diff --git a/mk/intrinsics.mk b/mk/intrinsics.mk
index 77442ec530a..2a43b7c1667 100644
--- a/mk/intrinsics.mk
+++ b/mk/intrinsics.mk
@@ -11,7 +11,7 @@ $(INTRINSICS_LL):  $(INTRINSICS_LL_IN) $(MKFILES)
 	@$(call E, mkdir: intrinsics)
 	$(Q)mkdir -p intrinsics
 	@$(call E, sed: $@)
-	$(Q)sed s/@CFG_LLVM_TRIPLE@/$(CFG_LLVM_TRIPLE)/g $< > $@
+	$(Q)sed s/@CFG_TARGET_TRIPLE@/$(CFG_LLVM_TRIPLE)/g $< > $@
 
 $(INTRINSICS_BC):   $(INTRINSICS_LL) $(MKFILES)
 	@$(call E, llvm-as: $@)
diff --git a/mk/rt.mk b/mk/rt.mk
index ff6d24def1c..a486edd6ea8 100644
--- a/mk/rt.mk
+++ b/mk/rt.mk
@@ -103,6 +103,8 @@ RUNTIME_INCS := -I $(S)src/rt/isaac -I $(S)src/rt/uthash \
 RUNTIME_OBJS := $(RUNTIME_CS:.cpp=.o) $(RUNTIME_LL:.ll=.o) $(RUNTIME_S:.S=.o)
 RUNTIME_LIBS := $(LIBUV_LIB)
 
+RT_COMPILE_C := $(call CFG_COMPILE_C, $(0), $(1) -I $(S)src/rt/arch/$(2))
+
 rt/%.o: rt/%.cpp $(MKFILES)
 	@$(call E, compile: $@)
 	$(Q)$(call CFG_COMPILE_C, $@, $(RUNTIME_INCS)) $<
diff --git a/mk/target.mk b/mk/target.mk
index 9f0739bfb7a..1b1d177e20f 100644
--- a/mk/target.mk
+++ b/mk/target.mk
@@ -7,9 +7,19 @@
 
 define TARGET_STAGE_N
 
-$$(TARGET_LIB$(1)$(2))/intrinsics.bc: $$(INTRINSICS_BC)
-	@$$(call E, cp: $$@)
-	$$(Q)cp $$< $$@
+TARGET_HOST    := $$(word 1,$$(subst -, ,$(2)))
+
+# For some reason there is (sometimes) a mismatch here between i686, i386, etc
+INTR_HOST := $$(subst i686,i386,$$(TARGET_HOST))
+
+$$(TARGET_LIB$(1)$(2))/intrinsics.ll: \
+		$$(S)src/rt/intrinsics/intrinsics.$$(INTR_HOST).ll.in
+	@$$(call E, sed: $$@)
+	sed s/@CFG_TARGET_TRIPLE@/$(2)/ $$< > $$@
+
+$$(TARGET_LIB$(1)$(2))/intrinsics.bc: $$(TARGET_LIB$(1)$(2))/intrinsics.ll
+	@$$(call E, llvms-as: $$@)
+	$$(LLVM_AS) -o $$@ $$<
 
