forked from OSchip/llvm-project
2220 lines
90 KiB
LLVM
2220 lines
90 KiB
LLVM
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
|
|
; RUN: llc -mtriple=amdgcn -mcpu=gfx700 -verify-machineinstrs < %s | FileCheck -check-prefixes=GFX7 %s
|
|
; RUN: llc -mtriple=amdgcn -mcpu=gfx803 -verify-machineinstrs < %s | FileCheck -check-prefixes=GFX8 %s
|
|
; RUN: llc -mtriple=amdgcn -mcpu=gfx900 -verify-machineinstrs < %s | FileCheck -check-prefixes=GFX9-NODL %s
|
|
; RUN: llc -mtriple=amdgcn -mcpu=gfx906 -verify-machineinstrs < %s | FileCheck -check-prefixes=GFX9-DL %s
|
|
; RUN: llc -mtriple=amdgcn -mcpu=gfx1011 -verify-machineinstrs < %s | FileCheck -check-prefixes=GFX10-DL %s
|
|
; RUN: llc -mtriple=amdgcn -mcpu=gfx1012 -verify-machineinstrs < %s | FileCheck -check-prefixes=GFX10-DL %s
|
|
|
|
define amdgpu_kernel void @udot4_acc32(<4 x i8> addrspace(1)* %src1,
|
|
; GFX7-LABEL: udot4_acc32:
|
|
; GFX7: ; %bb.0: ; %entry
|
|
; GFX7-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x9
|
|
; GFX7-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0xd
|
|
; GFX7-NEXT: s_movk_i32 s8, 0xff
|
|
; GFX7-NEXT: s_mov_b32 s3, 0xf000
|
|
; GFX7-NEXT: s_mov_b32 s2, -1
|
|
; GFX7-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX7-NEXT: s_load_dword s4, s[4:5], 0x0
|
|
; GFX7-NEXT: s_load_dword s5, s[6:7], 0x0
|
|
; GFX7-NEXT: s_load_dword s12, s[0:1], 0x0
|
|
; GFX7-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX7-NEXT: s_and_b32 s6, s4, s8
|
|
; GFX7-NEXT: s_and_b32 s7, s5, s8
|
|
; GFX7-NEXT: s_bfe_u32 s9, s5, 0x80008
|
|
; GFX7-NEXT: v_mov_b32_e32 v0, s7
|
|
; GFX7-NEXT: v_mov_b32_e32 v1, s12
|
|
; GFX7-NEXT: s_bfe_u32 s11, s5, 0x80010
|
|
; GFX7-NEXT: v_mad_u32_u24 v0, s6, v0, v1
|
|
; GFX7-NEXT: s_bfe_u32 s8, s4, 0x80008
|
|
; GFX7-NEXT: v_mov_b32_e32 v1, s9
|
|
; GFX7-NEXT: s_bfe_u32 s10, s4, 0x80010
|
|
; GFX7-NEXT: v_mad_u32_u24 v0, s8, v1, v0
|
|
; GFX7-NEXT: v_mov_b32_e32 v1, s11
|
|
; GFX7-NEXT: s_lshr_b32 s5, s5, 24
|
|
; GFX7-NEXT: v_mad_u32_u24 v0, s10, v1, v0
|
|
; GFX7-NEXT: s_lshr_b32 s4, s4, 24
|
|
; GFX7-NEXT: v_mov_b32_e32 v1, s5
|
|
; GFX7-NEXT: v_mad_u32_u24 v0, s4, v1, v0
|
|
; GFX7-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; GFX7-NEXT: s_endpgm
|
|
;
|
|
; GFX8-LABEL: udot4_acc32:
|
|
; GFX8: ; %bb.0: ; %entry
|
|
; GFX8-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x24
|
|
; GFX8-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x34
|
|
; GFX8-NEXT: s_movk_i32 s2, 0xff
|
|
; GFX8-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX8-NEXT: s_load_dword s3, s[4:5], 0x0
|
|
; GFX8-NEXT: s_load_dword s4, s[6:7], 0x0
|
|
; GFX8-NEXT: s_load_dword s10, s[0:1], 0x0
|
|
; GFX8-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX8-NEXT: s_and_b32 s5, s3, s2
|
|
; GFX8-NEXT: s_and_b32 s2, s4, s2
|
|
; GFX8-NEXT: s_bfe_u32 s7, s4, 0x80008
|
|
; GFX8-NEXT: v_mov_b32_e32 v0, s2
|
|
; GFX8-NEXT: v_mov_b32_e32 v1, s10
|
|
; GFX8-NEXT: s_bfe_u32 s9, s4, 0x80010
|
|
; GFX8-NEXT: v_mad_u32_u24 v0, s5, v0, v1
|
|
; GFX8-NEXT: s_bfe_u32 s6, s3, 0x80008
|
|
; GFX8-NEXT: v_mov_b32_e32 v1, s7
|
|
; GFX8-NEXT: s_bfe_u32 s8, s3, 0x80010
|
|
; GFX8-NEXT: v_mad_u32_u24 v0, s6, v1, v0
|
|
; GFX8-NEXT: v_mov_b32_e32 v1, s9
|
|
; GFX8-NEXT: s_lshr_b32 s4, s4, 24
|
|
; GFX8-NEXT: v_mad_u32_u24 v0, s8, v1, v0
|
|
; GFX8-NEXT: s_lshr_b32 s3, s3, 24
|
|
; GFX8-NEXT: v_mov_b32_e32 v1, s4
|
|
; GFX8-NEXT: v_mad_u32_u24 v2, s3, v1, v0
|
|
; GFX8-NEXT: v_mov_b32_e32 v0, s0
|
|
; GFX8-NEXT: v_mov_b32_e32 v1, s1
|
|
; GFX8-NEXT: flat_store_dword v[0:1], v2
|
|
; GFX8-NEXT: s_endpgm
|
|
;
|
|
; GFX9-NODL-LABEL: udot4_acc32:
|
|
; GFX9-NODL: ; %bb.0: ; %entry
|
|
; GFX9-NODL-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x24
|
|
; GFX9-NODL-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x34
|
|
; GFX9-NODL-NEXT: s_movk_i32 s2, 0xff
|
|
; GFX9-NODL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX9-NODL-NEXT: s_load_dword s3, s[4:5], 0x0
|
|
; GFX9-NODL-NEXT: s_load_dword s4, s[6:7], 0x0
|
|
; GFX9-NODL-NEXT: s_load_dword s10, s[0:1], 0x0
|
|
; GFX9-NODL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX9-NODL-NEXT: s_and_b32 s5, s3, s2
|
|
; GFX9-NODL-NEXT: s_and_b32 s2, s4, s2
|
|
; GFX9-NODL-NEXT: s_bfe_u32 s7, s4, 0x80008
|
|
; GFX9-NODL-NEXT: v_mov_b32_e32 v0, s2
|
|
; GFX9-NODL-NEXT: v_mov_b32_e32 v1, s10
|
|
; GFX9-NODL-NEXT: s_bfe_u32 s9, s4, 0x80010
|
|
; GFX9-NODL-NEXT: v_mad_u32_u24 v0, s5, v0, v1
|
|
; GFX9-NODL-NEXT: s_bfe_u32 s6, s3, 0x80008
|
|
; GFX9-NODL-NEXT: v_mov_b32_e32 v1, s7
|
|
; GFX9-NODL-NEXT: s_bfe_u32 s8, s3, 0x80010
|
|
; GFX9-NODL-NEXT: v_mad_u32_u24 v0, s6, v1, v0
|
|
; GFX9-NODL-NEXT: v_mov_b32_e32 v1, s9
|
|
; GFX9-NODL-NEXT: s_lshr_b32 s4, s4, 24
|
|
; GFX9-NODL-NEXT: v_mad_u32_u24 v0, s8, v1, v0
|
|
; GFX9-NODL-NEXT: s_lshr_b32 s3, s3, 24
|
|
; GFX9-NODL-NEXT: v_mov_b32_e32 v1, s4
|
|
; GFX9-NODL-NEXT: v_mad_u32_u24 v2, s3, v1, v0
|
|
; GFX9-NODL-NEXT: v_mov_b32_e32 v0, s0
|
|
; GFX9-NODL-NEXT: v_mov_b32_e32 v1, s1
|
|
; GFX9-NODL-NEXT: global_store_dword v[0:1], v2, off
|
|
; GFX9-NODL-NEXT: s_endpgm
|
|
;
|
|
; GFX9-DL-LABEL: udot4_acc32:
|
|
; GFX9-DL: ; %bb.0: ; %entry
|
|
; GFX9-DL-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x24
|
|
; GFX9-DL-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x34
|
|
; GFX9-DL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX9-DL-NEXT: s_load_dword s2, s[6:7], 0x0
|
|
; GFX9-DL-NEXT: s_load_dword s3, s[0:1], 0x0
|
|
; GFX9-DL-NEXT: s_load_dword s4, s[4:5], 0x0
|
|
; GFX9-DL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX9-DL-NEXT: v_mov_b32_e32 v0, s2
|
|
; GFX9-DL-NEXT: v_mov_b32_e32 v1, s3
|
|
; GFX9-DL-NEXT: v_dot4_u32_u8 v2, s4, v0, v1
|
|
; GFX9-DL-NEXT: v_mov_b32_e32 v0, s0
|
|
; GFX9-DL-NEXT: v_mov_b32_e32 v1, s1
|
|
; GFX9-DL-NEXT: global_store_dword v[0:1], v2, off
|
|
; GFX9-DL-NEXT: s_endpgm
|
|
;
|
|
; GFX10-DL-LABEL: udot4_acc32:
|
|
; GFX10-DL: ; %bb.0: ; %entry
|
|
; GFX10-DL-NEXT: s_clause 0x1
|
|
; GFX10-DL-NEXT: s_load_dwordx2 s[4:5], s[0:1], 0x34
|
|
; GFX10-DL-NEXT: s_load_dwordx4 s[0:3], s[0:1], 0x24
|
|
; GFX10-DL-NEXT: ; implicit-def: $vcc_hi
|
|
; GFX10-DL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX10-DL-NEXT: s_load_dword s6, s[4:5], 0x0
|
|
; GFX10-DL-NEXT: s_load_dword s0, s[0:1], 0x0
|
|
; GFX10-DL-NEXT: s_load_dword s1, s[2:3], 0x0
|
|
; GFX10-DL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX10-DL-NEXT: v_mov_b32_e32 v0, s6
|
|
; GFX10-DL-NEXT: v_dot4_u32_u8 v2, s0, s1, v0
|
|
; GFX10-DL-NEXT: v_mov_b32_e32 v0, s4
|
|
; GFX10-DL-NEXT: v_mov_b32_e32 v1, s5
|
|
; GFX10-DL-NEXT: global_store_dword v[0:1], v2, off
|
|
; GFX10-DL-NEXT: s_endpgm
|
|
<4 x i8> addrspace(1)* %src2,
|
|
i32 addrspace(1)* nocapture %dst) {
|
|
entry:
|
|
%vec1 = load <4 x i8>, <4 x i8> addrspace(1)* %src1
|
|
%vec2 = load <4 x i8>, <4 x i8> addrspace(1)* %src2
|
|
|
|
%v1e0 = extractelement <4 x i8> %vec1, i64 0
|
|
%cv1e0 = zext i8 %v1e0 to i32
|
|
%v2e0 = extractelement <4 x i8> %vec2, i64 0
|
|
%cv2e0 = zext i8 %v2e0 to i32
|
|
%mul1 = mul nuw nsw i32 %cv1e0, %cv2e0
|
|
|
|
%v1e1 = extractelement <4 x i8> %vec1, i64 1
|
|
%cv1e1 = zext i8 %v1e1 to i32
|
|
%v2e1 = extractelement <4 x i8> %vec2, i64 1
|
|
%cv2e1 = zext i8 %v2e1 to i32
|
|
%mul2 = mul nuw nsw i32 %cv1e1, %cv2e1
|
|
|
|
%v1e2 = extractelement <4 x i8> %vec1, i64 2
|
|
%cv1e2 = zext i8 %v1e2 to i32
|
|
%v2e2 = extractelement <4 x i8> %vec2, i64 2
|
|
%cv2e2 = zext i8 %v2e2 to i32
|
|
%mul3 = mul nuw nsw i32 %cv1e2, %cv2e2
|
|
|
|
%v1e3 = extractelement <4 x i8> %vec1, i64 3
|
|
%cv1e3 = zext i8 %v1e3 to i32
|
|
%v2e3 = extractelement <4 x i8> %vec2, i64 3
|
|
%cv2e3 = zext i8 %v2e3 to i32
|
|
%mul4 = mul nuw nsw i32 %cv1e3, %cv2e3
|
|
|
|
%acc = load i32, i32 addrspace(1)* %dst, align 4
|
|
%mad1 = add i32 %mul1, %acc
|
|
%mad2 = add i32 %mad1, %mul2
|
|
%mad3 = add i32 %mad2, %mul3
|
|
%mad4 = add i32 %mad3, %mul4
|
|
|
|
store i32 %mad4, i32 addrspace(1)* %dst, align 4
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_kernel void @udot4_acc16(<4 x i8> addrspace(1)* %src1,
|
|
; GFX7-LABEL: udot4_acc16:
|
|
; GFX7: ; %bb.0: ; %entry
|
|
; GFX7-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x9
|
|
; GFX7-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0xd
|
|
; GFX7-NEXT: s_mov_b32 s3, 0xf000
|
|
; GFX7-NEXT: s_mov_b32 s2, -1
|
|
; GFX7-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX7-NEXT: s_load_dword s6, s[6:7], 0x0
|
|
; GFX7-NEXT: buffer_load_ushort v0, off, s[0:3], 0
|
|
; GFX7-NEXT: s_load_dword s4, s[4:5], 0x0
|
|
; GFX7-NEXT: s_movk_i32 s5, 0xff
|
|
; GFX7-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX7-NEXT: s_and_b32 s7, s6, s5
|
|
; GFX7-NEXT: s_bfe_u32 s8, s6, 0x80008
|
|
; GFX7-NEXT: s_and_b32 s5, s4, s5
|
|
; GFX7-NEXT: v_mov_b32_e32 v1, s7
|
|
; GFX7-NEXT: s_bfe_u32 s10, s6, 0x80010
|
|
; GFX7-NEXT: s_bfe_u32 s9, s4, 0x80008
|
|
; GFX7-NEXT: v_mov_b32_e32 v2, s8
|
|
; GFX7-NEXT: s_bfe_u32 s11, s4, 0x80010
|
|
; GFX7-NEXT: s_lshr_b32 s6, s6, 24
|
|
; GFX7-NEXT: v_mov_b32_e32 v3, s10
|
|
; GFX7-NEXT: s_lshr_b32 s4, s4, 24
|
|
; GFX7-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX7-NEXT: v_mad_u32_u24 v0, s5, v1, v0
|
|
; GFX7-NEXT: v_mad_u32_u24 v0, s9, v2, v0
|
|
; GFX7-NEXT: v_mad_u32_u24 v0, s11, v3, v0
|
|
; GFX7-NEXT: v_mov_b32_e32 v1, s6
|
|
; GFX7-NEXT: v_mad_u32_u24 v0, s4, v1, v0
|
|
; GFX7-NEXT: buffer_store_short v0, off, s[0:3], 0
|
|
; GFX7-NEXT: s_endpgm
|
|
;
|
|
; GFX8-LABEL: udot4_acc16:
|
|
; GFX8: ; %bb.0: ; %entry
|
|
; GFX8-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x24
|
|
; GFX8-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x34
|
|
; GFX8-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX8-NEXT: s_load_dword s2, s[6:7], 0x0
|
|
; GFX8-NEXT: v_mov_b32_e32 v0, s0
|
|
; GFX8-NEXT: v_mov_b32_e32 v1, s1
|
|
; GFX8-NEXT: flat_load_ushort v2, v[0:1]
|
|
; GFX8-NEXT: s_load_dword s1, s[4:5], 0x0
|
|
; GFX8-NEXT: s_movk_i32 s0, 0xff
|
|
; GFX8-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX8-NEXT: s_bfe_u32 s5, s2, 0x80008
|
|
; GFX8-NEXT: s_bfe_u32 s7, s2, 0x80010
|
|
; GFX8-NEXT: v_mov_b32_e32 v4, s5
|
|
; GFX8-NEXT: s_and_b32 s3, s1, s0
|
|
; GFX8-NEXT: s_and_b32 s0, s2, s0
|
|
; GFX8-NEXT: v_mov_b32_e32 v3, s0
|
|
; GFX8-NEXT: s_bfe_u32 s4, s1, 0x80008
|
|
; GFX8-NEXT: s_bfe_u32 s6, s1, 0x80010
|
|
; GFX8-NEXT: s_lshr_b32 s2, s2, 24
|
|
; GFX8-NEXT: v_mov_b32_e32 v5, s7
|
|
; GFX8-NEXT: s_lshr_b32 s1, s1, 24
|
|
; GFX8-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX8-NEXT: v_mad_u32_u24 v2, s3, v3, v2
|
|
; GFX8-NEXT: v_mad_u32_u24 v2, s4, v4, v2
|
|
; GFX8-NEXT: v_mad_u32_u24 v2, s6, v5, v2
|
|
; GFX8-NEXT: v_mov_b32_e32 v3, s2
|
|
; GFX8-NEXT: v_mad_u32_u24 v2, s1, v3, v2
|
|
; GFX8-NEXT: flat_store_short v[0:1], v2
|
|
; GFX8-NEXT: s_endpgm
|
|
;
|
|
; GFX9-NODL-LABEL: udot4_acc16:
|
|
; GFX9-NODL: ; %bb.0: ; %entry
|
|
; GFX9-NODL-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x24
|
|
; GFX9-NODL-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x34
|
|
; GFX9-NODL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX9-NODL-NEXT: s_load_dword s2, s[6:7], 0x0
