s390x: Implement bitrev (#4617)

Since we do not have an instruction for this, this is a simple
open-coded implementation.

Needed by the cg_clif frontend.

cranelift/filetests/filetests/isa/s390x/bitops.clif

@@ -1,40 +1,163 @@
 test compile precise-output
 target s390x
 
-;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
-;; BITREV
-;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
+function %bitrev_i128(i128) -> i128 {
+block0(v0: i128):
+    v1 = bitrev v0
+    return v1
+}
 
-; FIXME: bitrev not yet implemented
+; block0:
+;   vl %v0, 0(%r3)
+;   vrepib %v5, 170
+;   vrepib %v7, 1
+;   vsl %v17, %v0, %v7
+;   vsrl %v19, %v0, %v7
+;   vsel %v21, %v17, %v19, %v5
+;   vrepib %v23, 204
+;   vrepib %v25, 2
+;   vsl %v27, %v21, %v25
+;   vsrl %v29, %v21, %v25
+;   vsel %v31, %v27, %v29, %v23
+;   vrepib %v1, 240
+;   vrepib %v3, 4
+;   vsl %v5, %v31, %v3
+;   vsrl %v7, %v31, %v3
+;   vsel %v17, %v5, %v7, %v1
+;   bras %r1, 20 ; data.u128 0x0f0e0d0c0b0a09080706050403020100 ; vl %v19, 0(%r1)
+;   vperm %v21, %v17, %v17, %v19
+;   vst %v21, 0(%r2)
+;   br %r14
 
-;function %bitrev_i64(i64) -> i64 {
-;block0(v0: i64):
-;    v1 = bitrev v0
-;    return v1
-;}
-;
-;function %bitrev_i32(i32) -> i32 {
-;block0(v0: i32):
-;    v1 = bitrev v0
-;    return v1
-;}
-;
-;function %bitrev_i16(i16) -> i16 {
-;block0(v0: i16):
-;    v1 = bitrev v0
-;    return v1
-;}
-;
-;function %bitrev_i8(i8) -> i8 {
-;block0(v0: i8):
-;    v1 = bitrev v0
-;    return v1
-;}
-;
+function %bitrev_i64(i64) -> i64 {
+block0(v0: i64):
+    v1 = bitrev v0
+    return v1
+}
 
-;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
-;; CLZ
-;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
+; block0:
+;   llihf %r5, 2863311530
+;   iilf %r5, 2863311530
+;   sllg %r4, %r2, 1
+;   srlg %r2, %r2, 1
+;   ngr %r4, %r5
+;   xilf %r5, 4294967295
+;   xihf %r5, 4294967295
+;   ngrk %r5, %r2, %r5
+;   ogrk %r2, %r4, %r5
+;   llihf %r4, 3435973836
+;   iilf %r4, 3435973836
+;   sllg %r3, %r2, 2
+;   srlg %r5, %r2, 2
+;   ngr %r3, %r4
+;   xilf %r4, 4294967295
+;   xihf %r4, 4294967295
+;   ngrk %r4, %r5, %r4
+;   ogrk %r5, %r3, %r4
+;   llihf %r3, 4042322160
+;   iilf %r3, 4042322160
+;   sllg %r2, %r5, 4
+;   srlg %r4, %r5, 4
+;   ngr %r2, %r3
+;   xilf %r3, 4294967295
+;   xihf %r3, 4294967295
+;   ngrk %r3, %r4, %r3
+;   ogrk %r4, %r2, %r3
+;   lrvgr %r2, %r4
+;   br %r14
+
+function %bitrev_i32(i32) -> i32 {
+block0(v0: i32):
+    v1 = bitrev v0
+    return v1
+}
+
+; block0:
+;   iilf %r5, 2863311530
+;   sllk %r3, %r2, 1
+;   srlk %r2, %r2, 1
+;   nr %r3, %r5
+;   xilf %r5, 4294967295
+;   nrk %r4, %r2, %r5
+;   ork %r2, %r3, %r4
+;   iilf %r4, 3435973836
+;   sllk %r3, %r2, 2
+;   srlk %r5, %r2, 2
+;   nrk %r2, %r3, %r4
+;   xilf %r4, 4294967295
+;   nrk %r3, %r5, %r4
+;   ork %r5, %r2, %r3
+;   iilf %r3, 4042322160
+;   sllk %r2, %r5, 4
+;   srlk %r4, %r5, 4
+;   nrk %r5, %r2, %r3
+;   xilf %r3, 4294967295
+;   nrk %r2, %r4, %r3
+;   ork %r4, %r5, %r2
+;   lrvr %r2, %r4
+;   br %r14
+
+function %bitrev_i16(i16) -> i16 {
+block0(v0: i16):
+    v1 = bitrev v0
+    return v1
+}
+
+; block0:
+;   lhi %r5, -21846
+;   sllk %r3, %r2, 1
+;   srlk %r2, %r2, 1
+;   nr %r3, %r5
+;   xilf %r5, 4294967295
+;   nrk %r4, %r2, %r5
+;   ork %r2, %r3, %r4
+;   lhi %r4, -13108
+;   sllk %r3, %r2, 2
+;   srlk %r5, %r2, 2
+;   nrk %r2, %r3, %r4
+;   xilf %r4, 4294967295
+;   nrk %r3, %r5, %r4
+;   ork %r5, %r2, %r3
+;   lhi %r3, -3856
+;   sllk %r2, %r5, 4
+;   srlk %r4, %r5, 4
+;   nrk %r5, %r2, %r3
+;   xilf %r3, 4294967295
+;   nrk %r2, %r4, %r3
+;   ork %r4, %r5, %r2
+;   lrvr %r2, %r4
+;   srlk %r2, %r2, 16
+;   br %r14
+
+function %bitrev_i8(i8) -> i8 {
+block0(v0: i8):
+    v1 = bitrev v0
+    return v1
+}
+
+; block0:
+;   lhi %r5, -21846
+;   sllk %r3, %r2, 1
+;   srlk %r2, %r2, 1
+;   nr %r3, %r5
+;   xilf %r5, 4294967295
+;   nrk %r4, %r2, %r5
+;   ork %r2, %r3, %r4
+;   lhi %r4, -13108
+;   sllk %r3, %r2, 2
+;   srlk %r5, %r2, 2
+;   nrk %r2, %r3, %r4
+;   xilf %r4, 4294967295
+;   nrk %r3, %r5, %r4
+;   ork %r5, %r2, %r3
+;   lhi %r3, -3856
+;   sllk %r2, %r5, 4
+;   srlk %r4, %r5, 4
+;   nrk %r5, %r2, %r3
+;   xilf %r3, 4294967295
+;   nrk %r2, %r4, %r3
+;   ork %r2, %r5, %r2
+;   br %r14
 
 function %clz_i128(i128) -> i128 {
 block0(v0: i128):

cranelift/filetests/filetests/runtests/bitrev.clif

@@ -1,6 +1,7 @@
 test interpret
 test run
 target aarch64
+target s390x
 target x86_64
 
 function %bitrev_i8(i8) -> i8 {

cranelift/filetests/filetests/runtests/i128-bitrev.clif

@@ -1,6 +1,7 @@
 test run
 set enable_llvm_abi_extensions=true
 target aarch64
+target s390x
 target x86_64
 
 function %reverse_bits_zero() -> b1 {