 $$(TARGET_LIB$(1)$(2))/$$(CFG_STDLIB): \
 	$$(STDLIB_CRATE) $$(STDLIB_INPUTS) \
diff --git a/src/etc/gen-intrinsics b/src/etc/gen-intrinsics
new file mode 100755
index 00000000000..45fd5482f4a
--- /dev/null
+++ b/src/etc/gen-intrinsics
@@ -0,0 +1,19 @@
+#!/bin/sh
+
+# This script generates new definitions for the intrinsics using
+# clang. This is not currently in the Makefile to avoid any dependency
+# on clang.
+
+for ARCH in i386 x86_64
+do
+    clang++ -emit-llvm -S -arch $ARCH -O3 -Isrc/rt/isaac -Isrc/rt/uthash \
+        -Isrc/rt/arch/$ARCH -fno-stack-protector \
+        -o src/rt/intrinsics/intrinsics.$ARCH.ll.in \
+        src/rt/intrinsics/intrinsics.cpp
+    sed -i "" \
+        -e 's/^target datalayout =/; target datalayout =/' \
+        src/rt/intrinsics/intrinsics.$ARCH.ll.in
+    sed -i "" \
+        -e 's/^target triple = "[^"]*"/target triple = "@CFG_TARGET_TRIPLE@"/' \
+        src/rt/intrinsics/intrinsics.$ARCH.ll.in
+done
\ No newline at end of file
diff --git a/src/rt/intrinsics/intrinsics.ll.in b/src/rt/intrinsics/intrinsics.i386.ll.in
similarity index 81%
rename from src/rt/intrinsics/intrinsics.ll.in
rename to src/rt/intrinsics/intrinsics.i386.ll.in
index aabc5bcd245..93cc32fe995 100644
--- a/src/rt/intrinsics/intrinsics.ll.in
+++ b/src/rt/intrinsics/intrinsics.i386.ll.in
@@ -1,5 +1,6 @@
-; ModuleID = 'intrinsics.cpp'
-target triple = "@CFG_LLVM_TRIPLE@"
+; ModuleID = 'src/rt/intrinsics/intrinsics.cpp'
+; target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-f32:32:32-f64:32:64-v64:64:64-v128:128:128-a0:0:64-f80:128:128-n8:16:32"
+target triple = "@CFG_TARGET_TRIPLE@"
 
 %0 = type { i32, %"struct.memory_region::alloc_header"**, i32 }
 %1 = type { i32, %struct.rust_scheduler**, i32 }
@@ -52,10 +53,10 @@ target triple = "@CFG_LLVM_TRIPLE@"
 %struct.rust_vec = type { i32, i32, [0 x i8] }
 %"struct.std::_Rb_tree<void *, std::pair<void *const, const type_desc *>, std::_Select1st<std::pair<void *const, const type_desc *> >, std::less<void *>, std::allocator<std::pair<void *const, const type_desc *> > >::_Rb_tree_impl" = type { %struct.rust_cond, %"struct.std::_Rb_tree_node_base", i32 }
 %"struct.std::_Rb_tree_node_base" = type { i32, %"struct.std::_Rb_tree_node_base"*, %"struct.std::_Rb_tree_node_base"*, %"struct.std::_Rb_tree_node_base"* }
-%struct.type_desc = type { %struct.type_desc**, i32, i32, void (i8*, i8*, %struct.type_desc**, i8*)*, void (i8*, i8*, %struct.type_desc**, i8*)*, void (i8*, i8*, %struct.type_desc**, i8*)*, i8*, void (i8*, i8*, %struct.type_desc**, i8*)*, void (i8*, i8*, %struct.type_desc**, i8*)*, i32, void (i8*, i8*, %struct.type_desc**, i8*, i8*, i8)*, i8*, %struct.rust_shape_tables*, i32, i32, %struct.UT_hash_handle, i32, [0 x %struct.type_desc*] }
+%struct.type_desc = type { %struct.type_desc**, i32, i32, void (i8*, %struct.rust_task*, i8*, %struct.type_desc**, i8*)*, void (i8*, %struct.rust_task*, i8*, %struct.type_desc**, i8*)*, void (i8*, %struct.rust_task*, i8*, %struct.type_desc**, i8*)*, i8*, void (i8*, %struct.rust_task*, i8*, %struct.type_desc**, i8*)*, void (i8*, %struct.rust_task*, i8*, %struct.type_desc**, i8*)*, i32, void (i8*, %struct.rust_task*, i8*, %struct.type_desc**, i8*, i8*, i8)*, i8*, %struct.rust_shape_tables*, i32, i32, %struct.UT_hash_handle, i32, [0 x %struct.type_desc*] }
 
