CPUEquivalence: Dump every core's end state, not only the interpreter's

So a diff between hosts covers the JITs too, without minimizing every case.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

docs/cpu-equivalence.md

@@ -30,7 +30,7 @@ Environment variables:
 | `PPSSPP_CPUFUZZ_SKIP` | cores to skip, by digit: 1 IR interpreter, 2 JIT, 3 JIT IR |
 | `PPSSPP_CPUFUZZ_REPORTS` | how many differences to minimize and print (default 20) |
 | `PPSSPP_CPUFUZZ_VERBOSE` | print every case before running it, to find one that crashes |
-| `PPSSPP_CPUFUZZ_DUMP` | write the interpreter's end state of every case to a file |
+| `PPSSPP_CPUFUZZ_DUMP` | write every core's end state of every case to a file |
 | `PPSSPP_CPUFUZZ_STRICT` | fail on any difference, not only on a crash |
 
 Always set `CI=1` when running it from a script on Windows. Otherwise a JIT assert opens a message
@@ -41,8 +41,9 @@ finds it and reduces it to the instructions still needed for the crash. A crash
 after other cases have run usually means uninitialized or stale state, not the reduced code.
 
 To compare hosts: run with the same seed and count and `PPSSPP_CPUFUZZ_DUMP` on each, and diff the
-files. A Windows-on-ARM machine can run the x64 build under emulation, which is enough for the
-host-dependent float behavior of the interpreter. It isn't for timing, of course.
+files. Each line starts with the core and the seed. A Windows-on-ARM machine can run the x64 build
+under emulation, which is enough for the host-dependent float behavior of the cores. It isn't for
+timing, of course.
 
 ## Fixed
 

unittest/JitHarness.cpp

@@ -376,7 +376,7 @@ bool TestJitReplacementHooks() {
 // Environment variables:
 //   PPSSPP_CPUFUZZ_COUNT    number of cases (default 2000)
 //   PPSSPP_CPUFUZZ_SEED     first case's seed (default 1). Case i always uses seed + i.
-//   PPSSPP_CPUFUZZ_DUMP     file to write the interpreter's end state of every case to, for diffing
+//   PPSSPP_CPUFUZZ_DUMP     file to write every core's end state of every case to, for diffing
 //                           the results of different host architectures.
 //   PPSSPP_CPUFUZZ_CLASSES  instruction classes to generate, as a weighted string of "aifvm" (alu,
 //                           imm, fpu, vfpu, memory).
@@ -754,10 +754,10 @@ static int DiffFuzzStates(const FuzzState &ref, const FuzzState &other, std::str
 	return diffs;
 }
 
-static void DumpFuzzState(FILE *f, u64 seed, const FuzzState &s) {
+static void DumpFuzzState(FILE *f, const char *core, u64 seed, const FuzzState &s) {
 	// All NaNs print the same, so their payloads don't show up as differences.
 	auto fl = [](u32 v) { return IsNaNBits(v) ? 0x7FFFFFFFU : v; };
-	fprintf(f, "%llu r", (unsigned long long)seed);
+	fprintf(f, "%s %llu r", core, (unsigned long long)seed);
 	for (int i = 0; i < 32; ++i) {
 		fprintf(f, " %08x", s.r[i]);
 	}
@@ -942,8 +942,10 @@ bool TestCPUEquivalence() {
 	if (dumpPath) {
 		FILE *f = fopen(dumpPath, "w");
 		if (f) {
-			for (int i = 0; i < count; ++i) {
-				DumpFuzzState(f, firstSeed + i, results[(size_t)i * numCores]);
+			for (int c = 0; c < numCores; ++c) {
+				for (int i = 0; i < count; ++i) {
+					DumpFuzzState(f, CoreName(cores[c]), firstSeed + i, results[(size_t)i * numCores + c]);
+				}
 			}
 			fclose(f);
 		} else {