7 files changed, 101 insertions, 3 deletions
diff --git a/target/hexagon/arch.c b/target/hexagon/arch.c
index 40b6e3d..46edf45 100644
--- a/target/hexagon/arch.c
+++ b/target/hexagon/arch.c
@@ -181,12 +181,13 @@ int arch_sf_recip_common(float32 *Rs, float32 *Rt, float32 *Rd, int *adjust,
         /* or put Inf in num fixup? */
         uint8_t RsV_sign = float32_is_neg(RsV);
         uint8_t RtV_sign = float32_is_neg(RtV);
+        /* Check that RsV is NOT infinite before we overwrite it */
+        if (!float32_is_infinity(RsV)) {
+            float_raise(float_flag_divbyzero, fp_status);
+        }
         RsV = infinite_float32(RsV_sign ^ RtV_sign);
         RtV = float32_one;
         RdV = float32_one;
-        if (float32_is_infinity(RsV)) {
-            float_raise(float_flag_divbyzero, fp_status);
-        }
     } else if (float32_is_infinity(RtV)) {
         RsV = make_float32(0x80000000 & (RsV ^ RtV));
         RtV = float32_one;
@@ -279,3 +280,22 @@ int arch_sf_invsqrt_common(float32 *Rs, float32 *Rd, int *adjust,
     *adjust = PeV;
     return ret;
 }
+
+const uint8_t recip_lookup_table[128] = {
+    0x0fe, 0x0fa, 0x0f6, 0x0f2, 0x0ef, 0x0eb, 0x0e7, 0x0e4,
+    0x0e0, 0x0dd, 0x0d9, 0x0d6, 0x0d2, 0x0cf, 0x0cc, 0x0c9,
+    0x0c6, 0x0c2, 0x0bf, 0x0bc, 0x0b9, 0x0b6, 0x0b3, 0x0b1,
+    0x0ae, 0x0ab, 0x0a8, 0x0a5, 0x0a3, 0x0a0, 0x09d, 0x09b,
+    0x098, 0x096, 0x093, 0x091, 0x08e, 0x08c, 0x08a, 0x087,
+    0x085, 0x083, 0x080, 0x07e, 0x07c, 0x07a, 0x078, 0x075,
+    0x073, 0x071, 0x06f, 0x06d, 0x06b, 0x069, 0x067, 0x065,
+    0x063, 0x061, 0x05f, 0x05e, 0x05c, 0x05a, 0x058, 0x056,
+    0x054, 0x053, 0x051, 0x04f, 0x04e, 0x04c, 0x04a, 0x049,
+    0x047, 0x045, 0x044, 0x042, 0x040, 0x03f, 0x03d, 0x03c,
+    0x03a, 0x039, 0x037, 0x036, 0x034, 0x033, 0x032, 0x030,
+    0x02f, 0x02d, 0x02c, 0x02b, 0x029, 0x028, 0x027, 0x025,
+    0x024, 0x023, 0x021, 0x020, 0x01f, 0x01e, 0x01c, 0x01b,
+    0x01a, 0x019, 0x017, 0x016, 0x015, 0x014, 0x013, 0x012,
+    0x011, 0x00f, 0x00e, 0x00d, 0x00c, 0x00b, 0x00a, 0x009,
+    0x008, 0x007, 0x006, 0x005, 0x004, 0x003, 0x002, 0x000,
+};
diff --git a/target/hexagon/arch.h b/target/hexagon/arch.h
index 6e0b0d9..b6634e9 100644
--- a/target/hexagon/arch.h
+++ b/target/hexagon/arch.h
@@ -30,4 +30,6 @@ int arch_sf_recip_common(float32 *Rs, float32 *Rt, float32 *Rd,
 int arch_sf_invsqrt_common(float32 *Rs, float32 *Rd, int *adjust,
                           float_status *fp_status);
 
+extern const uint8_t recip_lookup_table[128];
+
 #endif
diff --git a/target/hexagon/gen_tcg.h b/target/hexagon/gen_tcg.h
index a30048e..428a670 100644
--- a/target/hexagon/gen_tcg.h
+++ b/target/hexagon/gen_tcg.h
@@ -195,6 +195,27 @@
 #define fGEN_TCG_S4_stored_locked(SHORTCODE) \
     do { SHORTCODE; READ_PREG(PdV, PdN); } while (0)
 
+/*
+ * Mathematical operations with more than one definition require
+ * special handling
+ */
+
+/*
+ * Approximate reciprocal
+ * r3,p1 = sfrecipa(r0, r1)
+ *
+ * The helper packs the 2 32-bit results into a 64-bit value,
+ * so unpack them into the proper results.
+ */
+#define fGEN_TCG_F2_sfrecipa(SHORTCODE) \
+    do { \
+        TCGv_i64 tmp = tcg_temp_new_i64(); \
+        gen_helper_sfrecipa(tmp, cpu_env, RsV, RtV);  \
+        tcg_gen_extrh_i64_i32(RdV, tmp); \
+        tcg_gen_extrl_i64_i32(PeV, tmp); \
+        tcg_temp_free_i64(tmp); \
+    } while (0)
+
 /* Floating point */
 #define fGEN_TCG_F2_conv_sf2df(SHORTCODE) \
     gen_helper_conv_sf2df(RddV, cpu_env, RsV)
diff --git a/target/hexagon/helper.h b/target/hexagon/helper.h
index efe6069..b377293 100644
--- a/target/hexagon/helper.h
+++ b/target/hexagon/helper.h
@@ -24,6 +24,7 @@ DEF_HELPER_FLAGS_3(debug_check_store_width, TCG_CALL_NO_WG, void, env, int, int)
 DEF_HELPER_FLAGS_3(debug_commit_end, TCG_CALL_NO_WG, void, env, int, int)
 DEF_HELPER_2(commit_store, void, env, int)
 DEF_HELPER_FLAGS_4(fcircadd, TCG_CALL_NO_RWG_SE, s32, s32, s32, s32, s32)
+DEF_HELPER_3(sfrecipa, i64, env, f32, f32)
 
