Hexagon (target/hexagon) add A4_addp_c/A4_subp_c

Rdd32 = add(Rss32, Rtt32, Px4):carry
    Add with carry
Rdd32 = sub(Rss32, Rtt32, Px4):carry
    Sub with carry

Test cases in tests/tcg/hexagon/multi_result.c

Signed-off-by: Taylor Simpson <tsimpson@quicinc.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <1617930474-31979-22-git-send-email-tsimpson@quicinc.com>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

4 files changed, 65 insertions, 0 deletions

diff --git a/target/hexagon/gen_tcg.h b/target/hexagon/gen_tcg.h
index aea0c55..6bc578d 100644
--- a/target/hexagon/gen_tcg.h
+++ b/target/hexagon/gen_tcg.h
@@ -238,6 +238,43 @@
     } while (0)
 
 /*
+ * Add or subtract with carry.
+ * Predicate register is used as an extra input and output.
+ * r5:4 = add(r1:0, r3:2, p1):carry
+ */
+#define fGEN_TCG_A4_addp_c(SHORTCODE) \
+    do { \
+        TCGv_i64 carry = tcg_temp_new_i64(); \
+        TCGv_i64 zero = tcg_const_i64(0); \
+        tcg_gen_extu_i32_i64(carry, PxV); \
+        tcg_gen_andi_i64(carry, carry, 1); \
+        tcg_gen_add2_i64(RddV, carry, RssV, zero, carry, zero); \
+        tcg_gen_add2_i64(RddV, carry, RddV, carry, RttV, zero); \
+        tcg_gen_extrl_i64_i32(PxV, carry); \
+        gen_8bitsof(PxV, PxV); \
+        tcg_temp_free_i64(carry); \
+        tcg_temp_free_i64(zero); \
+    } while (0)
+
+/* r5:4 = sub(r1:0, r3:2, p1):carry */
+#define fGEN_TCG_A4_subp_c(SHORTCODE) \
+    do { \
+        TCGv_i64 carry = tcg_temp_new_i64(); \
+        TCGv_i64 zero = tcg_const_i64(0); \
+        TCGv_i64 not_RttV = tcg_temp_new_i64(); \
+        tcg_gen_extu_i32_i64(carry, PxV); \
+        tcg_gen_andi_i64(carry, carry, 1); \
+        tcg_gen_not_i64(not_RttV, RttV); \
+        tcg_gen_add2_i64(RddV, carry, RssV, zero, carry, zero); \
+        tcg_gen_add2_i64(RddV, carry, RddV, carry, not_RttV, zero); \
+        tcg_gen_extrl_i64_i32(PxV, carry); \
+        gen_8bitsof(PxV, PxV); \
+        tcg_temp_free_i64(carry); \
+        tcg_temp_free_i64(zero); \
+        tcg_temp_free_i64(not_RttV); \
+    } while (0)
+
+/*
  * Compare each of the 8 unsigned bytes
  * The minimum is placed in each byte of the destination.
  * Each bit of the predicate is set true if the bit from the first operand
diff --git a/target/hexagon/genptr.c b/target/hexagon/genptr.c
index 9dbebc6..333f7d7 100644
--- a/target/hexagon/genptr.c
+++ b/target/hexagon/genptr.c
@@ -361,5 +361,16 @@ static inline void gen_store_conditional8(CPUHexagonState *env,
     tcg_gen_movi_tl(hex_llsc_addr, ~0);
 }
 
+static TCGv gen_8bitsof(TCGv result, TCGv value)
+{
+    TCGv zero = tcg_const_tl(0);
+    TCGv ones = tcg_const_tl(0xff);
+    tcg_gen_movcond_tl(TCG_COND_NE, result, value, zero, ones, zero);
+    tcg_temp_free(zero);
+    tcg_temp_free(ones);
+
+    return result;
+}
+
 #include "tcg_funcs_generated.c.inc"
 #include "tcg_func_table_generated.c.inc"
diff --git a/target/hexagon/imported/alu.idef b/target/hexagon/imported/alu.idef
index f0c9bb4..58477ae 100644
--- a/target/hexagon/imported/alu.idef
+++ b/target/hexagon/imported/alu.idef
@@ -153,6 +153,21 @@ Q6INSN(A2_subp,"Rdd32=sub(Rtt32,Rss32)",ATTRIBS(),
 "Sub",
 { RddV=RttV-RssV;})
 
+/* 64-bit with carry */
+
+Q6INSN(A4_addp_c,"Rdd32=add(Rss32,Rtt32,Px4):carry",ATTRIBS(),"Add with Carry",
+{
+  RddV = RssV + RttV + fLSBOLD(PxV);
+  PxV = f8BITSOF(fCARRY_FROM_ADD(RssV,RttV,fLSBOLD(PxV)));
+})
+
+Q6INSN(A4_subp_c,"Rdd32=sub(Rss32,Rtt32,Px4):carry",ATTRIBS(),"Sub with Carry",
+{
+  RddV = RssV + ~RttV + fLSBOLD(PxV);
+  PxV = f8BITSOF(fCARRY_FROM_ADD(RssV,~RttV,fLSBOLD(PxV)));
+})
+
+
 /* NEG and ABS */
 
 Q6INSN(A2_negsat,"Rd32=neg(Rs32):sat",ATTRIBS(),
diff --git a/target/hexagon/imported/encode_pp.def b/target/hexagon/imported/encode_pp.def
index 4619398..514c240 100644
--- a/target/hexagon/imported/encode_pp.def
+++ b/target/hexagon/imported/encode_pp.def
@@ -1749,6 +1749,8 @@ SH_RRR_ENC(S4_extractp_rp,      "0001","11-","-","10-","ddddd")
 DEF_FIELDROW_DESC32(ICLASS_S3op" 0010 -------- PP------ --------","[#2] Rdd=(Rss,Rtt,Pu)")
 SH_RRR_ENC(S2_valignrb,         "0010","0--","-","-uu","ddddd")
 SH_RRR_ENC(S2_vsplicerb,        "0010","100","-","-uu","ddddd")
+SH_RRR_ENC(A4_addp_c,           "0010","110","-","-xx","ddddd")
+SH_RRR_ENC(A4_subp_c,           "0010","111","-","-xx","ddddd")
 
 
 DEF_FIELDROW_DESC32(ICLASS_S3op" 0011 -------- PP------ --------","[#3] Rdd=(Rss,Rt)")