|
|
; GFX9-NODL-NEXT: v_mov_b32_e32 v0, s0
|
|
; GFX9-NODL-NEXT: v_mov_b32_e32 v1, s1
|
|
; GFX9-NODL-NEXT: global_load_ushort v2, v[0:1], off
|
|
; GFX9-NODL-NEXT: s_load_dword s1, s[4:5], 0x0
|
|
; GFX9-NODL-NEXT: s_movk_i32 s0, 0xff
|
|
; GFX9-NODL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX9-NODL-NEXT: s_bfe_u32 s5, s2, 0x80008
|
|
; GFX9-NODL-NEXT: s_bfe_u32 s7, s2, 0x80010
|
|
; GFX9-NODL-NEXT: v_mov_b32_e32 v4, s5
|
|
; GFX9-NODL-NEXT: s_and_b32 s3, s1, s0
|
|
; GFX9-NODL-NEXT: s_and_b32 s0, s2, s0
|
|
; GFX9-NODL-NEXT: v_mov_b32_e32 v3, s0
|
|
; GFX9-NODL-NEXT: s_bfe_u32 s4, s1, 0x80008
|
|
; GFX9-NODL-NEXT: s_bfe_u32 s6, s1, 0x80010
|
|
; GFX9-NODL-NEXT: s_lshr_b32 s2, s2, 24
|
|
; GFX9-NODL-NEXT: v_mov_b32_e32 v5, s7
|
|
; GFX9-NODL-NEXT: s_lshr_b32 s1, s1, 24
|
|
; GFX9-NODL-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX9-NODL-NEXT: v_mad_u32_u24 v2, s3, v3, v2
|
|
; GFX9-NODL-NEXT: v_mad_u32_u24 v2, s4, v4, v2
|
|
; GFX9-NODL-NEXT: v_mad_u32_u24 v2, s6, v5, v2
|
|
; GFX9-NODL-NEXT: v_mov_b32_e32 v3, s2
|
|
; GFX9-NODL-NEXT: v_mad_u32_u24 v2, s1, v3, v2
|
|
; GFX9-NODL-NEXT: global_store_short v[0:1], v2, off
|
|
; GFX9-NODL-NEXT: s_endpgm
|
|
;
|
|
; GFX9-DL-LABEL: udot4_acc16:
|
|
; GFX9-DL: ; %bb.0: ; %entry
|
|
; GFX9-DL-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x24
|
|
; GFX9-DL-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x34
|
|
; GFX9-DL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX9-DL-NEXT: s_load_dword s2, s[4:5], 0x0
|
|
; GFX9-DL-NEXT: s_load_dword s3, s[6:7], 0x0
|
|
; GFX9-DL-NEXT: v_mov_b32_e32 v0, s0
|
|
; GFX9-DL-NEXT: v_mov_b32_e32 v1, s1
|
|
; GFX9-DL-NEXT: global_load_ushort v2, v[0:1], off
|
|
; GFX9-DL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX9-DL-NEXT: v_mov_b32_e32 v3, s3
|
|
; GFX9-DL-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX9-DL-NEXT: v_dot4_u32_u8 v2, s2, v3, v2
|
|
; GFX9-DL-NEXT: global_store_short v[0:1], v2, off
|
|
; GFX9-DL-NEXT: s_endpgm
|
|
;
|
|
; GFX10-DL-LABEL: udot4_acc16:
|
|
; GFX10-DL: ; %bb.0: ; %entry
|
|
; GFX10-DL-NEXT: s_load_dwordx2 s[2:3], s[0:1], 0x34
|
|
; GFX10-DL-NEXT: ; implicit-def: $vcc_hi
|
|
; GFX10-DL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX10-DL-NEXT: v_mov_b32_e32 v0, s2
|
|
; GFX10-DL-NEXT: v_mov_b32_e32 v1, s3
|
|
; GFX10-DL-NEXT: s_load_dwordx4 s[0:3], s[0:1], 0x24
|
|
; GFX10-DL-NEXT: global_load_ushort v2, v[0:1], off
|
|
; GFX10-DL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX10-DL-NEXT: s_load_dword s0, s[0:1], 0x0
|
|
; GFX10-DL-NEXT: s_load_dword s1, s[2:3], 0x0
|
|
; GFX10-DL-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
|
|
; GFX10-DL-NEXT: v_dot4_u32_u8 v2, s0, s1, v2
|
|
; GFX10-DL-NEXT: global_store_short v[0:1], v2, off
|
|
; GFX10-DL-NEXT: s_endpgm
|
|
<4 x i8> addrspace(1)* %src2,
|
|
i16 addrspace(1)* nocapture %dst) {
|
|
entry:
|
|
%vec1 = load <4 x i8>, <4 x i8> addrspace(1)* %src1
|
|
%vec2 = load <4 x i8>, <4 x i8> addrspace(1)* %src2
|
|
|
|
%v1e0 = extractelement <4 x i8> %vec1, i64 0
|
|
%cv1e0 = zext i8 %v1e0 to i16
|
|
%v2e0 = extractelement <4 x i8> %vec2, i64 0
|
|
%cv2e0 = zext i8 %v2e0 to i16
|
|
%mul1 = mul nuw nsw i16 %cv1e0, %cv2e0
|
|
|
|
%v1e1 = extractelement <4 x i8> %vec1, i64 1
|
|
%cv1e1 = zext i8 %v1e1 to i16
|
|
%v2e1 = extractelement <4 x i8> %vec2, i64 1
|
|
%cv2e1 = zext i8 %v2e1 to i16
|
|
%mul2 = mul nuw nsw i16 %cv1e1, %cv2e1
|
|
|
|
%v1e2 = extractelement <4 x i8> %vec1, i64 2
|
|
%cv1e2 = zext i8 %v1e2 to i16
|
|
%v2e2 = extractelement <4 x i8> %vec2, i64 2
|
|
%cv2e2 = zext i8 %v2e2 to i16
|
|
%mul3 = mul nuw nsw i16 %cv1e2, %cv2e2
|
|
|
|
%v1e3 = extractelement <4 x i8> %vec1, i64 3
|
|
%cv1e3 = zext i8 %v1e3 to i16
|
|
%v2e3 = extractelement <4 x i8> %vec2, i64 3
|
|
%cv2e3 = zext i8 %v2e3 to i16
|
|
%mul4 = mul nuw nsw i16 %cv1e3, %cv2e3
|
|
|
|
%acc = load i16, i16 addrspace(1)* %dst, align 2
|
|
%mad1 = add i16 %mul1, %acc
|
|
%mad2 = add i16 %mad1, %mul2
|
|
%mad3 = add i16 %mad2, %mul3
|
|
%mad4 = add i16 %mad3, %mul4
|
|
|
|
store i16 %mad4, i16 addrspace(1)* %dst, align 2
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_kernel void @udot4_acc8(<4 x i8> addrspace(1)* %src1,
|
|
; GFX7-LABEL: udot4_acc8:
|
|
; GFX7: ; %bb.0: ; %entry
|
|
; GFX7-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x9
|
|
; GFX7-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0xd
|
|
; GFX7-NEXT: s_mov_b32 s3, 0xf000
|
|
; GFX7-NEXT: s_mov_b32 s2, -1
|
|
; GFX7-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX7-NEXT: s_load_dword s6, s[6:7], 0x0
|
|
; GFX7-NEXT: buffer_load_ubyte v0, off, s[0:3], 0
|
|
; GFX7-NEXT: s_load_dword s4, s[4:5], 0x0
|
|
; GFX7-NEXT: s_movk_i32 s5, 0xff
|
|
; GFX7-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX7-NEXT: s_and_b32 s7, s6, s5
|
|
; GFX7-NEXT: s_bfe_u32 s8, s6, 0x80008
|
|
; GFX7-NEXT: s_and_b32 s5, s4, s5
|
|
; GFX7-NEXT: v_mov_b32_e32 v1, s7
|
|
; GFX7-NEXT: s_bfe_u32 s10, s6, 0x80010
|
|
; GFX7-NEXT: s_bfe_u32 s9, s4, 0x80008
|
|
; GFX7-NEXT: v_mov_b32_e32 v2, s8
|
|
; GFX7-NEXT: s_bfe_u32 s11, s4, 0x80010
|
|
; GFX7-NEXT: s_lshr_b32 s6, s6, 24
|
|
; GFX7-NEXT: v_mov_b32_e32 v3, s10
|
|
; GFX7-NEXT: s_lshr_b32 s4, s4, 24
|
|
; GFX7-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX7-NEXT: v_mad_u32_u24 v0, s5, v1, v0
|
|
; GFX7-NEXT: v_mad_u32_u24 v0, s9, v2, v0
|
|
; GFX7-NEXT: v_mad_u32_u24 v0, s11, v3, v0
|
|
; GFX7-NEXT: v_mov_b32_e32 v1, s6
|
|
; GFX7-NEXT: v_mad_u32_u24 v0, s4, v1, v0
|
|
; GFX7-NEXT: buffer_store_byte v0, off, s[0:3], 0
|
|
; GFX7-NEXT: s_endpgm
|
|
;
|
|
; GFX8-LABEL: udot4_acc8:
|
|
; GFX8: ; %bb.0: ; %entry
|
|
; GFX8-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x24
|
|
; GFX8-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x34
|
|
; GFX8-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX8-NEXT: s_load_dword s2, s[4:5], 0x0
|
|
; GFX8-NEXT: v_mov_b32_e32 v0, s0
|
|
; GFX8-NEXT: v_mov_b32_e32 v1, s1
|
|
; GFX8-NEXT: flat_load_ubyte v2, v[0:1]
|
|
; GFX8-NEXT: s_load_dword s1, s[6:7], 0x0
|
|
; GFX8-NEXT: s_movk_i32 s0, 0xff
|
|
; GFX8-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX8-NEXT: s_bfe_u32 s5, s2, 0x80008
|
|
; GFX8-NEXT: s_bfe_u32 s7, s2, 0x80010
|
|
; GFX8-NEXT: s_and_b32 s3, s1, s0
|
|
; GFX8-NEXT: s_and_b32 s0, s2, s0
|
|
; GFX8-NEXT: s_bfe_u32 s4, s1, 0x80008
|
|
; GFX8-NEXT: v_mov_b32_e32 v3, s3
|
|
; GFX8-NEXT: s_bfe_u32 s6, s1, 0x80010
|
|
; GFX8-NEXT: v_mov_b32_e32 v4, s4
|
|
; GFX8-NEXT: s_lshr_b32 s1, s1, 24
|
|
; GFX8-NEXT: v_mov_b32_e32 v5, s6
|
|
; GFX8-NEXT: s_lshr_b32 s2, s2, 24
|
|
; GFX8-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX8-NEXT: v_mad_u32_u24 v2, s0, v3, v2
|
|
; GFX8-NEXT: v_mad_u32_u24 v2, s5, v4, v2
|
|
; GFX8-NEXT: v_mad_u32_u24 v2, s7, v5, v2
|
|
; GFX8-NEXT: v_mov_b32_e32 v3, s1
|
|
; GFX8-NEXT: v_mad_u32_u24 v2, s2, v3, v2
|
|
; GFX8-NEXT: flat_store_byte v[0:1], v2
|
|
; GFX8-NEXT: s_endpgm
|
|
;
|
|
; GFX9-NODL-LABEL: udot4_acc8:
|
|
; GFX9-NODL: ; %bb.0: ; %entry
|
|
; GFX9-NODL-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x24
|
|
; GFX9-NODL-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x34
|
|
; GFX9-NODL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX9-NODL-NEXT: s_load_dword s2, s[4:5], 0x0
|
|
; GFX9-NODL-NEXT: v_mov_b32_e32 v0, s0
|
|
; GFX9-NODL-NEXT: v_mov_b32_e32 v1, s1
|
|
; GFX9-NODL-NEXT: global_load_ubyte v2, v[0:1], off
|
|
; GFX9-NODL-NEXT: s_load_dword s1, s[6:7], 0x0
|
|
; GFX9-NODL-NEXT: s_movk_i32 s0, 0xff
|
|
; GFX9-NODL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX9-NODL-NEXT: s_bfe_u32 s5, s2, 0x80008
|
|
; GFX9-NODL-NEXT: s_bfe_u32 s7, s2, 0x80010
|
|
; GFX9-NODL-NEXT: s_and_b32 s3, s1, s0
|
|
; GFX9-NODL-NEXT: s_and_b32 s0, s2, s0
|
|
; GFX9-NODL-NEXT: s_bfe_u32 s4, s1, 0x80008
|
|
; GFX9-NODL-NEXT: v_mov_b32_e32 v3, s3
|
|
; GFX9-NODL-NEXT: s_bfe_u32 s6, s1, 0x80010
|
|
; GFX9-NODL-NEXT: v_mov_b32_e32 v4, s4
|
|
; GFX9-NODL-NEXT: s_lshr_b32 s1, s1, 24
|
|
; GFX9-NODL-NEXT: v_mov_b32_e32 v5, s6
|
|
; GFX9-NODL-NEXT: s_lshr_b32 s2, s2, 24
|
|
; GFX9-NODL-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX9-NODL-NEXT: v_mad_u32_u24 v2, s0, v3, v2
|
|
; GFX9-NODL-NEXT: v_mad_u32_u24 v2, s5, v4, v2
|
|
; GFX9-NODL-NEXT: v_mad_u32_u24 v2, s7, v5, v2
|
|
; GFX9-NODL-NEXT: v_mov_b32_e32 v3, s1
|
|
; GFX9-NODL-NEXT: v_mad_u32_u24 v2, s2, v3, v2
|
|
; GFX9-NODL-NEXT: global_store_byte v[0:1], v2, off
|
|
; GFX9-NODL-NEXT: s_endpgm
|
|
;
|
|
; GFX9-DL-LABEL: udot4_acc8:
|
|
; GFX9-DL: ; %bb.0: ; %entry
|
|
; GFX9-DL-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x24
|
|
; GFX9-DL-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x34
|
|
; GFX9-DL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX9-DL-NEXT: s_load_dword s2, s[4:5], 0x0
|
|
; GFX9-DL-NEXT: s_load_dword s3, s[6:7], 0x0
|
|
; GFX9-DL-NEXT: v_mov_b32_e32 v0, s0
|
|
; GFX9-DL-NEXT: v_mov_b32_e32 v1, s1
|
|
; GFX9-DL-NEXT: global_load_ubyte v2, v[0:1], off
|
|
; GFX9-DL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX9-DL-NEXT: v_mov_b32_e32 v3, s3
|
|
; GFX9-DL-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX9-DL-NEXT: v_dot4_u32_u8 v2, s2, v3, v2
|
|
; GFX9-DL-NEXT: global_store_byte v[0:1], v2, off
|
|
; GFX9-DL-NEXT: s_endpgm
|
|
;
|
|
; GFX10-DL-LABEL: udot4_acc8:
|
|
; GFX10-DL: ; %bb.0: ; %entry
|
|
; GFX10-DL-NEXT: s_load_dwordx2 s[2:3], s[0:1], 0x34
|
|
; GFX10-DL-NEXT: ; implicit-def: $vcc_hi
|
|
; GFX10-DL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX10-DL-NEXT: v_mov_b32_e32 v0, s2
|
|
; GFX10-DL-NEXT: v_mov_b32_e32 v1, s3
|
|
; GFX10-DL-NEXT: s_load_dwordx4 s[0:3], s[0:1], 0x24
|
|
; GFX10-DL-NEXT: global_load_ubyte v2, v[0:1], off
|
|
; GFX10-DL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX10-DL-NEXT: s_load_dword s0, s[0:1], 0x0
|
|
; GFX10-DL-NEXT: s_load_dword s1, s[2:3], 0x0
|
|
; GFX10-DL-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
|
|
; GFX10-DL-NEXT: v_dot4_u32_u8 v2, s0, s1, v2
|
|
; GFX10-DL-NEXT: global_store_byte v[0:1], v2, off
|
|
; GFX10-DL-NEXT: s_endpgm
|
|
<4 x i8> addrspace(1)* %src2,
|
|
i8 addrspace(1)* nocapture %dst) {
|
|
entry:
|
|
%vec1 = load <4 x i8>, <4 x i8> addrspace(1)* %src1
|
|
%vec2 = load <4 x i8>, <4 x i8> addrspace(1)* %src2
|
|
|
|
%v1e0 = extractelement <4 x i8> %vec1, i64 0
|
|
%v2e0 = extractelement <4 x i8> %vec2, i64 0
|
|
%mul1 = mul nuw nsw i8 %v1e0, %v2e0
|
|
|
|
%v1e1 = extractelement <4 x i8> %vec1, i64 1
|
|
%v2e1 = extractelement <4 x i8> %vec2, i64 1
|
|
%mul2 = mul nuw nsw i8 %v1e1, %v2e1
|
|
|
|
%v1e2 = extractelement <4 x i8> %vec1, i64 2
|
|
%v2e2 = extractelement <4 x i8> %vec2, i64 2
|
|
%mul3 = mul nuw nsw i8 %v1e2, %v2e2
|
|
|
|
%v1e3 = extractelement <4 x i8> %vec1, i64 3
|
|
%v2e3 = extractelement <4 x i8> %vec2, i64 3
|
|
%mul4 = mul nuw nsw i8 %v1e3, %v2e3
|
|
|
|
%acc = load i8, i8 addrspace(1)* %dst, align 2
|
|
%mad1 = add i8 %mul1, %acc
|
|
%mad2 = add i8 %mad1, %mul2
|
|
%mad3 = add i8 %mad2, %mul3
|
|
%mad4 = add i8 %mad3, %mul4
|
|
|
|
store i8 %mad4, i8 addrspace(1)* %dst, align 2
|
|
ret void
|
|
}
|
|
|
|
; TODO: Generate udot4?
