1
0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-11-23 11:13:28 +01:00
llvm-mirror/lib/Target/AMDGPU/SIIntrinsics.td
Matt Arsenault ef8518e8b1 AMDGPU: Allow some control flow intrinsics to be CSEd
These clean up some unnecessary or instructions in
cases with complex loops.

In the original testcase I noticed this, the same
or with exec was repeated 5 or 6 times in a row. With
this only one is emitted or sometimes a copy.

llvm-svn: 281786
2016-09-16 22:11:18 +00:00

197 lines
7.6 KiB
TableGen

//===-- SIIntrinsics.td - SI Intrinsic defs ----------------*- tablegen -*-===//
//
// The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
//
// Backend internal SI Intrinsic Definitions. User code should not
// directly use these.
//
//===----------------------------------------------------------------------===//
let TargetPrefix = "SI", isTarget = 1 in {
def int_SI_packf16 : Intrinsic <[llvm_i32_ty], [llvm_float_ty, llvm_float_ty], [IntrNoMem]>;
def int_SI_export : Intrinsic <[], [llvm_i32_ty, llvm_i32_ty, llvm_i32_ty, llvm_i32_ty, llvm_i32_ty, llvm_float_ty, llvm_float_ty, llvm_float_ty, llvm_float_ty], []>;
def int_SI_load_const : Intrinsic <[llvm_float_ty], [llvm_anyint_ty, llvm_i32_ty], [IntrNoMem]>;
def int_SI_vs_load_input : Intrinsic <[llvm_v4f32_ty], [llvm_anyint_ty, llvm_i16_ty, llvm_i32_ty], [IntrNoMem]> ;
// Fully-flexible TBUFFER_STORE_FORMAT_* except for the ADDR64 bit, which is not exposed
def int_SI_tbuffer_store : Intrinsic <
[],
[llvm_anyint_ty, // rsrc(SGPR)
llvm_anyint_ty, // vdata(VGPR), overloaded for types i32, v2i32, v4i32
llvm_i32_ty, // num_channels(imm), selects opcode suffix: 1=X, 2=XY, 3=XYZ, 4=XYZW
llvm_i32_ty, // vaddr(VGPR)
llvm_i32_ty, // soffset(SGPR)
llvm_i32_ty, // inst_offset(imm)
llvm_i32_ty, // dfmt(imm)
llvm_i32_ty, // nfmt(imm)
llvm_i32_ty, // offen(imm)
llvm_i32_ty, // idxen(imm)
llvm_i32_ty, // glc(imm)
llvm_i32_ty, // slc(imm)
llvm_i32_ty], // tfe(imm)
[]>;
// Fully-flexible BUFFER_LOAD_DWORD_* except for the ADDR64 bit, which is not exposed
def int_SI_buffer_load_dword : Intrinsic <
[llvm_anyint_ty], // vdata(VGPR), overloaded for types i32, v2i32, v4i32
[llvm_anyint_ty, // rsrc(SGPR)
llvm_anyint_ty, // vaddr(VGPR)
llvm_i32_ty, // soffset(SGPR)
llvm_i32_ty, // inst_offset(imm)
llvm_i32_ty, // offen(imm)
llvm_i32_ty, // idxen(imm)
llvm_i32_ty, // glc(imm)
llvm_i32_ty, // slc(imm)
llvm_i32_ty], // tfe(imm)
[IntrReadMem, IntrArgMemOnly]>;
def int_SI_sendmsg : Intrinsic <[], [llvm_i32_ty, llvm_i32_ty], []>;
// Fully-flexible SAMPLE instruction.
class SampleRaw : Intrinsic <
[llvm_v4f32_ty], // vdata(VGPR)
[llvm_anyint_ty, // vaddr(VGPR)
llvm_v8i32_ty, // rsrc(SGPR)
llvm_v4i32_ty, // sampler(SGPR)
llvm_i32_ty, // dmask(imm)
llvm_i32_ty, // unorm(imm)
llvm_i32_ty, // r128(imm)
llvm_i32_ty, // da(imm)
llvm_i32_ty, // glc(imm)
llvm_i32_ty, // slc(imm)
llvm_i32_ty, // tfe(imm)
llvm_i32_ty], // lwe(imm)
[IntrNoMem]>;
// Image instruction without a sampler.
class Image : Intrinsic <
[llvm_v4f32_ty], // vdata(VGPR)
[llvm_anyint_ty, // vaddr(VGPR)
llvm_v8i32_ty, // rsrc(SGPR)
llvm_i32_ty, // dmask(imm)
llvm_i32_ty, // unorm(imm)
llvm_i32_ty, // r128(imm)
llvm_i32_ty, // da(imm)
llvm_i32_ty, // glc(imm)
llvm_i32_ty, // slc(imm)
llvm_i32_ty, // tfe(imm)
llvm_i32_ty], // lwe(imm)
[IntrNoMem]>;
// Basic sample
def int_SI_image_sample : SampleRaw;
def int_SI_image_sample_cl : SampleRaw;
def int_SI_image_sample_d : SampleRaw;
def int_SI_image_sample_d_cl : SampleRaw;
def int_SI_image_sample_l : SampleRaw;
def int_SI_image_sample_b : SampleRaw;
def int_SI_image_sample_b_cl : SampleRaw;
def int_SI_image_sample_lz : SampleRaw;
def int_SI_image_sample_cd : SampleRaw;
def int_SI_image_sample_cd_cl : SampleRaw;
// Sample with comparison
