RISC-V port and assorted changes

This commits adds code generation for the RISC-V architecture, both in 32- and 64-bit modes.

The generated code was lightly tested using the simulator and cross-binutils from https://riscv.org/software-tools/

This port required the following additional changes:

- Integers: More properties about shrx

- SelectOp: now provides smart constructors for mulhs and mulhu

- SelectDiv, 32-bit integer division and modulus: implement constant propagation, use the new smart constructors mulhs and mulhu.

- Runtime library: if no asm implementation is provided, run the reference C implementation through CompCert.  Since CompCert rejects the definitions of names of special functions such as __i64_shl, the reference implementation now uses "i64_" names, e.g. "i64_shl", and a renaming "i64_ -> __i64_" is performed over the generated assembly file, before assembling and building the runtime library.

- test/: add SIMU make variable to run tests through a simulator

- test/regression/alignas.c: make sure _Alignas and _Alignof are not #define'd by C headers

commit da14495c01cf4f66a928c2feff5c53f09bde837f
Author: Xavier Leroy <xavier.leroy@inria.fr>
Date:   Thu Apr 13 17:36:10 2017 +0200

    RISC-V port, continued

    Now working on Asmgen.

commit 36f36eb3a5abfbb8805960443d087b6a83e86005
Author: Xavier Leroy <xavier.leroy@inria.fr>
Date:   Wed Apr 12 17:26:39 2017 +0200

    RISC-V port, first steps

    This port is based on Prashanth Mundkur's experimental RV32 port and brings it up to date with CompCert, and adds 64-bit support (RV64).  Work in progress.

14 files changed, 66 insertions, 14 deletions

diff --git a/test/Makefile b/test/Makefile
index b469eec2..5e9e0555 100644
--- a/test/Makefile
+++ b/test/Makefile
@@ -4,7 +4,7 @@ all:
 	for i in $(DIRS); do $(MAKE) -C $$i all; done
 
 test:
-	set -e; for i in $(DIRS); do $(MAKE) -C $$i test; done
+	set -e; for i in $(DIRS); do $(MAKE) SIMU='$(SIMU)' -C $$i test; done
 
 bench:
 	for i in $(DIRS); do $(MAKE) -C $$i bench; done
diff --git a/test/c/Makefile b/test/c/Makefile
index 5979dfd4..94feb993 100644
--- a/test/c/Makefile
+++ b/test/c/Makefile
@@ -32,7 +32,7 @@ all_gcc: $(PROGS:%=%.gcc)
 
 test:
 	@for i in $(PROGS); do \
-	   if ./$$i.compcert | cmp -s - Results/$$i; \
+	   if $(SIMU) ./$$i.compcert | cmp -s - Results/$$i; \
            then echo "$$i: passed"; \
            else echo "$$i: FAILED"; exit 2; \
 	   fi; \
diff --git a/test/c/aes.c b/test/c/aes.c
index 053324f3..5bd57cbc 100644
--- a/test/c/aes.c
+++ b/test/c/aes.c
@@ -38,7 +38,8 @@ typedef unsigned int	u32;
 
 #if defined(__ppc__) || defined(__PPC__) || defined(__ARMEB__)
 #define ARCH_BIG_ENDIAN
-#elif defined(__i386__) || defined(__x86_64__) || defined(__ARMEL__)
+#elif defined(__i386__) || defined(__x86_64__) || defined(__ARMEL__) \
+   || defined(__riscv)
 #undef ARCH_BIG_ENDIAN
 #else
 #error "unknown endianness"
diff --git a/test/compression/Makefile b/test/compression/Makefile
index 784f7e73..fa73f0e6 100644
--- a/test/compression/Makefile
+++ b/test/compression/Makefile
@@ -35,9 +35,9 @@ test:
 	@echo "Test data: $(TESTFILE)"
 	@for i in $(EXE); do \
           echo "$$i: compression..."; \
-          ./$$i -c -i $(TESTFILE) -o $(TESTCOMPR); \
+          $(SIMU) ./$$i -c -i $(TESTFILE) -o $(TESTCOMPR); \
           echo "$$i: decompression..."; \
-          ./$$i -d -i $(TESTCOMPR) -o $(TESTEXPND); \
+          $(SIMU) ./$$i -d -i $(TESTCOMPR) -o $(TESTEXPND); \
           if cmp $(TESTFILE) $(TESTEXPND); \
           then echo "$$i: passed"; \
           else echo "$$i: FAILED"; exit 2; \
diff --git a/test/raytracer/Makefile b/test/raytracer/Makefile
index dc007fd7..c481ff86 100644
--- a/test/raytracer/Makefile
+++ b/test/raytracer/Makefile
@@ -23,7 +23,7 @@ depend:
 	gcc -MM *.c > .depend
 