|
|
define amdgpu_kernel void @udot2_8(<4 x i8> addrspace(1)* %src1,
|
|
; GFX7-LABEL: udot2_8:
|
|
; GFX7: ; %bb.0: ; %entry
|
|
; GFX7-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x9
|
|
; GFX7-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0xd
|
|
; GFX7-NEXT: s_mov_b32 s3, 0xf000
|
|
; GFX7-NEXT: s_mov_b32 s2, -1
|
|
; GFX7-NEXT: s_movk_i32 s8, 0xff
|
|
; GFX7-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX7-NEXT: buffer_load_ubyte v0, off, s[0:3], 0
|
|
; GFX7-NEXT: s_load_dword s4, s[4:5], 0x0
|
|
; GFX7-NEXT: s_load_dword s5, s[6:7], 0x0
|
|
; GFX7-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX7-NEXT: s_and_b32 s7, s4, s8
|
|
; GFX7-NEXT: s_and_b32 s6, s5, s8
|
|
; GFX7-NEXT: v_mov_b32_e32 v1, s6
|
|
; GFX7-NEXT: s_bfe_u32 s5, s5, 0x80008
|
|
; GFX7-NEXT: s_bfe_u32 s4, s4, 0x80008
|
|
; GFX7-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX7-NEXT: v_mad_u32_u24 v0, s7, v1, v0
|
|
; GFX7-NEXT: v_mov_b32_e32 v1, s5
|
|
; GFX7-NEXT: v_mad_u32_u24 v0, s4, v1, v0
|
|
; GFX7-NEXT: buffer_store_byte v0, off, s[0:3], 0
|
|
; GFX7-NEXT: s_endpgm
|
|
;
|
|
; GFX8-LABEL: udot2_8:
|
|
; GFX8: ; %bb.0: ; %entry
|
|
; GFX8-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x24
|
|
; GFX8-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x34
|
|
; GFX8-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX8-NEXT: v_mov_b32_e32 v0, s0
|
|
; GFX8-NEXT: v_mov_b32_e32 v1, s1
|
|
; GFX8-NEXT: flat_load_ubyte v2, v[0:1]
|
|
; GFX8-NEXT: s_load_dword s1, s[4:5], 0x0
|
|
; GFX8-NEXT: s_load_dword s2, s[6:7], 0x0
|
|
; GFX8-NEXT: s_movk_i32 s0, 0xff
|
|
; GFX8-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX8-NEXT: s_and_b32 s3, s2, s0
|
|
; GFX8-NEXT: s_and_b32 s0, s1, s0
|
|
; GFX8-NEXT: v_mov_b32_e32 v3, s3
|
|
; GFX8-NEXT: s_bfe_u32 s2, s2, 0x80008
|
|
; GFX8-NEXT: s_bfe_u32 s1, s1, 0x80008
|
|
; GFX8-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX8-NEXT: v_mad_u32_u24 v2, s0, v3, v2
|
|
; GFX8-NEXT: v_mov_b32_e32 v3, s2
|
|
; GFX8-NEXT: v_mad_u32_u24 v2, s1, v3, v2
|
|
; GFX8-NEXT: flat_store_byte v[0:1], v2
|
|
; GFX8-NEXT: s_endpgm
|
|
;
|
|
; GFX9-NODL-LABEL: udot2_8:
|
|
; GFX9-NODL: ; %bb.0: ; %entry
|
|
; GFX9-NODL-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x24
|
|
; GFX9-NODL-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x34
|
|
; GFX9-NODL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX9-NODL-NEXT: v_mov_b32_e32 v0, s0
|
|
; GFX9-NODL-NEXT: v_mov_b32_e32 v1, s1
|
|
; GFX9-NODL-NEXT: global_load_ubyte v2, v[0:1], off
|
|
; GFX9-NODL-NEXT: s_load_dword s1, s[4:5], 0x0
|
|
; GFX9-NODL-NEXT: s_load_dword s2, s[6:7], 0x0
|
|
; GFX9-NODL-NEXT: s_movk_i32 s0, 0xff
|
|
; GFX9-NODL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX9-NODL-NEXT: s_and_b32 s3, s2, s0
|
|
; GFX9-NODL-NEXT: s_and_b32 s0, s1, s0
|
|
; GFX9-NODL-NEXT: v_mov_b32_e32 v3, s3
|
|
; GFX9-NODL-NEXT: s_bfe_u32 s2, s2, 0x80008
|
|
; GFX9-NODL-NEXT: s_bfe_u32 s1, s1, 0x80008
|
|
; GFX9-NODL-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX9-NODL-NEXT: v_mad_u32_u24 v2, s0, v3, v2
|
|
; GFX9-NODL-NEXT: v_mov_b32_e32 v3, s2
|
|
; GFX9-NODL-NEXT: v_mad_u32_u24 v2, s1, v3, v2
|
|
; GFX9-NODL-NEXT: global_store_byte v[0:1], v2, off
|
|
; GFX9-NODL-NEXT: s_endpgm
|
|
;
|
|
; GFX9-DL-LABEL: udot2_8:
|
|
; GFX9-DL: ; %bb.0: ; %entry
|
|
; GFX9-DL-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x24
|
|
; GFX9-DL-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x34
|
|
; GFX9-DL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX9-DL-NEXT: v_mov_b32_e32 v0, s0
|
|
; GFX9-DL-NEXT: v_mov_b32_e32 v1, s1
|
|
; GFX9-DL-NEXT: global_load_ubyte v2, v[0:1], off
|
|
; GFX9-DL-NEXT: s_load_dword s1, s[4:5], 0x0
|
|
; GFX9-DL-NEXT: s_load_dword s2, s[6:7], 0x0
|
|
; GFX9-DL-NEXT: s_movk_i32 s0, 0xff
|
|
; GFX9-DL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX9-DL-NEXT: s_and_b32 s3, s2, s0
|
|
; GFX9-DL-NEXT: s_and_b32 s0, s1, s0
|
|
; GFX9-DL-NEXT: v_mov_b32_e32 v3, s3
|
|
; GFX9-DL-NEXT: s_bfe_u32 s2, s2, 0x80008
|
|
; GFX9-DL-NEXT: s_bfe_u32 s1, s1, 0x80008
|
|
; GFX9-DL-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX9-DL-NEXT: v_mad_u32_u24 v2, s0, v3, v2
|
|
; GFX9-DL-NEXT: v_mov_b32_e32 v3, s2
|
|
; GFX9-DL-NEXT: v_mad_u32_u24 v2, s1, v3, v2
|
|
; GFX9-DL-NEXT: global_store_byte v[0:1], v2, off
|
|
; GFX9-DL-NEXT: s_endpgm
|
|
;
|
|
; GFX10-DL-LABEL: udot2_8:
|
|
; GFX10-DL: ; %bb.0: ; %entry
|
|
; GFX10-DL-NEXT: s_load_dwordx2 s[2:3], s[0:1], 0x34
|
|
; GFX10-DL-NEXT: ; implicit-def: $vcc_hi
|
|
; GFX10-DL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX10-DL-NEXT: v_mov_b32_e32 v0, s2
|
|
; GFX10-DL-NEXT: v_mov_b32_e32 v1, s3
|
|
; GFX10-DL-NEXT: s_load_dwordx4 s[0:3], s[0:1], 0x24
|
|
; GFX10-DL-NEXT: global_load_ubyte v2, v[0:1], off
|
|
; GFX10-DL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX10-DL-NEXT: s_load_dword s2, s[2:3], 0x0
|
|
; GFX10-DL-NEXT: s_load_dword s0, s[0:1], 0x0
|
|
; GFX10-DL-NEXT: s_movk_i32 s1, 0xff
|
|
; GFX10-DL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX10-DL-NEXT: s_and_b32 s3, s2, s1
|
|
; GFX10-DL-NEXT: s_and_b32 s1, s0, s1
|
|
; GFX10-DL-NEXT: s_bfe_u32 s0, s0, 0x80008
|
|
; GFX10-DL-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX10-DL-NEXT: v_mad_u32_u24 v2, s1, s3, v2
|
|
; GFX10-DL-NEXT: s_bfe_u32 s1, s2, 0x80008
|
|
; GFX10-DL-NEXT: v_mad_u32_u24 v2, s0, s1, v2
|
|
; GFX10-DL-NEXT: global_store_byte v[0:1], v2, off
|
|
; GFX10-DL-NEXT: s_endpgm
|
|
<4 x i8> addrspace(1)* %src2,
|
|
i8 addrspace(1)* nocapture %dst) {
|
|
entry:
|
|
%vec1 = load <4 x i8>, <4 x i8> addrspace(1)* %src1
|
|
%vec2 = load <4 x i8>, <4 x i8> addrspace(1)* %src2
|
|
|
|
%v1e0 = extractelement <4 x i8> %vec1, i64 0
|
|
%v2e0 = extractelement <4 x i8> %vec2, i64 0
|
|
%mul1 = mul nuw nsw i8 %v1e0, %v2e0
|
|
|
|
%v1e1 = extractelement <4 x i8> %vec1, i64 1
|
|
%v2e1 = extractelement <4 x i8> %vec2, i64 1
|
|
%mul2 = mul nuw nsw i8 %v1e1, %v2e1
|
|
|
|
%acc = load i8, i8 addrspace(1)* %dst, align 2
|
|
%mad1 = add i8 %mul1, %acc
|
|
%mad2 = add i8 %mad1, %mul2
|
|
store i8 %mad2, i8 addrspace(1)* %dst, align 2
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_kernel void @udot4_CommutationInsideMAD(<4 x i8> addrspace(1)* %src1,
|
|
; GFX7-LABEL: udot4_CommutationInsideMAD:
|
|
; GFX7: ; %bb.0: ; %entry
|
|
; GFX7-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x9
|
|
; GFX7-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0xd
|
|
; GFX7-NEXT: s_mov_b32 s3, 0xf000
|
|
; GFX7-NEXT: s_mov_b32 s2, -1
|
|
; GFX7-NEXT: s_movk_i32 s8, 0xff
|
|
; GFX7-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX7-NEXT: buffer_load_ubyte v0, off, s[0:3], 0
|
|
; GFX7-NEXT: s_load_dword s4, s[4:5], 0x0
|
|
; GFX7-NEXT: s_load_dword s5, s[6:7], 0x0
|
|
; GFX7-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX7-NEXT: s_and_b32 s6, s4, s8
|
|
; GFX7-NEXT: s_and_b32 s7, s5, s8
|
|
; GFX7-NEXT: s_bfe_u32 s8, s4, 0x80008
|
|
; GFX7-NEXT: v_mov_b32_e32 v1, s6
|
|
; GFX7-NEXT: s_bfe_u32 s10, s4, 0x80010
|
|
; GFX7-NEXT: s_bfe_u32 s9, s5, 0x80008
|
|
; GFX7-NEXT: v_mov_b32_e32 v2, s8
|
|
; GFX7-NEXT: s_bfe_u32 s11, s5, 0x80010
|
|
; GFX7-NEXT: s_lshr_b32 s4, s4, 24
|
|
; GFX7-NEXT: v_mov_b32_e32 v3, s10
|
|
; GFX7-NEXT: s_lshr_b32 s5, s5, 24
|
|
; GFX7-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX7-NEXT: v_mad_u32_u24 v0, s7, v1, v0
|
|
; GFX7-NEXT: v_mad_u32_u24 v0, s9, v2, v0
|
|
; GFX7-NEXT: v_mad_u32_u24 v0, s11, v3, v0
|
|
; GFX7-NEXT: v_mov_b32_e32 v1, s4
|
|
; GFX7-NEXT: v_mad_u32_u24 v0, s5, v1, v0
|
|
; GFX7-NEXT: buffer_store_byte v0, off, s[0:3], 0
|
|
; GFX7-NEXT: s_endpgm
|
|
;
|
|
; GFX8-LABEL: udot4_CommutationInsideMAD:
|
|
; GFX8: ; %bb.0: ; %entry
|
|
; GFX8-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x24
|
|
; GFX8-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x34
|
|
; GFX8-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX8-NEXT: v_mov_b32_e32 v0, s0
|
|
; GFX8-NEXT: v_mov_b32_e32 v1, s1
|
|
; GFX8-NEXT: flat_load_ubyte v2, v[0:1]
|
|
; GFX8-NEXT: s_load_dword s1, s[4:5], 0x0
|
|
; GFX8-NEXT: s_load_dword s2, s[6:7], 0x0
|
|
; GFX8-NEXT: s_movk_i32 s0, 0xff
|
|
; GFX8-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX8-NEXT: s_and_b32 s3, s1, s0
|
|
; GFX8-NEXT: s_bfe_u32 s4, s1, 0x80008
|
|
; GFX8-NEXT: s_and_b32 s0, s2, s0
|
|
; GFX8-NEXT: v_mov_b32_e32 v3, s3
|
|
; GFX8-NEXT: s_bfe_u32 s6, s1, 0x80010
|
|
; GFX8-NEXT: s_bfe_u32 s5, s2, 0x80008
|
|
; GFX8-NEXT: v_mov_b32_e32 v4, s4
|
|
; GFX8-NEXT: s_bfe_u32 s7, s2, 0x80010
|
|
; GFX8-NEXT: s_lshr_b32 s1, s1, 24
|
|
; GFX8-NEXT: v_mov_b32_e32 v5, s6
|
|
; GFX8-NEXT: s_lshr_b32 s2, s2, 24
|
|
; GFX8-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX8-NEXT: v_mad_u32_u24 v2, s0, v3, v2
|
|
; GFX8-NEXT: v_mad_u32_u24 v2, s5, v4, v2
|
|
; GFX8-NEXT: v_mad_u32_u24 v2, s7, v5, v2
|
|
; GFX8-NEXT: v_mov_b32_e32 v3, s1
|
|
; GFX8-NEXT: v_mad_u32_u24 v2, s2, v3, v2
|
|
; GFX8-NEXT: flat_store_byte v[0:1], v2
|
|
; GFX8-NEXT: s_endpgm
|
|
;
|
|
; GFX9-NODL-LABEL: udot4_CommutationInsideMAD:
|
|
; GFX9-NODL: ; %bb.0: ; %entry
|
|
; GFX9-NODL-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x24
|
|
; GFX9-NODL-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x34
|
|
; GFX9-NODL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX9-NODL-NEXT: v_mov_b32_e32 v0, s0
|
|
; GFX9-NODL-NEXT: v_mov_b32_e32 v1, s1
|
|
; GFX9-NODL-NEXT: global_load_ubyte v2, v[0:1], off
|
|
; GFX9-NODL-NEXT: s_load_dword s1, s[4:5], 0x0
|
|
; GFX9-NODL-NEXT: s_load_dword s2, s[6:7], 0x0
|
|
