Make improvements

- Emulator can now test the αcτµαlly pδrταblε εxεcµταblε bootloader

- Whipped up a webserver named redbean. It services 150k requests per
  second on a single core. Bundling assets inside zip enables extremely
  fast serving for two reasons. The first is that zip central directory
  lookups go faster than stat() system calls. The second is that both
  zip and gzip content-encoding use DEFLATE, therefore, compressed
  responses can be served via the sendfile() system call which does an
  in-kernel copy directly from the zip executable structure. Also note
  that red bean zip executables can be deployed easily to all platforms,
  since these native executables work on Linux, Mac, BSD, and Windows.

- Address sanitizer now works very well

tool/build/lib/fpu.c

@@ -31,7 +31,6 @@
 #include "tool/build/lib/modrm.h"
 #include "tool/build/lib/throw.h"
 #include "tool/build/lib/word.h"
-#include "tool/build/lib/x87.h"
 
 #define FPUREG 0
 #define MEMORY 1
@@ -153,6 +152,57 @@ static void FpuSetMemoryLdbl(struct Machine *m, long double f) {
   SetMemoryLdbl(m, m->fpu.dp, f);
 }
 
+static long ltruncl(long double x) {
+  return x;
+}
+
+static int ClearC2(int sw) {
+  return sw & ~FPU_C2;
+}
+
+static long double f2xm1(long double x) {
+  return exp2l(x) - 1;
+}
+
+static long double fyl2x(long double x, long double y) {
+  return y * log2l(x);
+}
+
+static long double fyl2xp1(long double x, long double y) {
+  return y * log2l(x + 1);
+}
+
+static long double fscale(long double significand, long double exponent) {
+  return scalbl(significand, exponent);
+}
+
+static long double x87remainder(long double x, long double y, uint32_t *sw,
+                                long double rem(long double, long double),
+                                long rnd(long double)) {
+  int s;
+  long q;
+  long double r;
+  s = 0;
+  r = rem(x, y);
+  q = rnd(x / y);
+  s &= ~FPU_C2; /* ty libm */
+  if (q & 0b001) s |= FPU_C1;
+  if (q & 0b010) s |= FPU_C3;
+  if (q & 0b100) s |= FPU_C0;
+  if (sw) *sw = s | (*sw & ~(FPU_C0 | FPU_C1 | FPU_C2 | FPU_C3));
+  return r;
+}
+
+static long double fprem(long double dividend, long double modulus,
+                         uint32_t *sw) {
+  return x87remainder(dividend, modulus, sw, fmodl, ltruncl);
+}
+
+static long double fprem1(long double dividend, long double modulus,
+                          uint32_t *sw) {
+  return x87remainder(dividend, modulus, sw, remainderl, lrintl);
+}
+
 static long double FpuAdd(struct Machine *m, long double x, long double y) {
   if (!isunordered(x, y)) {
     switch (isinf(y) << 1 | isinf(x)) {
@@ -263,7 +313,7 @@ static void FpuCompare(struct Machine *m, long double y) {
   }
 }
 
-void OpFxam(struct Machine *m) {
+static void OpFxam(struct Machine *m) {
   long double x;
   x = *FpuSt(m, 0);
   m->fpu.c1 = !!signbit(x);
@@ -741,7 +791,7 @@ static void OpFldConstant(struct Machine *m) {
       x = fldz();
       break;
     default:
-      OpUd(m);
+      OpUd(m, m->xedd->op.rde);
   }
   FpuPush(m, x);
 }
@@ -844,7 +894,7 @@ static void OpFfree(struct Machine *m) {
 }
 
 static void OpFfreep(struct Machine *m) {
-  OpFfree(m);
+  if (ModrmRm(m->xedd->op.rde)) OpFfree(m);
   FpuPop(m);
 }
 
@@ -932,7 +982,7 @@ void OpFinit(struct Machine *m) {
   m->fpu.tw = -1;
 }
 
-void OpFwait(struct Machine *m) {
+void OpFwait(struct Machine *m, uint32_t rde) {
   if ((m->fpu.ie & !m->fpu.im) | (m->fpu.de & !m->fpu.dm) |
       (m->fpu.ze & !m->fpu.zm) | (m->fpu.oe & !m->fpu.om) |
       (m->fpu.ue & !m->fpu.um) | (m->fpu.pe & !m->fpu.pm) |
@@ -986,15 +1036,15 @@ long double FpuPop(struct Machine *m) {
   return x;
 }
 
-void OpFpu(struct Machine *m) {
+void OpFpu(struct Machine *m, uint32_t rde) {
   unsigned op;
   bool ismemory;
   op = m->xedd->op.opcode & 0b111;
-  ismemory = ModrmMod(m->xedd->op.rde) != 0b11;
+  ismemory = ModrmMod(rde) != 0b11;
   m->fpu.ip = m->ip - m->xedd->length;
-  m->fpu.op = op << 8 | m->xedd->op.modrm;
-  m->fpu.dp = ismemory ? ComputeAddress(m, m->xedd->op.rde) : 0;
-  switch (DISP(op, ismemory, m->xedd->op.reg)) {
+  m->fpu.op = op << 8 | ModrmMod(rde) << 6 | ModrmReg(rde) << 3 | ModrmRm(rde);
+  m->fpu.dp = ismemory ? ComputeAddress(m, rde) : 0;
+  switch (DISP(op, ismemory, ModrmReg(rde))) {
     CASE(DISP(0xD8, FPUREG, 0), OpFaddStEst(m));
     CASE(DISP(0xD8, FPUREG, 1), OpFmulStEst(m));
     CASE(DISP(0xD8, FPUREG, 2), OpFcom(m));
@@ -1106,17 +1156,17 @@ void OpFpu(struct Machine *m) {
     CASE(DISP(0xDF, MEMORY, 5), OpFildll(m));
     CASE(DISP(0xDF, MEMORY, 7), OpFistpll(m));
     case DISP(0xD9, FPUREG, 4):
-      switch (ModrmRm(m->xedd->op.rde)) {
+      switch (ModrmRm(rde)) {
         CASE(0, OpFchs(m));
         CASE(1, OpFabs(m));
         CASE(4, OpFtst(m));
         CASE(5, OpFxam(m));
         default:
-          OpUd(m);
+          OpUd(m, rde);
       }
       break;
     case DISP(0xD9, FPUREG, 6):
-      switch (ModrmRm(m->xedd->op.rde)) {
+      switch (ModrmRm(rde)) {
         CASE(0, OpF2xm1(m));
         CASE(1, OpFyl2x(m));
         CASE(2, OpFptan(m));
@@ -1130,7 +1180,7 @@ void OpFpu(struct Machine *m) {
       }
       break;
     case DISP(0xD9, FPUREG, 7):
-      switch (ModrmRm(m->xedd->op.rde)) {
+      switch (ModrmRm(rde)) {
         CASE(0, OpFprem(m));
         CASE(1, OpFyl2xp1(m));
         CASE(2, OpFsqrt(m));
@@ -1144,14 +1194,14 @@ void OpFpu(struct Machine *m) {
       }
       break;
     case DISP(0xDb, FPUREG, 4):
-      switch (ModrmRm(m->xedd->op.rde)) {
+      switch (ModrmRm(rde)) {
         CASE(2, OpFnclex(m));
         CASE(3, OpFinit(m));
         default:
-          OpUd(m);
+          OpUd(m, rde);
       }
       break;
     default:
-      OpUd(m);
+      OpUd(m, rde);
   }
 }