 test:
-	./render < kal.gml
+	$(SIMU) ./render < kal.gml
 	@if cmp kal.ppm Results/kal.ppm; \
         then echo "raytracer: passed"; \
         else echo "raytracer: FAILED"; exit 2; \
diff --git a/test/regression/Makefile b/test/regression/Makefile
index 5def966b..54745863 100644
--- a/test/regression/Makefile
+++ b/test/regression/Makefile
@@ -71,15 +71,15 @@ clean:
 test:
 	@echo "----------- Compiled tests -------------"
 	@for i in $(TESTS) $(TESTS_COMP); do \
-	   ./Runtest $$i ./$$i.compcert; \
+	   SIMU='$(SIMU)' ./Runtest $$i ./$$i.compcert; \
          done
 	@echo "----------- Interpreted tests -------------"
 	@for i in $(TESTS); do \
-           ./Runtest $$i $(CCOMP) -fall -interp -quiet $$i.c; \
+           SIMU='' ./Runtest $$i $(CCOMP) -fall -interp -quiet $$i.c; \
          done
 	@for i in $(TESTS_DIFF); do \
            if $(CCOMP) -fall -interp -quiet $$i.c > _cinterp.log; then \
-             if ./$$i.compcert | cmp -s _cinterp.log -; \
+             if $(SIMU) ./$$i.compcert | cmp -s _cinterp.log -; \
              then echo "$$i: compiler and interpreter agree"; \
              else echo "$$i: compiler and interpreter DISAGREE"; \
              fi; \
diff --git a/test/regression/Results/builtins-riscV b/test/regression/Results/builtins-riscV
new file mode 100644
index 00000000..1576b252
--- /dev/null
+++ b/test/regression/Results/builtins-riscV
@@ -0,0 +1,12 @@
+bswap16(1234) = 3412
+bswap32(12345678) = 78563412
+bswap64(123456789abcdef0) = f0debc9a78563412
+fmadd(3.141590, 2.718000, 1.414000) = 9.952842
+fmsub(3.141590, 2.718000, 1.414000) = 7.124842
+fnmadd(3.141590, 2.718000, 1.414000) = -9.952842
+fnmsub(3.141590, 2.718000, 1.414000) = -7.124842
+fabs(3.141590) = 3.141590
+fabs(-3.141590) = 3.141590
+fsqrt(3.141590) = 1.772453
+fmax(3.141590, 2.718000) = 3.141590
+fmin(3.141590, 2.718000) = 2.718000
diff --git a/test/regression/Runtest b/test/regression/Runtest
index 9051b5b7..ad2a58f1 100755
--- a/test/regression/Runtest
+++ b/test/regression/Runtest
@@ -27,7 +27,7 @@ else
 fi
 
 # Administer the test
-if $* > $out
+if $SIMU $* > $out
 then
   if test -n "$ref"; then
     if cmp -s "$out" "$ref"; then
diff --git a/test/regression/alignas.c b/test/regression/alignas.c
index a6a2e690..b3754039 100644
--- a/test/regression/alignas.c
+++ b/test/regression/alignas.c
@@ -7,6 +7,13 @@
 #define _Alignof(x) __alignof__(x)
 #endif
 
