Hexagon (target/hexagon) Add v69 HVX instructions

The following instructions are added
    V6_vasrvuhubrndsat
    V6_vasrvuhubsat
    V6_vasrvwuhrndsat
    V6_vasrvwuhsat
    V6_vassign_tmp
    V6_vcombine_tmp
    V6_vmpyuhvs

Signed-off-by: Taylor Simpson <tsimpson@quicinc.com>
Reviewed-by: Anton Johansson <anjo@rev.ng>
Message-Id: <20230427224057.3766963-7-tsimpson@quicinc.com>

4 files changed, 68 insertions, 0 deletions

diff --git a/target/hexagon/attribs_def.h.inc b/target/hexagon/attribs_def.h.inc
index 0ddfb45..3bef60b 100644
--- a/target/hexagon/attribs_def.h.inc
+++ b/target/hexagon/attribs_def.h.inc
@@ -69,11 +69,13 @@ DEF_ATTRIB(CVI_VP_VS, "Double vector permute/shft insn executes on HVX", "", "")
 DEF_ATTRIB(CVI_VX, "Multiply instruction executes on HVX", "", "")
 DEF_ATTRIB(CVI_VX_DV, "Double vector multiply insn executes on HVX", "", "")
 DEF_ATTRIB(CVI_VS, "Shift instruction executes on HVX", "", "")
+DEF_ATTRIB(CVI_VS_3SRC, "This shift needs to borrow a source register", "", "")
 DEF_ATTRIB(CVI_VS_VX, "Permute/shift and multiply insn executes on HVX", "", "")
 DEF_ATTRIB(CVI_VA, "ALU instruction executes on HVX", "", "")
 DEF_ATTRIB(CVI_VA_DV, "Double vector alu instruction executes on HVX", "", "")
 DEF_ATTRIB(CVI_4SLOT, "Consumes all the vector execution resources", "", "")
 DEF_ATTRIB(CVI_TMP, "Transient Memory Load not written to register", "", "")
+DEF_ATTRIB(CVI_REMAP, "Register Renaming not written to register file", "", "")
 DEF_ATTRIB(CVI_GATHER, "CVI Gather operation", "", "")
 DEF_ATTRIB(CVI_SCATTER, "CVI Scatter operation", "", "")
 DEF_ATTRIB(CVI_SCATTER_RELEASE, "CVI Store Release for scatter", "", "")
@@ -147,6 +149,8 @@ DEF_ATTRIB(L2FETCH, "Instruction is l2fetch type", "", "")
 DEF_ATTRIB(ICINVA, "icinva", "", "")
 DEF_ATTRIB(DCCLEANINVA, "dccleaninva", "", "")
 
+DEF_ATTRIB(NO_INTRINSIC, "Don't generate an intrisic", "", "")
+
 /* Documentation Notes */
 DEF_ATTRIB(NOTE_CONDITIONAL, "can be conditionally executed", "", "")
 DEF_ATTRIB(NOTE_NEWVAL_SLOT0, "New-value oprnd must execute on slot 0", "", "")
@@ -155,7 +159,11 @@ DEF_ATTRIB(NOTE_NOPACKET, "solo instruction", "", "")
 DEF_ATTRIB(NOTE_AXOK, "May only be grouped with ALU32 or non-FP XTYPE.", "", "")
 DEF_ATTRIB(NOTE_LATEPRED, "The predicate can not be used as a .new", "", "")
 DEF_ATTRIB(NOTE_NVSLOT0, "Can execute only in slot 0 (ST)", "", "")
+DEF_ATTRIB(NOTE_NOVP, "Cannot be paired with a HVX permute instruction", "", "")
+DEF_ATTRIB(NOTE_VA_UNARY, "Combined with HVX ALU op (must be unary)", "", "")
 