 @.str = private unnamed_addr constant [42 x i8] c"attempt to cast values of differing sizes\00"
-@.str1 = private unnamed_addr constant [15 x i8] c"intrinsics.cpp\00"
+@.str1 = private unnamed_addr constant [33 x i8] c"src/rt/intrinsics/intrinsics.cpp\00"
 
 define void @rust_intrinsic_vec_len(%struct.rust_task* nocapture %task, i32* nocapture %retptr, %struct.type_desc* nocapture %ty, %struct.rust_vec** nocapture %vp) nounwind {
   %1 = load %struct.rust_vec** %vp, align 4
@@ -77,7 +78,7 @@ define void @rust_intrinsic_ptr_offset(%struct.rust_task* nocapture %task, i8**
   ret void
 }
 
-define void @rust_intrinsic_cast(%struct.rust_task* nocapture %task, i8* nocapture %retptr, %struct.type_desc* nocapture %t1, %struct.type_desc* nocapture %t2, i8* nocapture %src) {
+define void @rust_intrinsic_cast(%struct.rust_task* %task, i8* nocapture %retptr, %struct.type_desc* nocapture %t1, %struct.type_desc* nocapture %t2, i8* nocapture %src) {
   %1 = getelementptr inbounds %struct.type_desc* %t1, i32 0, i32 1
   %2 = load i32* %1, align 4
   %3 = getelementptr inbounds %struct.type_desc* %t2, i32 0, i32 1
@@ -86,7 +87,7 @@ define void @rust_intrinsic_cast(%struct.rust_task* nocapture %task, i8* nocaptu
   br i1 %5, label %7, label %6
 
 ; <label>:6                                       ; preds = %0
-  tail call void @upcall_fail(i8* getelementptr inbounds ([42 x i8]* @.str, i32 0, i32 0), i8* getelementptr inbounds ([15 x i8]* @.str1, i32 0, i32 0), i32 32)
+  tail call void @upcall_fail(%struct.rust_task* %task, i8* getelementptr inbounds ([42 x i8]* @.str, i32 0, i32 0), i8* getelementptr inbounds ([33 x i8]* @.str1, i32 0, i32 0), i32 32)
   br label %8
 
 ; <label>:7                                       ; preds = %0
@@ -97,7 +98,7 @@ define void @rust_intrinsic_cast(%struct.rust_task* nocapture %task, i8* nocaptu
   ret void
 }
 
-declare void @upcall_fail(i8*, i8*, i32)
+declare void @upcall_fail(%struct.rust_task*, i8*, i8*, i32)
 
 declare void @llvm.memmove.p0i8.p0i8.i32(i8* nocapture, i8* nocapture, i32, i32, i1) nounwind
 
@@ -106,16 +107,11 @@ define void @rust_intrinsic_addr_of(%struct.rust_task* nocapture %task, i8** noc
   ret void
 }
 
-define void @rust_intrinsic_recv(%struct.rust_task* nocapture %task, i8** %retptr, %struct.type_desc* nocapture %ty, %class.rust_port* %port) {
-  %1 = bitcast i8** %retptr to i32*
-  tail call void @port_recv(i32* %1, %class.rust_port* %port)
+define void @rust_intrinsic_recv(%struct.rust_task* %task, i8** %retptr, %struct.type_desc* nocapture %ty, %class.rust_port* %port) {
+  %1 = bitcast %struct.rust_task* %task to i8*
+  %2 = bitcast i8** %retptr to i32*
+  tail call void @port_recv(i8* %1, i32* %2, %class.rust_port* %port)
   ret void
 }
 
