[AArch64][SME]Update intrinsic interface for read/write (llvm#65594)

The new ACLE PR#225[1] now combines the slice parameters for some builtins. This patch is the #2 of 3 patches to update the interface. Slice specifies the ZA slice number directly and needs to be explicity implemented by the "user" with the base register plus the immediate offset [1]https://github.com/ARM-software/acle/pull/225/files
GPUOpen-Drivers · Sep 12, 2023 · dc8d2ec · dc8d2ec
1 parent 4d5906e
commit dc8d2ec
Show file tree

Hide file tree

Showing 5 changed files with 304 additions and 303 deletions.
diff --git a/clang/include/clang/Basic/arm_sme.td b/clang/include/clang/Basic/arm_sme.td
@@ -96,42 +96,42 @@ def SVSTR_ZA : MInst<"svstr_za", "vm%", "",
 
 multiclass ZARead<string n_suffix, string t, string i_prefix, list<ImmCheck> ch> {
   let TargetGuard = "sme" in {
-    def NAME # _H : SInst<"svread_hor_" # n_suffix # "[_{d}]", "ddPimi", t,
+    def NAME # _H : SInst<"svread_hor_" # n_suffix # "[_{d}]", "ddPim", t,
                           MergeOp1, i_prefix # "_horiz",
                           [IsReadZA, IsStreaming, IsSharedZA, IsPreservesZA], ch>;
 
-    def NAME # _V : SInst<"svread_ver_" # n_suffix # "[_{d}]", "ddPimi", t,
+    def NAME # _V : SInst<"svread_ver_" # n_suffix # "[_{d}]", "ddPim", t,
                           MergeOp1, i_prefix # "_vert",
                           [IsReadZA, IsStreaming, IsSharedZA, IsPreservesZA], ch>;
   }
 }
 
-defm SVREAD_ZA8 : ZARead<"za8", "cUc", "aarch64_sme_read", [ImmCheck<2, ImmCheck0_0>, ImmCheck<4, ImmCheck0_15>]>;
-defm SVREAD_ZA16 : ZARead<"za16", "sUshb", "aarch64_sme_read", [ImmCheck<2, ImmCheck0_1>, ImmCheck<4, ImmCheck0_7>]>;
-defm SVREAD_ZA32 : ZARead<"za32", "iUif", "aarch64_sme_read", [ImmCheck<2, ImmCheck0_3>, ImmCheck<4, ImmCheck0_3>]>;
-defm SVREAD_ZA64 : ZARead<"za64", "lUld", "aarch64_sme_read", [ImmCheck<2, ImmCheck0_7>, ImmCheck<4, ImmCheck0_1>]>;
-defm SVREAD_ZA128 : ZARead<"za128", "csilUcUsUiUlhbfd", "aarch64_sme_readq", [ImmCheck<2, ImmCheck0_15>, ImmCheck<4, ImmCheck0_0>]>;
+defm SVREAD_ZA8 : ZARead<"za8", "cUc", "aarch64_sme_read", [ImmCheck<2, ImmCheck0_0>]>;
+defm SVREAD_ZA16 : ZARead<"za16", "sUshb", "aarch64_sme_read", [ImmCheck<2, ImmCheck0_1>]>;
+defm SVREAD_ZA32 : ZARead<"za32", "iUif", "aarch64_sme_read", [ImmCheck<2, ImmCheck0_3>]>;
+defm SVREAD_ZA64 : ZARead<"za64", "lUld", "aarch64_sme_read", [ImmCheck<2, ImmCheck0_7>]>;
+defm SVREAD_ZA128 : ZARead<"za128", "csilUcUsUiUlhbfd", "aarch64_sme_readq", [ImmCheck<2, ImmCheck0_15>]>;
 
