aboutsummaryrefslogtreecommitdiff
path: root/llvm/lib/Target/AMDGPU/SIInstrInfo.td
diff options
context:
space:
mode:
Diffstat (limited to 'llvm/lib/Target/AMDGPU/SIInstrInfo.td')
-rw-r--r--llvm/lib/Target/AMDGPU/SIInstrInfo.td28
1 files changed, 28 insertions, 0 deletions
diff --git a/llvm/lib/Target/AMDGPU/SIInstrInfo.td b/llvm/lib/Target/AMDGPU/SIInstrInfo.td
index bd4995b..38b609c 100644
--- a/llvm/lib/Target/AMDGPU/SIInstrInfo.td
+++ b/llvm/lib/Target/AMDGPU/SIInstrInfo.td
@@ -1313,6 +1313,10 @@ def MatrixBFMT : CustomOperand<i32, 1, "MatrixBFMT">;
def MatrixAReuse : NamedBitOperand<"matrix_a_reuse">;
def MatrixBReuse : NamedBitOperand<"matrix_b_reuse">;
+def ScaleSel : NamedIntOperand<"scale_sel"> {
+ let Validator = "isUInt<3>";
+}
+
class KImmFPOperand<ValueType vt> : ImmOperand<vt> {
let OperandNamespace = "AMDGPU";
let OperandType = "OPERAND_KIMM"#vt.Size;
@@ -1662,6 +1666,8 @@ def VOP3OpSelMods : ComplexPattern<untyped, 2, "SelectVOP3OpSelMods">;
def VOP3PMadMixModsExt : ComplexPattern<untyped, 2, "SelectVOP3PMadMixModsExt">;
def VOP3PMadMixMods : ComplexPattern<untyped, 2, "SelectVOP3PMadMixMods">;
+def VOP3PMadMixBF16ModsExt : ComplexPattern<untyped, 2, "SelectVOP3PMadMixBF16ModsExt">;
+def VOP3PMadMixBF16Mods : ComplexPattern<untyped, 2, "SelectVOP3PMadMixBF16Mods">;
def VINTERPMods : ComplexPattern<untyped, 2, "SelectVINTERPMods">;
def VINTERPModsHi : ComplexPattern<untyped, 2, "SelectVINTERPModsHi">;
@@ -2857,15 +2863,18 @@ def VOP_I16_F16 : VOPProfile <[i16, f16, untyped, untyped]>;
def VOP_I16_I16 : VOPProfile <[i16, i16, untyped, untyped]>;
def VOP_BF16_BF16 : VOPProfile<[bf16, bf16, untyped, untyped]>;
def VOP1_I16_I32 : VOPProfile<[i16, i32, untyped, untyped]>;
+def VOP_I16_V2F16 : VOPProfile<[i16, v2f16, untyped, untyped]>;
def VOP_F16_F16_F16 : VOPProfile <[f16, f16, f16, untyped]>;
def VOP_F16_F16_I16 : VOPProfile <[f16, f16, i16, untyped]>;
def VOP_F16_F16_I32 : VOPProfile <[f16, f16, i32, untyped]>;
def VOP_I16_I16_I16 : VOPProfile <[i16, i16, i16, untyped]>;
def VOP_I16_I16_I16_ARITH : VOPProfile <[i16, i16, i16, untyped], /*EnableClamp=*/1>;
+def VOP_BF16_BF16_BF16 : VOPProfile <[bf16, bf16, bf16, untyped]>;
def VOP_I16_I16_I16_I16 : VOPProfile <[i16, i16, i16, i16, untyped]>;
def VOP_F16_F16_F16_F16 : VOPProfile <[f16, f16, f16, f16, untyped]>;
+def VOP_BF16_BF16_BF16_BF16 : VOPProfile <[bf16, bf16, bf16, bf16, untyped]>;
def VOP_I32_I16_I16_I32 : VOPProfile <[i32, i16, i16, i32, untyped]>;
def VOP_I32_I16 : VOPProfile <[i32, i16, untyped, untyped]>;
@@ -2873,10 +2882,12 @@ def VOP_I16_I32 : VOPProfile <[i16, i32, untyped, untyped]>;
def VOP_V2F16_V2F16_V2F16 : VOPProfile <[v2f16, v2f16, v2f16, untyped]>;
def VOP_V2I16_V2I16_V2I16 : VOPProfile <[v2i16, v2i16, v2i16, untyped]>;
+def VOP_V2BF16_V2BF16_V2BF16 : VOPProfile <[v2bf16, v2bf16, v2bf16, untyped]>;
def VOP_B32_F16_F16 : VOPProfile <[i32, f16, f16, untyped]>;