+/* V6 MMVector Notes for Documentation */
+DEF_ATTRIB(NOTE_SHIFT_RESOURCE, "Uses the HVX shift resource.", "", "")
 /* Restrictions to make note of */
 DEF_ATTRIB(RESTRICT_NOSLOT1_STORE, "Packet must not have slot 1 store", "", "")
 DEF_ATTRIB(RESTRICT_LATEPRED, "Predicate can not be used as a .new.", "", "")
diff --git a/target/hexagon/gen_tcg_hvx.h b/target/hexagon/gen_tcg_hvx.h
index d4aefe8..8dceead 100644
--- a/target/hexagon/gen_tcg_hvx.h
+++ b/target/hexagon/gen_tcg_hvx.h
@@ -128,6 +128,18 @@ static inline void assert_vhist_tmp(DisasContext *ctx)
     tcg_gen_gvec_mov(MO_64, VdV_off, VuV_off, \
                      sizeof(MMVector), sizeof(MMVector))
 
+#define fGEN_TCG_V6_vassign_tmp(SHORTCODE) \
+    tcg_gen_gvec_mov(MO_64, VdV_off, VuV_off, \
+                     sizeof(MMVector), sizeof(MMVector))
+
+#define fGEN_TCG_V6_vcombine_tmp(SHORTCODE) \
+    do { \
+        tcg_gen_gvec_mov(MO_64, VddV_off, VvV_off, \
+                         sizeof(MMVector), sizeof(MMVector)); \
+        tcg_gen_gvec_mov(MO_64, VddV_off + sizeof(MMVector), VuV_off, \
+                         sizeof(MMVector), sizeof(MMVector)); \
+    } while (0)
+
 /* Vector conditional move */
 #define fGEN_TCG_VEC_CMOV(PRED) \
     do { \
diff --git a/target/hexagon/imported/mmvec/encode_ext.def b/target/hexagon/imported/mmvec/encode_ext.def
index b9b62fe..402438f 100644
--- a/target/hexagon/imported/mmvec/encode_ext.def
+++ b/target/hexagon/imported/mmvec/encode_ext.def
@@ -257,6 +257,11 @@ DEF_ENC(V6_vasruhubrndsat,         ICLASS_CJ" 1 000 vvv vvttt PP 0 uuuuu 111 ddd
 DEF_ENC(V6_vasruwuhsat,         ICLASS_CJ" 1 000 vvv vvttt PP 1 uuuuu 100 ddddd") //
 DEF_ENC(V6_vasruhubsat,            ICLASS_CJ" 1 000 vvv vvttt PP 1 uuuuu 101 ddddd") //
 
+DEF_ENC(V6_vasrvuhubrndsat,"00011101000vvvvvPP0uuuuu011ddddd")
+DEF_ENC(V6_vasrvuhubsat,"00011101000vvvvvPP0uuuuu010ddddd")
+DEF_ENC(V6_vasrvwuhrndsat,"00011101000vvvvvPP0uuuuu001ddddd")
+DEF_ENC(V6_vasrvwuhsat,"00011101000vvvvvPP0uuuuu000ddddd")
+
 /***************************************************************
 *
 *  Group #1, Uses Q6 Rt32
@@ -716,6 +721,7 @@ DEF_ENC(V6_vaddclbw,    ICLASS_CJ" 1 111 000 vvvvv PP 1 uuuuu 001 ddddd") //
 
 DEF_ENC(V6_vavguw,        ICLASS_CJ" 1 111 000 vvvvv PP 1 uuuuu 010 ddddd") //
 DEF_ENC(V6_vavguwrnd,    ICLASS_CJ" 1 111 000 vvvvv PP 1 uuuuu 011 ddddd") //
+DEF_ENC(V6_vassign_tmp,"00011110--0---01PP0uuuuu110ddddd")
 DEF_ENC(V6_vavgb,        ICLASS_CJ" 1 111 000 vvvvv PP 1 uuuuu 100 ddddd") //
 DEF_ENC(V6_vavgbrnd,    ICLASS_CJ" 1 111 000 vvvvv PP 1 uuuuu 101 ddddd") //
 DEF_ENC(V6_vnavgb,        ICLASS_CJ" 1 111 000 vvvvv PP 1 uuuuu 110 ddddd") //
@@ -741,6 +747,7 @@ DEF_ENC(V6_vshufoh,      ICLASS_CJ" 1 111 010 vvvvv PP 0 uuuuu 100 ddddd") //
 DEF_ENC(V6_vshufoeh,      ICLASS_CJ" 1 111 010 vvvvv PP 0 uuuuu 101 ddddd") //
 DEF_ENC(V6_vshufoeb,      ICLASS_CJ" 1 111 010 vvvvv PP 0 uuuuu 110 ddddd") //
 DEF_ENC(V6_vcombine,     ICLASS_CJ" 1 111 010 vvvvv PP 0 uuuuu 111 ddddd") //
+DEF_ENC(V6_vcombine_tmp,"00011110101vvvvvPP0uuuuu111ddddd")
 
