R600 -> AMDGPU rename

[oota-llvm.git] / lib / Target / AMDGPU / VIInstructions.td

diff --git a/lib/Target/AMDGPU/VIInstructions.td b/lib/Target/AMDGPU/VIInstructions.td
new file mode 100644 (file)
index 0000000..5bf86e6
--- /dev/null
+++ b/lib/Target/AMDGPU/VIInstructions.td
@@ -0,0 +1,106 @@
+//===-- VIInstructions.td - VI Instruction Defintions ---------------------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+// Instruction definitions for VI and newer.
+//===----------------------------------------------------------------------===//
+
+let SIAssemblerPredicate = DisableInst, SubtargetPredicate = isVI in {
+
+//===----------------------------------------------------------------------===//
+// VOP1 Instructions
+//===----------------------------------------------------------------------===//
+
+defm V_CVT_F16_U16 : VOP1Inst <vop1<0, 0x39>, "v_cvt_f16_u16", VOP_F16_I16>;
+defm V_CVT_F16_I16 : VOP1Inst <vop1<0, 0x3a>, "v_cvt_f16_i16", VOP_F16_I16>;
+defm V_CVT_U16_F16 : VOP1Inst <vop1<0, 0x3b>, "v_cvt_u16_f16", VOP_I16_F16>;
+defm V_CVT_I16_F16 : VOP1Inst <vop1<0, 0x3c>, "v_cvt_i16_f16", VOP_I16_F16>;
+defm V_RCP_F16 : VOP1Inst <vop1<0, 0x3d>, "v_rcp_f16", VOP_F16_F16>;
+defm V_SQRT_F16 : VOP1Inst <vop1<0, 0x3e>, "v_sqrt_f16", VOP_F16_F16>;
+defm V_RSQ_F16 : VOP1Inst <vop1<0, 0x3f>, "v_rsq_f16", VOP_F16_F16>;
+defm V_LOG_F16 : VOP1Inst <vop1<0, 0x40>, "v_log_f16", VOP_F16_F16>;
+defm V_EXP_F16 : VOP1Inst <vop1<0, 0x41>, "v_exp_f16", VOP_F16_F16>;
+defm V_FREXP_MANT_F16 : VOP1Inst <vop1<0, 0x42>, "v_frexp_mant_f16",
+  VOP_F16_F16
+>;
+defm V_FREXP_EXP_I16_F16 : VOP1Inst <vop1<0, 0x43>, "v_frexp_exp_i16_f16",
+  VOP_I16_F16
+>;
+defm V_FLOOR_F16 : VOP1Inst <vop1<0, 0x44>, "v_floor_f16", VOP_F16_F16>;
+defm V_CEIL_F16 : VOP1Inst <vop1<0, 0x45>, "v_ceil_f16", VOP_F16_F16>;
+defm V_TRUNC_F16 : VOP1Inst <vop1<0, 0x46>, "v_trunc_f16", VOP_F16_F16>;
+defm V_RNDNE_F16 : VOP1Inst <vop1<0, 0x47>, "v_rndne_f16", VOP_F16_F16>;
+defm V_FRACT_F16 : VOP1Inst <vop1<0, 0x48>, "v_fract_f16", VOP_F16_F16>;
+defm V_SIN_F16 : VOP1Inst <vop1<0, 0x49>, "v_sin_f16", VOP_F16_F16>;
+defm V_COS_F16 : VOP1Inst <vop1<0, 0x4a>, "v_cos_f16", VOP_F16_F16>;
+
+//===----------------------------------------------------------------------===//
+// VOP2 Instructions
+//===----------------------------------------------------------------------===//
+
+let isCommutable = 1 in {
+
+defm V_ADD_F16 : VOP2Inst <vop2<0, 0x1f>, "v_add_f16", VOP_F16_F16_F16>;
+defm V_SUB_F16 : VOP2Inst <vop2<0, 0x20>, "v_sub_f16", VOP_F16_F16_F16>;
+defm V_SUBREV_F16 : VOP2Inst <vop2<0, 0x21>, "v_subrev_f16", VOP_F16_F16_F16,
+  null_frag, "v_sub_f16"
+>;