def VOP_V2F16_V2F16_V2F16_V2F16 : VOPProfile <[v2f16, v2f16, v2f16, v2f16]>;
def VOP_V2I16_V2I16_V2I16_V2I16 : VOPProfile <[v2i16, v2i16, v2i16, v2i16]>;
+def VOP_V2BF16_V2BF16_V2BF16_V2BF16 : VOPProfile <[v2bf16, v2bf16, v2bf16, v2bf16]>;
def VOP_V2I16_F32_F32 : VOPProfile <[v2i16, f32, f32, untyped]>;
def VOP_V2I16_I32_I32 : VOPProfile <[v2i16, i32, i32, untyped]>;
@@ -2912,12 +2923,16 @@ def VOP_I32_F32_I32 : VOPProfile <[i32, f32, i32, untyped]>;
def VOP_I32_I32_I32 : VOPProfile <[i32, i32, i32, untyped]>;
def VOP_I16_F32_F32 : VOPProfile <[i16, f32, f32, untyped]>;
def VOP_I32_I32_I32_ARITH : VOPProfile <[i32, i32, i32, untyped], /*EnableClamp=*/1>;
+def VOP_I64_I64_I64_ARITH : VOPProfile <[i64, i64, i64, untyped], /*EnableClamp=*/1>;
def VOP_V2F16_F32_F32 : VOPProfile <[v2f16, f32, f32, untyped]>;
def VOP_F32_F16_F16_F16 : VOPProfile <[f32, f16, f16, f16]>;
+def VOP_F32_BF16_BF16_BF16 : VOPProfile <[f32, bf16, bf16, bf16]>;
def VOP_V2BF16_F32_F32 : VOPProfile <[v2bf16, f32, f32, untyped]>;
def VOP_V32F32_V6I32_F32 : VOPProfile <[v32f32, v6i32, f32, untyped]>;
def VOP_V32F16_V6I32_F32 : VOPProfile <[v32f16, v6i32, f32, untyped]>;
def VOP_V32BF16_V6I32_F32 : VOPProfile <[v32bf16, v6i32, f32, untyped]>;
+def VOP_V2BF16_F32_F32_I32 : VOPProfile <[v2bf16, f32, f32, i32]>;
+def VOP_V2F16_F32_F32_I32 : VOPProfile <[v2f16, f32, f32, i32]>;
def VOP_V6I32_V32F16_F32 : VOPProfile<[v6i32, v32f16, f32, untyped]>;
def VOP_V6I32_V32BF16_F32 : VOPProfile<[v6i32, v32bf16, f32, untyped]>;
def VOP_V6I32_V16F32_V16F32_F32 : VOPProfile<[v6i32, v16f32, v16f32, f32]>;
@@ -2933,6 +2948,19 @@ def VOP_BF16_F32_I32 : VOPProfile<[bf16, f32, i32, untyped]>;
def VOP_F16_F32_I32 : VOPProfile<[f16, f32, i32, untyped]>;
def VOP_I32_BF16_I32_F32 : VOPProfile<[i32, bf16, i32, f32]>;
def VOP_I32_F16_I32_F32 : VOPProfile<[i32, f16, i32, f32]>;
+def VOP_V8F16_V2I32_I32 : VOPProfile<[v8f16, v2i32, i32, untyped]>;
+def VOP_V8BF16_V2I32_I32 : VOPProfile<[v8bf16, v2i32, i32, untyped]>;
+def VOP_V8F16_I32_I32 : VOPProfile<[v8f16, i32, i32, untyped]>;
+def VOP_V8BF16_I32_I32 : VOPProfile<[v8bf16, i32, i32, untyped]>;
+def VOP_V16F32_V3I32_I32 : VOPProfile<[v16f32, v3i32, i32, untyped]>;
+def VOP_V8F32_V2I32_I32 : VOPProfile<[v8f32, v2i32, i32, untyped]>;
+def VOP_V8F32_I32_I32 : VOPProfile<[v8f32, i32, i32, untyped]>;
+def VOP_V2I32_V8BF16_F32 : VOPProfile<[v2i32, v8bf16, f32, untyped]>;
+def VOP_V2I32_V8F16_F32 : VOPProfile<[v2i32, v8f16, f32, untyped]>;
+def VOP_V2I32_V8F32_F32 : VOPProfile<[v2i32, v8f32, f32, untyped]>;
+def VOP_I32_V8F32_F32 : VOPProfile<[i32, v8f32, f32, untyped]>;
+def VOP_I32_V8F16_F32 : VOPProfile<[i32, v8f16, f32, untyped]>;
+def VOP_I32_V8BF16_F32 : VOPProfile<[i32, v8bf16, f32, untyped]>;
def VOP_I32_F32_I32_F32 : VOPProfile<[i32, f32, i32, f32]>;
def VOP_V6I32_V32BF16_I32_F32 : VOPProfile<[v6i32, v32bf16, i32, f32]>;