; GFX9-NODL-NEXT: s_movk_i32 s0, 0xff
|
|
; GFX9-NODL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX9-NODL-NEXT: s_and_b32 s3, s1, s0
|
|
; GFX9-NODL-NEXT: s_bfe_u32 s4, s1, 0x80008
|
|
; GFX9-NODL-NEXT: s_and_b32 s0, s2, s0
|
|
; GFX9-NODL-NEXT: v_mov_b32_e32 v3, s3
|
|
; GFX9-NODL-NEXT: s_bfe_u32 s6, s1, 0x80010
|
|
; GFX9-NODL-NEXT: s_bfe_u32 s5, s2, 0x80008
|
|
; GFX9-NODL-NEXT: v_mov_b32_e32 v4, s4
|
|
; GFX9-NODL-NEXT: s_bfe_u32 s7, s2, 0x80010
|
|
; GFX9-NODL-NEXT: s_lshr_b32 s1, s1, 24
|
|
; GFX9-NODL-NEXT: v_mov_b32_e32 v5, s6
|
|
; GFX9-NODL-NEXT: s_lshr_b32 s2, s2, 24
|
|
; GFX9-NODL-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX9-NODL-NEXT: v_mad_u32_u24 v2, s0, v3, v2
|
|
; GFX9-NODL-NEXT: v_mad_u32_u24 v2, s5, v4, v2
|
|
; GFX9-NODL-NEXT: v_mad_u32_u24 v2, s7, v5, v2
|
|
; GFX9-NODL-NEXT: v_mov_b32_e32 v3, s1
|
|
; GFX9-NODL-NEXT: v_mad_u32_u24 v2, s2, v3, v2
|
|
; GFX9-NODL-NEXT: global_store_byte v[0:1], v2, off
|
|
; GFX9-NODL-NEXT: s_endpgm
|
|
;
|
|
; GFX9-DL-LABEL: udot4_CommutationInsideMAD:
|
|
; GFX9-DL: ; %bb.0: ; %entry
|
|
; GFX9-DL-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x24
|
|
; GFX9-DL-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x34
|
|
; GFX9-DL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX9-DL-NEXT: s_load_dword s2, s[4:5], 0x0
|
|
; GFX9-DL-NEXT: s_load_dword s3, s[6:7], 0x0
|
|
; GFX9-DL-NEXT: v_mov_b32_e32 v0, s0
|
|
; GFX9-DL-NEXT: v_mov_b32_e32 v1, s1
|
|
; GFX9-DL-NEXT: global_load_ubyte v2, v[0:1], off
|
|
; GFX9-DL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX9-DL-NEXT: v_mov_b32_e32 v3, s2
|
|
; GFX9-DL-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX9-DL-NEXT: v_dot4_u32_u8 v2, s3, v3, v2
|
|
; GFX9-DL-NEXT: global_store_byte v[0:1], v2, off
|
|
; GFX9-DL-NEXT: s_endpgm
|
|
;
|
|
; GFX10-DL-LABEL: udot4_CommutationInsideMAD:
|
|
; GFX10-DL: ; %bb.0: ; %entry
|
|
; GFX10-DL-NEXT: s_load_dwordx2 s[2:3], s[0:1], 0x34
|
|
; GFX10-DL-NEXT: ; implicit-def: $vcc_hi
|
|
; GFX10-DL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX10-DL-NEXT: v_mov_b32_e32 v0, s2
|
|
; GFX10-DL-NEXT: v_mov_b32_e32 v1, s3
|
|
; GFX10-DL-NEXT: s_load_dwordx4 s[0:3], s[0:1], 0x24
|
|
; GFX10-DL-NEXT: global_load_ubyte v2, v[0:1], off
|
|
; GFX10-DL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX10-DL-NEXT: s_load_dword s0, s[0:1], 0x0
|
|
; GFX10-DL-NEXT: s_load_dword s1, s[2:3], 0x0
|
|
; GFX10-DL-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
|
|
; GFX10-DL-NEXT: v_dot4_u32_u8 v2, s1, s0, v2
|
|
; GFX10-DL-NEXT: global_store_byte v[0:1], v2, off
|
|
; GFX10-DL-NEXT: s_endpgm
|
|
<4 x i8> addrspace(1)* %src2,
|
|
i8 addrspace(1)* nocapture %dst) {
|
|
entry:
|
|
%vec1 = load <4 x i8>, <4 x i8> addrspace(1)* %src1
|
|
%vec2 = load <4 x i8>, <4 x i8> addrspace(1)* %src2
|
|
|
|
%v1e0 = extractelement <4 x i8> %vec1, i64 0
|
|
%v2e0 = extractelement <4 x i8> %vec2, i64 0
|
|
%mul1 = mul nuw nsw i8 %v2e0, %v1e0
|
|
|
|
%v1e1 = extractelement <4 x i8> %vec1, i64 1
|
|
%v2e1 = extractelement <4 x i8> %vec2, i64 1
|
|
%mul2 = mul nuw nsw i8 %v2e1, %v1e1
|
|
|
|
%v1e2 = extractelement <4 x i8> %vec1, i64 2
|
|
%v2e2 = extractelement <4 x i8> %vec2, i64 2
|
|
%mul3 = mul nuw nsw i8 %v2e2, %v1e2
|
|
|
|
%v1e3 = extractelement <4 x i8> %vec1, i64 3
|
|
%v2e3 = extractelement <4 x i8> %vec2, i64 3
|
|
%mul4 = mul nuw nsw i8 %v2e3, %v1e3
|
|
|
|
%acc = load i8, i8 addrspace(1)* %dst, align 2
|
|
%mad1 = add i8 %acc, %mul1
|
|
%mad2 = add i8 %mul2, %mad1
|
|
%mad3 = add i8 %mul3, %mad2
|
|
%mad4 = add i8 %mul4, %mad3
|
|
|
|
store i8 %mad4, i8 addrspace(1)* %dst, align 2
|
|
ret void
|
|
}
|
|
|
|
; TODO: Support commutation accross the adds.
|
|
define amdgpu_kernel void @udot4_CommutationAccrossMADs(<4 x i8> addrspace(1)* %src1,
|
|
; GFX7-LABEL: udot4_CommutationAccrossMADs:
|
|
; GFX7: ; %bb.0: ; %entry
|
|
; GFX7-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x9
|
|
; GFX7-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0xd
|
|
; GFX7-NEXT: s_mov_b32 s3, 0xf000
|
|
; GFX7-NEXT: s_mov_b32 s2, -1
|
|
; GFX7-NEXT: s_movk_i32 s8, 0xff
|
|
; GFX7-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX7-NEXT: buffer_load_ubyte v0, off, s[0:3], 0
|
|
; GFX7-NEXT: s_load_dword s4, s[4:5], 0x0
|
|
; GFX7-NEXT: s_load_dword s5, s[6:7], 0x0
|
|
; GFX7-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX7-NEXT: s_and_b32 s6, s4, s8
|
|
; GFX7-NEXT: s_and_b32 s7, s5, s8
|
|
; GFX7-NEXT: s_bfe_u32 s8, s4, 0x80008
|
|
; GFX7-NEXT: s_bfe_u32 s9, s5, 0x80008
|
|
; GFX7-NEXT: v_mov_b32_e32 v1, s8
|
|
; GFX7-NEXT: s_bfe_u32 s10, s4, 0x80010
|
|
; GFX7-NEXT: v_mov_b32_e32 v2, s6
|
|
; GFX7-NEXT: s_bfe_u32 s11, s5, 0x80010
|
|
; GFX7-NEXT: s_lshr_b32 s4, s4, 24
|
|
; GFX7-NEXT: v_mov_b32_e32 v3, s10
|
|
; GFX7-NEXT: s_lshr_b32 s5, s5, 24
|
|
; GFX7-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX7-NEXT: v_mad_u32_u24 v0, s9, v1, v0
|
|
; GFX7-NEXT: v_mad_u32_u24 v0, s7, v2, v0
|
|
; GFX7-NEXT: v_mad_u32_u24 v0, s11, v3, v0
|
|
; GFX7-NEXT: v_mov_b32_e32 v1, s4
|
|
; GFX7-NEXT: v_mad_u32_u24 v0, s5, v1, v0
|
|
; GFX7-NEXT: buffer_store_byte v0, off, s[0:3], 0
|
|
; GFX7-NEXT: s_endpgm
|
|
;
|
|
; GFX8-LABEL: udot4_CommutationAccrossMADs:
|
|
; GFX8: ; %bb.0: ; %entry
|
|
; GFX8-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x24
|
|
; GFX8-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x34
|
|
; GFX8-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX8-NEXT: v_mov_b32_e32 v0, s0
|
|
; GFX8-NEXT: v_mov_b32_e32 v1, s1
|
|
; GFX8-NEXT: flat_load_ubyte v2, v[0:1]
|
|
; GFX8-NEXT: s_load_dword s1, s[4:5], 0x0
|
|
; GFX8-NEXT: s_load_dword s2, s[6:7], 0x0
|
|
; GFX8-NEXT: s_movk_i32 s0, 0xff
|
|
; GFX8-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX8-NEXT: s_bfe_u32 s4, s1, 0x80008
|
|
; GFX8-NEXT: s_and_b32 s3, s1, s0
|
|
; GFX8-NEXT: s_bfe_u32 s5, s2, 0x80008
|
|
; GFX8-NEXT: v_mov_b32_e32 v3, s4
|
|
; GFX8-NEXT: s_bfe_u32 s6, s1, 0x80010
|
|
; GFX8-NEXT: s_and_b32 s0, s2, s0
|
|
; GFX8-NEXT: v_mov_b32_e32 v4, s3
|
|
; GFX8-NEXT: s_bfe_u32 s7, s2, 0x80010
|
|
; GFX8-NEXT: s_lshr_b32 s1, s1, 24
|
|
; GFX8-NEXT: v_mov_b32_e32 v5, s6
|
|
; GFX8-NEXT: s_lshr_b32 s2, s2, 24
|
|
; GFX8-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX8-NEXT: v_mad_u32_u24 v2, s5, v3, v2
|
|
; GFX8-NEXT: v_mad_u32_u24 v2, s0, v4, v2
|
|
; GFX8-NEXT: v_mad_u32_u24 v2, s7, v5, v2
|
|
; GFX8-NEXT: v_mov_b32_e32 v3, s1
|
|
; GFX8-NEXT: v_mad_u32_u24 v2, s2, v3, v2
|
|
; GFX8-NEXT: flat_store_byte v[0:1], v2
|
|
; GFX8-NEXT: s_endpgm
|
|
;
|
|
; GFX9-NODL-LABEL: udot4_CommutationAccrossMADs:
|
|
; GFX9-NODL: ; %bb.0: ; %entry
|
|
; GFX9-NODL-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x24
|
|
; GFX9-NODL-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x34
|
|
; GFX9-NODL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX9-NODL-NEXT: v_mov_b32_e32 v0, s0
|
|
; GFX9-NODL-NEXT: v_mov_b32_e32 v1, s1
|
|
; GFX9-NODL-NEXT: global_load_ubyte v2, v[0:1], off
|
|
; GFX9-NODL-NEXT: s_load_dword s1, s[4:5], 0x0
|
|
; GFX9-NODL-NEXT: s_load_dword s2, s[6:7], 0x0
|
|
; GFX9-NODL-NEXT: s_movk_i32 s0, 0xff
|
|
; GFX9-NODL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX9-NODL-NEXT: s_bfe_u32 s4, s1, 0x80008
|
|
; GFX9-NODL-NEXT: s_and_b32 s3, s1, s0
|
|
; GFX9-NODL-NEXT: s_bfe_u32 s5, s2, 0x80008
|
|
; GFX9-NODL-NEXT: v_mov_b32_e32 v3, s4
|
|
; GFX9-NODL-NEXT: s_bfe_u32 s6, s1, 0x80010
|
|
; GFX9-NODL-NEXT: s_and_b32 s0, s2, s0
|
|
; GFX9-NODL-NEXT: v_mov_b32_e32 v4, s3
|
|
; GFX9-NODL-NEXT: s_bfe_u32 s7, s2, 0x80010
|
|
; GFX9-NODL-NEXT: s_lshr_b32 s1, s1, 24
|
|
; GFX9-NODL-NEXT: v_mov_b32_e32 v5, s6
|
|
; GFX9-NODL-NEXT: s_lshr_b32 s2, s2, 24
|
|
; GFX9-NODL-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX9-NODL-NEXT: v_mad_u32_u24 v2, s5, v3, v2
|
|
; GFX9-NODL-NEXT: v_mad_u32_u24 v2, s0, v4, v2
|
|
; GFX9-NODL-NEXT: v_mad_u32_u24 v2, s7, v5, v2
|
|
; GFX9-NODL-NEXT: v_mov_b32_e32 v3, s1
|
|
; GFX9-NODL-NEXT: v_mad_u32_u24 v2, s2, v3, v2
|
|
; GFX9-NODL-NEXT: global_store_byte v[0:1], v2, off
|
|
; GFX9-NODL-NEXT: s_endpgm
|
|
;
|
|
; GFX9-DL-LABEL: udot4_CommutationAccrossMADs:
|
|
; GFX9-DL: ; %bb.0: ; %entry
|
|
; GFX9-DL-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x24
|
|
; GFX9-DL-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x34
|
|
; GFX9-DL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX9-DL-NEXT: v_mov_b32_e32 v0, s0
|
|
; GFX9-DL-NEXT: v_mov_b32_e32 v1, s1
|
|
; GFX9-DL-NEXT: global_load_ubyte v2, v[0:1], off
|
|
; GFX9-DL-NEXT: s_load_dword s1, s[4:5], 0x0
|
|
; GFX9-DL-NEXT: s_load_dword s2, s[6:7], 0x0
|
|
; GFX9-DL-NEXT: s_movk_i32 s0, 0xff
|
|
; GFX9-DL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX9-DL-NEXT: s_bfe_u32 s4, s1, 0x80008
|
|
; GFX9-DL-NEXT: s_and_b32 s3, s1, s0
|
|
; GFX9-DL-NEXT: s_bfe_u32 s5, s2, 0x80008
|
|
; GFX9-DL-NEXT: v_mov_b32_e32 v3, s4
|
|
; GFX9-DL-NEXT: s_bfe_u32 s6, s1, 0x80010
|
|
; GFX9-DL-NEXT: s_and_b32 s0, s2, s0
|
|
; GFX9-DL-NEXT: v_mov_b32_e32 v4, s3
|
|
; GFX9-DL-NEXT: s_bfe_u32 s7, s2, 0x80010
|
|
; GFX9-DL-NEXT: s_lshr_b32 s1, s1, 24
|
|
; GFX9-DL-NEXT: v_mov_b32_e32 v5, s6
|
|
; GFX9-DL-NEXT: s_lshr_b32 s2, s2, 24
|
|
; GFX9-DL-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX9-DL-NEXT: v_mad_u32_u24 v2, s5, v3, v2
|
|
; GFX9-DL-NEXT: v_mad_u32_u24 v2, s0, v4, v2
|
|
; GFX9-DL-NEXT: v_mad_u32_u24 v2, s7, v5, v2
|
|
; GFX9-DL-NEXT: v_mov_b32_e32 v3, s1
|
|
; GFX9-DL-NEXT: v_mad_u32_u24 v2, s2, v3, v2
|
|
; GFX9-DL-NEXT: global_store_byte v[0:1], v2, off
|
|
; GFX9-DL-NEXT: s_endpgm
|