+defm V_MUL_F16 : VOP2Inst <vop2<0, 0x22>, "v_mul_f16", VOP_F16_F16_F16>;
+defm V_MAC_F16 : VOP2Inst <vop2<0, 0x23>, "v_mac_f16", VOP_F16_F16_F16>;
+} // End isCommutable = 1
+defm V_MADMK_F16 : VOP2MADK <vop2<0,0x24>, "v_madmk_f16">;
+let isCommutable = 1 in {
+defm V_MADAK_F16 : VOP2MADK <vop2<0,0x25>, "v_madak_f16">;
+defm V_ADD_U16 : VOP2Inst <vop2<0,0x26>, "v_add_u16", VOP_I16_I16_I16>;
+defm V_SUB_U16 : VOP2Inst <vop2<0,0x27>, "v_sub_u16" , VOP_I16_I16_I16>;
+defm V_SUBREV_U16 : VOP2Inst <vop2<0,0x28>, "v_subrev_u16", VOP_I16_I16_I16>;
+defm V_MUL_LO_U16 : VOP2Inst <vop2<0,0x29>, "v_mul_lo_u16", VOP_I16_I16_I16>;
+} // End isCommutable = 1
+defm V_LSHLREV_B16 : VOP2Inst <vop2<0,0x2a>, "v_lshlrev_b16", VOP_I16_I16_I16>;
+defm V_LSHRREV_B16 : VOP2Inst <vop2<0,0x2b>, "v_lshrrev_b16", VOP_I16_I16_I16>;
+defm V_ASHRREV_B16 : VOP2Inst <vop2<0,0x2c>, "v_ashrrev_b16", VOP_I16_I16_I16>;
+let isCommutable = 1 in {
+defm V_MAX_F16 : VOP2Inst <vop2<0,0x2d>, "v_max_f16", VOP_F16_F16_F16>;
+defm V_MIN_F16 : VOP2Inst <vop2<0,0x2e>, "v_min_f16", VOP_F16_F16_F16>;
+defm V_MAX_U16 : VOP2Inst <vop2<0,0x2f>, "v_max_u16", VOP_I16_I16_I16>;
+defm V_MAX_I16 : VOP2Inst <vop2<0,0x30>, "v_max_i16", VOP_I16_I16_I16>;
+defm V_MIN_U16 : VOP2Inst <vop2<0,0x31>, "v_min_u16", VOP_I16_I16_I16>;
+defm V_MIN_I16 : VOP2Inst <vop2<0,0x32>, "v_min_i16", VOP_I16_I16_I16>;
+} // End isCommutable = 1
+defm V_LDEXP_F16 : VOP2Inst <vop2<0,0x33>, "v_ldexp_f16", VOP_F16_F16_I16>;
+
+// Aliases to simplify matching of floating-pint instructions that are VOP2 on
+// SI and VOP3 on VI.
+
+class SI2_VI3Alias <string name, Instruction inst> : InstAlias <
+  name#" $dst, $src0, $src1",
+  (inst VGPR_32:$dst, 0, VCSrc_32:$src0, 0, VCSrc_32:$src1, 0, 0)
+>, PredicateControl {
+  let UseInstAsmMatchConverter = 0;
+}
+
+def : SI2_VI3Alias <"v_ldexp_f32", V_LDEXP_F32_e64_vi>;
+def : SI2_VI3Alias <"v_cvt_pkaccum_u8_f32", V_CVT_PKACCUM_U8_F32_e64_vi>;
+def : SI2_VI3Alias <"v_cvt_pknorm_i16_f32", V_CVT_PKNORM_I16_F32_e64_vi>;
+def : SI2_VI3Alias <"v_cvt_pknorm_u16_f32", V_CVT_PKNORM_U16_F32_e64_vi>;
+def : SI2_VI3Alias <"v_cvt_pkrtz_f16_f32", V_CVT_PKRTZ_F16_F32_e64_vi>;
+
+} // End SIAssemblerPredicate = DisableInst, SubtargetPredicate = isVI
+
+//===----------------------------------------------------------------------===//
+// SMEM Patterns
+//===----------------------------------------------------------------------===//
+
+let Predicates = [isVI] in {
+
+// 1. Offset as 20bit DWORD immediate
+def : Pat <
+  (SIload_constant v4i32:$sbase, IMM20bit:$offset),
+  (S_BUFFER_LOAD_DWORD_IMM $sbase, (as_i32imm $offset))
+>;
+
+} // End Predicates = [isVI]