
Update to v094r05 release. byuu says: Commands can be prefixed with: (cpu|smp|ppu|dsp|apu|vram|oam|cgram)/ to set their source. Eg "vram/hex 0800" or "smp/breakpoints.append execute ffc0"; default is cpu. These overlap a little bit in odd ways, but that's just the way the SNES works: it's not a very orthogonal system. CPU is both a processor and the main bus (ROM, RAM, WRAM, etc), APU is the shared memory by the SMP+DSP (eg use it to catch writes from either chip); PPU probably won't ever be used since it's broken down into three separate buses (VRAM, OAM, CGRAM), but DSP could be useful for tracking bugs like we found in Koushien 2 with the DSP echo buffer corrupting SMP opcodes. Technically the PPU memory pools are only ever tripped by the CPU poking at them, as the PPU doesn't ever write. I now have run.for, run.to, step.for, step.to. The difference is that run only prints the next instruction after running, whereas step prints all of the instructions along the way as well. run.to acts the same as "step over" here. Although it's not quite as nice, since you have to specify the address of the next instruction. Logging the Field/Vcounter/Hcounter on instruction listings now, good for timing information. Added in the tracer mask, as well as memory export, as well as VRAM/OAM/CGRAM/SMP read/write/execute breakpoints, as well as an APU usage map (it tracks DSP reads/writes separately, although I don't currently have debugger callbacks on DSP accesses just yet.) Have not hooked up actual SMP debugging just yet, but I plan to soon. Still thinking about how I want to allow / block interleaving of instructions (terminal output and tracing.) So ... remaining tasks at this point: - full SMP debugging - CPU+SMP interleave support - aliases - hotkeys - save states (will be kind of tricky ... will have to suppress breakpoints during synchronization, or abort a save in a break event.) - keep track of window geometry between runs
2014-02-05 11:30:08 +00:00
return;
}
Update to v094r05 release. byuu says: Commands can be prefixed with: (cpu|smp|ppu|dsp|apu|vram|oam|cgram)/ to set their source. Eg "vram/hex 0800" or "smp/breakpoints.append execute ffc0"; default is cpu. These overlap a little bit in odd ways, but that's just the way the SNES works: it's not a very orthogonal system. CPU is both a processor and the main bus (ROM, RAM, WRAM, etc), APU is the shared memory by the SMP+DSP (eg use it to catch writes from either chip); PPU probably won't ever be used since it's broken down into three separate buses (VRAM, OAM, CGRAM), but DSP could be useful for tracking bugs like we found in Koushien 2 with the DSP echo buffer corrupting SMP opcodes. Technically the PPU memory pools are only ever tripped by the CPU poking at them, as the PPU doesn't ever write. I now have run.for, run.to, step.for, step.to. The difference is that run only prints the next instruction after running, whereas step prints all of the instructions along the way as well. run.to acts the same as "step over" here. Although it's not quite as nice, since you have to specify the address of the next instruction. Logging the Field/Vcounter/Hcounter on instruction listings now, good for timing information. Added in the tracer mask, as well as memory export, as well as VRAM/OAM/CGRAM/SMP read/write/execute breakpoints, as well as an APU usage map (it tracks DSP reads/writes separately, although I don't currently have debugger callbacks on DSP accesses just yet.) Have not hooked up actual SMP debugging just yet, but I plan to soon. Still thinking about how I want to allow / block interleaving of instructions (terminal output and tracing.) So ... remaining tasks at this point: - full SMP debugging - CPU+SMP interleave support - aliases - hotkeys - save states (will be kind of tricky ... will have to suppress breakpoints during synchronization, or abort a save in a break event.) - keep track of window geometry between runs
2014-02-05 11:30:08 +00:00
emulator->run();
}
Update to v094r05 release. byuu says: Commands can be prefixed with: (cpu|smp|ppu|dsp|apu|vram|oam|cgram)/ to set their source. Eg "vram/hex 0800" or "smp/breakpoints.append execute ffc0"; default is cpu. These overlap a little bit in odd ways, but that's just the way the SNES works: it's not a very orthogonal system. CPU is both a processor and the main bus (ROM, RAM, WRAM, etc), APU is the shared memory by the SMP+DSP (eg use it to catch writes from either chip); PPU probably won't ever be used since it's broken down into three separate buses (VRAM, OAM, CGRAM), but DSP could be useful for tracking bugs like we found in Koushien 2 with the DSP echo buffer corrupting SMP opcodes. Technically the PPU memory pools are only ever tripped by the CPU poking at them, as the PPU doesn't ever write. I now have run.for, run.to, step.for, step.to. The difference is that run only prints the next instruction after running, whereas step prints all of the instructions along the way as well. run.to acts the same as "step over" here. Although it's not quite as nice, since you have to specify the address of the next instruction. Logging the Field/Vcounter/Hcounter on instruction listings now, good for timing information. Added in the tracer mask, as well as memory export, as well as VRAM/OAM/CGRAM/SMP read/write/execute breakpoints, as well as an APU usage map (it tracks DSP reads/writes separately, although I don't currently have debugger callbacks on DSP accesses just yet.) Have not hooked up actual SMP debugging just yet, but I plan to soon. Still thinking about how I want to allow / block interleaving of instructions (terminal output and tracing.) So ... remaining tasks at this point: - full SMP debugging - CPU+SMP interleave support - aliases - hotkeys - save states (will be kind of tricky ... will have to suppress breakpoints during synchronization, or abort a save in a break event.) - keep track of window geometry between runs
2014-02-05 11:30:08 +00:00
void Debugger::run() {
running = true;
}
void Debugger::stop() {
Update to v094r05 release. byuu says: Commands can be prefixed with: (cpu|smp|ppu|dsp|apu|vram|oam|cgram)/ to set their source. Eg "vram/hex 0800" or "smp/breakpoints.append execute ffc0"; default is cpu. These overlap a little bit in odd ways, but that's just the way the SNES works: it's not a very orthogonal system. CPU is both a processor and the main bus (ROM, RAM, WRAM, etc), APU is the shared memory by the SMP+DSP (eg use it to catch writes from either chip); PPU probably won't ever be used since it's broken down into three separate buses (VRAM, OAM, CGRAM), but DSP could be useful for tracking bugs like we found in Koushien 2 with the DSP echo buffer corrupting SMP opcodes. Technically the PPU memory pools are only ever tripped by the CPU poking at them, as the PPU doesn't ever write. I now have run.for, run.to, step.for, step.to. The difference is that run only prints the next instruction after running, whereas step prints all of the instructions along the way as well. run.to acts the same as "step over" here. Although it's not quite as nice, since you have to specify the address of the next instruction. Logging the Field/Vcounter/Hcounter on instruction listings now, good for timing information. Added in the tracer mask, as well as memory export, as well as VRAM/OAM/CGRAM/SMP read/write/execute breakpoints, as well as an APU usage map (it tracks DSP reads/writes separately, although I don't currently have debugger callbacks on DSP accesses just yet.) Have not hooked up actual SMP debugging just yet, but I plan to soon. Still thinking about how I want to allow / block interleaving of instructions (terminal output and tracing.) So ... remaining tasks at this point: - full SMP debugging - CPU+SMP interleave support - aliases - hotkeys - save states (will be kind of tricky ... will have to suppress breakpoints during synchronization, or abort a save in a break event.) - keep track of window geometry between runs
2014-02-05 11:30:08 +00:00
running = false;
cpuRunFor = nothing;
cpuRunTo = nothing;
cpuStepFor = nothing;
cpuStepTo = nothing;
smpRunFor = nothing;
smpRunTo = nothing;
smpStepFor = nothing;
smpStepTo = nothing;
Update to v094r05 release. byuu says: Commands can be prefixed with: (cpu|smp|ppu|dsp|apu|vram|oam|cgram)/ to set their source. Eg "vram/hex 0800" or "smp/breakpoints.append execute ffc0"; default is cpu. These overlap a little bit in odd ways, but that's just the way the SNES works: it's not a very orthogonal system. CPU is both a processor and the main bus (ROM, RAM, WRAM, etc), APU is the shared memory by the SMP+DSP (eg use it to catch writes from either chip); PPU probably won't ever be used since it's broken down into three separate buses (VRAM, OAM, CGRAM), but DSP could be useful for tracking bugs like we found in Koushien 2 with the DSP echo buffer corrupting SMP opcodes. Technically the PPU memory pools are only ever tripped by the CPU poking at them, as the PPU doesn't ever write. I now have run.for, run.to, step.for, step.to. The difference is that run only prints the next instruction after running, whereas step prints all of the instructions along the way as well. run.to acts the same as "step over" here. Although it's not quite as nice, since you have to specify the address of the next instruction. Logging the Field/Vcounter/Hcounter on instruction listings now, good for timing information. Added in the tracer mask, as well as memory export, as well as VRAM/OAM/CGRAM/SMP read/write/execute breakpoints, as well as an APU usage map (it tracks DSP reads/writes separately, although I don't currently have debugger callbacks on DSP accesses just yet.) Have not hooked up actual SMP debugging just yet, but I plan to soon. Still thinking about how I want to allow / block interleaving of instructions (terminal output and tracing.) So ... remaining tasks at this point: - full SMP debugging - CPU+SMP interleave support - aliases - hotkeys - save states (will be kind of tricky ... will have to suppress breakpoints during synchronization, or abort a save in a break event.) - keep track of window geometry between runs
2014-02-05 11:30:08 +00:00
}
void Debugger::leave() {
stop();
SFC::scheduler.debug();
}
bool Debugger::breakpointTest(Source source, Breakpoint::Mode mode, unsigned addr, uint8 data) {
if(savingState) return false;