|
;
|
|
; GFX10-DL-LABEL: udot4_CommutationAccrossMADs:
|
|
; GFX10-DL: ; %bb.0: ; %entry
|
|
; GFX10-DL-NEXT: s_load_dwordx2 s[2:3], s[0:1], 0x34
|
|
; GFX10-DL-NEXT: s_movk_i32 s4, 0xff
|
|
; GFX10-DL-NEXT: ; implicit-def: $vcc_hi
|
|
; GFX10-DL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX10-DL-NEXT: v_mov_b32_e32 v0, s2
|
|
; GFX10-DL-NEXT: v_mov_b32_e32 v1, s3
|
|
; GFX10-DL-NEXT: s_load_dwordx4 s[0:3], s[0:1], 0x24
|
|
; GFX10-DL-NEXT: global_load_ubyte v2, v[0:1], off
|
|
; GFX10-DL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX10-DL-NEXT: s_load_dword s0, s[0:1], 0x0
|
|
; GFX10-DL-NEXT: s_load_dword s1, s[2:3], 0x0
|
|
; GFX10-DL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX10-DL-NEXT: s_bfe_u32 s2, s0, 0x80008
|
|
; GFX10-DL-NEXT: s_bfe_u32 s3, s1, 0x80008
|
|
; GFX10-DL-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX10-DL-NEXT: v_mad_u32_u24 v2, s3, s2, v2
|
|
; GFX10-DL-NEXT: s_and_b32 s2, s0, s4
|
|
; GFX10-DL-NEXT: s_and_b32 s3, s1, s4
|
|
; GFX10-DL-NEXT: v_mad_u32_u24 v2, s3, s2, v2
|
|
; GFX10-DL-NEXT: s_bfe_u32 s2, s0, 0x80010
|
|
; GFX10-DL-NEXT: s_bfe_u32 s3, s1, 0x80010
|
|
; GFX10-DL-NEXT: s_lshr_b32 s0, s0, 24
|
|
; GFX10-DL-NEXT: s_lshr_b32 s1, s1, 24
|
|
; GFX10-DL-NEXT: v_mad_u32_u24 v2, s3, s2, v2
|
|
; GFX10-DL-NEXT: v_mad_u32_u24 v2, s1, s0, v2
|
|
; GFX10-DL-NEXT: global_store_byte v[0:1], v2, off
|
|
; GFX10-DL-NEXT: s_endpgm
|
|
<4 x i8> addrspace(1)* %src2,
|
|
i8 addrspace(1)* nocapture %dst) {
|
|
entry:
|
|
%vec1 = load <4 x i8>, <4 x i8> addrspace(1)* %src1
|
|
%vec2 = load <4 x i8>, <4 x i8> addrspace(1)* %src2
|
|
|
|
%v1e0 = extractelement <4 x i8> %vec1, i64 0
|
|
%v2e0 = extractelement <4 x i8> %vec2, i64 0
|
|
%mul1 = mul nuw nsw i8 %v2e0, %v1e0
|
|
|
|
%v1e1 = extractelement <4 x i8> %vec1, i64 1
|
|
%v2e1 = extractelement <4 x i8> %vec2, i64 1
|
|
%mul2 = mul nuw nsw i8 %v2e1, %v1e1
|
|
|
|
%v1e2 = extractelement <4 x i8> %vec1, i64 2
|
|
%v2e2 = extractelement <4 x i8> %vec2, i64 2
|
|
%mul3 = mul nuw nsw i8 %v2e2, %v1e2
|
|
|
|
%v1e3 = extractelement <4 x i8> %vec1, i64 3
|
|
%v2e3 = extractelement <4 x i8> %vec2, i64 3
|
|
%mul4 = mul nuw nsw i8 %v2e3, %v1e3
|
|
|
|
%acc = load i8, i8 addrspace(1)* %dst, align 2
|
|
%mad1 = add i8 %acc, %mul2
|
|
%mad2 = add i8 %mad1, %mul1
|
|
%mad3 = add i8 %mad2, %mul3
|
|
%mad4 = add i8 %mad3, %mul4
|
|
|
|
store i8 %mad4, i8 addrspace(1)* %dst, align 2
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_kernel void @udot4_multiuse_mul1(<4 x i8> addrspace(1)* %src1,
|
|
; GFX7-LABEL: udot4_multiuse_mul1:
|
|
; GFX7: ; %bb.0: ; %entry
|
|
; GFX7-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x9
|
|
; GFX7-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0xd
|
|
; GFX7-NEXT: s_movk_i32 s8, 0xff
|
|
; GFX7-NEXT: s_mov_b32 s3, 0xf000
|
|
; GFX7-NEXT: s_mov_b32 s2, -1
|
|
; GFX7-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX7-NEXT: s_load_dword s4, s[4:5], 0x0
|
|
; GFX7-NEXT: s_load_dword s5, s[6:7], 0x0
|
|
; GFX7-NEXT: s_load_dword s12, s[0:1], 0x0
|
|
; GFX7-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX7-NEXT: s_and_b32 s6, s4, s8
|
|
; GFX7-NEXT: s_and_b32 s7, s5, s8
|
|
; GFX7-NEXT: s_bfe_u32 s9, s5, 0x80008
|
|
; GFX7-NEXT: v_mov_b32_e32 v0, s7
|
|
; GFX7-NEXT: v_mov_b32_e32 v1, s12
|
|
; GFX7-NEXT: s_bfe_u32 s8, s4, 0x80008
|
|
; GFX7-NEXT: v_mad_u32_u24 v1, s6, v0, v1
|
|
; GFX7-NEXT: v_mov_b32_e32 v2, s9
|
|
; GFX7-NEXT: s_bfe_u32 s11, s5, 0x80010
|
|
; GFX7-NEXT: v_mad_u32_u24 v1, s8, v2, v1
|
|
; GFX7-NEXT: s_bfe_u32 s10, s4, 0x80010
|
|
; GFX7-NEXT: v_mad_u32_u24 v0, s6, v0, v1
|
|
; GFX7-NEXT: v_mov_b32_e32 v1, s11
|
|
; GFX7-NEXT: s_lshr_b32 s5, s5, 24
|
|
; GFX7-NEXT: v_mad_u32_u24 v0, s10, v1, v0
|
|
; GFX7-NEXT: s_lshr_b32 s4, s4, 24
|
|
; GFX7-NEXT: v_mov_b32_e32 v1, s5
|
|
; GFX7-NEXT: v_mad_u32_u24 v0, s4, v1, v0
|
|
; GFX7-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; GFX7-NEXT: s_endpgm
|
|
;
|
|
; GFX8-LABEL: udot4_multiuse_mul1:
|
|
; GFX8: ; %bb.0: ; %entry
|
|
; GFX8-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x24
|
|
; GFX8-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x34
|
|
; GFX8-NEXT: s_movk_i32 s2, 0xff
|
|
; GFX8-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX8-NEXT: s_load_dword s3, s[4:5], 0x0
|
|
; GFX8-NEXT: s_load_dword s4, s[6:7], 0x0
|
|
; GFX8-NEXT: s_load_dword s10, s[0:1], 0x0
|
|
; GFX8-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX8-NEXT: s_and_b32 s5, s3, s2
|
|
; GFX8-NEXT: s_and_b32 s2, s4, s2
|
|
; GFX8-NEXT: s_bfe_u32 s7, s4, 0x80008
|
|
; GFX8-NEXT: v_mov_b32_e32 v0, s2
|
|
; GFX8-NEXT: v_mov_b32_e32 v1, s10
|
|
; GFX8-NEXT: s_bfe_u32 s6, s3, 0x80008
|
|
; GFX8-NEXT: v_mad_u32_u24 v1, s5, v0, v1
|
|
; GFX8-NEXT: v_mov_b32_e32 v2, s7
|
|
; GFX8-NEXT: s_bfe_u32 s9, s4, 0x80010
|
|
; GFX8-NEXT: v_mad_u32_u24 v1, s6, v2, v1
|
|
; GFX8-NEXT: s_bfe_u32 s8, s3, 0x80010
|
|
; GFX8-NEXT: v_mad_u32_u24 v0, s5, v0, v1
|
|
; GFX8-NEXT: v_mov_b32_e32 v1, s9
|
|
; GFX8-NEXT: s_lshr_b32 s4, s4, 24
|
|
; GFX8-NEXT: v_mad_u32_u24 v0, s8, v1, v0
|
|
; GFX8-NEXT: s_lshr_b32 s3, s3, 24
|
|
; GFX8-NEXT: v_mov_b32_e32 v1, s4
|
|
; GFX8-NEXT: v_mad_u32_u24 v2, s3, v1, v0
|
|
; GFX8-NEXT: v_mov_b32_e32 v0, s0
|
|
; GFX8-NEXT: v_mov_b32_e32 v1, s1
|
|
; GFX8-NEXT: flat_store_dword v[0:1], v2
|
|
; GFX8-NEXT: s_endpgm
|
|
;
|
|
; GFX9-NODL-LABEL: udot4_multiuse_mul1:
|
|
; GFX9-NODL: ; %bb.0: ; %entry
|
|
; GFX9-NODL-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x24
|
|
; GFX9-NODL-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x34
|
|
; GFX9-NODL-NEXT: s_movk_i32 s2, 0xff
|
|
; GFX9-NODL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX9-NODL-NEXT: s_load_dword s3, s[4:5], 0x0
|
|
; GFX9-NODL-NEXT: s_load_dword s4, s[6:7], 0x0
|
|
; GFX9-NODL-NEXT: s_load_dword s10, s[0:1], 0x0
|
|
; GFX9-NODL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX9-NODL-NEXT: s_and_b32 s5, s3, s2
|
|
; GFX9-NODL-NEXT: s_and_b32 s2, s4, s2
|
|
; GFX9-NODL-NEXT: s_bfe_u32 s7, s4, 0x80008
|
|
; GFX9-NODL-NEXT: v_mov_b32_e32 v0, s2
|
|
; GFX9-NODL-NEXT: v_mov_b32_e32 v1, s10
|
|
; GFX9-NODL-NEXT: s_bfe_u32 s6, s3, 0x80008
|
|
; GFX9-NODL-NEXT: v_mad_u32_u24 v1, s5, v0, v1
|
|
; GFX9-NODL-NEXT: v_mov_b32_e32 v2, s7
|
|
; GFX9-NODL-NEXT: s_bfe_u32 s9, s4, 0x80010
|
|
; GFX9-NODL-NEXT: v_mad_u32_u24 v1, s6, v2, v1
|
|
; GFX9-NODL-NEXT: s_bfe_u32 s8, s3, 0x80010
|
|
; GFX9-NODL-NEXT: v_mad_u32_u24 v0, s5, v0, v1
|
|
; GFX9-NODL-NEXT: v_mov_b32_e32 v1, s9
|
|
; GFX9-NODL-NEXT: s_lshr_b32 s4, s4, 24
|
|
; GFX9-NODL-NEXT: v_mad_u32_u24 v0, s8, v1, v0
|
|
; GFX9-NODL-NEXT: s_lshr_b32 s3, s3, 24
|
|
; GFX9-NODL-NEXT: v_mov_b32_e32 v1, s4
|
|
; GFX9-NODL-NEXT: v_mad_u32_u24 v2, s3, v1, v0
|
|
; GFX9-NODL-NEXT: v_mov_b32_e32 v0, s0
|
|
; GFX9-NODL-NEXT: v_mov_b32_e32 v1, s1
|
|
; GFX9-NODL-NEXT: global_store_dword v[0:1], v2, off
|
|
; GFX9-NODL-NEXT: s_endpgm
|
|
;
|
|
; GFX9-DL-LABEL: udot4_multiuse_mul1:
|
|
; GFX9-DL: ; %bb.0: ; %entry
|
|
; GFX9-DL-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x24
|
|
; GFX9-DL-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x34
|
|
; GFX9-DL-NEXT: s_movk_i32 s2, 0xff
|
|
; GFX9-DL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX9-DL-NEXT: s_load_dword s3, s[4:5], 0x0
|
|
; GFX9-DL-NEXT: s_load_dword s4, s[6:7], 0x0
|
|
; GFX9-DL-NEXT: s_load_dword s10, s[0:1], 0x0
|
|
; GFX9-DL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX9-DL-NEXT: s_and_b32 s5, s3, s2
|
|
; GFX9-DL-NEXT: s_and_b32 s2, s4, s2
|
|
; GFX9-DL-NEXT: s_bfe_u32 s7, s4, 0x80008
|
|
; GFX9-DL-NEXT: v_mov_b32_e32 v0, s2
|
|
; GFX9-DL-NEXT: v_mov_b32_e32 v1, s10
|
|
; GFX9-DL-NEXT: s_bfe_u32 s6, s3, 0x80008
|
|
; GFX9-DL-NEXT: v_mad_u32_u24 v1, s5, v0, v1
|
|
; GFX9-DL-NEXT: v_mov_b32_e32 v2, s7
|
|
; GFX9-DL-NEXT: s_bfe_u32 s9, s4, 0x80010
|
|
; GFX9-DL-NEXT: v_mad_u32_u24 v1, s6, v2, v1
|
|
; GFX9-DL-NEXT: s_bfe_u32 s8, s3, 0x80010
|
|
; GFX9-DL-NEXT: v_mad_u32_u24 v0, s5, v0, v1
|
|
; GFX9-DL-NEXT: v_mov_b32_e32 v1, s9
|
|
; GFX9-DL-NEXT: s_lshr_b32 s4, s4, 24
|
|
; GFX9-DL-NEXT: v_mad_u32_u24 v0, s8, v1, v0
|
|
; GFX9-DL-NEXT: s_lshr_b32 s3, s3, 24
|
|
; GFX9-DL-NEXT: v_mov_b32_e32 v1, s4
|
|
; GFX9-DL-NEXT: v_mad_u32_u24 v2, s3, v1, v0
|
|
; GFX9-DL-NEXT: v_mov_b32_e32 v0, s0
|
|
; GFX9-DL-NEXT: v_mov_b32_e32 v1, s1
|
|
; GFX9-DL-NEXT: global_store_dword v[0:1], v2, off
|
|
; GFX9-DL-NEXT: s_endpgm
|
|
;
|
|
; GFX10-DL-LABEL: udot4_multiuse_mul1:
|
|
; GFX10-DL: ; %bb.0: ; %entry
|
|
; GFX10-DL-NEXT: s_clause 0x1
|
|
; GFX10-DL-NEXT: s_load_dwordx2 s[4:5], s[0:1], 0x34
|
|
; GFX10-DL-NEXT: s_load_dwordx4 s[0:3], s[0:1], 0x24
|
|
; GFX10-DL-NEXT: ; implicit-def: $vcc_hi
|
|
; GFX10-DL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX10-DL-NEXT: s_load_dword s6, s[4:5], 0x0
|
|
; GFX10-DL-NEXT: s_load_dword s0, s[0:1], 0x0
|
|
; GFX10-DL-NEXT: s_load_dword s1, s[2:3], 0x0
|
|
; GFX10-DL-NEXT: s_movk_i32 s2, 0xff
|
|
; GFX10-DL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX10-DL-NEXT: v_mov_b32_e32 v0, s6
|
|
; GFX10-DL-NEXT: s_and_b32 s3, s0, s2
|
|
; GFX10-DL-NEXT: s_and_b32 s2, s1, s2
|
|
; GFX10-DL-NEXT: s_bfe_u32 s6, s0, 0x80008
|
|
; GFX10-DL-NEXT: s_bfe_u32 s7, s1, 0x80008
|
|
; GFX10-DL-NEXT: v_mad_u32_u24 v0, s3, s2, v0
|
|
; GFX10-DL-NEXT: v_mad_u32_u24 v0, s6, s7, v0
|
|
; GFX10-DL-NEXT: v_mad_u32_u24 v0, s3, s2, v0
|
|
; GFX10-DL-NEXT: s_bfe_u32 s2, s0, 0x80010
|
|
; GFX10-DL-NEXT: s_bfe_u32 s3, s1, 0x80010
|
|
; GFX10-DL-NEXT: s_lshr_b32 s0, s0, 24
|
|
; GFX10-DL-NEXT: s_lshr_b32 s1, s1, 24
|
|