-declare void @port_recv(i32*, %class.rust_port*)
-
-define void @rust_intrinsic_get_type_desc(%struct.rust_task* nocapture %task, i8** nocapture %retptr, %struct.type_desc* %ty) nounwind {
-  %ty.c = bitcast %struct.type_desc* %ty to i8*
-  store i8* %ty.c, i8** %retptr, align 4
-  ret void
-}
+declare void @port_recv(i8*, i32*, %class.rust_port*)
diff --git a/src/rt/intrinsics/intrinsics.x86_64.ll.in b/src/rt/intrinsics/intrinsics.x86_64.ll.in
new file mode 100644
index 00000000000..50e2e9178d4
--- /dev/null
+++ b/src/rt/intrinsics/intrinsics.x86_64.ll.in
@@ -0,0 +1,117 @@
+; ModuleID = 'src/rt/intrinsics/intrinsics.cpp'
+; target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64"
+target triple = "@CFG_TARGET_TRIPLE@"
+
+%0 = type { i64, %"struct.memory_region::alloc_header"**, i64 }
+%1 = type { i64, %struct.rust_scheduler**, i64 }
+%2 = type { %"struct.hash_map<long, rust_task *>::map_entry"* }
+%3 = type { %struct.rust_task*, i64, i64, %class.rust_chan** }
+%class.array_list = type { i64, %struct.rust_task**, i64 }
+%class.circular_buffer = type { %class.rust_kernel*, i64, i64, i64, i64, i8* }
+%class.context = type { %struct.registers_t, %class.context* }
+%"class.debug::task_debug_info" = type { %"class.std::map" }
+%class.hash_map = type { %"struct.hash_map<long, rust_port *>::map_entry"* }
+%class.indexed_list = type { i32 (...)**, %class.array_list }
+%class.lock_and_signal = type { i32 (...)**, %struct._opaque_pthread_cond_t, %struct._opaque_pthread_attr_t, %struct._opaque_pthread_t*, i8, i8 }
+%class.memory_region = type { i32 (...)**, %class.rust_srv*, %class.memory_region*, i32, %0, i8, i8, %class.lock_and_signal }
+%class.ptr_vec = type { %struct.rust_task*, i64, i64, %struct.rust_token** }
+%class.rust_chan = type { i64, %class.rust_kernel*, %struct.rust_task*, %class.rust_port*, i64, %class.circular_buffer }
+%class.rust_crate_cache = type { %struct.type_desc*, %struct.rust_scheduler*, i64 }
+%class.rust_kernel = type { i32 (...)**, %class.memory_region, %class.rust_log, %class.rust_srv*, %class.lock_and_signal, %1, %struct.randctx, i64, %2, i64, i32, i32, %struct.rust_env* }
+%class.rust_log = type { i32 (...)**, %class.rust_srv*, %struct.rust_scheduler*, i8 }
+%class.rust_obstack = type { %struct.rust_obstack_chunk*, %struct.rust_task* }
+%class.rust_port = type { i64, i64, %class.rust_kernel*, %struct.rust_task*, %class.rust_chan*, i64, %class.ptr_vec, %3, %class.lock_and_signal }
+%class.rust_srv = type { i32 (...)**, %struct.rust_env*, %class.memory_region }
+%"class.rust_task::wakeup_callback" = type { i32 (...)** }
+%class.rust_task_list = type { %class.indexed_list, %struct.rust_scheduler*, i8* }
+%class.rust_thread = type { i32 (...)**, i8, %struct._opaque_pthread_t* }
+%"class.std::_Rb_tree" = type { %"struct.std::_Rb_tree<void *, std::pair<void *const, const type_desc *>, std::_Select1st<std::pair<void *const, const type_desc *> >, std::less<void *>, std::allocator<std::pair<void *const, const type_desc *> > >::_Rb_tree_impl" }
+%"class.std::map" = type { %"class.std::_Rb_tree" }