Update to v094r05 release. byuu says: Commands can be prefixed with: (cpu|smp|ppu|dsp|apu|vram|oam|cgram)/ to set their source. Eg "vram/hex 0800" or "smp/breakpoints.append execute ffc0"; default is cpu. These overlap a little bit in odd ways, but that's just the way the SNES works: it's not a very orthogonal system. CPU is both a processor and the main bus (ROM, RAM, WRAM, etc), APU is the shared memory by the SMP+DSP (eg use it to catch writes from either chip); PPU probably won't ever be used since it's broken down into three separate buses (VRAM, OAM, CGRAM), but DSP could be useful for tracking bugs like we found in Koushien 2 with the DSP echo buffer corrupting SMP opcodes. Technically the PPU memory pools are only ever tripped by the CPU poking at them, as the PPU doesn't ever write. I now have run.for, run.to, step.for, step.to. The difference is that run only prints the next instruction after running, whereas step prints all of the instructions along the way as well. run.to acts the same as "step over" here. Although it's not quite as nice, since you have to specify the address of the next instruction. Logging the Field/Vcounter/Hcounter on instruction listings now, good for timing information. Added in the tracer mask, as well as memory export, as well as VRAM/OAM/CGRAM/SMP read/write/execute breakpoints, as well as an APU usage map (it tracks DSP reads/writes separately, although I don't currently have debugger callbacks on DSP accesses just yet.) Have not hooked up actual SMP debugging just yet, but I plan to soon. Still thinking about how I want to allow / block interleaving of instructions (terminal output and tracing.) So ... remaining tasks at this point: - full SMP debugging - CPU+SMP interleave support - aliases - hotkeys - save states (will be kind of tricky ... will have to suppress breakpoints during synchronization, or abort a save in a break event.) - keep track of window geometry between runs
2014-02-05 11:30:08 +00:00
for(unsigned n = 0; n < breakpoints.size(); n++) {
auto& bp = breakpoints[n];
if(bp.source != source) continue;
if(bp.mode != mode) continue;
if(bp.addr != addr) continue;
if(bp.mode != Breakpoint::Mode::Execute && bp.data && bp.data() != data) continue;
string output = {"Breakpoint #", n, " hit"};
if(bp.mode == Breakpoint::Mode::Read ) output.append("; read ", hex<2>(data));
if(bp.mode == Breakpoint::Mode::Write) output.append("; wrote ", hex<2>(data));
output.append("; triggered: ", ++bp.triggered);
echo(output, "\n");
Update to v094r05 release. byuu says: Commands can be prefixed with: (cpu|smp|ppu|dsp|apu|vram|oam|cgram)/ to set their source. Eg "vram/hex 0800" or "smp/breakpoints.append execute ffc0"; default is cpu. These overlap a little bit in odd ways, but that's just the way the SNES works: it's not a very orthogonal system. CPU is both a processor and the main bus (ROM, RAM, WRAM, etc), APU is the shared memory by the SMP+DSP (eg use it to catch writes from either chip); PPU probably won't ever be used since it's broken down into three separate buses (VRAM, OAM, CGRAM), but DSP could be useful for tracking bugs like we found in Koushien 2 with the DSP echo buffer corrupting SMP opcodes. Technically the PPU memory pools are only ever tripped by the CPU poking at them, as the PPU doesn't ever write. I now have run.for, run.to, step.for, step.to. The difference is that run only prints the next instruction after running, whereas step prints all of the instructions along the way as well. run.to acts the same as "step over" here. Although it's not quite as nice, since you have to specify the address of the next instruction. Logging the Field/Vcounter/Hcounter on instruction listings now, good for timing information. Added in the tracer mask, as well as memory export, as well as VRAM/OAM/CGRAM/SMP read/write/execute breakpoints, as well as an APU usage map (it tracks DSP reads/writes separately, although I don't currently have debugger callbacks on DSP accesses just yet.) Have not hooked up actual SMP debugging just yet, but I plan to soon. Still thinking about how I want to allow / block interleaving of instructions (terminal output and tracing.) So ... remaining tasks at this point: - full SMP debugging - CPU+SMP interleave support - aliases - hotkeys - save states (will be kind of tricky ... will have to suppress breakpoints during synchronization, or abort a save in a break event.) - keep track of window geometry between runs
2014-02-05 11:30:08 +00:00
return true;
}
return false;
}
string Debugger::cpuDisassemble() {
char text[4096];
SFC::cpu.disassemble_opcode(text);
return {text, " F:", (unsigned)SFC::cpu.field(), " V:", format<3>(SFC::cpu.vcounter()), " H:", format<4>(SFC::cpu.hcounter())};
}
string Debugger::cpuDisassemble(unsigned addr, bool e, bool m, bool x) {
char text[4096];
SFC::cpu.disassemble_opcode(text, addr, e, m, x);
return {text, " F:", (unsigned)SFC::cpu.field(), " V:", format<3>(SFC::cpu.vcounter()), " H:", format<4>(SFC::cpu.hcounter())};
}
void Debugger::cpuExec(uint24 addr) {
cpuUsage[addr] |= Usage::Execute;
if(SFC::cpu.regs.e == 0) cpuUsage[addr] &= ~Usage::FlagE;
if(SFC::cpu.regs.p.m == 0) cpuUsage[addr] &= ~Usage::FlagM;
if(SFC::cpu.regs.p.x == 0) cpuUsage[addr] &= ~Usage::FlagX;
if(SFC::cpu.regs.e == 1) cpuUsage[addr] |= Usage::FlagE;
if(SFC::cpu.regs.p.m == 1) cpuUsage[addr] |= Usage::FlagM;
if(SFC::cpu.regs.p.x == 1) cpuUsage[addr] |= Usage::FlagX;
cpuInstructionCounter++;
if(cpuTracerFile.open()) {
if(!cpuTracerMask || cpuTracerMask[addr] == false) {
if(cpuTracerMask) cpuTracerMask[addr] = true;
cpuTracerFile.print(cpuDisassemble(), "\n");
Update to v094r05 release. byuu says: Commands can be prefixed with: (cpu|smp|ppu|dsp|apu|vram|oam|cgram)/ to set their source. Eg "vram/hex 0800" or "smp/breakpoints.append execute ffc0"; default is cpu. These overlap a little bit in odd ways, but that's just the way the SNES works: it's not a very orthogonal system. CPU is both a processor and the main bus (ROM, RAM, WRAM, etc), APU is the shared memory by the SMP+DSP (eg use it to catch writes from either chip); PPU probably won't ever be used since it's broken down into three separate buses (VRAM, OAM, CGRAM), but DSP could be useful for tracking bugs like we found in Koushien 2 with the DSP echo buffer corrupting SMP opcodes. Technically the PPU memory pools are only ever tripped by the CPU poking at them, as the PPU doesn't ever write. I now have run.for, run.to, step.for, step.to. The difference is that run only prints the next instruction after running, whereas step prints all of the instructions along the way as well. run.to acts the same as "step over" here. Although it's not quite as nice, since you have to specify the address of the next instruction. Logging the Field/Vcounter/Hcounter on instruction listings now, good for timing information. Added in the tracer mask, as well as memory export, as well as VRAM/OAM/CGRAM/SMP read/write/execute breakpoints, as well as an APU usage map (it tracks DSP reads/writes separately, although I don't currently have debugger callbacks on DSP accesses just yet.) Have not hooked up actual SMP debugging just yet, but I plan to soon. Still thinking about how I want to allow / block interleaving of instructions (terminal output and tracing.) So ... remaining tasks at this point: - full SMP debugging - CPU+SMP interleave support - aliases - hotkeys - save states (will be kind of tricky ... will have to suppress breakpoints during synchronization, or abort a save in a break event.) - keep track of window geometry between runs
2014-02-05 11:30:08 +00:00
}
}
if(savingState) return;
Update to v094r05 release. byuu says: Commands can be prefixed with: (cpu|smp|ppu|dsp|apu|vram|oam|cgram)/ to set their source. Eg "vram/hex 0800" or "smp/breakpoints.append execute ffc0"; default is cpu. These overlap a little bit in odd ways, but that's just the way the SNES works: it's not a very orthogonal system. CPU is both a processor and the main bus (ROM, RAM, WRAM, etc), APU is the shared memory by the SMP+DSP (eg use it to catch writes from either chip); PPU probably won't ever be used since it's broken down into three separate buses (VRAM, OAM, CGRAM), but DSP could be useful for tracking bugs like we found in Koushien 2 with the DSP echo buffer corrupting SMP opcodes. Technically the PPU memory pools are only ever tripped by the CPU poking at them, as the PPU doesn't ever write. I now have run.for, run.to, step.for, step.to. The difference is that run only prints the next instruction after running, whereas step prints all of the instructions along the way as well. run.to acts the same as "step over" here. Although it's not quite as nice, since you have to specify the address of the next instruction. Logging the Field/Vcounter/Hcounter on instruction listings now, good for timing information. Added in the tracer mask, as well as memory export, as well as VRAM/OAM/CGRAM/SMP read/write/execute breakpoints, as well as an APU usage map (it tracks DSP reads/writes separately, although I don't currently have debugger callbacks on DSP accesses just yet.) Have not hooked up actual SMP debugging just yet, but I plan to soon. Still thinking about how I want to allow / block interleaving of instructions (terminal output and tracing.) So ... remaining tasks at this point: - full SMP debugging - CPU+SMP interleave support - aliases - hotkeys - save states (will be kind of tricky ... will have to suppress breakpoints during synchronization, or abort a save in a break event.) - keep track of window geometry between runs
2014-02-05 11:30:08 +00:00
if(breakpointTest(Source::CPU, Breakpoint::Mode::Execute, addr)) {
echo(cpuDisassemble(), "\n");
Update to v094r05 release. byuu says: Commands can be prefixed with: (cpu|smp|ppu|dsp|apu|vram|oam|cgram)/ to set their source. Eg "vram/hex 0800" or "smp/breakpoints.append execute ffc0"; default is cpu. These overlap a little bit in odd ways, but that's just the way the SNES works: it's not a very orthogonal system. CPU is both a processor and the main bus (ROM, RAM, WRAM, etc), APU is the shared memory by the SMP+DSP (eg use it to catch writes from either chip); PPU probably won't ever be used since it's broken down into three separate buses (VRAM, OAM, CGRAM), but DSP could be useful for tracking bugs like we found in Koushien 2 with the DSP echo buffer corrupting SMP opcodes. Technically the PPU memory pools are only ever tripped by the CPU poking at them, as the PPU doesn't ever write. I now have run.for, run.to, step.for, step.to. The difference is that run only prints the next instruction after running, whereas step prints all of the instructions along the way as well. run.to acts the same as "step over" here. Although it's not quite as nice, since you have to specify the address of the next instruction. Logging the Field/Vcounter/Hcounter on instruction listings now, good for timing information. Added in the tracer mask, as well as memory export, as well as VRAM/OAM/CGRAM/SMP read/write/execute breakpoints, as well as an APU usage map (it tracks DSP reads/writes separately, although I don't currently have debugger callbacks on DSP accesses just yet.) Have not hooked up actual SMP debugging just yet, but I plan to soon. Still thinking about how I want to allow / block interleaving of instructions (terminal output and tracing.) So ... remaining tasks at this point: - full SMP debugging - CPU+SMP interleave support - aliases - hotkeys - save states (will be kind of tricky ... will have to suppress breakpoints during synchronization, or abort a save in a break event.) - keep track of window geometry between runs
2014-02-05 11:30:08 +00:00
return leave();
}