; GFX10-DL-NEXT: v_mad_u32_u24 v0, s2, s3, v0
|
|
; GFX10-DL-NEXT: v_mad_u32_u24 v2, s0, s1, v0
|
|
; GFX10-DL-NEXT: v_mov_b32_e32 v0, s4
|
|
; GFX10-DL-NEXT: v_mov_b32_e32 v1, s5
|
|
; GFX10-DL-NEXT: global_store_dword v[0:1], v2, off
|
|
; GFX10-DL-NEXT: s_endpgm
|
|
<4 x i8> addrspace(1)* %src2,
|
|
i32 addrspace(1)* nocapture %dst) {
|
|
entry:
|
|
%vec1 = load <4 x i8>, <4 x i8> addrspace(1)* %src1
|
|
%vec2 = load <4 x i8>, <4 x i8> addrspace(1)* %src2
|
|
|
|
%v1e0 = extractelement <4 x i8> %vec1, i64 0
|
|
%cv1e0 = zext i8 %v1e0 to i32
|
|
%v2e0 = extractelement <4 x i8> %vec2, i64 0
|
|
%cv2e0 = zext i8 %v2e0 to i32
|
|
%mul1 = mul nuw nsw i32 %cv1e0, %cv2e0
|
|
|
|
%v1e1 = extractelement <4 x i8> %vec1, i64 1
|
|
%cv1e1 = zext i8 %v1e1 to i32
|
|
%v2e1 = extractelement <4 x i8> %vec2, i64 1
|
|
%cv2e1 = zext i8 %v2e1 to i32
|
|
%mul2 = mul nuw nsw i32 %cv1e1, %cv2e1
|
|
|
|
%v1e2 = extractelement <4 x i8> %vec1, i64 2
|
|
%cv1e2 = zext i8 %v1e2 to i32
|
|
%v2e2 = extractelement <4 x i8> %vec2, i64 2
|
|
%cv2e2 = zext i8 %v2e2 to i32
|
|
%mul3 = mul nuw nsw i32 %cv1e2, %cv2e2
|
|
|
|
%v1e3 = extractelement <4 x i8> %vec1, i64 3
|
|
%cv1e3 = zext i8 %v1e3 to i32
|
|
%v2e3 = extractelement <4 x i8> %vec2, i64 3
|
|
%cv2e3 = zext i8 %v2e3 to i32
|
|
%mul4 = mul nuw nsw i32 %cv1e3, %cv2e3
|
|
|
|
%acc = load i32, i32 addrspace(1)* %dst, align 4
|
|
%add = add i32 %mul1, %acc
|
|
%add1 = add i32 %mul2, %add
|
|
%add2 = add i32 %add1, %mul1
|
|
%add3 = add i32 %add2, %mul3
|
|
%add4 = add i32 %add3, %mul4
|
|
|
|
store i32 %add4, i32 addrspace(1)* %dst, align 4
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_kernel void @udot4_multiuse_add1(<4 x i8> addrspace(1)* %src1,
|
|
; GFX7-LABEL: udot4_multiuse_add1:
|
|
; GFX7: ; %bb.0: ; %entry
|
|
; GFX7-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x9
|
|
; GFX7-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0xd
|
|
; GFX7-NEXT: s_movk_i32 s8, 0xff
|
|
; GFX7-NEXT: s_mov_b32 s3, 0xf000
|
|
; GFX7-NEXT: s_mov_b32 s2, -1
|
|
; GFX7-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX7-NEXT: s_load_dword s4, s[4:5], 0x0
|
|
; GFX7-NEXT: s_load_dword s5, s[6:7], 0x0
|
|
; GFX7-NEXT: s_load_dword s12, s[0:1], 0x0
|
|
; GFX7-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX7-NEXT: s_and_b32 s6, s4, s8
|
|
; GFX7-NEXT: s_bfe_u32 s9, s5, 0x80008
|
|
; GFX7-NEXT: s_and_b32 s7, s5, s8
|
|
; GFX7-NEXT: s_bfe_u32 s8, s4, 0x80008
|
|
; GFX7-NEXT: v_mov_b32_e32 v0, s9
|
|
; GFX7-NEXT: v_mov_b32_e32 v1, s12
|
|
; GFX7-NEXT: v_mad_u32_u24 v0, s8, v0, v1
|
|
; GFX7-NEXT: s_bfe_u32 s11, s5, 0x80010
|
|
; GFX7-NEXT: v_mov_b32_e32 v2, s7
|
|
; GFX7-NEXT: s_bfe_u32 s10, s4, 0x80010
|
|
; GFX7-NEXT: v_add_i32_e32 v1, vcc, s12, v0
|
|
; GFX7-NEXT: v_mad_u32_u24 v0, s6, v2, v0
|
|
; GFX7-NEXT: v_mov_b32_e32 v2, s11
|
|
; GFX7-NEXT: s_lshr_b32 s5, s5, 24
|
|
; GFX7-NEXT: v_mad_u32_u24 v0, s10, v2, v0
|
|
; GFX7-NEXT: s_lshr_b32 s4, s4, 24
|
|
; GFX7-NEXT: v_mov_b32_e32 v2, s5
|
|
; GFX7-NEXT: v_mad_u32_u24 v0, s4, v2, v0
|
|
; GFX7-NEXT: v_add_i32_e32 v0, vcc, v0, v1
|
|
; GFX7-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; GFX7-NEXT: s_endpgm
|
|
;
|
|
; GFX8-LABEL: udot4_multiuse_add1:
|
|
; GFX8: ; %bb.0: ; %entry
|
|
; GFX8-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x24
|
|
; GFX8-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x34
|
|
; GFX8-NEXT: s_movk_i32 s2, 0xff
|
|
; GFX8-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX8-NEXT: s_load_dword s3, s[4:5], 0x0
|
|
; GFX8-NEXT: s_load_dword s4, s[6:7], 0x0
|
|
; GFX8-NEXT: s_load_dword s10, s[0:1], 0x0
|
|
; GFX8-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX8-NEXT: s_and_b32 s5, s3, s2
|
|
; GFX8-NEXT: s_bfe_u32 s7, s4, 0x80008
|
|
; GFX8-NEXT: s_and_b32 s2, s4, s2
|
|
; GFX8-NEXT: s_bfe_u32 s6, s3, 0x80008
|
|
; GFX8-NEXT: v_mov_b32_e32 v0, s7
|
|
; GFX8-NEXT: v_mov_b32_e32 v1, s10
|
|
; GFX8-NEXT: v_mad_u32_u24 v0, s6, v0, v1
|
|
; GFX8-NEXT: s_bfe_u32 s9, s4, 0x80010
|
|
; GFX8-NEXT: v_mov_b32_e32 v2, s2
|
|
; GFX8-NEXT: s_bfe_u32 s8, s3, 0x80010
|
|
; GFX8-NEXT: v_add_u32_e32 v1, vcc, s10, v0
|
|
; GFX8-NEXT: v_mad_u32_u24 v0, s5, v2, v0
|
|
; GFX8-NEXT: v_mov_b32_e32 v2, s9
|
|
; GFX8-NEXT: s_lshr_b32 s4, s4, 24
|
|
; GFX8-NEXT: v_mad_u32_u24 v0, s8, v2, v0
|
|
; GFX8-NEXT: s_lshr_b32 s3, s3, 24
|
|
; GFX8-NEXT: v_mov_b32_e32 v2, s4
|
|
; GFX8-NEXT: v_mad_u32_u24 v0, s3, v2, v0
|
|
; GFX8-NEXT: v_add_u32_e32 v2, vcc, v0, v1
|
|
; GFX8-NEXT: v_mov_b32_e32 v0, s0
|
|
; GFX8-NEXT: v_mov_b32_e32 v1, s1
|
|
; GFX8-NEXT: flat_store_dword v[0:1], v2
|
|
; GFX8-NEXT: s_endpgm
|
|
;
|
|
; GFX9-NODL-LABEL: udot4_multiuse_add1:
|
|
; GFX9-NODL: ; %bb.0: ; %entry
|
|
; GFX9-NODL-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x24
|
|
; GFX9-NODL-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x34
|
|
; GFX9-NODL-NEXT: s_movk_i32 s2, 0xff
|
|
; GFX9-NODL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX9-NODL-NEXT: s_load_dword s3, s[4:5], 0x0
|
|
; GFX9-NODL-NEXT: s_load_dword s4, s[6:7], 0x0
|
|
; GFX9-NODL-NEXT: s_load_dword s10, s[0:1], 0x0
|
|
; GFX9-NODL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX9-NODL-NEXT: s_and_b32 s5, s3, s2
|
|
; GFX9-NODL-NEXT: s_bfe_u32 s7, s4, 0x80008
|
|
; GFX9-NODL-NEXT: s_and_b32 s2, s4, s2
|
|
; GFX9-NODL-NEXT: s_bfe_u32 s6, s3, 0x80008
|
|
; GFX9-NODL-NEXT: v_mov_b32_e32 v0, s7
|
|
; GFX9-NODL-NEXT: v_mov_b32_e32 v1, s10
|
|
; GFX9-NODL-NEXT: v_mad_u32_u24 v0, s6, v0, v1
|
|
; GFX9-NODL-NEXT: s_bfe_u32 s9, s4, 0x80010
|
|
; GFX9-NODL-NEXT: v_mov_b32_e32 v2, s2
|
|
; GFX9-NODL-NEXT: s_bfe_u32 s8, s3, 0x80010
|
|
; GFX9-NODL-NEXT: v_add_u32_e32 v1, s10, v0
|
|
; GFX9-NODL-NEXT: v_mad_u32_u24 v0, s5, v2, v0
|
|
; GFX9-NODL-NEXT: v_mov_b32_e32 v2, s9
|
|
; GFX9-NODL-NEXT: s_lshr_b32 s4, s4, 24
|
|
; GFX9-NODL-NEXT: v_mad_u32_u24 v0, s8, v2, v0
|
|
; GFX9-NODL-NEXT: s_lshr_b32 s3, s3, 24
|
|
; GFX9-NODL-NEXT: v_mov_b32_e32 v2, s4
|
|
; GFX9-NODL-NEXT: v_mad_u32_u24 v0, s3, v2, v0
|
|
; GFX9-NODL-NEXT: v_add_u32_e32 v2, v0, v1
|
|
; GFX9-NODL-NEXT: v_mov_b32_e32 v0, s0
|
|
; GFX9-NODL-NEXT: v_mov_b32_e32 v1, s1
|
|
; GFX9-NODL-NEXT: global_store_dword v[0:1], v2, off
|
|
; GFX9-NODL-NEXT: s_endpgm
|
|
;
|
|
; GFX9-DL-LABEL: udot4_multiuse_add1:
|
|
; GFX9-DL: ; %bb.0: ; %entry
|
|
; GFX9-DL-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x24
|
|
; GFX9-DL-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x34
|
|
; GFX9-DL-NEXT: s_movk_i32 s2, 0xff
|
|
; GFX9-DL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX9-DL-NEXT: s_load_dword s3, s[4:5], 0x0
|
|
; GFX9-DL-NEXT: s_load_dword s4, s[6:7], 0x0
|
|
; GFX9-DL-NEXT: s_load_dword s10, s[0:1], 0x0
|
|
; GFX9-DL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX9-DL-NEXT: s_and_b32 s5, s3, s2
|
|
; GFX9-DL-NEXT: s_bfe_u32 s7, s4, 0x80008
|
|
; GFX9-DL-NEXT: s_and_b32 s2, s4, s2
|
|
; GFX9-DL-NEXT: s_bfe_u32 s6, s3, 0x80008
|
|
; GFX9-DL-NEXT: v_mov_b32_e32 v0, s7
|
|
; GFX9-DL-NEXT: v_mov_b32_e32 v1, s10
|
|
; GFX9-DL-NEXT: v_mad_u32_u24 v0, s6, v0, v1
|
|
; GFX9-DL-NEXT: s_bfe_u32 s9, s4, 0x80010
|
|
; GFX9-DL-NEXT: v_mov_b32_e32 v2, s2
|
|
; GFX9-DL-NEXT: s_bfe_u32 s8, s3, 0x80010
|
|
; GFX9-DL-NEXT: v_add_u32_e32 v1, s10, v0
|
|
; GFX9-DL-NEXT: v_mad_u32_u24 v0, s5, v2, v0
|
|
; GFX9-DL-NEXT: v_mov_b32_e32 v2, s9
|
|
; GFX9-DL-NEXT: s_lshr_b32 s4, s4, 24
|
|
; GFX9-DL-NEXT: v_mad_u32_u24 v0, s8, v2, v0
|
|
; GFX9-DL-NEXT: s_lshr_b32 s3, s3, 24
|
|
; GFX9-DL-NEXT: v_mov_b32_e32 v2, s4
|
|
; GFX9-DL-NEXT: v_mad_u32_u24 v0, s3, v2, v0
|
|
; GFX9-DL-NEXT: v_add_u32_e32 v2, v0, v1
|
|
; GFX9-DL-NEXT: v_mov_b32_e32 v0, s0
|
|
; GFX9-DL-NEXT: v_mov_b32_e32 v1, s1
|
|
; GFX9-DL-NEXT: global_store_dword v[0:1], v2, off
|
|
; GFX9-DL-NEXT: s_endpgm
|
|
;
|
|
; GFX10-DL-LABEL: udot4_multiuse_add1:
|
|
; GFX10-DL: ; %bb.0: ; %entry
|
|
; GFX10-DL-NEXT: s_clause 0x1
|
|
; GFX10-DL-NEXT: s_load_dwordx2 s[4:5], s[0:1], 0x34
|
|
; GFX10-DL-NEXT: s_load_dwordx4 s[0:3], s[0:1], 0x24
|
|
; GFX10-DL-NEXT: s_movk_i32 s7, 0xff
|
|
; GFX10-DL-NEXT: ; implicit-def: $vcc_hi
|
|
; GFX10-DL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX10-DL-NEXT: s_load_dword s6, s[4:5], 0x0
|
|
; GFX10-DL-NEXT: s_load_dword s0, s[0:1], 0x0
|
|
; GFX10-DL-NEXT: s_load_dword s1, s[2:3], 0x0
|
|
; GFX10-DL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX10-DL-NEXT: v_mov_b32_e32 v0, s6
|
|
; GFX10-DL-NEXT: s_bfe_u32 s2, s0, 0x80008
|
|
; GFX10-DL-NEXT: s_bfe_u32 s3, s1, 0x80008
|
|
; GFX10-DL-NEXT: v_mad_u32_u24 v0, s2, s3, v0
|
|
; GFX10-DL-NEXT: s_and_b32 s2, s0, s7
|
|
; GFX10-DL-NEXT: s_and_b32 s3, s1, s7
|
|
; GFX10-DL-NEXT: v_mad_u32_u24 v1, s2, s3, v0
|
|
; GFX10-DL-NEXT: s_bfe_u32 s2, s0, 0x80010
|
|
; GFX10-DL-NEXT: s_bfe_u32 s3, s1, 0x80010
|
|
; GFX10-DL-NEXT: s_lshr_b32 s0, s0, 24
|
|
; GFX10-DL-NEXT: s_lshr_b32 s1, s1, 24
|
|
; GFX10-DL-NEXT: v_mad_u32_u24 v1, s2, s3, v1
|
|
; GFX10-DL-NEXT: v_add_nc_u32_e32 v0, s6, v0
|
|
; GFX10-DL-NEXT: v_mad_u32_u24 v1, s0, s1, v1
|
|
; GFX10-DL-NEXT: v_add_nc_u32_e32 v2, v1, v0
|
|
; GFX10-DL-NEXT: v_mov_b32_e32 v0, s4
|
|
; GFX10-DL-NEXT: v_mov_b32_e32 v1, s5
|
|
; GFX10-DL-NEXT: global_store_dword v[0:1], v2, off
|
|
; GFX10-DL-NEXT: s_endpgm
|
|
<4 x i8> addrspace(1)* %src2,
|
|
i32 addrspace(1)* nocapture %dst) {
|
|
entry:
|
|
%vec1 = load <4 x i8>, <4 x i8> addrspace(1)* %src1
|
|
%vec2 = load <4 x i8>, <4 x i8> addrspace(1)* %src2
|
|
|
|
%v1e0 = extractelement <4 x i8> %vec1, i64 0
|
|
%cv1e0 = zext i8 %v1e0 to i32
|
|
%v2e0 = extractelement <4 x i8> %vec2, i64 0
|
|
%cv2e0 = zext i8 %v2e0 to i32