+%class.timer = type { i32 (...)**, i64, i64 }
+%struct.UT_hash_bucket = type { %struct.UT_hash_handle*, i32, i32 }
+%struct.UT_hash_handle = type { %struct.UT_hash_table*, i8*, i8*, %struct.UT_hash_handle*, %struct.UT_hash_handle*, i8*, i32, i32 }
+%struct.UT_hash_table = type { %struct.UT_hash_bucket*, i32, i32, i32, %struct.UT_hash_handle*, i64, i32, i32, i32, i32 }
+%struct.__darwin_pthread_handler_rec = type { void (i8*)*, i8*, %struct.__darwin_pthread_handler_rec* }
+%struct._opaque_pthread_attr_t = type { i64, [56 x i8] }
+%struct._opaque_pthread_cond_t = type { i64, [40 x i8] }
+%struct._opaque_pthread_t = type { i64, %struct.__darwin_pthread_handler_rec*, [1168 x i8] }
+%struct.chan_handle = type { i64, i64 }
+%"struct.hash_map<long, rust_port *>::map_entry" = type opaque
+%"struct.hash_map<long, rust_task *>::map_entry" = type opaque
+%"struct.memory_region::alloc_header" = type { i32, i32, i8*, i32, [0 x i8] }
+%struct.randctx = type { i64, [256 x i64], [256 x i64], i64, i64, i64 }
+%struct.registers_t = type { [7 x i64], [6 x i64], i64 }
+%struct.rust_cond = type { i8 }
+%struct.rust_env = type { i64, i64, i8*, i8, i8, i8* }
+%struct.rust_obstack_chunk = type { %struct.rust_obstack_chunk*, i64, i64, i64, [0 x i8] }
+%struct.rust_scheduler = type { %class.rust_thread, i64, i64, %class.rust_log, i32, %class.rust_srv*, i8*, %class.rust_task_list, %class.rust_task_list, %class.rust_task_list, %class.rust_task_list, %class.rust_crate_cache, %struct.randctx, %class.rust_kernel*, i32, i32, %class.lock_and_signal, i64, %struct._opaque_pthread_attr_t, %struct.rust_env*, %class.context }
+%struct.rust_shape_tables = type { i8*, i8* }
+%struct.rust_task = type { %struct.rust_task_user, i64, %class.context, %struct.stk_seg*, i64, %struct.rust_scheduler*, %class.rust_crate_cache*, %class.rust_kernel*, i8*, %class.rust_task_list*, %struct.rust_cond*, i8*, %struct.rust_task*, i32, i64, %class.timer, i64*, %class.array_list, i32, i32, %class.memory_region, %"class.rust_task::wakeup_callback"*, i8, i8, i8, %class.lock_and_signal, %class.hash_map, %class.rust_obstack, %"class.std::map", i32, %"class.debug::task_debug_info" }
+%struct.rust_task_user = type { i64, i32, %struct.chan_handle, i64 }
+%struct.rust_token = type opaque
+%struct.rust_vec = type { i64, i64, [0 x i8] }
+%"struct.std::_Rb_tree<void *, std::pair<void *const, const type_desc *>, std::_Select1st<std::pair<void *const, const type_desc *> >, std::less<void *>, std::allocator<std::pair<void *const, const type_desc *> > >::_Rb_tree_impl" = type { %struct.rust_cond, %"struct.std::_Rb_tree_node_base", i64 }
+%"struct.std::_Rb_tree_node_base" = type { i32, %"struct.std::_Rb_tree_node_base"*, %"struct.std::_Rb_tree_node_base"*, %"struct.std::_Rb_tree_node_base"* }
+%struct.stk_seg = type { i32, i64, [0 x i8] }