Update to v094r05 release. byuu says: Commands can be prefixed with: (cpu|smp|ppu|dsp|apu|vram|oam|cgram)/ to set their source. Eg "vram/hex 0800" or "smp/breakpoints.append execute ffc0"; default is cpu. These overlap a little bit in odd ways, but that's just the way the SNES works: it's not a very orthogonal system. CPU is both a processor and the main bus (ROM, RAM, WRAM, etc), APU is the shared memory by the SMP+DSP (eg use it to catch writes from either chip); PPU probably won't ever be used since it's broken down into three separate buses (VRAM, OAM, CGRAM), but DSP could be useful for tracking bugs like we found in Koushien 2 with the DSP echo buffer corrupting SMP opcodes. Technically the PPU memory pools are only ever tripped by the CPU poking at them, as the PPU doesn't ever write. I now have run.for, run.to, step.for, step.to. The difference is that run only prints the next instruction after running, whereas step prints all of the instructions along the way as well. run.to acts the same as "step over" here. Although it's not quite as nice, since you have to specify the address of the next instruction. Logging the Field/Vcounter/Hcounter on instruction listings now, good for timing information. Added in the tracer mask, as well as memory export, as well as VRAM/OAM/CGRAM/SMP read/write/execute breakpoints, as well as an APU usage map (it tracks DSP reads/writes separately, although I don't currently have debugger callbacks on DSP accesses just yet.) Have not hooked up actual SMP debugging just yet, but I plan to soon. Still thinking about how I want to allow / block interleaving of instructions (terminal output and tracing.) So ... remaining tasks at this point: - full SMP debugging - CPU+SMP interleave support - aliases - hotkeys - save states (will be kind of tricky ... will have to suppress breakpoints during synchronization, or abort a save in a break event.) - keep track of window geometry between runs
2014-02-05 11:30:08 +00:00
if(cpuRunFor) {
if(--cpuRunFor() == 0) {
echo(cpuDisassemble(), "\n");
return leave();
}
Update to v094r05 release. byuu says: Commands can be prefixed with: (cpu|smp|ppu|dsp|apu|vram|oam|cgram)/ to set their source. Eg "vram/hex 0800" or "smp/breakpoints.append execute ffc0"; default is cpu. These overlap a little bit in odd ways, but that's just the way the SNES works: it's not a very orthogonal system. CPU is both a processor and the main bus (ROM, RAM, WRAM, etc), APU is the shared memory by the SMP+DSP (eg use it to catch writes from either chip); PPU probably won't ever be used since it's broken down into three separate buses (VRAM, OAM, CGRAM), but DSP could be useful for tracking bugs like we found in Koushien 2 with the DSP echo buffer corrupting SMP opcodes. Technically the PPU memory pools are only ever tripped by the CPU poking at them, as the PPU doesn't ever write. I now have run.for, run.to, step.for, step.to. The difference is that run only prints the next instruction after running, whereas step prints all of the instructions along the way as well. run.to acts the same as "step over" here. Although it's not quite as nice, since you have to specify the address of the next instruction. Logging the Field/Vcounter/Hcounter on instruction listings now, good for timing information. Added in the tracer mask, as well as memory export, as well as VRAM/OAM/CGRAM/SMP read/write/execute breakpoints, as well as an APU usage map (it tracks DSP reads/writes separately, although I don't currently have debugger callbacks on DSP accesses just yet.) Have not hooked up actual SMP debugging just yet, but I plan to soon. Still thinking about how I want to allow / block interleaving of instructions (terminal output and tracing.) So ... remaining tasks at this point: - full SMP debugging - CPU+SMP interleave support - aliases - hotkeys - save states (will be kind of tricky ... will have to suppress breakpoints during synchronization, or abort a save in a break event.) - keep track of window geometry between runs
2014-02-05 11:30:08 +00:00
}
Update to v094r05 release. byuu says: Commands can be prefixed with: (cpu|smp|ppu|dsp|apu|vram|oam|cgram)/ to set their source. Eg "vram/hex 0800" or "smp/breakpoints.append execute ffc0"; default is cpu. These overlap a little bit in odd ways, but that's just the way the SNES works: it's not a very orthogonal system. CPU is both a processor and the main bus (ROM, RAM, WRAM, etc), APU is the shared memory by the SMP+DSP (eg use it to catch writes from either chip); PPU probably won't ever be used since it's broken down into three separate buses (VRAM, OAM, CGRAM), but DSP could be useful for tracking bugs like we found in Koushien 2 with the DSP echo buffer corrupting SMP opcodes. Technically the PPU memory pools are only ever tripped by the CPU poking at them, as the PPU doesn't ever write. I now have run.for, run.to, step.for, step.to. The difference is that run only prints the next instruction after running, whereas step prints all of the instructions along the way as well. run.to acts the same as "step over" here. Although it's not quite as nice, since you have to specify the address of the next instruction. Logging the Field/Vcounter/Hcounter on instruction listings now, good for timing information. Added in the tracer mask, as well as memory export, as well as VRAM/OAM/CGRAM/SMP read/write/execute breakpoints, as well as an APU usage map (it tracks DSP reads/writes separately, although I don't currently have debugger callbacks on DSP accesses just yet.) Have not hooked up actual SMP debugging just yet, but I plan to soon. Still thinking about how I want to allow / block interleaving of instructions (terminal output and tracing.) So ... remaining tasks at this point: - full SMP debugging - CPU+SMP interleave support - aliases - hotkeys - save states (will be kind of tricky ... will have to suppress breakpoints during synchronization, or abort a save in a break event.) - keep track of window geometry between runs
2014-02-05 11:30:08 +00:00
if(cpuRunTo) {
if(addr == cpuRunTo()) {
echo(cpuDisassemble(), "\n");
return leave();
}
Update to v094r05 release. byuu says: Commands can be prefixed with: (cpu|smp|ppu|dsp|apu|vram|oam|cgram)/ to set their source. Eg "vram/hex 0800" or "smp/breakpoints.append execute ffc0"; default is cpu. These overlap a little bit in odd ways, but that's just the way the SNES works: it's not a very orthogonal system. CPU is both a processor and the main bus (ROM, RAM, WRAM, etc), APU is the shared memory by the SMP+DSP (eg use it to catch writes from either chip); PPU probably won't ever be used since it's broken down into three separate buses (VRAM, OAM, CGRAM), but DSP could be useful for tracking bugs like we found in Koushien 2 with the DSP echo buffer corrupting SMP opcodes. Technically the PPU memory pools are only ever tripped by the CPU poking at them, as the PPU doesn't ever write. I now have run.for, run.to, step.for, step.to. The difference is that run only prints the next instruction after running, whereas step prints all of the instructions along the way as well. run.to acts the same as "step over" here. Although it's not quite as nice, since you have to specify the address of the next instruction. Logging the Field/Vcounter/Hcounter on instruction listings now, good for timing information. Added in the tracer mask, as well as memory export, as well as VRAM/OAM/CGRAM/SMP read/write/execute breakpoints, as well as an APU usage map (it tracks DSP reads/writes separately, although I don't currently have debugger callbacks on DSP accesses just yet.) Have not hooked up actual SMP debugging just yet, but I plan to soon. Still thinking about how I want to allow / block interleaving of instructions (terminal output and tracing.) So ... remaining tasks at this point: - full SMP debugging - CPU+SMP interleave support - aliases - hotkeys - save states (will be kind of tricky ... will have to suppress breakpoints during synchronization, or abort a save in a break event.) - keep track of window geometry between runs
2014-02-05 11:30:08 +00:00
}
Update to v094r05 release. byuu says: Commands can be prefixed with: (cpu|smp|ppu|dsp|apu|vram|oam|cgram)/ to set their source. Eg "vram/hex 0800" or "smp/breakpoints.append execute ffc0"; default is cpu. These overlap a little bit in odd ways, but that's just the way the SNES works: it's not a very orthogonal system. CPU is both a processor and the main bus (ROM, RAM, WRAM, etc), APU is the shared memory by the SMP+DSP (eg use it to catch writes from either chip); PPU probably won't ever be used since it's broken down into three separate buses (VRAM, OAM, CGRAM), but DSP could be useful for tracking bugs like we found in Koushien 2 with the DSP echo buffer corrupting SMP opcodes. Technically the PPU memory pools are only ever tripped by the CPU poking at them, as the PPU doesn't ever write. I now have run.for, run.to, step.for, step.to. The difference is that run only prints the next instruction after running, whereas step prints all of the instructions along the way as well. run.to acts the same as "step over" here. Although it's not quite as nice, since you have to specify the address of the next instruction. Logging the Field/Vcounter/Hcounter on instruction listings now, good for timing information. Added in the tracer mask, as well as memory export, as well as VRAM/OAM/CGRAM/SMP read/write/execute breakpoints, as well as an APU usage map (it tracks DSP reads/writes separately, although I don't currently have debugger callbacks on DSP accesses just yet.) Have not hooked up actual SMP debugging just yet, but I plan to soon. Still thinking about how I want to allow / block interleaving of instructions (terminal output and tracing.) So ... remaining tasks at this point: - full SMP debugging - CPU+SMP interleave support - aliases - hotkeys - save states (will be kind of tricky ... will have to suppress breakpoints during synchronization, or abort a save in a break event.) - keep track of window geometry between runs
2014-02-05 11:30:08 +00:00
if(cpuStepFor) {
echo(cpuDisassemble(), "\n");
if(--cpuStepFor() == 0) return leave();
}
Update to v094r05 release. byuu says: Commands can be prefixed with: (cpu|smp|ppu|dsp|apu|vram|oam|cgram)/ to set their source. Eg "vram/hex 0800" or "smp/breakpoints.append execute ffc0"; default is cpu. These overlap a little bit in odd ways, but that's just the way the SNES works: it's not a very orthogonal system. CPU is both a processor and the main bus (ROM, RAM, WRAM, etc), APU is the shared memory by the SMP+DSP (eg use it to catch writes from either chip); PPU probably won't ever be used since it's broken down into three separate buses (VRAM, OAM, CGRAM), but DSP could be useful for tracking bugs like we found in Koushien 2 with the DSP echo buffer corrupting SMP opcodes. Technically the PPU memory pools are only ever tripped by the CPU poking at them, as the PPU doesn't ever write. I now have run.for, run.to, step.for, step.to. The difference is that run only prints the next instruction after running, whereas step prints all of the instructions along the way as well. run.to acts the same as "step over" here. Although it's not quite as nice, since you have to specify the address of the next instruction. Logging the Field/Vcounter/Hcounter on instruction listings now, good for timing information. Added in the tracer mask, as well as memory export, as well as VRAM/OAM/CGRAM/SMP read/write/execute breakpoints, as well as an APU usage map (it tracks DSP reads/writes separately, although I don't currently have debugger callbacks on DSP accesses just yet.) Have not hooked up actual SMP debugging just yet, but I plan to soon. Still thinking about how I want to allow / block interleaving of instructions (terminal output and tracing.) So ... remaining tasks at this point: - full SMP debugging - CPU+SMP interleave support - aliases - hotkeys - save states (will be kind of tricky ... will have to suppress breakpoints during synchronization, or abort a save in a break event.) - keep track of window geometry between runs