 DEF_ENC(V6_v6mpyvubs10,  ICLASS_CJ" 1 111 010 vvvvv PP 1 uuuuu 0ii ddddd")
 DEF_ENC(V6_v6mpyhubs10,  ICLASS_CJ" 1 111 010 vvvvv PP 1 uuuuu 1ii ddddd")
@@ -795,6 +802,7 @@ DEF_ENC(V6_vrounduhub,     ICLASS_CJ" 1 111 111 vvvvv PP 0 uuuuu 011 ddddd") //
 DEF_ENC(V6_vrounduwuh,     ICLASS_CJ" 1 111 111 vvvvv PP 0 uuuuu 100 ddddd") //
 DEF_ENC(V6_vmpyewuh,    ICLASS_CJ" 1 111 111 vvvvv PP 0 uuuuu 101 ddddd")
 DEF_ENC(V6_vmpyowh,        ICLASS_CJ" 1 111 111 vvvvv PP 0 uuuuu 111 ddddd")
+DEF_ENC(V6_vmpyuhvs,"00011111110vvvvvPP1uuuuu111ddddd")
 
 
 #endif /* NO MMVEC */
diff --git a/target/hexagon/imported/mmvec/ext.idef b/target/hexagon/imported/mmvec/ext.idef
index c0d169f..ead32c2 100644
--- a/target/hexagon/imported/mmvec/ext.idef
+++ b/target/hexagon/imported/mmvec/ext.idef
@@ -62,6 +62,9 @@ EXTINSN(V6_##TAG, SYNTAX, ATTRIBS(A_EXTENSION,A_CVI,A_CVI_VS),  \
 DESCR, DO_FOR_EACH_CODE(WIDTH, CODE))
 
 
+#define ITERATOR_INSN_SHIFT3_SLOT(WIDTH,TAG,SYNTAX,DESCR,CODE) \
+EXTINSN(V6_##TAG, SYNTAX, ATTRIBS(A_EXTENSION,A_CVI,A_CVI_VS,A_CVI_VS_3SRC,A_NOTE_SHIFT_RESOURCE,A_NOTE_NOVP,A_NOTE_VA_UNARY),  \
+DESCR, DO_FOR_EACH_CODE(WIDTH, CODE))
 
 #define ITERATOR_INSN_SHIFT_SLOT_VV_LATE(WIDTH,TAG,SYNTAX,DESCR,CODE) \
 EXTINSN(V6_##TAG, SYNTAX, ATTRIBS(A_EXTENSION,A_CVI,A_CVI_VS),  \
@@ -980,6 +983,22 @@ NARROWING_SHIFT(16,vasrhubrndsat,fSETBYTE,ub,h,:rnd:sat,fVSATUB,fVROUND,0x7)
 NARROWING_SHIFT(16,vasrhbsat,fSETBYTE,b,h,:sat,fVSATB,fVNOROUND,0x7)
 NARROWING_SHIFT(16,vasrhbrndsat,fSETBYTE,b,h,:rnd:sat,fVSATB,fVROUND,0x7)
 