+%struct.type_desc = type { %struct.type_desc**, i64, i64, void (i8*, %struct.rust_task*, i8*, %struct.type_desc**, i8*)*, void (i8*, %struct.rust_task*, i8*, %struct.type_desc**, i8*)*, void (i8*, %struct.rust_task*, i8*, %struct.type_desc**, i8*)*, i8*, void (i8*, %struct.rust_task*, i8*, %struct.type_desc**, i8*)*, void (i8*, %struct.rust_task*, i8*, %struct.type_desc**, i8*)*, i64, void (i8*, %struct.rust_task*, i8*, %struct.type_desc**, i8*, i8*, i8)*, i8*, %struct.rust_shape_tables*, i64, i64, %struct.UT_hash_handle, i64, [0 x %struct.type_desc*] }
+
+@.str = private unnamed_addr constant [42 x i8] c"attempt to cast values of differing sizes\00"
+@.str1 = private unnamed_addr constant [33 x i8] c"src/rt/intrinsics/intrinsics.cpp\00"
+
+define void @rust_intrinsic_vec_len(%struct.rust_task* nocapture %task, i64* nocapture %retptr, %struct.type_desc* nocapture %ty, %struct.rust_vec** nocapture %vp) nounwind {
+  %1 = load %struct.rust_vec** %vp, align 8
+  %2 = getelementptr inbounds %struct.rust_vec* %1, i64 0, i32 0
+  %3 = load i64* %2, align 8
+  %4 = getelementptr inbounds %struct.type_desc* %ty, i64 0, i32 1
+  %5 = load i64* %4, align 8
+  %6 = udiv i64 %3, %5
+  store i64 %6, i64* %retptr, align 8
+  ret void
+}
+
+define void @rust_intrinsic_ptr_offset(%struct.rust_task* nocapture %task, i8** nocapture %retptr, %struct.type_desc* nocapture %ty, i8* %ptr, i64 %count) nounwind {
+  %1 = getelementptr inbounds %struct.type_desc* %ty, i64 0, i32 1
+  %2 = load i64* %1, align 8
+  %3 = mul i64 %2, %count
+  %4 = getelementptr inbounds i8* %ptr, i64 %3
+  store i8* %4, i8** %retptr, align 8
+  ret void
+}
+
+define void @rust_intrinsic_cast(%struct.rust_task* %task, i8* nocapture %retptr, %struct.type_desc* nocapture %t1, %struct.type_desc* nocapture %t2, i8* nocapture %src) {
+  %1 = getelementptr inbounds %struct.type_desc* %t1, i64 0, i32 1
+  %2 = load i64* %1, align 8
+  %3 = getelementptr inbounds %struct.type_desc* %t2, i64 0, i32 1
+  %4 = load i64* %3, align 8
+  %5 = icmp eq i64 %2, %4
+  br i1 %5, label %7, label %6
+
+; <label>:6                                       ; preds = %0
+  tail call void @upcall_fail(%struct.rust_task* %task, i8* getelementptr inbounds ([42 x i8]* @.str, i64 0, i64 0), i8* getelementptr inbounds ([33 x i8]* @.str1, i64 0, i64 0), i64 32)
+  br label %8
+
+; <label>:7                                       ; preds = %0
+  tail call void @llvm.memmove.p0i8.p0i8.i64(i8* %retptr, i8* %src, i64 %2, i32 1, i1 false)
+  br label %8
+
+; <label>:8                                       ; preds = %7, %6
+  ret void
+}
+
+declare void @upcall_fail(%struct.rust_task*, i8*, i8*, i64)
+
+declare void @llvm.memmove.p0i8.p0i8.i64(i8* nocapture, i8* nocapture, i64, i32, i1) nounwind
+
+define void @rust_intrinsic_addr_of(%struct.rust_task* nocapture %task, i8** nocapture %retptr, %struct.type_desc* nocapture %ty, i8* %valptr) nounwind {
+  store i8* %valptr, i8** %retptr, align 8
+  ret void
+}
+
+define void @rust_intrinsic_recv(%struct.rust_task* %task, i8** %retptr, %struct.type_desc* nocapture %ty, %class.rust_port* %port) {
+  %1 = bitcast %struct.rust_task* %task to i8*
+  %2 = bitcast i8** %retptr to i64*
+  tail call void @port_recv(i8* %1, i64* %2, %class.rust_port* %port)
+  ret void
+}
+
+declare void @port_recv(i8*, i64*, %class.rust_port*)