2014-02-05 11:30:08 +00:00
if(cpuStepTo) {
echo(cpuDisassemble(), "\n");
Update to v094r05 release. byuu says: Commands can be prefixed with: (cpu|smp|ppu|dsp|apu|vram|oam|cgram)/ to set their source. Eg "vram/hex 0800" or "smp/breakpoints.append execute ffc0"; default is cpu. These overlap a little bit in odd ways, but that's just the way the SNES works: it's not a very orthogonal system. CPU is both a processor and the main bus (ROM, RAM, WRAM, etc), APU is the shared memory by the SMP+DSP (eg use it to catch writes from either chip); PPU probably won't ever be used since it's broken down into three separate buses (VRAM, OAM, CGRAM), but DSP could be useful for tracking bugs like we found in Koushien 2 with the DSP echo buffer corrupting SMP opcodes. Technically the PPU memory pools are only ever tripped by the CPU poking at them, as the PPU doesn't ever write. I now have run.for, run.to, step.for, step.to. The difference is that run only prints the next instruction after running, whereas step prints all of the instructions along the way as well. run.to acts the same as "step over" here. Although it's not quite as nice, since you have to specify the address of the next instruction. Logging the Field/Vcounter/Hcounter on instruction listings now, good for timing information. Added in the tracer mask, as well as memory export, as well as VRAM/OAM/CGRAM/SMP read/write/execute breakpoints, as well as an APU usage map (it tracks DSP reads/writes separately, although I don't currently have debugger callbacks on DSP accesses just yet.) Have not hooked up actual SMP debugging just yet, but I plan to soon. Still thinking about how I want to allow / block interleaving of instructions (terminal output and tracing.) So ... remaining tasks at this point: - full SMP debugging - CPU+SMP interleave support - aliases - hotkeys - save states (will be kind of tricky ... will have to suppress breakpoints during synchronization, or abort a save in a break event.) - keep track of window geometry between runs
2014-02-05 11:30:08 +00:00
if(addr == cpuStepTo()) return leave();
}
}
void Debugger::cpuRead(uint24 addr, uint8 data) {
cpuUsage[addr] |= Usage::Read;
Update to v094r05 release. byuu says: Commands can be prefixed with: (cpu|smp|ppu|dsp|apu|vram|oam|cgram)/ to set their source. Eg "vram/hex 0800" or "smp/breakpoints.append execute ffc0"; default is cpu. These overlap a little bit in odd ways, but that's just the way the SNES works: it's not a very orthogonal system. CPU is both a processor and the main bus (ROM, RAM, WRAM, etc), APU is the shared memory by the SMP+DSP (eg use it to catch writes from either chip); PPU probably won't ever be used since it's broken down into three separate buses (VRAM, OAM, CGRAM), but DSP could be useful for tracking bugs like we found in Koushien 2 with the DSP echo buffer corrupting SMP opcodes. Technically the PPU memory pools are only ever tripped by the CPU poking at them, as the PPU doesn't ever write. I now have run.for, run.to, step.for, step.to. The difference is that run only prints the next instruction after running, whereas step prints all of the instructions along the way as well. run.to acts the same as "step over" here. Although it's not quite as nice, since you have to specify the address of the next instruction. Logging the Field/Vcounter/Hcounter on instruction listings now, good for timing information. Added in the tracer mask, as well as memory export, as well as VRAM/OAM/CGRAM/SMP read/write/execute breakpoints, as well as an APU usage map (it tracks DSP reads/writes separately, although I don't currently have debugger callbacks on DSP accesses just yet.) Have not hooked up actual SMP debugging just yet, but I plan to soon. Still thinking about how I want to allow / block interleaving of instructions (terminal output and tracing.) So ... remaining tasks at this point: - full SMP debugging - CPU+SMP interleave support - aliases - hotkeys - save states (will be kind of tricky ... will have to suppress breakpoints during synchronization, or abort a save in a break event.) - keep track of window geometry between runs
2014-02-05 11:30:08 +00:00
if(breakpointTest(Source::CPU, Breakpoint::Mode::Read, addr, data)) leave();
}
void Debugger::cpuWrite(uint24 addr, uint8 data) {
cpuUsage[addr] |= Usage::Write;
Update to v094r05 release. byuu says: Commands can be prefixed with: (cpu|smp|ppu|dsp|apu|vram|oam|cgram)/ to set their source. Eg "vram/hex 0800" or "smp/breakpoints.append execute ffc0"; default is cpu. These overlap a little bit in odd ways, but that's just the way the SNES works: it's not a very orthogonal system. CPU is both a processor and the main bus (ROM, RAM, WRAM, etc), APU is the shared memory by the SMP+DSP (eg use it to catch writes from either chip); PPU probably won't ever be used since it's broken down into three separate buses (VRAM, OAM, CGRAM), but DSP could be useful for tracking bugs like we found in Koushien 2 with the DSP echo buffer corrupting SMP opcodes. Technically the PPU memory pools are only ever tripped by the CPU poking at them, as the PPU doesn't ever write. I now have run.for, run.to, step.for, step.to. The difference is that run only prints the next instruction after running, whereas step prints all of the instructions along the way as well. run.to acts the same as "step over" here. Although it's not quite as nice, since you have to specify the address of the next instruction. Logging the Field/Vcounter/Hcounter on instruction listings now, good for timing information. Added in the tracer mask, as well as memory export, as well as VRAM/OAM/CGRAM/SMP read/write/execute breakpoints, as well as an APU usage map (it tracks DSP reads/writes separately, although I don't currently have debugger callbacks on DSP accesses just yet.) Have not hooked up actual SMP debugging just yet, but I plan to soon. Still thinking about how I want to allow / block interleaving of instructions (terminal output and tracing.) So ... remaining tasks at this point: - full SMP debugging - CPU+SMP interleave support - aliases - hotkeys - save states (will be kind of tricky ... will have to suppress breakpoints during synchronization, or abort a save in a break event.) - keep track of window geometry between runs
2014-02-05 11:30:08 +00:00
if(breakpointTest(Source::CPU, Breakpoint::Mode::Write, addr, data)) leave();
}
void Debugger::echoBreakpoints() {
if(breakpoints.size() == 0) return;
Update to v094r05 release. byuu says: Commands can be prefixed with: (cpu|smp|ppu|dsp|apu|vram|oam|cgram)/ to set their source. Eg "vram/hex 0800" or "smp/breakpoints.append execute ffc0"; default is cpu. These overlap a little bit in odd ways, but that's just the way the SNES works: it's not a very orthogonal system. CPU is both a processor and the main bus (ROM, RAM, WRAM, etc), APU is the shared memory by the SMP+DSP (eg use it to catch writes from either chip); PPU probably won't ever be used since it's broken down into three separate buses (VRAM, OAM, CGRAM), but DSP could be useful for tracking bugs like we found in Koushien 2 with the DSP echo buffer corrupting SMP opcodes. Technically the PPU memory pools are only ever tripped by the CPU poking at them, as the PPU doesn't ever write. I now have run.for, run.to, step.for, step.to. The difference is that run only prints the next instruction after running, whereas step prints all of the instructions along the way as well. run.to acts the same as "step over" here. Although it's not quite as nice, since you have to specify the address of the next instruction. Logging the Field/Vcounter/Hcounter on instruction listings now, good for timing information. Added in the tracer mask, as well as memory export, as well as VRAM/OAM/CGRAM/SMP read/write/execute breakpoints, as well as an APU usage map (it tracks DSP reads/writes separately, although I don't currently have debugger callbacks on DSP accesses just yet.) Have not hooked up actual SMP debugging just yet, but I plan to soon. Still thinking about how I want to allow / block interleaving of instructions (terminal output and tracing.) So ... remaining tasks at this point: - full SMP debugging - CPU+SMP interleave support - aliases - hotkeys - save states (will be kind of tricky ... will have to suppress breakpoints during synchronization, or abort a save in a break event.) - keep track of window geometry between runs
2014-02-05 11:30:08 +00:00
echo("# source type addr data triggered\n");
echo("--- ------ -------- ------ ---- ---------\n");
for(unsigned n = 0; n < breakpoints.size(); n++) {
auto& bp = breakpoints[n];
string output = {format<-3>(n), " "};
Update to v094r05 release. byuu says: Commands can be prefixed with: (cpu|smp|ppu|dsp|apu|vram|oam|cgram)/ to set their source. Eg "vram/hex 0800" or "smp/breakpoints.append execute ffc0"; default is cpu. These overlap a little bit in odd ways, but that's just the way the SNES works: it's not a very orthogonal system. CPU is both a processor and the main bus (ROM, RAM, WRAM, etc), APU is the shared memory by the SMP+DSP (eg use it to catch writes from either chip); PPU probably won't ever be used since it's broken down into three separate buses (VRAM, OAM, CGRAM), but DSP could be useful for tracking bugs like we found in Koushien 2 with the DSP echo buffer corrupting SMP opcodes. Technically the PPU memory pools are only ever tripped by the CPU poking at them, as the PPU doesn't ever write. I now have run.for, run.to, step.for, step.to. The difference is that run only prints the next instruction after running, whereas step prints all of the instructions along the way as well. run.to acts the same as "step over" here. Although it's not quite as nice, since you have to specify the address of the next instruction. Logging the Field/Vcounter/Hcounter on instruction listings now, good for timing information. Added in the tracer mask, as well as memory export, as well as VRAM/OAM/CGRAM/SMP read/write/execute breakpoints, as well as an APU usage map (it tracks DSP reads/writes separately, although I don't currently have debugger callbacks on DSP accesses just yet.) Have not hooked up actual SMP debugging just yet, but I plan to soon. Still thinking about how I want to allow / block interleaving of instructions (terminal output and tracing.) So ... remaining tasks at this point: - full SMP debugging - CPU+SMP interleave support - aliases - hotkeys - save states (will be kind of tricky ... will have to suppress breakpoints during synchronization, or abort a save in a break event.) - keep track of window geometry between runs