+#define NARROWING_VECTOR_SHIFT(ITERSIZE,TAG,DSTM,DSTTYPE,SRCTYPE,SRCTYPE2,SYNOPTS,SATFUNC,RNDFUNC,SHAMTMASK) \
+ITERATOR_INSN_SHIFT3_SLOT(ITERSIZE,TAG, \
+"Vd32." #DSTTYPE "=vasr(Vuu32." #SRCTYPE ",Vv32." #SRCTYPE2 ")" #SYNOPTS, \
+"Vector shift by vector right and shuffle", \
+    fHIDE(int )shamt = VvV.SRCTYPE2[2*i+0] & SHAMTMASK; \
+    DSTM(0,VdV.SRCTYPE[i],SATFUNC(RNDFUNC(VuuV.v[0].SRCTYPE[i],shamt) >> shamt)); \
+    shamt = VvV.SRCTYPE2[2*i+1] & SHAMTMASK; \
+    DSTM(1,VdV.SRCTYPE[i],SATFUNC(RNDFUNC(VuuV.v[1].SRCTYPE[i],shamt) >> shamt)))
+
+/* WORD TO HALF*/
+NARROWING_VECTOR_SHIFT(32,vasrvwuhsat,fSETHALF,uh,w,uh,:sat,fVSATUH,fVNOROUND,0xF)
+NARROWING_VECTOR_SHIFT(32,vasrvwuhrndsat,fSETHALF,uh,w,uh,:rnd:sat,fVSATUH,fVROUND,0xF)
+/* HALF TO BYTE*/
+NARROWING_VECTOR_SHIFT(16,vasrvuhubsat,fSETBYTE,ub,uh,ub,:sat,fVSATUB,fVNOROUND,0x7)
+NARROWING_VECTOR_SHIFT(16,vasrvuhubrndsat,fSETBYTE,ub,uh,ub,:rnd:sat,fVSATUB,fVROUND,0x7)
+
 NARROWING_SHIFT_NOV1(16,vasruhubsat,fSETBYTE,ub,uh,:sat,fVSATUB,fVNOROUND,0x7)
 NARROWING_SHIFT_NOV1(16,vasruhubrndsat,fSETBYTE,ub,uh,:rnd:sat,fVSATUB,fVROUND,0x7)
 
@@ -1364,6 +1383,9 @@ ITERATOR_INSN2_MPY_SLOT_DOUBLE_VEC(16,vmpyhvsrs,"Vd32=vmpyh(Vu32,Vv32):<<1:rnd:s
 
 
 
+ITERATOR_INSN_MPY_SLOT(16,vmpyuhvs, "Vd32.uh=vmpy(Vu32.uh,Vv32.uh):>>16",
+"Vector by Vector Unsigned Halfword Multiply with 16 bit rightshift",
+    VdV.uh[i] = fGETUHALF(1,fMPY16UU(VuV.uh[i],VvV.uh[i])))
 
 
 ITERATOR_INSN2_MPY_SLOT_DOUBLE_VEC(32,vmpyhus, "Vdd32=vmpyhus(Vu32,Vv32)","Vdd32.w=vmpy(Vu32.h,Vv32.uh)",
@@ -2042,6 +2064,24 @@ ITERATOR_INSN_ANY_SLOT_DOUBLE_VEC(8,vcombine,"Vdd32=vcombine(Vu32,Vv32)",
 
 ///////////////////////////////////////////////////////////////////////////
 
+EXTINSN(V6_vcombine_tmp, "Vdd32.tmp=vcombine(Vu32,Vv32)",    ATTRIBS(A_EXTENSION,A_CVI,A_CVI_REMAP,A_CVI_TMP,A_NO_INTRINSIC),
+"Vector assign tmp, Any two to Vector Pair ",
+{
+   fHIDE(int i;)
+    fVFOREACH(8, i) {
+           VddV.v[0].ub[i] = VvV.ub[i];
+           VddV.v[1].ub[i] = VuV.ub[i];
+    }
+})
+
+EXTINSN(V6_vassign_tmp, "Vd32.tmp=Vu32",    ATTRIBS(A_EXTENSION,A_CVI,A_CVI_REMAP,A_CVI_TMP,A_NO_INTRINSIC),
+"Vector assign tmp, Any two to Vector Pair ",
+{
+   fHIDE(int i;)
+    fVFOREACH(32, i) {
+           VdV.w[i]=VuV.w[i];
+    }
+})
 
 /*********************************************************
 * GENERAL PERMUTE NETWORKS