|
|
%mul1 = mul nuw nsw i32 %cv1e0, %cv2e0
|
|
|
|
%v1e1 = extractelement <4 x i8> %vec1, i64 1
|
|
%cv1e1 = zext i8 %v1e1 to i32
|
|
%v2e1 = extractelement <4 x i8> %vec2, i64 1
|
|
%cv2e1 = zext i8 %v2e1 to i32
|
|
%mul2 = mul nuw nsw i32 %cv1e1, %cv2e1
|
|
|
|
%v1e2 = extractelement <4 x i8> %vec1, i64 2
|
|
%cv1e2 = zext i8 %v1e2 to i32
|
|
%v2e2 = extractelement <4 x i8> %vec2, i64 2
|
|
%cv2e2 = zext i8 %v2e2 to i32
|
|
%mul3 = mul nuw nsw i32 %cv1e2, %cv2e2
|
|
|
|
%v1e3 = extractelement <4 x i8> %vec1, i64 3
|
|
%cv1e3 = zext i8 %v1e3 to i32
|
|
%v2e3 = extractelement <4 x i8> %vec2, i64 3
|
|
%cv2e3 = zext i8 %v2e3 to i32
|
|
%mul4 = mul nuw nsw i32 %cv1e3, %cv2e3
|
|
|
|
%acc = load i32, i32 addrspace(1)* %dst, align 4
|
|
%add1 = add i32 %mul2, %acc
|
|
%add = add i32 %add1, %acc
|
|
%add2 = add i32 %add1, %mul1
|
|
%add3 = add i32 %add2, %mul3
|
|
%add4 = add i32 %add3, %mul4
|
|
%res = add i32 %add4, %add
|
|
store i32 %res, i32 addrspace(1)* %dst, align 4
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_kernel void @notdot4_mixedtypes(<4 x i8> addrspace(1)* %src1,
|
|
; GFX7-LABEL: notdot4_mixedtypes:
|
|
; GFX7: ; %bb.0: ; %entry
|
|
; GFX7-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x9
|
|
; GFX7-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0xd
|
|
; GFX7-NEXT: s_mov_b32 s3, 0xf000
|
|
; GFX7-NEXT: s_mov_b32 s2, -1
|
|
; GFX7-NEXT: s_mov_b32 s8, 0xffff
|
|
; GFX7-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX7-NEXT: s_load_dword s6, s[6:7], 0x0
|
|
; GFX7-NEXT: buffer_load_ushort v0, off, s[0:3], 0
|
|
; GFX7-NEXT: s_load_dword s4, s[4:5], 0x0
|
|
; GFX7-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX7-NEXT: s_sext_i32_i8 s7, s6
|
|
; GFX7-NEXT: s_bfe_u32 s9, s6, 0x80008
|
|
; GFX7-NEXT: s_sext_i32_i8 s5, s4
|
|
; GFX7-NEXT: s_and_b32 s7, s7, s8
|
|
; GFX7-NEXT: s_bfe_u32 s10, s4, 0x80008
|
|
; GFX7-NEXT: v_mov_b32_e32 v1, s9
|
|
; GFX7-NEXT: s_bfe_u32 s11, s6, 0x80010
|
|
; GFX7-NEXT: s_and_b32 s5, s5, s8
|
|
; GFX7-NEXT: v_mov_b32_e32 v3, s7
|
|
; GFX7-NEXT: s_bfe_u32 s12, s4, 0x80010
|
|
; GFX7-NEXT: s_lshr_b32 s6, s6, 24
|
|
; GFX7-NEXT: v_mov_b32_e32 v2, s11
|
|
; GFX7-NEXT: s_lshr_b32 s4, s4, 24
|
|
; GFX7-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX7-NEXT: v_mad_u32_u24 v0, s10, v1, v0
|
|
; GFX7-NEXT: v_mad_u32_u24 v0, s5, v3, v0
|
|
; GFX7-NEXT: v_mad_u32_u24 v0, s12, v2, v0
|
|
; GFX7-NEXT: v_mov_b32_e32 v1, s6
|
|
; GFX7-NEXT: v_mad_u32_u24 v0, s4, v1, v0
|
|
; GFX7-NEXT: buffer_store_short v0, off, s[0:3], 0
|
|
; GFX7-NEXT: s_endpgm
|
|
;
|
|
; GFX8-LABEL: notdot4_mixedtypes:
|
|
; GFX8: ; %bb.0: ; %entry
|
|
; GFX8-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x24
|
|
; GFX8-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x34
|
|
; GFX8-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX8-NEXT: s_load_dword s2, s[6:7], 0x0
|
|
; GFX8-NEXT: v_mov_b32_e32 v0, s0
|
|
; GFX8-NEXT: v_mov_b32_e32 v1, s1
|
|
; GFX8-NEXT: flat_load_ushort v2, v[0:1]
|
|
; GFX8-NEXT: s_load_dword s0, s[4:5], 0x0
|
|
; GFX8-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX8-NEXT: s_bfe_u32 s5, s2, 0x80008
|
|
; GFX8-NEXT: s_sext_i32_i8 s3, s2
|
|
; GFX8-NEXT: v_mov_b32_e32 v3, s5
|
|
; GFX8-NEXT: s_bfe_u32 s7, s2, 0x80010
|
|
; GFX8-NEXT: s_bfe_u32 s4, s0, 0x80008
|
|
; GFX8-NEXT: s_sext_i32_i8 s1, s0
|
|
; GFX8-NEXT: v_mov_b32_e32 v4, s3
|
|
; GFX8-NEXT: s_bfe_u32 s6, s0, 0x80010
|
|
; GFX8-NEXT: s_lshr_b32 s2, s2, 24
|
|
; GFX8-NEXT: v_mov_b32_e32 v5, s7
|
|
; GFX8-NEXT: s_lshr_b32 s0, s0, 24
|
|
; GFX8-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX8-NEXT: v_mad_u32_u24 v2, s4, v3, v2
|
|
; GFX8-NEXT: v_mad_i32_i24 v2, s1, v4, v2
|
|
; GFX8-NEXT: v_mad_u32_u24 v2, s6, v5, v2
|
|
; GFX8-NEXT: v_mov_b32_e32 v3, s2
|
|
; GFX8-NEXT: v_mad_u32_u24 v2, s0, v3, v2
|
|
; GFX8-NEXT: flat_store_short v[0:1], v2
|
|
; GFX8-NEXT: s_endpgm
|
|
;
|
|
; GFX9-NODL-LABEL: notdot4_mixedtypes:
|
|
; GFX9-NODL: ; %bb.0: ; %entry
|
|
; GFX9-NODL-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x24
|
|
; GFX9-NODL-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x34
|
|
; GFX9-NODL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX9-NODL-NEXT: s_load_dword s2, s[6:7], 0x0
|
|
; GFX9-NODL-NEXT: v_mov_b32_e32 v0, s0
|
|
; GFX9-NODL-NEXT: v_mov_b32_e32 v1, s1
|
|
; GFX9-NODL-NEXT: global_load_ushort v2, v[0:1], off
|
|
; GFX9-NODL-NEXT: s_load_dword s0, s[4:5], 0x0
|
|
; GFX9-NODL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX9-NODL-NEXT: s_bfe_u32 s5, s2, 0x80008
|
|
; GFX9-NODL-NEXT: s_sext_i32_i8 s3, s2
|
|
; GFX9-NODL-NEXT: v_mov_b32_e32 v3, s5
|
|
; GFX9-NODL-NEXT: s_bfe_u32 s7, s2, 0x80010
|
|
; GFX9-NODL-NEXT: s_bfe_u32 s4, s0, 0x80008
|
|
; GFX9-NODL-NEXT: s_sext_i32_i8 s1, s0
|
|
; GFX9-NODL-NEXT: v_mov_b32_e32 v4, s3
|
|
; GFX9-NODL-NEXT: s_bfe_u32 s6, s0, 0x80010
|
|
; GFX9-NODL-NEXT: s_lshr_b32 s2, s2, 24
|
|
; GFX9-NODL-NEXT: v_mov_b32_e32 v5, s7
|
|
; GFX9-NODL-NEXT: s_lshr_b32 s0, s0, 24
|
|
; GFX9-NODL-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX9-NODL-NEXT: v_mad_u32_u24 v2, s4, v3, v2
|
|
; GFX9-NODL-NEXT: v_mad_i32_i24 v2, s1, v4, v2
|
|
; GFX9-NODL-NEXT: v_mad_u32_u24 v2, s6, v5, v2
|
|
; GFX9-NODL-NEXT: v_mov_b32_e32 v3, s2
|
|
; GFX9-NODL-NEXT: v_mad_u32_u24 v2, s0, v3, v2
|
|
; GFX9-NODL-NEXT: global_store_short v[0:1], v2, off
|
|
; GFX9-NODL-NEXT: s_endpgm
|
|
;
|
|
; GFX9-DL-LABEL: notdot4_mixedtypes:
|
|
; GFX9-DL: ; %bb.0: ; %entry
|
|
; GFX9-DL-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x24
|
|
; GFX9-DL-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x34
|
|
; GFX9-DL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX9-DL-NEXT: s_load_dword s2, s[6:7], 0x0
|
|
; GFX9-DL-NEXT: v_mov_b32_e32 v0, s0
|
|
; GFX9-DL-NEXT: v_mov_b32_e32 v1, s1
|
|
; GFX9-DL-NEXT: global_load_ushort v2, v[0:1], off
|
|
; GFX9-DL-NEXT: s_load_dword s0, s[4:5], 0x0
|
|
; GFX9-DL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX9-DL-NEXT: s_bfe_u32 s5, s2, 0x80008
|
|
; GFX9-DL-NEXT: s_sext_i32_i8 s3, s2
|
|
; GFX9-DL-NEXT: v_mov_b32_e32 v3, s5
|
|
; GFX9-DL-NEXT: s_bfe_u32 s7, s2, 0x80010
|
|
; GFX9-DL-NEXT: s_bfe_u32 s4, s0, 0x80008
|
|
; GFX9-DL-NEXT: s_sext_i32_i8 s1, s0
|
|
; GFX9-DL-NEXT: v_mov_b32_e32 v4, s3
|
|
; GFX9-DL-NEXT: s_bfe_u32 s6, s0, 0x80010
|
|
; GFX9-DL-NEXT: s_lshr_b32 s2, s2, 24
|
|
; GFX9-DL-NEXT: v_mov_b32_e32 v5, s7
|
|
; GFX9-DL-NEXT: s_lshr_b32 s0, s0, 24
|
|
; GFX9-DL-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX9-DL-NEXT: v_mad_u32_u24 v2, s4, v3, v2
|
|
; GFX9-DL-NEXT: v_mad_i32_i24 v2, s1, v4, v2
|
|
; GFX9-DL-NEXT: v_mad_u32_u24 v2, s6, v5, v2
|
|
; GFX9-DL-NEXT: v_mov_b32_e32 v3, s2
|
|
; GFX9-DL-NEXT: v_mad_u32_u24 v2, s0, v3, v2
|
|
; GFX9-DL-NEXT: global_store_short v[0:1], v2, off
|
|
; GFX9-DL-NEXT: s_endpgm
|
|
;
|
|
; GFX10-DL-LABEL: notdot4_mixedtypes:
|
|
; GFX10-DL: ; %bb.0: ; %entry
|
|
; GFX10-DL-NEXT: s_load_dwordx2 s[2:3], s[0:1], 0x34
|
|
; GFX10-DL-NEXT: ; implicit-def: $vcc_hi
|
|
; GFX10-DL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX10-DL-NEXT: v_mov_b32_e32 v0, s2
|
|
; GFX10-DL-NEXT: v_mov_b32_e32 v1, s3
|
|
; GFX10-DL-NEXT: s_load_dwordx4 s[0:3], s[0:1], 0x24
|
|
; GFX10-DL-NEXT: global_load_ushort v2, v[0:1], off
|
|
; GFX10-DL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX10-DL-NEXT: s_load_dword s0, s[0:1], 0x0
|
|
; GFX10-DL-NEXT: s_load_dword s1, s[2:3], 0x0
|
|
; GFX10-DL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX10-DL-NEXT: s_bfe_u32 s2, s0, 0x80008
|
|
; GFX10-DL-NEXT: s_bfe_u32 s3, s1, 0x80008
|
|
; GFX10-DL-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX10-DL-NEXT: v_mad_u32_u24 v2, s2, s3, v2
|
|
; GFX10-DL-NEXT: s_sext_i32_i8 s2, s0
|
|
; GFX10-DL-NEXT: s_sext_i32_i8 s3, s1
|
|
; GFX10-DL-NEXT: v_mad_i32_i24 v2, s2, s3, v2
|
|
; GFX10-DL-NEXT: s_bfe_u32 s2, s0, 0x80010
|
|
; GFX10-DL-NEXT: s_bfe_u32 s3, s1, 0x80010
|
|
; GFX10-DL-NEXT: s_lshr_b32 s0, s0, 24
|
|
; GFX10-DL-NEXT: s_lshr_b32 s1, s1, 24
|
|
; GFX10-DL-NEXT: v_mad_u32_u24 v2, s2, s3, v2
|
|
; GFX10-DL-NEXT: v_mad_u32_u24 v2, s0, s1, v2
|
|
; GFX10-DL-NEXT: global_store_short v[0:1], v2, off
|
|
; GFX10-DL-NEXT: s_endpgm
|
|
<4 x i8> addrspace(1)* %src2,
|
|
i16 addrspace(1)* nocapture %dst) {
|
|
entry:
|
|
%vec1 = load <4 x i8>, <4 x i8> addrspace(1)* %src1
|
|
%vec2 = load <4 x i8>, <4 x i8> addrspace(1)* %src2
|
|
|
|
%v1e0 = extractelement <4 x i8> %vec1, i64 0
|
|
%cv1e0 = sext i8 %v1e0 to i16
|
|
%v2e0 = extractelement <4 x i8> %vec2, i64 0
|
|
%cv2e0 = sext i8 %v2e0 to i16
|
|
%mul1 = mul nuw nsw i16 %cv1e0, %cv2e0
|
|
|
|
%v1e1 = extractelement <4 x i8> %vec1, i64 1
|
|
%cv1e1 = zext i8 %v1e1 to i16
|
|
%v2e1 = extractelement <4 x i8> %vec2, i64 1
|
|
%cv2e1 = zext i8 %v2e1 to i16
|
|
%mul2 = mul nuw nsw i16 %cv1e1, %cv2e1
|
|
|
|
%v1e2 = extractelement <4 x i8> %vec1, i64 2
|
|
%cv1e2 = zext i8 %v1e2 to i16
|
|
%v2e2 = extractelement <4 x i8> %vec2, i64 2
|
|
%cv2e2 = zext i8 %v2e2 to i16
|
|
%mul3 = mul nuw nsw i16 %cv1e2, %cv2e2
|
|
|
|
%v1e3 = extractelement <4 x i8> %vec1, i64 3
|
|
%cv1e3 = zext i8 %v1e3 to i16
|
|
%v2e3 = extractelement <4 x i8> %vec2, i64 3
|
|
%cv2e3 = zext i8 %v2e3 to i16
|
|
%mul4 = mul nuw nsw i16 %cv1e3, %cv2e3
|
|
|
|
%acc = load i16, i16 addrspace(1)* %dst, align 2
|
|
%add1 = add i16 %mul2, %acc
|
|
%add2 = add i16 %add1, %mul1
|
|
%add3 = add i16 %add2, %mul3
|
|
%add4 = add i16 %add3, %mul4
|
|
|
|
store i16 %add4, i16 addrspace(1)* %dst, align 2
|
|
ret void
|
|
}
|
|
|
|
; TODO: cleanup s_lshr_b32 and support this pattern.