2014-02-05 11:30:08 +00:00
output.append(format<-6>(sourceName(bp.source)), " ");
if(bp.mode == Breakpoint::Mode::Disabled) output.append("disabled ");
if(bp.mode == Breakpoint::Mode::Read ) output.append("read ");
if(bp.mode == Breakpoint::Mode::Write ) output.append("write ");
if(bp.mode == Breakpoint::Mode::Execute ) output.append("execute ");
output.append(hex<6>(bp.addr), " ");
output.append(bp.data ? hex<2>(bp.data()) : " ", " ");
output.append(format<-9>(bp.triggered));
echo(output, "\n");
}
}
void Debugger::echoDisassemble(Source source, unsigned addr, signed size) {
if(source != Source::CPU && source != Source::SMP) return;
const unsigned maximumDisplacement = (source == Source::CPU ? 5 : 4); //maximum opcode length
uint8* usage = (source == Source::CPU ? cpuUsage : apuUsage);
if(!(usage[addr] & Usage::Execute)) return echo("No usage data available for ", sourceName(source), "/", hex<6>(addr), "\n");
while(size > 0) {
string text;
if(source == Source::CPU) {
text = cpuDisassemble(addr, usage[addr] & Usage::FlagE, usage[addr] & Usage::FlagM, usage[addr] & Usage::FlagX);
}
if(source == Source::SMP) {
text = smpDisassemble(addr, usage[addr] & Usage::FlagP);
}
Update to v094r05 release. byuu says: Commands can be prefixed with: (cpu|smp|ppu|dsp|apu|vram|oam|cgram)/ to set their source. Eg "vram/hex 0800" or "smp/breakpoints.append execute ffc0"; default is cpu. These overlap a little bit in odd ways, but that's just the way the SNES works: it's not a very orthogonal system. CPU is both a processor and the main bus (ROM, RAM, WRAM, etc), APU is the shared memory by the SMP+DSP (eg use it to catch writes from either chip); PPU probably won't ever be used since it's broken down into three separate buses (VRAM, OAM, CGRAM), but DSP could be useful for tracking bugs like we found in Koushien 2 with the DSP echo buffer corrupting SMP opcodes. Technically the PPU memory pools are only ever tripped by the CPU poking at them, as the PPU doesn't ever write. I now have run.for, run.to, step.for, step.to. The difference is that run only prints the next instruction after running, whereas step prints all of the instructions along the way as well. run.to acts the same as "step over" here. Although it's not quite as nice, since you have to specify the address of the next instruction. Logging the Field/Vcounter/Hcounter on instruction listings now, good for timing information. Added in the tracer mask, as well as memory export, as well as VRAM/OAM/CGRAM/SMP read/write/execute breakpoints, as well as an APU usage map (it tracks DSP reads/writes separately, although I don't currently have debugger callbacks on DSP accesses just yet.) Have not hooked up actual SMP debugging just yet, but I plan to soon. Still thinking about how I want to allow / block interleaving of instructions (terminal output and tracing.) So ... remaining tasks at this point: - full SMP debugging - CPU+SMP interleave support - aliases - hotkeys - save states (will be kind of tricky ... will have to suppress breakpoints during synchronization, or abort a save in a break event.) - keep track of window geometry between runs
2014-02-05 11:30:08 +00:00
text.resize(20); //remove register information
echo(text, "\n");
if(--size <= 0) break;
unsigned displacement = 1;
while(displacement < maximumDisplacement) { //maximum opcode length is four bytes
if(usage[addr + displacement] & Usage::Execute) break;
displacement++;
}
if(displacement >= maximumDisplacement) {
echo("...\n");
return;
}
addr += displacement;
}
}
Update to v094r05 release. byuu says: Commands can be prefixed with: (cpu|smp|ppu|dsp|apu|vram|oam|cgram)/ to set their source. Eg "vram/hex 0800" or "smp/breakpoints.append execute ffc0"; default is cpu. These overlap a little bit in odd ways, but that's just the way the SNES works: it's not a very orthogonal system. CPU is both a processor and the main bus (ROM, RAM, WRAM, etc), APU is the shared memory by the SMP+DSP (eg use it to catch writes from either chip); PPU probably won't ever be used since it's broken down into three separate buses (VRAM, OAM, CGRAM), but DSP could be useful for tracking bugs like we found in Koushien 2 with the DSP echo buffer corrupting SMP opcodes. Technically the PPU memory pools are only ever tripped by the CPU poking at them, as the PPU doesn't ever write. I now have run.for, run.to, step.for, step.to. The difference is that run only prints the next instruction after running, whereas step prints all of the instructions along the way as well. run.to acts the same as "step over" here. Although it's not quite as nice, since you have to specify the address of the next instruction. Logging the Field/Vcounter/Hcounter on instruction listings now, good for timing information. Added in the tracer mask, as well as memory export, as well as VRAM/OAM/CGRAM/SMP read/write/execute breakpoints, as well as an APU usage map (it tracks DSP reads/writes separately, although I don't currently have debugger callbacks on DSP accesses just yet.) Have not hooked up actual SMP debugging just yet, but I plan to soon. Still thinking about how I want to allow / block interleaving of instructions (terminal output and tracing.) So ... remaining tasks at this point: - full SMP debugging - CPU+SMP interleave support - aliases - hotkeys - save states (will be kind of tricky ... will have to suppress breakpoints during synchronization, or abort a save in a break event.) - keep track of window geometry between runs
2014-02-05 11:30:08 +00:00
void Debugger::echoHex(Source source, unsigned addr, signed size) {
if(memorySize(source) == 0) return; //not a valid memory pool
while(size > 0) {
string hexdata, asciidata;
for(unsigned n = 0; n < 16; n++) {
unsigned offset = addr;
Update to v094r05 release. byuu says: Commands can be prefixed with: (cpu|smp|ppu|dsp|apu|vram|oam|cgram)/ to set their source. Eg "vram/hex 0800" or "smp/breakpoints.append execute ffc0"; default is cpu. These overlap a little bit in odd ways, but that's just the way the SNES works: it's not a very orthogonal system. CPU is both a processor and the main bus (ROM, RAM, WRAM, etc), APU is the shared memory by the SMP+DSP (eg use it to catch writes from either chip); PPU probably won't ever be used since it's broken down into three separate buses (VRAM, OAM, CGRAM), but DSP could be useful for tracking bugs like we found in Koushien 2 with the DSP echo buffer corrupting SMP opcodes. Technically the PPU memory pools are only ever tripped by the CPU poking at them, as the PPU doesn't ever write. I now have run.for, run.to, step.for, step.to. The difference is that run only prints the next instruction after running, whereas step prints all of the instructions along the way as well. run.to acts the same as "step over" here. Although it's not quite as nice, since you have to specify the address of the next instruction. Logging the Field/Vcounter/Hcounter on instruction listings now, good for timing information. Added in the tracer mask, as well as memory export, as well as VRAM/OAM/CGRAM/SMP read/write/execute breakpoints, as well as an APU usage map (it tracks DSP reads/writes separately, although I don't currently have debugger callbacks on DSP accesses just yet.) Have not hooked up actual SMP debugging just yet, but I plan to soon. Still thinking about how I want to allow / block interleaving of instructions (terminal output and tracing.) So ... remaining tasks at this point: - full SMP debugging - CPU+SMP interleave support - aliases - hotkeys - save states (will be kind of tricky ... will have to suppress breakpoints during synchronization, or abort a save in a break event.) - keep track of window geometry between runs
2014-02-05 11:30:08 +00:00
if(source == Source::CPU && ((offset & 0x40e000) == 0x002000 || (offset & 0x40e000) == 0x004000)) {
//$00-3f,80-bf:2000-5fff
//reading MMIO registers can negatively impact emulation, so disallow these reads
hexdata.append("?? ");
asciidata.append("?");
} else {
Update to v094r05 release. byuu says: Commands can be prefixed with: (cpu|smp|ppu|dsp|apu|vram|oam|cgram)/ to set their source. Eg "vram/hex 0800" or "smp/breakpoints.append execute ffc0"; default is cpu. These overlap a little bit in odd ways, but that's just the way the SNES works: it's not a very orthogonal system. CPU is both a processor and the main bus (ROM, RAM, WRAM, etc), APU is the shared memory by the SMP+DSP (eg use it to catch writes from either chip); PPU probably won't ever be used since it's broken down into three separate buses (VRAM, OAM, CGRAM), but DSP could be useful for tracking bugs like we found in Koushien 2 with the DSP echo buffer corrupting SMP opcodes. Technically the PPU memory pools are only ever tripped by the CPU poking at them, as the PPU doesn't ever write. I now have run.for, run.to, step.for, step.to. The difference is that run only prints the next instruction after running, whereas step prints all of the instructions along the way as well. run.to acts the same as "step over" here. Although it's not quite as nice, since you have to specify the address of the next instruction. Logging the Field/Vcounter/Hcounter on instruction listings now, good for timing information. Added in the tracer mask, as well as memory export, as well as VRAM/OAM/CGRAM/SMP read/write/execute breakpoints, as well as an APU usage map (it tracks DSP reads/writes separately, although I don't currently have debugger callbacks on DSP accesses just yet.) Have not hooked up actual SMP debugging just yet, but I plan to soon. Still thinking about how I want to allow / block interleaving of instructions (terminal output and tracing.) So ... remaining tasks at this point: - full SMP debugging - CPU+SMP interleave support - aliases - hotkeys - save states (will be kind of tricky ... will have to suppress breakpoints during synchronization, or abort a save in a break event.) - keep track of window geometry between runs
2014-02-05 11:30:08 +00:00
uint8 byte = memoryRead(source, addr + n);
hexdata.append(hex<2>(byte), " ");
asciidata.append(byte >= 0x20 && byte <= 0x7e ? (char)byte : '.');
}
}
