1
0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-11-26 12:43:36 +01:00
llvm-mirror/test/CodeGen/AMDGPU/inline-constraints.ll
Daniil Fukalov 1d375d9d5b [AMDGPU] Inline asm - added i16, half and i128 types support
AMDGPU inline assembler support i16, half and i128 typed variables in constraints, but they were reported as error.
Needed to fix https://github.com/RadeonOpenCompute/ROCm/issues/341,
e.g. to be able to load with global_load_dwordx4 to a 128bit integer variable

Differential Revision: https://reviews.llvm.org/D44920

llvm-svn: 334301
2018-06-08 16:29:04 +00:00

78 lines
3.6 KiB
LLVM

; RUN: llc < %s -march=amdgcn -mcpu=bonaire -verify-machineinstrs | FileCheck --check-prefix=GCN %s
; RUN: llc < %s -march=amdgcn -mcpu=tonga -verify-machineinstrs | FileCheck --check-prefix=GCN %s
; GCN-LABEL: {{^}}inline_reg_constraints:
; GCN: flat_load_dword v{{[0-9]+}}, v[{{[0-9]+:[0-9]+}}]
; GCN: flat_load_dwordx2 v[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}]
; GCN: flat_load_dwordx2 v[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}]
; GCN: flat_load_dwordx4 v[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}]
; GCN: flat_load_dwordx4 v[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}]
; GCN: s_load_dword s{{[0-9]+}}, s[{{[0-9]+:[0-9]+}}]
; GCN: s_load_dwordx2 s[{{[0-9]+:[0-9]+}}], s[{{[0-9]+:[0-9]+}}]
; GCN: s_load_dwordx2 s[{{[0-9]+:[0-9]+}}], s[{{[0-9]+:[0-9]+}}]
; GCN: s_load_dwordx4 s[{{[0-9]+:[0-9]+}}], s[{{[0-9]+:[0-9]+}}]
; GCN: s_load_dwordx4 s[{{[0-9]+:[0-9]+}}], s[{{[0-9]+:[0-9]+}}]
; GCN: s_load_dwordx8 s[{{[0-9]+:[0-9]+}}], s[{{[0-9]+:[0-9]+}}]
define amdgpu_kernel void @inline_reg_constraints(i32 addrspace(1)* %ptr) {
entry:
%v32 = tail call i32 asm sideeffect "flat_load_dword $0, $1", "=v,v"(i32 addrspace(1)* %ptr)
%v2_32 = tail call <2 x i32> asm sideeffect "flat_load_dwordx2 $0, $1", "=v,v"(i32 addrspace(1)* %ptr)
%v64 = tail call i64 asm sideeffect "flat_load_dwordx2 $0, $1", "=v,v"(i32 addrspace(1)* %ptr)
%v4_32 = tail call <4 x i32> asm sideeffect "flat_load_dwordx4 $0, $1", "=v,v"(i32 addrspace(1)* %ptr)
%v128 = tail call i128 asm sideeffect "flat_load_dwordx4 $0, $1", "=v,v"(i32 addrspace(1)* %ptr)
%s32 = tail call i32 asm sideeffect "s_load_dword $0, $1", "=s,s"(i32 addrspace(1)* %ptr)
%s32_2 = tail call <2 x i32> asm sideeffect "s_load_dwordx2 $0, $1", "=s,s"(i32 addrspace(1)* %ptr)
%s64 = tail call i64 asm sideeffect "s_load_dwordx2 $0, $1", "=s,s"(i32 addrspace(1)* %ptr)
%s4_32 = tail call <4 x i32> asm sideeffect "s_load_dwordx4 $0, $1", "=s,s"(i32 addrspace(1)* %ptr)
%s128 = tail call i128 asm sideeffect "s_load_dwordx4 $0, $1", "=s,s"(i32 addrspace(1)* %ptr)
%s256 = tail call <8 x i32> asm sideeffect "s_load_dwordx8 $0, $1", "=s,s"(i32 addrspace(1)* %ptr)
ret void
}
; FIXME: Should be able to avoid copy
; GCN-LABEL: {{^}}inline_sreg_constraint_m0:
; GCN: s_mov_b32 m0, -1
; GCN: s_mov_b32 [[COPY_M0:s[0-9]+]], m0
; GCN: ; use [[COPY_M0]]
define amdgpu_kernel void @inline_sreg_constraint_m0() {
%m0 = tail call i32 asm sideeffect "s_mov_b32 m0, -1", "={M0}"()
tail call void asm sideeffect "; use $0", "s"(i32 %m0)
ret void
}
; GCN-LABEL: {{^}}inline_sreg_constraint_imm_i32:
; GCN: s_mov_b32 [[REG:s[0-9]+]], 32
; GCN: ; use [[REG]]
define amdgpu_kernel void @inline_sreg_constraint_imm_i32() {
tail call void asm sideeffect "; use $0", "s"(i32 32)
ret void
}
; GCN-LABEL: {{^}}inline_sreg_constraint_imm_f32:
; GCN: s_mov_b32 [[REG:s[0-9]+]], 1.0
; GCN: ; use [[REG]]
define amdgpu_kernel void @inline_sreg_constraint_imm_f32() {
tail call void asm sideeffect "; use $0", "s"(float 1.0)
ret void
}
; FIXME: Should be able to use s_mov_b64
; GCN-LABEL: {{^}}inline_sreg_constraint_imm_i64:
; GCN-DAG: s_mov_b32 s[[REG_LO:[0-9]+]], -4{{$}}
; GCN-DAG: s_mov_b32 s[[REG_HI:[0-9]+]], -1{{$}}
; GCN: ; use s{{\[}}[[REG_LO]]:[[REG_HI]]{{\]}}
define amdgpu_kernel void @inline_sreg_constraint_imm_i64() {
tail call void asm sideeffect "; use $0", "s"(i64 -4)
ret void
}
; GCN-LABEL: {{^}}inline_sreg_constraint_imm_f64:
; GCN-DAG: s_mov_b32 s[[REG_LO:[0-9]+]], 0{{$}}
; GCN-DAG: s_mov_b32 s[[REG_HI:[0-9]+]], 0x3ff00000{{$}}
; GCN: ; use s{{\[}}[[REG_LO]]:[[REG_HI]]{{\]}}
define amdgpu_kernel void @inline_sreg_constraint_imm_f64() {
tail call void asm sideeffect "; use $0", "s"(double 1.0)
ret void
}