diff --git a/README.md b/README.md
index 88bacd2..ba946dd 100644
--- a/README.md
+++ b/README.md
@@ -8,7 +8,9 @@ PoC by Shiz, bugfixing and 64-bit version by PoroCYon.
 
 ```sh
 ./smol.py -lfoo -lbar input.o... smol-output.asm
-nasm -I src/ [-DUSE_NX] [-DUSE_DL_FINI] -o nasm-output.o smol-output.asm
+nasm -I src/ [-DUSE_INTERP] [-DALIGN_STACK] [-DUSE_NX] [-DUSE_DL_FINI] \
+    -o nasm-output.o smol-output.asm
+# -DALIGN_STACK: 64-bit only.
 ld -T ld/link.ld -o binary nasm-output.o input.o...
 ```
 
diff --git a/smol.py b/smol.py
index c9eb7ca..324a164 100755
--- a/smol.py
+++ b/smol.py
@@ -52,13 +52,13 @@ def main():
     libs = list(find_libs(spaths, libnames))
     symbols = {}
 
-    for symbol in syms:
+    for symbol, reloc in syms:
         library = find_symbol(args.scanelf, libs, libnames, symbol)
         if not library:
             eprintf("could not find symbol: {}".format(symbol))
             sys.exit(1)
         symbols.setdefault(library, [])
-        symbols[library].append(symbol)
+        symbols[library].append((symbol, reloc))
 
     output(arch, symbols, args.output)
 
diff --git a/smolemit.py b/smolemit.py
index 8c5111a..190bbab 100644
--- a/smolemit.py
+++ b/smolemit.py
@@ -15,11 +15,20 @@ def output_x86(libraries, outf):
         outf.write('dd (_symbols.{} - _symbols)\n'.format(shorts[library]))
     outf.write('.dynamic.end:\n')
 
+#    if needgot:
+#        outf.write('global _GLOBAL_OFFSET_TABLE_\n')
+#        outf.write('_GLOBAL_OFFSET_TABLE_:\n')
+#        outf.write('dd dynamic\n')
     outf.write('_symbols:\n')
-    for library, symbols in libraries.items():
+    for library, symrels in libraries.items():
         outf.write('\t_symbols.{}: db "{}",0\n'.format(shorts[library], library))
 
-        for sym in symbols:
+        for sym, reloc in symrels:
+            # meh
+            if reloc != 'R_386_PC32':
+                eprintf('Relocation type ' + reloc + ' of symbol ' + sym + ' unsupported!')
+                sys.exit(1)
+
             hash = hash_djb2(sym)
             outf.write("""
 \t\tglobal {name}
@@ -46,11 +55,25 @@ def output_amd64(libraries, outf):
     outf.write('dynamic.end:\n')
 
     outf.write('[section .data.smolgot]\n')
+#    if needgot:
+#        outf.write('global _GLOBAL_OFFSET_TABLE_\n')
+#        outf.write('_GLOBAL_OFFSET_TABLE_:\n')
+#        outf.write('dq dynamic\n')
     outf.write('_symbols:\n')
-    for library, symbols in libraries.items():
+    for library, symrels in libraries.items():
         outf.write('\t_symbols.{}: db "{}",0\n'.format(shorts[library], library))
 
-        for sym in symbols:
+        for sym, reloc in symrels:
+            if reloc != 'R_X86_64_PLT32' and reloc != 'R_X86_64_GOTPCRELX':
+                eprintf('Relocation type ' + reloc + ' of symbol ' + sym + ' unsupported!')
+                sys.exit(1)
+
+            if reloc == 'R_X86_64_GOTPCRELX':
+                outf.write("""
+global {name}
+{name}:
+""".format(name=sym).lstrip('\n'))
+
             hash = hash_djb2(sym)
             outf.write('\t\t_symbols.{lib}.{name}: dq 0x{hash:x}\n'\
                        .format(lib=shorts[library],name=sym,hash=hash))
@@ -60,9 +83,10 @@ def output_amd64(libraries, outf):
     outf.write('_symbols.end:\n')
 
     outf.write('_smolplt:\n')
-    for library, symbols in libraries.items():
-        for sym in symbols:
-            outf.write("""
+    for library, symrels in libraries.items():
+        for sym, reloc in symrels:
+            if reloc == 'R_X86_64_PLT32':
+                outf.write("""
 [section .text.smolplt.{name}]
 global {name}
 {name}:
diff --git a/smolparse.py b/smolparse.py
index 43f1cd4..ab6048e 100644
--- a/smolparse.py
+++ b/smolparse.py
@@ -31,18 +31,45 @@ def decide_arch(inpfiles):
 
     return archmagic[archn]
 
+def build_reloc_typ_table(reo):
+    relocs = dict({})
+
+    for s in reo.decode('utf-8').splitlines():
+        stuff = s.split()
+
+        # prolly a 'header' line
+        if len(stuff) < 5:
+            continue
+
+        # yes, we're assuming every reference to the same symbol will use the
+        # same relocation type. if this isn't the case, your compiler flags are
+        # stupid
+        relocs[stuff[4]] = stuff[2]
+
+    return relocs
+
 def get_needed_syms(readelf_bin, inpfiles):
     output = subprocess.check_output([readelf_bin, '-s', '-W']+inpfiles,
                                      stderr=subprocess.DEVNULL)
+    outrel = subprocess.check_output([readelf_bin, '-r', '-W']+inpfiles,
+                                     stderr=subprocess.DEVNULL)
+
+    relocs = build_reloc_typ_table(outrel)
 
     syms=set({})
     for entry in output.decode('utf-8').splitlines():
         stuff = entry.split()
         if len(stuff)<8: continue
-        if stuff[4] == "GLOBAL" and stuff[6] == "UND" and len(stuff[7])>0:
-            syms.add(stuff[7])
+        if stuff[4] == "GLOBAL" and stuff[6] == "UND" and len(stuff[7])>0 \
+                and stuff[7] in relocs:
+            syms.add((stuff[7], relocs[stuff[7]]))
 
-    return syms
+    #needgot = False
+    #if "_GLOBAL_OFFSET_TABLE_" in syms:
+    #    needgot = True
+    #    syms.remove("_GLOBAL_OFFSET_TABLE_")
+
+    return syms#, needgot
 
 def get_cc_paths(cc_bin):
     output = subprocess.check_output([cc_bin, '-print-search-dirs'],
diff --git a/src/loader64.asm b/src/loader64.asm
index ad955ee..7a6b125 100644
--- a/src/loader64.asm
+++ b/src/loader64.asm
@@ -44,7 +44,7 @@ _smol_start:
     sub rsi, LF_ENTRY_OFF+8
    xchg rbx, rsi
 
-    mov rsi, _symbols
+    mov esi, _symbols
 
             ; for (rsi = (uint8_t*)_symbols; *rsi; ++rsi) {
      .next_needed:
@@ -159,7 +159,9 @@ _smol_start:
 .needed_end:
    ;xor rbp, rbp ; still 0 from _dl_start_user
     mov rdi, rsp
+%ifdef ALIGN_STACK
    push rax
+%endif
 %ifdef USE_DL_FINI
    xchg rsi, r13 ; _dl_fini
 %endif