Update to v094r05 release. byuu says: Commands can be prefixed with: (cpu|smp|ppu|dsp|apu|vram|oam|cgram)/ to set their source. Eg "vram/hex 0800" or "smp/breakpoints.append execute ffc0"; default is cpu. These overlap a little bit in odd ways, but that's just the way the SNES works: it's not a very orthogonal system. CPU is both a processor and the main bus (ROM, RAM, WRAM, etc), APU is the shared memory by the SMP+DSP (eg use it to catch writes from either chip); PPU probably won't ever be used since it's broken down into three separate buses (VRAM, OAM, CGRAM), but DSP could be useful for tracking bugs like we found in Koushien 2 with the DSP echo buffer corrupting SMP opcodes. Technically the PPU memory pools are only ever tripped by the CPU poking at them, as the PPU doesn't ever write. I now have run.for, run.to, step.for, step.to. The difference is that run only prints the next instruction after running, whereas step prints all of the instructions along the way as well. run.to acts the same as "step over" here. Although it's not quite as nice, since you have to specify the address of the next instruction. Logging the Field/Vcounter/Hcounter on instruction listings now, good for timing information. Added in the tracer mask, as well as memory export, as well as VRAM/OAM/CGRAM/SMP read/write/execute breakpoints, as well as an APU usage map (it tracks DSP reads/writes separately, although I don't currently have debugger callbacks on DSP accesses just yet.) Have not hooked up actual SMP debugging just yet, but I plan to soon. Still thinking about how I want to allow / block interleaving of instructions (terminal output and tracing.) So ... remaining tasks at this point: - full SMP debugging - CPU+SMP interleave support - aliases - hotkeys - save states (will be kind of tricky ... will have to suppress breakpoints during synchronization, or abort a save in a break event.) - keep track of window geometry between runs
2014-02-05 11:30:08 +00:00
echo(hex<6>(addr % memorySize(source)), " [ ", hexdata, "] ", asciidata, "\n");
addr += 16, size -= 16;
}
}
Update to v094r05 release. byuu says: Commands can be prefixed with: (cpu|smp|ppu|dsp|apu|vram|oam|cgram)/ to set their source. Eg "vram/hex 0800" or "smp/breakpoints.append execute ffc0"; default is cpu. These overlap a little bit in odd ways, but that's just the way the SNES works: it's not a very orthogonal system. CPU is both a processor and the main bus (ROM, RAM, WRAM, etc), APU is the shared memory by the SMP+DSP (eg use it to catch writes from either chip); PPU probably won't ever be used since it's broken down into three separate buses (VRAM, OAM, CGRAM), but DSP could be useful for tracking bugs like we found in Koushien 2 with the DSP echo buffer corrupting SMP opcodes. Technically the PPU memory pools are only ever tripped by the CPU poking at them, as the PPU doesn't ever write. I now have run.for, run.to, step.for, step.to. The difference is that run only prints the next instruction after running, whereas step prints all of the instructions along the way as well. run.to acts the same as "step over" here. Although it's not quite as nice, since you have to specify the address of the next instruction. Logging the Field/Vcounter/Hcounter on instruction listings now, good for timing information. Added in the tracer mask, as well as memory export, as well as VRAM/OAM/CGRAM/SMP read/write/execute breakpoints, as well as an APU usage map (it tracks DSP reads/writes separately, although I don't currently have debugger callbacks on DSP accesses just yet.) Have not hooked up actual SMP debugging just yet, but I plan to soon. Still thinking about how I want to allow / block interleaving of instructions (terminal output and tracing.) So ... remaining tasks at this point: - full SMP debugging - CPU+SMP interleave support - aliases - hotkeys - save states (will be kind of tricky ... will have to suppress breakpoints during synchronization, or abort a save in a break event.) - keep track of window geometry between runs
2014-02-05 11:30:08 +00:00
void Debugger::memoryExport(Source source, string filename) {
file fp;
if(fp.open(filename, file::mode::write)) {
unsigned size = memorySize(source);
for(unsigned addr = 0; addr < size; addr++) {
fp.write(memoryRead(source, addr));
}
echo("Exported memory to ", notdir(filename), "\n");
}
}
Update to v094r05 release. byuu says: Commands can be prefixed with: (cpu|smp|ppu|dsp|apu|vram|oam|cgram)/ to set their source. Eg "vram/hex 0800" or "smp/breakpoints.append execute ffc0"; default is cpu. These overlap a little bit in odd ways, but that's just the way the SNES works: it's not a very orthogonal system. CPU is both a processor and the main bus (ROM, RAM, WRAM, etc), APU is the shared memory by the SMP+DSP (eg use it to catch writes from either chip); PPU probably won't ever be used since it's broken down into three separate buses (VRAM, OAM, CGRAM), but DSP could be useful for tracking bugs like we found in Koushien 2 with the DSP echo buffer corrupting SMP opcodes. Technically the PPU memory pools are only ever tripped by the CPU poking at them, as the PPU doesn't ever write. I now have run.for, run.to, step.for, step.to. The difference is that run only prints the next instruction after running, whereas step prints all of the instructions along the way as well. run.to acts the same as "step over" here. Although it's not quite as nice, since you have to specify the address of the next instruction. Logging the Field/Vcounter/Hcounter on instruction listings now, good for timing information. Added in the tracer mask, as well as memory export, as well as VRAM/OAM/CGRAM/SMP read/write/execute breakpoints, as well as an APU usage map (it tracks DSP reads/writes separately, although I don't currently have debugger callbacks on DSP accesses just yet.) Have not hooked up actual SMP debugging just yet, but I plan to soon. Still thinking about how I want to allow / block interleaving of instructions (terminal output and tracing.) So ... remaining tasks at this point: - full SMP debugging - CPU+SMP interleave support - aliases - hotkeys - save states (will be kind of tricky ... will have to suppress breakpoints during synchronization, or abort a save in a break event.) - keep track of window geometry between runs
2014-02-05 11:30:08 +00:00
uint8 Debugger::memoryRead(Source source, unsigned addr) {
if(source == Source::CPU) {
return SFC::bus.read(addr & 0xffffff);
}
if(source == Source::APU) {
return SFC::smp.apuram[addr & 0xffff];
}
if(source == Source::WRAM) {
return SFC::cpu.wram[addr & 0x1ffff];
}
Update to v094r05 release. byuu says: Commands can be prefixed with: (cpu|smp|ppu|dsp|apu|vram|oam|cgram)/ to set their source. Eg "vram/hex 0800" or "smp/breakpoints.append execute ffc0"; default is cpu. These overlap a little bit in odd ways, but that's just the way the SNES works: it's not a very orthogonal system. CPU is both a processor and the main bus (ROM, RAM, WRAM, etc), APU is the shared memory by the SMP+DSP (eg use it to catch writes from either chip); PPU probably won't ever be used since it's broken down into three separate buses (VRAM, OAM, CGRAM), but DSP could be useful for tracking bugs like we found in Koushien 2 with the DSP echo buffer corrupting SMP opcodes. Technically the PPU memory pools are only ever tripped by the CPU poking at them, as the PPU doesn't ever write. I now have run.for, run.to, step.for, step.to. The difference is that run only prints the next instruction after running, whereas step prints all of the instructions along the way as well. run.to acts the same as "step over" here. Although it's not quite as nice, since you have to specify the address of the next instruction. Logging the Field/Vcounter/Hcounter on instruction listings now, good for timing information. Added in the tracer mask, as well as memory export, as well as VRAM/OAM/CGRAM/SMP read/write/execute breakpoints, as well as an APU usage map (it tracks DSP reads/writes separately, although I don't currently have debugger callbacks on DSP accesses just yet.) Have not hooked up actual SMP debugging just yet, but I plan to soon. Still thinking about how I want to allow / block interleaving of instructions (terminal output and tracing.) So ... remaining tasks at this point: - full SMP debugging - CPU+SMP interleave support - aliases - hotkeys - save states (will be kind of tricky ... will have to suppress breakpoints during synchronization, or abort a save in a break event.) - keep track of window geometry between runs
2014-02-05 11:30:08 +00:00
if(source == Source::VRAM) {
return SFC::ppu.vram[addr & 0xffff];
}
if(source == Source::OAM) {
return SFC::ppu.oam[addr % 544];
}
if(source == Source::CGRAM) {
return SFC::ppu.cgram[addr & 511];
}
return 0x00;
}
unsigned Debugger::memorySize(Source source) {
switch(source) {
case Source::CPU: return 0x1000000;
case Source::APU: return 0x10000;
case Source::WRAM: return 0x20000;
Update to v094r05 release. byuu says: Commands can be prefixed with: (cpu|smp|ppu|dsp|apu|vram|oam|cgram)/ to set their source. Eg "vram/hex 0800" or "smp/breakpoints.append execute ffc0"; default is cpu. These overlap a little bit in odd ways, but that's just the way the SNES works: it's not a very orthogonal system. CPU is both a processor and the main bus (ROM, RAM, WRAM, etc), APU is the shared memory by the SMP+DSP (eg use it to catch writes from either chip); PPU probably won't ever be used since it's broken down into three separate buses (VRAM, OAM, CGRAM), but DSP could be useful for tracking bugs like we found in Koushien 2 with the DSP echo buffer corrupting SMP opcodes. Technically the PPU memory pools are only ever tripped by the CPU poking at them, as the PPU doesn't ever write. I now have run.for, run.to, step.for, step.to. The difference is that run only prints the next instruction after running, whereas step prints all of the instructions along the way as well. run.to acts the same as "step over" here. Although it's not quite as nice, since you have to specify the address of the next instruction. Logging the Field/Vcounter/Hcounter on instruction listings now, good for timing information. Added in the tracer mask, as well as memory export, as well as VRAM/OAM/CGRAM/SMP read/write/execute breakpoints, as well as an APU usage map (it tracks DSP reads/writes separately, although I don't currently have debugger callbacks on DSP accesses just yet.) Have not hooked up actual SMP debugging just yet, but I plan to soon. Still thinking about how I want to allow / block interleaving of instructions (terminal output and tracing.) So ... remaining tasks at this point: - full SMP debugging - CPU+SMP interleave support - aliases - hotkeys - save states (will be kind of tricky ... will have to suppress breakpoints during synchronization, or abort a save in a break event.) - keep track of window geometry between runs
2014-02-05 11:30:08 +00:00
case Source::VRAM: return 0x10000;
case Source::OAM: return 544;
case Source::CGRAM: return 512;
}
return 0;