 ////////////////////////////////////////////////////////////////////////////////
 // Write horizontal/vertical ZA slices
 
 multiclass ZAWrite<string n_suffix, string t, string i_prefix, list<ImmCheck> ch> {
   let TargetGuard = "sme" in {
-    def NAME # _H : SInst<"svwrite_hor_" # n_suffix # "[_{d}]", "vimiPd", t,
+    def NAME # _H : SInst<"svwrite_hor_" # n_suffix # "[_{d}]", "vimPd", t,
                           MergeOp1, i_prefix # "_horiz",
                           [IsWriteZA, IsStreaming, IsSharedZA], ch>;
 
-    def NAME # _V : SInst<"svwrite_ver_" # n_suffix # "[_{d}]", "vimiPd", t,
+    def NAME # _V : SInst<"svwrite_ver_" # n_suffix # "[_{d}]", "vimPd", t,
                           MergeOp1, i_prefix # "_vert",
                           [IsWriteZA, IsStreaming, IsSharedZA], ch>;
   }
 }
 
-defm SVWRITE_ZA8 : ZAWrite<"za8", "cUc", "aarch64_sme_write", [ImmCheck<0, ImmCheck0_0>, ImmCheck<2, ImmCheck0_15>]>;
-defm SVWRITE_ZA16 : ZAWrite<"za16", "sUshb", "aarch64_sme_write", [ImmCheck<0, ImmCheck0_1>, ImmCheck<2, ImmCheck0_7>]>;
-defm SVWRITE_ZA32 : ZAWrite<"za32", "iUif", "aarch64_sme_write", [ImmCheck<0, ImmCheck0_3>, ImmCheck<2, ImmCheck0_3>]>;
-defm SVWRITE_ZA64 : ZAWrite<"za64", "lUld", "aarch64_sme_write", [ImmCheck<0, ImmCheck0_7>, ImmCheck<2, ImmCheck0_1>]>;
-defm SVWRITE_ZA128 : ZAWrite<"za128", "csilUcUsUiUlhbfd", "aarch64_sme_writeq", [ImmCheck<0, ImmCheck0_15>, ImmCheck<2, ImmCheck0_0>]>;
+defm SVWRITE_ZA8 : ZAWrite<"za8", "cUc", "aarch64_sme_write", [ImmCheck<0, ImmCheck0_0>]>;
+defm SVWRITE_ZA16 : ZAWrite<"za16", "sUshb", "aarch64_sme_write", [ImmCheck<0, ImmCheck0_1>]>;
+defm SVWRITE_ZA32 : ZAWrite<"za32", "iUif", "aarch64_sme_write", [ImmCheck<0, ImmCheck0_3>]>;
+defm SVWRITE_ZA64 : ZAWrite<"za64", "lUld", "aarch64_sme_write", [ImmCheck<0, ImmCheck0_7>]>;
+defm SVWRITE_ZA128 : ZAWrite<"za128", "csilUcUsUiUlhbfd", "aarch64_sme_writeq", [ImmCheck<0, ImmCheck0_15>]>;
 
 ////////////////////////////////////////////////////////////////////////////////
 // SME - Zero

diff --git a/clang/lib/CodeGen/CGBuiltin.cpp b/clang/lib/CodeGen/CGBuiltin.cpp
@@ -9617,15 +9617,10 @@ Value *CodeGenFunction::EmitSMEReadWrite(const SVETypeFlags &TypeFlags,
                                          unsigned IntID) {
   auto *VecTy = getSVEType(TypeFlags);
   Function *F = CGM.getIntrinsic(IntID, VecTy);
-  if (TypeFlags.isReadZA()) {
+  if (TypeFlags.isReadZA())
     Ops[1] = EmitSVEPredicateCast(Ops[1], VecTy);
-    Ops[3] = EmitTileslice(Ops[4], Ops[3]);
-    Ops.erase(&Ops[4]);
-  } else if (TypeFlags.isWriteZA()) {
-    Ops[1] = EmitTileslice(Ops[2], Ops[1]);
-    Ops[2] = EmitSVEPredicateCast(Ops[3], VecTy);
-    Ops.erase(&Ops[3]);
-  }
+  else if (TypeFlags.isWriteZA())
+    Ops[2] = EmitSVEPredicateCast(Ops[2], VecTy);
   return Builder.CreateCall(F, Ops);
 }