+#ifdef _Alignas
+#undef _Alignas
+#endif
+#ifdef _Alignof
+#undef _Alignof
+#endif
+
 /* Base type */
 int _Alignas(16) a;
 char filler1;
diff --git a/test/regression/builtins-riscV.c b/test/regression/builtins-riscV.c
new file mode 100644
index 00000000..a302a6c4
--- /dev/null
+++ b/test/regression/builtins-riscV.c
@@ -0,0 +1,30 @@
+/* Fun with builtins */
+
+#include <stdio.h>
+
+int main(int argc, char ** argv)
+{
+  unsigned int x = 0x12345678;
+  unsigned short s = 0x1234;
+  unsigned long long zz = 0x123456789ABCDEF0ULL;
+  double a = 3.14159;
+  double b = 2.718;
+  double c = 1.414;
+
+  printf("bswap16(%x) = %x\n", s, __builtin_bswap16(s));
+  printf("bswap32(%x) = %x\n", x, __builtin_bswap32(x));
+  printf("bswap64(%llx) = %llx\n", zz, __builtin_bswap64(zz));
+  printf("fmadd(%f, %f, %f) = %f\n", a, b, c, __builtin_fmadd(a, b, c));
+  printf("fmsub(%f, %f, %f) = %f\n", a, b, c, __builtin_fmsub(a, b, c));
+  printf("fnmadd(%f, %f, %f) = %f\n", a, b, c, __builtin_fnmadd(a, b, c));
+  printf("fnmsub(%f, %f, %f) = %f\n", a, b, c, __builtin_fnmsub(a, b, c));
+  printf("fabs(%f) = %f\n", a, __builtin_fabs(a));
+  printf("fabs(%f) = %f\n", -a, __builtin_fabs(-a));
+  printf("fsqrt(%f) = %f\n", a, __builtin_fsqrt(a));
+  printf("fmax(%f, %f) = %f\n", a, b, __builtin_fmax(a, b));
+  printf("fmin(%f, %f) = %f\n", a, b, __builtin_fmin(a, b));
+  /* Make sure that ignoring the result of a builtin
+     doesn't cause an internal error */
+  (void) __builtin_fsqrt(a);
+  return 0;
+}
diff --git a/test/regression/extasm.c b/test/regression/extasm.c
index c0534047..a41c4202 100644
--- a/test/regression/extasm.c
+++ b/test/regression/extasm.c
@@ -18,7 +18,7 @@ int clobbers(int x, int z)
   return y + z;
 }
 
-#if defined(__x86_64__)
+#if defined(__x86_64__) || __riscv_xlen == 64
 #define SIXTYFOUR
 #else
 #undef SIXTYFOUR
diff --git a/test/regression/floats-basics.c b/test/regression/floats-basics.c
index 5aa91d14..a7ba3623 100644
--- a/test/regression/floats-basics.c
+++ b/test/regression/floats-basics.c
@@ -6,7 +6,8 @@
 
 #if defined(__ppc__) || defined(__PPC__) || defined(__ARMEB__)
 #define ARCH_BIG_ENDIAN
-#elif defined(__i386__) || defined(__x86_64__) || defined(__ARMEL__)
+#elif defined(__i386__) || defined(__x86_64__) || defined(__ARMEL__) \
+   || defined(__riscv)
 #undef ARCH_BIG_ENDIAN
 #else
 #error "unknown endianness"
diff --git a/test/regression/floats.c b/test/regression/floats.c
index 68d60f65..84c4e062 100644
--- a/test/regression/floats.c
+++ b/test/regression/floats.c
@@ -5,7 +5,8 @@
 
 #if defined(__ppc__) || defined(__PPC__) || defined(__ARMEB__)
 #define ARCH_BIG_ENDIAN
-#elif defined(__i386__) || defined(__x86_64__) || defined(__ARMEL__)
+#elif defined(__i386__) || defined(__x86_64__) || defined(__ARMEL__) \
+   || defined(__riscv)
 #undef ARCH_BIG_ENDIAN
 #else
 #error "unknown endianness"
diff --git a/test/spass/Makefile b/test/spass/Makefile
index f6acc551..110359ad 100644
--- a/test/spass/Makefile
+++ b/test/spass/Makefile
@@ -22,7 +22,7 @@ clean:
 	rm -f *.o *.s *.parsed.c *.light.c *.sdump
 
 test:
-	./spass small_problem.dfg | grep 'Proof found'
+	$(SIMU) ./spass small_problem.dfg | grep 'Proof found'
 
 TIME=xtime -o /dev/null     # Xavier's hack
 #TIME=time >/dev/null       # Otherwise