Update to v094r05 release. byuu says: Commands can be prefixed with: (cpu|smp|ppu|dsp|apu|vram|oam|cgram)/ to set their source. Eg "vram/hex 0800" or "smp/breakpoints.append execute ffc0"; default is cpu. These overlap a little bit in odd ways, but that's just the way the SNES works: it's not a very orthogonal system. CPU is both a processor and the main bus (ROM, RAM, WRAM, etc), APU is the shared memory by the SMP+DSP (eg use it to catch writes from either chip); PPU probably won't ever be used since it's broken down into three separate buses (VRAM, OAM, CGRAM), but DSP could be useful for tracking bugs like we found in Koushien 2 with the DSP echo buffer corrupting SMP opcodes. Technically the PPU memory pools are only ever tripped by the CPU poking at them, as the PPU doesn't ever write. I now have run.for, run.to, step.for, step.to. The difference is that run only prints the next instruction after running, whereas step prints all of the instructions along the way as well. run.to acts the same as "step over" here. Although it's not quite as nice, since you have to specify the address of the next instruction. Logging the Field/Vcounter/Hcounter on instruction listings now, good for timing information. Added in the tracer mask, as well as memory export, as well as VRAM/OAM/CGRAM/SMP read/write/execute breakpoints, as well as an APU usage map (it tracks DSP reads/writes separately, although I don't currently have debugger callbacks on DSP accesses just yet.) Have not hooked up actual SMP debugging just yet, but I plan to soon. Still thinking about how I want to allow / block interleaving of instructions (terminal output and tracing.) So ... remaining tasks at this point: - full SMP debugging - CPU+SMP interleave support - aliases - hotkeys - save states (will be kind of tricky ... will have to suppress breakpoints during synchronization, or abort a save in a break event.) - keep track of window geometry between runs
2014-02-05 11:30:08 +00:00
}
void Debugger::memoryWrite(Source source, unsigned addr, uint8 data) {
if(source == Source::CPU) {
SFC::bus.write(addr & 0xffffff, data);
return;
}
if(source == Source::APU) {
SFC::smp.apuram[addr & 0xffff] = data;
return;
}
if(source == Source::WRAM) {
SFC::cpu.wram[addr & 0x1ffff] = data;
return;
}
Update to v094r05 release. byuu says: Commands can be prefixed with: (cpu|smp|ppu|dsp|apu|vram|oam|cgram)/ to set their source. Eg "vram/hex 0800" or "smp/breakpoints.append execute ffc0"; default is cpu. These overlap a little bit in odd ways, but that's just the way the SNES works: it's not a very orthogonal system. CPU is both a processor and the main bus (ROM, RAM, WRAM, etc), APU is the shared memory by the SMP+DSP (eg use it to catch writes from either chip); PPU probably won't ever be used since it's broken down into three separate buses (VRAM, OAM, CGRAM), but DSP could be useful for tracking bugs like we found in Koushien 2 with the DSP echo buffer corrupting SMP opcodes. Technically the PPU memory pools are only ever tripped by the CPU poking at them, as the PPU doesn't ever write. I now have run.for, run.to, step.for, step.to. The difference is that run only prints the next instruction after running, whereas step prints all of the instructions along the way as well. run.to acts the same as "step over" here. Although it's not quite as nice, since you have to specify the address of the next instruction. Logging the Field/Vcounter/Hcounter on instruction listings now, good for timing information. Added in the tracer mask, as well as memory export, as well as VRAM/OAM/CGRAM/SMP read/write/execute breakpoints, as well as an APU usage map (it tracks DSP reads/writes separately, although I don't currently have debugger callbacks on DSP accesses just yet.) Have not hooked up actual SMP debugging just yet, but I plan to soon. Still thinking about how I want to allow / block interleaving of instructions (terminal output and tracing.) So ... remaining tasks at this point: - full SMP debugging - CPU+SMP interleave support - aliases - hotkeys - save states (will be kind of tricky ... will have to suppress breakpoints during synchronization, or abort a save in a break event.) - keep track of window geometry between runs
2014-02-05 11:30:08 +00:00
if(source == Source::VRAM) {
SFC::ppu.vram[addr & 0xffff] = data;
return;
}
if(source == Source::OAM) {
SFC::ppu.oam[addr % 544] = data;
SFC::ppu.sprite.update(addr % 544, data);
return;
}
if(source == Source::CGRAM) {
if(addr & 1) data &= 0x7f;
SFC::ppu.cgram[addr] = data;
return;
}
}
void Debugger::ppuCgramRead(uint16 addr, uint8 data) {
if(breakpointTest(Source::CGRAM, Breakpoint::Mode::Read, addr, data)) leave();
}
void Debugger::ppuCgramWrite(uint16 addr, uint8 data) {
if(breakpointTest(Source::CGRAM, Breakpoint::Mode::Write, addr, data)) leave();
}
void Debugger::ppuOamRead(uint16 addr, uint8 data) {
if(breakpointTest(Source::OAM, Breakpoint::Mode::Read, addr, data)) leave();
}
void Debugger::ppuOamWrite(uint16 addr, uint8 data) {
if(breakpointTest(Source::OAM, Breakpoint::Mode::Write, addr, data)) leave();
}
void Debugger::ppuVramRead(uint16 addr, uint8 data) {
if(breakpointTest(Source::VRAM, Breakpoint::Mode::Read, addr, data)) leave();
}
void Debugger::ppuVramWrite(uint16 addr, uint8 data) {
if(breakpointTest(Source::VRAM, Breakpoint::Mode::Write, addr, data)) leave();
}
string Debugger::smpDisassemble() {
return SFC::smp.disassemble_opcode(SFC::smp.regs.pc, SFC::smp.regs.p.p);
}
string Debugger::smpDisassemble(uint16 addr, bool p) {
return SFC::smp.disassemble_opcode(addr, p);
}
Update to v094r05 release. byuu says: Commands can be prefixed with: (cpu|smp|ppu|dsp|apu|vram|oam|cgram)/ to set their source. Eg "vram/hex 0800" or "smp/breakpoints.append execute ffc0"; default is cpu. These overlap a little bit in odd ways, but that's just the way the SNES works: it's not a very orthogonal system. CPU is both a processor and the main bus (ROM, RAM, WRAM, etc), APU is the shared memory by the SMP+DSP (eg use it to catch writes from either chip); PPU probably won't ever be used since it's broken down into three separate buses (VRAM, OAM, CGRAM), but DSP could be useful for tracking bugs like we found in Koushien 2 with the DSP echo buffer corrupting SMP opcodes. Technically the PPU memory pools are only ever tripped by the CPU poking at them, as the PPU doesn't ever write. I now have run.for, run.to, step.for, step.to. The difference is that run only prints the next instruction after running, whereas step prints all of the instructions along the way as well. run.to acts the same as "step over" here. Although it's not quite as nice, since you have to specify the address of the next instruction. Logging the Field/Vcounter/Hcounter on instruction listings now, good for timing information. Added in the tracer mask, as well as memory export, as well as VRAM/OAM/CGRAM/SMP read/write/execute breakpoints, as well as an APU usage map (it tracks DSP reads/writes separately, although I don't currently have debugger callbacks on DSP accesses just yet.) Have not hooked up actual SMP debugging just yet, but I plan to soon. Still thinking about how I want to allow / block interleaving of instructions (terminal output and tracing.) So ... remaining tasks at this point: - full SMP debugging - CPU+SMP interleave support - aliases - hotkeys - save states (will be kind of tricky ... will have to suppress breakpoints during synchronization, or abort a save in a break event.) - keep track of window geometry between runs
2014-02-05 11:30:08 +00:00
void Debugger::smpExec(uint16 addr) {
apuUsage[addr] |= Usage::Execute;
if(SFC::smp.regs.p.p == 0) apuUsage[addr] &= ~Usage::FlagP;
if(SFC::smp.regs.p.p == 1) apuUsage[addr] |= Usage::FlagP;
smpInstructionCounter++;
if(smpTracerFile.open()) {
if(!smpTracerMask || smpTracerMask[addr] == false) {
if(smpTracerMask) smpTracerMask[addr] = true;
smpTracerFile.print(smpDisassemble(), "\n");
}
}
Update to v094r05 release. byuu says: Commands can be prefixed with: (cpu|smp|ppu|dsp|apu|vram|oam|cgram)/ to set their source. Eg "vram/hex 0800" or "smp/breakpoints.append execute ffc0"; default is cpu. These overlap a little bit in odd ways, but that's just the way the SNES works: it's not a very orthogonal system. CPU is both a processor and the main bus (ROM, RAM, WRAM, etc), APU is the shared memory by the SMP+DSP (eg use it to catch writes from either chip); PPU probably won't ever be used since it's broken down into three separate buses (VRAM, OAM, CGRAM), but DSP could be useful for tracking bugs like we found in Koushien 2 with the DSP echo buffer corrupting SMP opcodes. Technically the PPU memory pools are only ever tripped by the CPU poking at them, as the PPU doesn't ever write. I now have run.for, run.to, step.for, step.to. The difference is that run only prints the next instruction after running, whereas step prints all of the instructions along the way as well. run.to acts the same as "step over" here. Although it's not quite as nice, since you have to specify the address of the next instruction. Logging the Field/Vcounter/Hcounter on instruction listings now, good for timing information. Added in the tracer mask, as well as memory export, as well as VRAM/OAM/CGRAM/SMP read/write/execute breakpoints, as well as an APU usage map (it tracks DSP reads/writes separately, although I don't currently have debugger callbacks on DSP accesses just yet.) Have not hooked up actual SMP debugging just yet, but I plan to soon. Still thinking about how I want to allow / block interleaving of instructions (terminal output and tracing.) So ... remaining tasks at this point: - full SMP debugging - CPU+SMP interleave support - aliases - hotkeys - save states (will be kind of tricky ... will have to suppress breakpoints during synchronization, or abort a save in a break event.) - keep track of window geometry between runs
2014-02-05 11:30:08 +00:00
if(savingState) return;
Update to v094r05 release. byuu says: Commands can be prefixed with: (cpu|smp|ppu|dsp|apu|vram|oam|cgram)/ to set their source. Eg "vram/hex 0800" or "smp/breakpoints.append execute ffc0"; default is cpu. These overlap a little bit in odd ways, but that's just the way the SNES works: it's not a very orthogonal system. CPU is both a processor and the main bus (ROM, RAM, WRAM, etc), APU is the shared memory by the SMP+DSP (eg use it to catch writes from either chip); PPU probably won't ever be used since it's broken down into three separate buses (VRAM, OAM, CGRAM), but DSP could be useful for tracking bugs like we found in Koushien 2 with the DSP echo buffer corrupting SMP opcodes. Technically the PPU memory pools are only ever tripped by the CPU poking at them, as the PPU doesn't ever write. I now have run.for, run.to, step.for, step.to. The difference is that run only prints the next instruction after running, whereas step prints all of the instructions along the way as well. run.to acts the same as "step over" here. Although it's not quite as nice, since you have to specify the address of the next instruction. Logging the Field/Vcounter/Hcounter on instruction listings now, good for timing information. Added in the tracer mask, as well as memory export, as well as VRAM/OAM/CGRAM/SMP read/write/execute breakpoints, as well as an APU usage map (it tracks DSP reads/writes separately, although I don't currently have debugger callbacks on DSP accesses just yet.) Have not hooked up actual SMP debugging just yet, but I plan to soon. Still thinking about how I want to allow / block interleaving of instructions (terminal output and tracing.) So ... remaining tasks at this point: - full SMP debugging - CPU+SMP interleave support - aliases - hotkeys - save states (will be kind of tricky ... will have to suppress breakpoints during synchronization, or abort a save in a break event.) - keep track of window geometry between runs
2014-02-05 11:30:08 +00:00