 /* Floating point */
 DEF_HELPER_2(conv_sf2df, f64, env, f32)
diff --git a/target/hexagon/imported/encode_pp.def b/target/hexagon/imported/encode_pp.def
index c21cb73..b01b4d7 100644
--- a/target/hexagon/imported/encode_pp.def
+++ b/target/hexagon/imported/encode_pp.def
@@ -1028,6 +1028,7 @@ MPY_ENC(F2_sfmin,            "1011","ddddd","0","0","0","1","01")
 MPY_ENC(F2_sfmpy,            "1011","ddddd","0","0","1","0","00")
 MPY_ENC(F2_sffixupn,         "1011","ddddd","0","0","1","1","00")
 MPY_ENC(F2_sffixupd,         "1011","ddddd","0","0","1","1","01")
+MPY_ENC(F2_sfrecipa,         "1011","ddddd","1","1","1","1","ee")
 
 DEF_FIELDROW_DESC32(ICLASS_M" 1100 -------- PP------ --------","[#12] Rd=(Rs,Rt)")
 DEF_FIELD32(ICLASS_M"         1100 -------- PP------ --!-----",Mc_tH,"Rt is High") /*Rt high */
diff --git a/target/hexagon/imported/float.idef b/target/hexagon/imported/float.idef
index 76cecfe..eb54158 100644
--- a/target/hexagon/imported/float.idef
+++ b/target/hexagon/imported/float.idef
@@ -146,6 +146,22 @@ Q6INSN(F2_sfimm_n,"Rd32=sfmake(#u10):neg",ATTRIBS(),
 })
 
 
+Q6INSN(F2_sfrecipa,"Rd32,Pe4=sfrecipa(Rs32,Rt32)",ATTRIBS(),
+"Reciprocal Approximation for Division",
+{
+    fHIDE(int idx;)
+    fHIDE(int adjust;)
+    fHIDE(int mant;)
+    fHIDE(int exp;)
+    if (fSF_RECIP_COMMON(RsV,RtV,RdV,adjust)) {
+        PeV = adjust;
+        idx = (RtV >> 16) & 0x7f;
+        mant = (fSF_RECIP_LOOKUP(idx) << 15) | 1;
+        exp = fSF_BIAS() - (fSF_GETEXP(RtV) - fSF_BIAS()) - 1;
+        RdV = fMAKESF(fGETBIT(31,RtV),exp,mant);
+    }
+})
+
 Q6INSN(F2_sffixupn,"Rd32=sffixupn(Rs32,Rt32)",ATTRIBS(),
 "Fix Up Numerator",
 {
diff --git a/target/hexagon/op_helper.c b/target/hexagon/op_helper.c
index 33b6713..75861e2 100644
--- a/target/hexagon/op_helper.c
+++ b/target/hexagon/op_helper.c
@@ -289,6 +289,43 @@ int32_t HELPER(fcircadd)(int32_t RxV, int32_t offset, int32_t M, int32_t CS)
     return new_ptr;
 }
 
+static float32 build_float32(uint8_t sign, uint32_t exp, uint32_t mant)
+{
+    return make_float32(
+        ((sign & 1) << 31) |
+        ((exp & 0xff) << SF_MANTBITS) |
+        (mant & ((1 << SF_MANTBITS) - 1)));
+}
+
+/*
+ * sfrecipa, sfinvsqrta have two 32-bit results
+ *     r0,p0=sfrecipa(r1,r2)
+ *     r0,p0=sfinvsqrta(r1)
+ *
+ * Since helpers can only return a single value, we pack the two results
+ * into a 64-bit value.
+ */
+uint64_t HELPER(sfrecipa)(CPUHexagonState *env, float32 RsV, float32 RtV)
+{
+    int32_t PeV = 0;
+    float32 RdV;
+    int idx;
+    int adjust;
+    int mant;
+    int exp;
+
+    arch_fpop_start(env);
+    if (arch_sf_recip_common(&RsV, &RtV, &RdV, &adjust, &env->fp_status)) {
+        PeV = adjust;
+        idx = (RtV >> 16) & 0x7f;
+        mant = (recip_lookup_table[idx] << 15) | 1;
+        exp = SF_BIAS - (float32_getexp(RtV) - SF_BIAS) - 1;
+        RdV = build_float32(extract32(RtV, 31, 1), exp, mant);
+    }
+    arch_fpop_end(env);
+    return ((uint64_t)RdV << 32) | PeV;
+}
+
 /*
  * mem_noshuf
  * Section 5.5 of the Hexagon V67 Programmer's Reference Manual