================
@@ -0,0 +1,42 @@
+; RUN: opt -S < %s | FileCheck %s
+; RUN: llvm-as %s -o - | llvm-dis | FileCheck %s
+
+target triple = "aarch64-linux"
+
+define void @ftmopa_za16_nxv16i8(<vscale x 16 x i8> %zn1, <vscale x 16 x i8>
%zn2, <vscale x 16 x i8> %zm, <vscale x 16 x i8> %zk) #0 {
+; CHECK-LABEL: @ftmopa_za16_nxv16i8
+; CHECK: call void @llvm.aarch64.sme.fp8.ftmopa.za16(i32 0, <vscale x 16 x i8>
%zn1, <vscale x 16 x i8> %zn2, <vscale x 16 x i8> %zm, <vscale x 16 x i8> %zk,
i32 0)
+ call void @llvm.aarch64.sme.ftmopa.za16.nxv16i8(i32 0, <vscale x 16 x i8>
%zn1, <vscale x 16 x i8> %zn2, <vscale x 16 x i8> %zm, <vscale x 16 x i8> %zk,
i32 0)
+ ret void
+}
+
+define void @ftmopa_za32_nxv16i8(<vscale x 16 x i8> %zn1, <vscale x 16 x i8>
%zn2, <vscale x 16 x i8> %zm, <vscale x 16 x i8> %zk) #0 {
+; CHECK-LABEL: @ftmopa_za32_nxv16i
+; CHECK: call void @llvm.aarch64.sme.fp8.ftmopa.za32(i32 0, <vscale x 16 x i8>
%zn1, <vscale x 16 x i8> %zn2, <vscale x 16 x i8> %zm, <vscale x 16 x i8> %zk,
i32 0)
+ call void @llvm.aarch64.sme.ftmopa.za32.nxv16i8(i32 0, <vscale x 16 x i8>
%zn1, <vscale x 16 x i8> %zn2, <vscale x 16 x i8> %zm, <vscale x 16 x i8> %zk,
i32 0)
+ ret void
+}
+
+
+define void @ftmopa_za16(<vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2,
<vscale x 16 x i8> %zm, <vscale x 16 x i8> %zk) #0 {
+; CHECK-LABEL: @ftmopa_za16
+; CHECK: call void @llvm.aarch64.sme.fp8.ftmopa.za16(i32 0, <vscale x 16 x i8>
%zn1, <vscale x 16 x i8> %zn2, <vscale x 16 x i8> %zm, <vscale x 16 x i8> %zk,
i32 0)
+ call void @llvm.aarch64.sme.ftmopa.za16(i32 0, <vscale x 16 x i8> %zn1,
<vscale x 16 x i8> %zn2, <vscale x 16 x i8> %zm, <vscale x 16 x i8> %zk, i32 0)
+ ret void
+}
+
+define void @ftmopa_za32(<vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2,
<vscale x 16 x i8> %zm, <vscale x 16 x i8> %zk) #0 {
+; CHECK-LABEL: @ftmopa_za32
+; CHECK: call void @llvm.aarch64.sme.fp8.ftmopa.za32(i32 0, <vscale x 16 x i8>
%zn1, <vscale x 16 x i8> %zn2, <vscale x 16 x i8> %zm, <vscale x 16 x i8> %zk,
i32 0)
+ call void @llvm.aarch64.sme.ftmopa.za32(i32 0, <vscale x 16 x i8> %zn1,
<vscale x 16 x i8> %zn2, <vscale x 16 x i8> %zm, <vscale x 16 x i8> %zk, i32 0)
+ ret void
+}
+
+define void @ftmopa_za32_wrong(<vscale x 16 x i8> %zn1, <vscale x 16 x i8>
%zn2, <vscale x 16 x i8> %zm, <vscale x 16 x i8> %zk) #0 {
----------------
Lukacma wrote:
Ah I see the difference. But this is wrong imo. We don't use
ftmopa.za32.nxv8i16 intrinsics from what I can see, so if it appears in
bytocode, the bytecode is invalid and should be treated as such. Or am I
missing smth ?
https://github.com/llvm/llvm-project/pull/203310
_______________________________________________
cfe-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