if(breakpointTest(Source::SMP, Breakpoint::Mode::Execute, addr)) {
echo(smpDisassemble(), "\n");
return leave();
}
if(smpRunFor) {
if(--smpRunFor() == 0) {
echo(smpDisassemble(), "\n");
return leave();
}
}
if(smpRunTo) {
if(addr == smpRunTo()) {
echo(smpDisassemble(), "\n");
return leave();
}
}
if(smpStepFor) {
echo(smpDisassemble(), "\n");
if(--smpStepFor() == 0) return leave();
}
if(smpStepTo) {
echo(smpDisassemble(), "\n");
if(addr == smpStepTo()) return leave();
Update to v094r05 release. byuu says: Commands can be prefixed with: (cpu|smp|ppu|dsp|apu|vram|oam|cgram)/ to set their source. Eg "vram/hex 0800" or "smp/breakpoints.append execute ffc0"; default is cpu. These overlap a little bit in odd ways, but that's just the way the SNES works: it's not a very orthogonal system. CPU is both a processor and the main bus (ROM, RAM, WRAM, etc), APU is the shared memory by the SMP+DSP (eg use it to catch writes from either chip); PPU probably won't ever be used since it's broken down into three separate buses (VRAM, OAM, CGRAM), but DSP could be useful for tracking bugs like we found in Koushien 2 with the DSP echo buffer corrupting SMP opcodes. Technically the PPU memory pools are only ever tripped by the CPU poking at them, as the PPU doesn't ever write. I now have run.for, run.to, step.for, step.to. The difference is that run only prints the next instruction after running, whereas step prints all of the instructions along the way as well. run.to acts the same as "step over" here. Although it's not quite as nice, since you have to specify the address of the next instruction. Logging the Field/Vcounter/Hcounter on instruction listings now, good for timing information. Added in the tracer mask, as well as memory export, as well as VRAM/OAM/CGRAM/SMP read/write/execute breakpoints, as well as an APU usage map (it tracks DSP reads/writes separately, although I don't currently have debugger callbacks on DSP accesses just yet.) Have not hooked up actual SMP debugging just yet, but I plan to soon. Still thinking about how I want to allow / block interleaving of instructions (terminal output and tracing.) So ... remaining tasks at this point: - full SMP debugging - CPU+SMP interleave support - aliases - hotkeys - save states (will be kind of tricky ... will have to suppress breakpoints during synchronization, or abort a save in a break event.) - keep track of window geometry between runs
2014-02-05 11:30:08 +00:00
}
}
void Debugger::smpRead(uint16 addr, uint8 data) {
apuUsage[addr] |= Usage::Read;
Update to v094r05 release. byuu says: Commands can be prefixed with: (cpu|smp|ppu|dsp|apu|vram|oam|cgram)/ to set their source. Eg "vram/hex 0800" or "smp/breakpoints.append execute ffc0"; default is cpu. These overlap a little bit in odd ways, but that's just the way the SNES works: it's not a very orthogonal system. CPU is both a processor and the main bus (ROM, RAM, WRAM, etc), APU is the shared memory by the SMP+DSP (eg use it to catch writes from either chip); PPU probably won't ever be used since it's broken down into three separate buses (VRAM, OAM, CGRAM), but DSP could be useful for tracking bugs like we found in Koushien 2 with the DSP echo buffer corrupting SMP opcodes. Technically the PPU memory pools are only ever tripped by the CPU poking at them, as the PPU doesn't ever write. I now have run.for, run.to, step.for, step.to. The difference is that run only prints the next instruction after running, whereas step prints all of the instructions along the way as well. run.to acts the same as "step over" here. Although it's not quite as nice, since you have to specify the address of the next instruction. Logging the Field/Vcounter/Hcounter on instruction listings now, good for timing information. Added in the tracer mask, as well as memory export, as well as VRAM/OAM/CGRAM/SMP read/write/execute breakpoints, as well as an APU usage map (it tracks DSP reads/writes separately, although I don't currently have debugger callbacks on DSP accesses just yet.) Have not hooked up actual SMP debugging just yet, but I plan to soon. Still thinking about how I want to allow / block interleaving of instructions (terminal output and tracing.) So ... remaining tasks at this point: - full SMP debugging - CPU+SMP interleave support - aliases - hotkeys - save states (will be kind of tricky ... will have to suppress breakpoints during synchronization, or abort a save in a break event.) - keep track of window geometry between runs
2014-02-05 11:30:08 +00:00
if(breakpointTest(Source::SMP, Breakpoint::Mode::Read, addr, data)) leave();
if(breakpointTest(Source::APU, Breakpoint::Mode::Read, addr, data)) leave();
}
void Debugger::smpWrite(uint16 addr, uint8 data) {
apuUsage[addr] |= Usage::Write;
Update to v094r05 release. byuu says: Commands can be prefixed with: (cpu|smp|ppu|dsp|apu|vram|oam|cgram)/ to set their source. Eg "vram/hex 0800" or "smp/breakpoints.append execute ffc0"; default is cpu. These overlap a little bit in odd ways, but that's just the way the SNES works: it's not a very orthogonal system. CPU is both a processor and the main bus (ROM, RAM, WRAM, etc), APU is the shared memory by the SMP+DSP (eg use it to catch writes from either chip); PPU probably won't ever be used since it's broken down into three separate buses (VRAM, OAM, CGRAM), but DSP could be useful for tracking bugs like we found in Koushien 2 with the DSP echo buffer corrupting SMP opcodes. Technically the PPU memory pools are only ever tripped by the CPU poking at them, as the PPU doesn't ever write. I now have run.for, run.to, step.for, step.to. The difference is that run only prints the next instruction after running, whereas step prints all of the instructions along the way as well. run.to acts the same as "step over" here. Although it's not quite as nice, since you have to specify the address of the next instruction. Logging the Field/Vcounter/Hcounter on instruction listings now, good for timing information. Added in the tracer mask, as well as memory export, as well as VRAM/OAM/CGRAM/SMP read/write/execute breakpoints, as well as an APU usage map (it tracks DSP reads/writes separately, although I don't currently have debugger callbacks on DSP accesses just yet.) Have not hooked up actual SMP debugging just yet, but I plan to soon. Still thinking about how I want to allow / block interleaving of instructions (terminal output and tracing.) So ... remaining tasks at this point: - full SMP debugging - CPU+SMP interleave support - aliases - hotkeys - save states (will be kind of tricky ... will have to suppress breakpoints during synchronization, or abort a save in a break event.) - keep track of window geometry between runs
2014-02-05 11:30:08 +00:00
if(breakpointTest(Source::SMP, Breakpoint::Mode::Write, addr, data)) leave();
if(breakpointTest(Source::APU, Breakpoint::Mode::Write, addr, data)) leave();
}
string Debugger::sourceName(Source source) {
switch(source) {
case Source::CPU: return "cpu";
case Source::SMP: return "smp";
case Source::PPU: return "ppu";
case Source::DSP: return "dsp";
case Source::APU: return "apu";
case Source::WRAM: return "wram";
Update to v094r05 release. byuu says: Commands can be prefixed with: (cpu|smp|ppu|dsp|apu|vram|oam|cgram)/ to set their source. Eg "vram/hex 0800" or "smp/breakpoints.append execute ffc0"; default is cpu. These overlap a little bit in odd ways, but that's just the way the SNES works: it's not a very orthogonal system. CPU is both a processor and the main bus (ROM, RAM, WRAM, etc), APU is the shared memory by the SMP+DSP (eg use it to catch writes from either chip); PPU probably won't ever be used since it's broken down into three separate buses (VRAM, OAM, CGRAM), but DSP could be useful for tracking bugs like we found in Koushien 2 with the DSP echo buffer corrupting SMP opcodes. Technically the PPU memory pools are only ever tripped by the CPU poking at them, as the PPU doesn't ever write. I now have run.for, run.to, step.for, step.to. The difference is that run only prints the next instruction after running, whereas step prints all of the instructions along the way as well. run.to acts the same as "step over" here. Although it's not quite as nice, since you have to specify the address of the next instruction. Logging the Field/Vcounter/Hcounter on instruction listings now, good for timing information. Added in the tracer mask, as well as memory export, as well as VRAM/OAM/CGRAM/SMP read/write/execute breakpoints, as well as an APU usage map (it tracks DSP reads/writes separately, although I don't currently have debugger callbacks on DSP accesses just yet.) Have not hooked up actual SMP debugging just yet, but I plan to soon. Still thinking about how I want to allow / block interleaving of instructions (terminal output and tracing.) So ... remaining tasks at this point: - full SMP debugging - CPU+SMP interleave support - aliases - hotkeys - save states (will be kind of tricky ... will have to suppress breakpoints during synchronization, or abort a save in a break event.) - keep track of window geometry between runs
2014-02-05 11:30:08 +00:00
case Source::VRAM: return "vram";
case Source::OAM: return "oam";
case Source::CGRAM: return "cgram";
}
return "none";
}
void Debugger::stateLoad(string filename) {
auto memory = file::read(filename);
if(memory.size() == 0) return echo("Error: state file ", notdir(filename), " not found\n");
serializer s(memory.data(), memory.size());
if(emulator->unserialize(s) == false) return echo("Error: failed to unserialize state from ", notdir(filename), "\n");
echo("State loaded from ", notdir(filename), "\n");
}
void Debugger::stateSave(string filename) {
savingState = true;
serializer s = emulator->serialize();
if(file::write(filename, s.data(), s.size())) {
echo("State saved to ", notdir(filename), "\n");
}
savingState = false;
}
void Debugger::tracerDisable(Source source) {
if(source != Source::CPU && source != Source::SMP) return;
file& tracerFile = (source == Source::CPU ? cpuTracerFile : smpTracerFile);
if(tracerFile.open() == false) return;
tracerFile.close();
echo(sourceName(source).upper(), " tracer disabled\n");
}
void Debugger::tracerEnable(Source source, string filename) {
if(source != Source::CPU && source != Source::SMP) return;
file& tracerFile = (source == Source::CPU ? cpuTracerFile : smpTracerFile);
if(tracerFile.open() == true) return;
if(tracerFile.open(filename, file::mode::write)) {
echo(sourceName(source).upper(), " tracer enabled\n");
}
}
void Debugger::tracerMaskDisable(Source source) {
if(source != Source::CPU && source != Source::SMP) return;
bitvector& tracerMask = (source == Source::CPU ? cpuTracerMask : smpTracerMask);
tracerMask.reset();
echo(sourceName(source).upper(), " tracer mask disabled\n");
}
void Debugger::tracerMaskEnable(Source source) {
if(source != Source::CPU && source != Source::SMP) return;
bitvector& tracerMask = (source == Source::CPU ? cpuTracerMask : smpTracerMask);
unsigned size = (source == Source::CPU ? 0x1000000 : 0x10000);
tracerMask.resize(size);
tracerMask.clear();
echo(sourceName(source).upper(), " tracer mask enabled\n");
}