def int_SI_image_sample_c : SampleRaw;
def int_SI_image_sample_c_cl : SampleRaw;
def int_SI_image_sample_c_d : SampleRaw;
def int_SI_image_sample_c_d_cl : SampleRaw;
def int_SI_image_sample_c_l : SampleRaw;
def int_SI_image_sample_c_b : SampleRaw;
def int_SI_image_sample_c_b_cl : SampleRaw;
def int_SI_image_sample_c_lz : SampleRaw;
def int_SI_image_sample_c_cd : SampleRaw;
def int_SI_image_sample_c_cd_cl : SampleRaw;
// Sample with offsets
def int_SI_image_sample_o : SampleRaw;
def int_SI_image_sample_cl_o : SampleRaw;
def int_SI_image_sample_d_o : SampleRaw;
def int_SI_image_sample_d_cl_o : SampleRaw;
def int_SI_image_sample_l_o : SampleRaw;
def int_SI_image_sample_b_o : SampleRaw;
def int_SI_image_sample_b_cl_o : SampleRaw;
def int_SI_image_sample_lz_o : SampleRaw;
def int_SI_image_sample_cd_o : SampleRaw;
def int_SI_image_sample_cd_cl_o : SampleRaw;
// Sample with comparison and offsets
def int_SI_image_sample_c_o : SampleRaw;
def int_SI_image_sample_c_cl_o : SampleRaw;
def int_SI_image_sample_c_d_o : SampleRaw;
def int_SI_image_sample_c_d_cl_o : SampleRaw;
def int_SI_image_sample_c_l_o : SampleRaw;
def int_SI_image_sample_c_b_o : SampleRaw;
def int_SI_image_sample_c_b_cl_o : SampleRaw;
def int_SI_image_sample_c_lz_o : SampleRaw;
def int_SI_image_sample_c_cd_o : SampleRaw;
def int_SI_image_sample_c_cd_cl_o : SampleRaw;
// Basic gather4
def int_SI_gather4 : SampleRaw;
def int_SI_gather4_cl : SampleRaw;
def int_SI_gather4_l : SampleRaw;
def int_SI_gather4_b : SampleRaw;
def int_SI_gather4_b_cl : SampleRaw;
def int_SI_gather4_lz : SampleRaw;
// Gather4 with comparison
def int_SI_gather4_c : SampleRaw;
def int_SI_gather4_c_cl : SampleRaw;
def int_SI_gather4_c_l : SampleRaw;
def int_SI_gather4_c_b : SampleRaw;
def int_SI_gather4_c_b_cl : SampleRaw;
def int_SI_gather4_c_lz : SampleRaw;
// Gather4 with offsets
def int_SI_gather4_o : SampleRaw;
def int_SI_gather4_cl_o : SampleRaw;
def int_SI_gather4_l_o : SampleRaw;
def int_SI_gather4_b_o : SampleRaw;
def int_SI_gather4_b_cl_o : SampleRaw;
def int_SI_gather4_lz_o : SampleRaw;
// Gather4 with comparison and offsets
def int_SI_gather4_c_o : SampleRaw;
def int_SI_gather4_c_cl_o : SampleRaw;
def int_SI_gather4_c_l_o : SampleRaw;
def int_SI_gather4_c_b_o : SampleRaw;
def int_SI_gather4_c_b_cl_o : SampleRaw;
def int_SI_gather4_c_lz_o : SampleRaw;
def int_SI_getlod : SampleRaw;
// Image instrinsics.
def int_SI_image_load : Image;
def int_SI_image_load_mip : Image;
def int_SI_getresinfo : Image;
/* Interpolation Intrinsics */
def int_SI_fs_constant : Intrinsic <[llvm_float_ty], [llvm_i32_ty, llvm_i32_ty, llvm_i32_ty], [IntrNoMem]>;
def int_SI_fs_interp : Intrinsic <[llvm_float_ty], [llvm_i32_ty, llvm_i32_ty, llvm_i32_ty, llvm_v2i32_ty], [IntrNoMem]>;
} // End TargetPrefix = "SI", isTarget = 1
let TargetPrefix = "amdgcn", isTarget = 1 in {
// Emit 2.5 ulp, no denormal division. Should only be inserted by
// pass based on !fpmath metadata.
def int_amdgcn_fdiv_fast : Intrinsic<
[llvm_float_ty], [llvm_float_ty], [IntrNoMem]
>;
/* Control flow Intrinsics */
def int_amdgcn_if : Intrinsic<[llvm_i64_ty], [llvm_i1_ty, llvm_empty_ty], [IntrConvergent]>;
def int_amdgcn_else : Intrinsic<[llvm_i64_ty], [llvm_i64_ty, llvm_empty_ty], [IntrConvergent]>;
def int_amdgcn_break : Intrinsic<[llvm_i64_ty], [llvm_i64_ty], [IntrNoMem, IntrConvergent]>;
def int_amdgcn_if_break : Intrinsic<[llvm_i64_ty], [llvm_i1_ty, llvm_i64_ty], [IntrNoMem, IntrConvergent]>;
def int_amdgcn_else_break : Intrinsic<[llvm_i64_ty], [llvm_i64_ty, llvm_i64_ty], [IntrNoMem, IntrConvergent]>;
def int_amdgcn_loop : Intrinsic<[], [llvm_i64_ty, llvm_empty_ty], [IntrConvergent]>;
def int_amdgcn_end_cf : Intrinsic<[], [llvm_i64_ty], [IntrConvergent]>;
}