|
|
define amdgpu_kernel void @udot4_acc32_vecMul(<4 x i8> addrspace(1)* %src1,
|
|
; GFX7-LABEL: udot4_acc32_vecMul:
|
|
; GFX7: ; %bb.0: ; %entry
|
|
; GFX7-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x9
|
|
; GFX7-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0xd
|
|
; GFX7-NEXT: s_movk_i32 s11, 0xff
|
|
; GFX7-NEXT: s_mov_b32 s3, 0xf000
|
|
; GFX7-NEXT: s_mov_b32 s2, -1
|
|
; GFX7-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX7-NEXT: s_load_dword s4, s[4:5], 0x0
|
|
; GFX7-NEXT: s_load_dword s5, s[6:7], 0x0
|
|
; GFX7-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX7-NEXT: s_lshr_b32 s6, s4, 24
|
|
; GFX7-NEXT: s_bfe_u32 s7, s4, 0x80008
|
|
; GFX7-NEXT: s_bfe_u32 s10, s4, 0x80010
|
|
; GFX7-NEXT: s_lshr_b32 s8, s5, 24
|
|
; GFX7-NEXT: s_bfe_u32 s9, s5, 0x80008
|
|
; GFX7-NEXT: s_bfe_u32 s12, s5, 0x80010
|
|
; GFX7-NEXT: s_and_b32 s5, s5, s11
|
|
; GFX7-NEXT: s_and_b32 s4, s4, s11
|
|
; GFX7-NEXT: s_load_dword s11, s[0:1], 0x0
|
|
; GFX7-NEXT: v_mov_b32_e32 v0, s5
|
|
; GFX7-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX7-NEXT: v_mov_b32_e32 v1, s11
|
|
; GFX7-NEXT: v_mad_u32_u24 v0, s4, v0, v1
|
|
; GFX7-NEXT: v_mov_b32_e32 v1, s9
|
|
; GFX7-NEXT: v_mad_u32_u24 v0, s7, v1, v0
|
|
; GFX7-NEXT: v_mov_b32_e32 v1, s12
|
|
; GFX7-NEXT: v_mad_u32_u24 v0, s10, v1, v0
|
|
; GFX7-NEXT: v_mov_b32_e32 v1, s8
|
|
; GFX7-NEXT: v_mad_u32_u24 v0, s6, v1, v0
|
|
; GFX7-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; GFX7-NEXT: s_endpgm
|
|
;
|
|
; GFX8-LABEL: udot4_acc32_vecMul:
|
|
; GFX8: ; %bb.0: ; %entry
|
|
; GFX8-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x24
|
|
; GFX8-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x34
|
|
; GFX8-NEXT: s_movk_i32 s2, 0xff
|
|
; GFX8-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX8-NEXT: s_load_dword s3, s[4:5], 0x0
|
|
; GFX8-NEXT: s_load_dword s4, s[6:7], 0x0
|
|
; GFX8-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX8-NEXT: s_lshr_b32 s5, s3, 24
|
|
; GFX8-NEXT: s_lshr_b32 s6, s4, 24
|
|
; GFX8-NEXT: s_bfe_u32 s7, s3, 0x80010
|
|
; GFX8-NEXT: v_lshrrev_b16_e64 v0, 8, s3
|
|
; GFX8-NEXT: s_and_b32 s3, s3, s2
|
|
; GFX8-NEXT: s_and_b32 s2, s4, s2
|
|
; GFX8-NEXT: s_bfe_u32 s8, s4, 0x80010
|
|
; GFX8-NEXT: v_lshrrev_b16_e64 v1, 8, s4
|
|
; GFX8-NEXT: s_load_dword s4, s[0:1], 0x0
|
|
; GFX8-NEXT: v_mov_b32_e32 v2, s2
|
|
; GFX8-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX8-NEXT: v_mov_b32_e32 v3, s4
|
|
; GFX8-NEXT: v_mad_u32_u24 v2, s3, v2, v3
|
|
; GFX8-NEXT: v_mad_u32_u24 v0, v0, v1, v2
|
|
; GFX8-NEXT: v_mov_b32_e32 v1, s8
|
|
; GFX8-NEXT: v_mad_u32_u24 v0, s7, v1, v0
|
|
; GFX8-NEXT: v_mov_b32_e32 v1, s6
|
|
; GFX8-NEXT: v_mad_u32_u24 v2, s5, v1, v0
|
|
; GFX8-NEXT: v_mov_b32_e32 v0, s0
|
|
; GFX8-NEXT: v_mov_b32_e32 v1, s1
|
|
; GFX8-NEXT: flat_store_dword v[0:1], v2
|
|
; GFX8-NEXT: s_endpgm
|
|
;
|
|
; GFX9-NODL-LABEL: udot4_acc32_vecMul:
|
|
; GFX9-NODL: ; %bb.0: ; %entry
|
|
; GFX9-NODL-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x24
|
|
; GFX9-NODL-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x34
|
|
; GFX9-NODL-NEXT: s_movk_i32 s2, 0xff
|
|
; GFX9-NODL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX9-NODL-NEXT: s_load_dword s3, s[4:5], 0x0
|
|
; GFX9-NODL-NEXT: s_load_dword s4, s[6:7], 0x0
|
|
; GFX9-NODL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX9-NODL-NEXT: s_lshr_b32 s5, s3, 24
|
|
; GFX9-NODL-NEXT: s_lshr_b32 s6, s4, 24
|
|
; GFX9-NODL-NEXT: s_bfe_u32 s7, s3, 0x80010
|
|
; GFX9-NODL-NEXT: v_lshrrev_b16_e64 v0, 8, s3
|
|
; GFX9-NODL-NEXT: s_and_b32 s3, s3, s2
|
|
; GFX9-NODL-NEXT: s_and_b32 s2, s4, s2
|
|
; GFX9-NODL-NEXT: s_bfe_u32 s8, s4, 0x80010
|
|
; GFX9-NODL-NEXT: v_lshrrev_b16_e64 v1, 8, s4
|
|
; GFX9-NODL-NEXT: s_load_dword s4, s[0:1], 0x0
|
|
; GFX9-NODL-NEXT: v_mov_b32_e32 v2, s2
|
|
; GFX9-NODL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX9-NODL-NEXT: v_mov_b32_e32 v3, s4
|
|
; GFX9-NODL-NEXT: v_mad_u32_u24 v2, s3, v2, v3
|
|
; GFX9-NODL-NEXT: v_mad_u32_u24 v0, v0, v1, v2
|
|
; GFX9-NODL-NEXT: v_mov_b32_e32 v1, s8
|
|
; GFX9-NODL-NEXT: v_mad_u32_u24 v0, s7, v1, v0
|
|
; GFX9-NODL-NEXT: v_mov_b32_e32 v1, s6
|
|
; GFX9-NODL-NEXT: v_mad_u32_u24 v2, s5, v1, v0
|
|
; GFX9-NODL-NEXT: v_mov_b32_e32 v0, s0
|
|
; GFX9-NODL-NEXT: v_mov_b32_e32 v1, s1
|
|
; GFX9-NODL-NEXT: global_store_dword v[0:1], v2, off
|
|
; GFX9-NODL-NEXT: s_endpgm
|
|
;
|
|
; GFX9-DL-LABEL: udot4_acc32_vecMul:
|
|
; GFX9-DL: ; %bb.0: ; %entry
|
|
; GFX9-DL-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x24
|
|
; GFX9-DL-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x34
|
|
; GFX9-DL-NEXT: s_movk_i32 s2, 0xff
|
|
; GFX9-DL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX9-DL-NEXT: s_load_dword s3, s[4:5], 0x0
|
|
; GFX9-DL-NEXT: s_load_dword s4, s[6:7], 0x0
|
|
; GFX9-DL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX9-DL-NEXT: s_lshr_b32 s5, s3, 24
|
|
; GFX9-DL-NEXT: s_lshr_b32 s6, s4, 24
|
|
; GFX9-DL-NEXT: s_bfe_u32 s7, s3, 0x80010
|
|
; GFX9-DL-NEXT: v_lshrrev_b16_e64 v0, 8, s3
|
|
; GFX9-DL-NEXT: s_and_b32 s3, s3, s2
|
|
; GFX9-DL-NEXT: s_and_b32 s2, s4, s2
|
|
; GFX9-DL-NEXT: s_bfe_u32 s8, s4, 0x80010
|
|
; GFX9-DL-NEXT: v_lshrrev_b16_e64 v1, 8, s4
|
|
; GFX9-DL-NEXT: s_load_dword s4, s[0:1], 0x0
|
|
; GFX9-DL-NEXT: v_mov_b32_e32 v2, s2
|
|
; GFX9-DL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX9-DL-NEXT: v_mov_b32_e32 v3, s4
|
|
; GFX9-DL-NEXT: v_mad_u32_u24 v2, s3, v2, v3
|
|
; GFX9-DL-NEXT: v_mad_u32_u24 v0, v0, v1, v2
|
|
; GFX9-DL-NEXT: v_mov_b32_e32 v1, s8
|
|
; GFX9-DL-NEXT: v_mad_u32_u24 v0, s7, v1, v0
|
|
; GFX9-DL-NEXT: v_mov_b32_e32 v1, s6
|
|
; GFX9-DL-NEXT: v_mad_u32_u24 v2, s5, v1, v0
|
|
; GFX9-DL-NEXT: v_mov_b32_e32 v0, s0
|
|
; GFX9-DL-NEXT: v_mov_b32_e32 v1, s1
|
|
; GFX9-DL-NEXT: global_store_dword v[0:1], v2, off
|
|
; GFX9-DL-NEXT: s_endpgm
|
|
;
|
|
; GFX10-DL-LABEL: udot4_acc32_vecMul:
|
|
; GFX10-DL: ; %bb.0: ; %entry
|
|
; GFX10-DL-NEXT: s_clause 0x1
|
|
; GFX10-DL-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x24
|
|
; GFX10-DL-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x34
|
|
; GFX10-DL-NEXT: ; implicit-def: $vcc_hi
|
|
; GFX10-DL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX10-DL-NEXT: s_load_dword s2, s[4:5], 0x0
|
|
; GFX10-DL-NEXT: s_load_dword s3, s[6:7], 0x0
|
|
; GFX10-DL-NEXT: s_load_dword s4, s[0:1], 0x0
|
|
; GFX10-DL-NEXT: s_movk_i32 s6, 0xff
|
|
; GFX10-DL-NEXT: s_mov_b32 s5, 0xffff
|
|
; GFX10-DL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX10-DL-NEXT: v_mov_b32_e32 v0, s2
|
|
; GFX10-DL-NEXT: v_mov_b32_e32 v1, s3
|
|
; GFX10-DL-NEXT: v_mov_b32_e32 v2, s4
|
|
; GFX10-DL-NEXT: s_and_b32 s4, s2, s6
|
|
; GFX10-DL-NEXT: s_and_b32 s6, s3, s6
|
|
; GFX10-DL-NEXT: v_and_b32_sdwa v0, s5, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_1
|
|
; GFX10-DL-NEXT: v_and_b32_sdwa v1, s5, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_1
|
|
; GFX10-DL-NEXT: v_mad_u32_u24 v2, s4, s6, v2
|
|
; GFX10-DL-NEXT: s_bfe_u32 s4, s2, 0x80010
|
|
; GFX10-DL-NEXT: s_bfe_u32 s5, s3, 0x80010
|
|
; GFX10-DL-NEXT: s_lshr_b32 s2, s2, 24
|
|
; GFX10-DL-NEXT: s_lshr_b32 s3, s3, 24
|
|
; GFX10-DL-NEXT: v_mad_u32_u24 v0, v0, v1, v2
|
|
; GFX10-DL-NEXT: v_mad_u32_u24 v0, s4, s5, v0
|
|
; GFX10-DL-NEXT: v_mad_u32_u24 v2, s2, s3, v0
|
|
; GFX10-DL-NEXT: v_mov_b32_e32 v0, s0
|
|
; GFX10-DL-NEXT: v_mov_b32_e32 v1, s1
|
|
; GFX10-DL-NEXT: global_store_dword v[0:1], v2, off
|
|
; GFX10-DL-NEXT: s_endpgm
|
|
<4 x i8> addrspace(1)* %src2,
|
|
i32 addrspace(1)* nocapture %dst) {
|
|
entry:
|
|
%vec1 = load <4 x i8>, <4 x i8> addrspace(1)* %src1
|
|
%vec2 = load <4 x i8>, <4 x i8> addrspace(1)* %src2
|
|
|
|
%cvec1 = zext <4 x i8> %vec1 to <4 x i32>
|
|
%cvec2 = zext <4 x i8> %vec2 to <4 x i32>
|
|
|
|
%mul = mul <4 x i32> %cvec1, %cvec2
|
|
%mul0 = extractelement <4 x i32> %mul, i64 0
|
|
%mul1 = extractelement <4 x i32> %mul, i64 1
|
|
%mul2 = extractelement <4 x i32> %mul, i64 2
|
|
%mul3 = extractelement <4 x i32> %mul, i64 3
|
|
|
|
%acc = load i32, i32 addrspace(1)* %dst, align 4
|
|
%add1 = add i32 %mul0, %acc
|
|
%add2 = add i32 %add1, %mul1
|
|
%add3 = add i32 %add2, %mul2
|
|
%add4 = add i32 %add3, %mul3
|
|
|
|
store i32 %add4, i32 addrspace(1)* %dst, align 4
|
|
ret void
|
|
}
|
|
|
|
; TODO: This pattern should be recognized.
|
|
define amdgpu_kernel void @udot4_acc16_vecMul(<4 x i8> addrspace(1)* %src1,
|
|
; GFX7-LABEL: udot4_acc16_vecMul:
|
|
; GFX7: ; %bb.0: ; %entry
|
|
; GFX7-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x9
|
|
; GFX7-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0xd
|
|
; GFX7-NEXT: s_mov_b32 s3, 0xf000
|
|
; GFX7-NEXT: s_mov_b32 s2, -1
|
|
; GFX7-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX7-NEXT: s_load_dword s6, s[6:7], 0x0
|
|
; GFX7-NEXT: buffer_load_ushort v0, off, s[0:3], 0
|
|
; GFX7-NEXT: s_load_dword s4, s[4:5], 0x0
|
|
; GFX7-NEXT: s_movk_i32 s7, 0xff
|
|
; GFX7-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX7-NEXT: s_bfe_u32 s10, s6, 0x80008
|
|
; GFX7-NEXT: s_bfe_u32 s12, s6, 0x80010
|
|
; GFX7-NEXT: s_lshr_b32 s9, s6, 24
|
|
; GFX7-NEXT: s_and_b32 s6, s6, s7
|
|
; GFX7-NEXT: s_lshr_b32 s5, s4, 24
|
|
; GFX7-NEXT: s_bfe_u32 s8, s4, 0x80008
|
|
; GFX7-NEXT: s_bfe_u32 s11, s4, 0x80010
|
|
; GFX7-NEXT: s_and_b32 s4, s4, s7
|
|
; GFX7-NEXT: v_mov_b32_e32 v1, s6
|
|
; GFX7-NEXT: v_mov_b32_e32 v2, s10
|
|
; GFX7-NEXT: v_mov_b32_e32 v3, s12
|
|
; GFX7-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX7-NEXT: v_mad_u32_u24 v0, s4, v1, v0
|
|
; GFX7-NEXT: v_mad_u32_u24 v0, s8, v2, v0
|
|
; GFX7-NEXT: v_mad_u32_u24 v0, s11, v3, v0
|
|
; GFX7-NEXT: v_mov_b32_e32 v1, s9
|
|
; GFX7-NEXT: v_mad_u32_u24 v0, s5, v1, v0
|
|
; GFX7-NEXT: buffer_store_short v0, off, s[0:3], 0
|
|
; GFX7-NEXT: s_endpgm
|
|
;
|
|
; GFX8-LABEL: udot4_acc16_vecMul:
|
|
; GFX8: ; %bb.0: ; %entry
|
|
; GFX8-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x24
|
|
; GFX8-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x34
|
|
; GFX8-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX8-NEXT: v_mov_b32_e32 v0, s0
|
|
; GFX8-NEXT: v_mov_b32_e32 v1, s1
|
|
; GFX8-NEXT: flat_load_ushort v2, v[0:1]
|
|
; GFX8-NEXT: s_load_dword s1, s[4:5], 0x0
|
|
; GFX8-NEXT: s_load_dword s2, s[6:7], 0x0
|
|
; GFX8-NEXT: s_movk_i32 s0, 0xff
|
|
; GFX8-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX8-NEXT: s_and_b32 s6, s1, s0
|
|
; GFX8-NEXT: s_and_b32 s0, s2, s0
|
|
; GFX8-NEXT: v_mov_b32_e32 v5, s0
|
|
; GFX8-NEXT: s_bfe_u32 s7, s2, 0x80010
|
|
; GFX8-NEXT: v_lshrrev_b16_e64 v3, 8, s2
|
|
; GFX8-NEXT: v_lshrrev_b16_e64 v4, 8, s1
|
|
; GFX8-NEXT: s_lshr_b32 s4, s2, 24
|
|
; GFX8-NEXT: s_bfe_u32 s5, s1, 0x80010
|
|
; GFX8-NEXT: v_mov_b32_e32 v6, s7
|
|
; GFX8-NEXT: s_lshr_b32 s3, s1, 24
|
|
; GFX8-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX8-NEXT: v_mad_u32_u24 v2, s6, v5, v2
|
|
; GFX8-NEXT: v_mad_u32_u24 v2, v4, v3, v2
|
|
; GFX8-NEXT: v_mad_u32_u24 v2, s5, v6, v2
|
|
; GFX8-NEXT: v_mov_b32_e32 v3, s4
|
|
; GFX8-NEXT: v_mad_u32_u24 v2, s3, v3, v2
|
|
; GFX8-NEXT: flat_store_short v[0:1], v2
|
|
; GFX8-NEXT: s_endpgm
|
|
;
|
|
; GFX9-NODL-LABEL: udot4_acc16_vecMul:
|
|
; GFX9-NODL: ; %bb.0: ; %entry
|
|
; GFX9-NODL-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x24
|
|
; GFX9-NODL-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x34
|
|
; GFX9-NODL-NEXT: v_mov_b32_e32 v0, 0xffff
|
|
; GFX9-NODL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX9-NODL-NEXT: s_load_dword s2, s[4:5], 0x0
|
|
; GFX9-NODL-NEXT: s_load_dword s3, s[6:7], 0x0
|
|
; GFX9-NODL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX9-NODL-NEXT: s_lshr_b32 s5, s2, 16
|
|
; GFX9-NODL-NEXT: s_lshr_b32 s7, s3, 16
|
|
; GFX9-NODL-NEXT: s_lshr_b32 s4, s2, 24
|
|
; GFX9-NODL-NEXT: v_and_b32_sdwa v4, v0, s5 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
|
|
; GFX9-NODL-NEXT: s_lshr_b32 s6, s3, 24
|
|
; GFX9-NODL-NEXT: v_and_b32_sdwa v3, v0, s7 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
|
|
; GFX9-NODL-NEXT: v_lshl_or_b32 v3, s6, 16, v3
|
|
; GFX9-NODL-NEXT: v_lshl_or_b32 v4, s4, 16, v4
|
|
; GFX9-NODL-NEXT: v_pk_mul_lo_u16 v3, v4, v3
|
|
; GFX9-NODL-NEXT: v_and_b32_sdwa v4, v0, s3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
|
|
; GFX9-NODL-NEXT: v_lshrrev_b16_e64 v2, 8, s3
|
|
; GFX9-NODL-NEXT: v_lshrrev_b16_e64 v1, 8, s2
|
|
; GFX9-NODL-NEXT: v_and_b32_sdwa v0, v0, s2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
|
|
; GFX9-NODL-NEXT: v_lshl_or_b32 v2, v2, 16, v4
|
|
; GFX9-NODL-NEXT: v_lshl_or_b32 v0, v1, 16, v0
|
|
; GFX9-NODL-NEXT: v_pk_mul_lo_u16 v2, v0, v2
|
|
; GFX9-NODL-NEXT: v_mov_b32_e32 v0, s0
|
|
; GFX9-NODL-NEXT: v_mov_b32_e32 v1, s1
|
|
; GFX9-NODL-NEXT: global_load_ushort v4, v[0:1], off
|
|
; GFX9-NODL-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX9-NODL-NEXT: v_add_u32_e32 v4, v2, v4
|
|
; GFX9-NODL-NEXT: v_add_u32_sdwa v2, v4, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
|
|
; GFX9-NODL-NEXT: v_add_u32_e32 v2, v2, v3
|
|
; GFX9-NODL-NEXT: v_add_u32_sdwa v2, v2, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
|
|
; GFX9-NODL-NEXT: global_store_short v[0:1], v2, off
|
|
; GFX9-NODL-NEXT: s_endpgm
|
|
;
|
|
; GFX9-DL-LABEL: udot4_acc16_vecMul:
|
|
; GFX9-DL: ; %bb.0: ; %entry
|
|
; GFX9-DL-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x24
|
|
; GFX9-DL-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x34
|
|
; GFX9-DL-NEXT: v_mov_b32_e32 v0, 0xffff
|
|
; GFX9-DL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX9-DL-NEXT: s_load_dword s2, s[4:5], 0x0
|
|
; GFX9-DL-NEXT: s_load_dword s3, s[6:7], 0x0
|
|
; GFX9-DL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX9-DL-NEXT: s_lshr_b32 s5, s2, 16
|
|
; GFX9-DL-NEXT: s_lshr_b32 s7, s3, 16
|
|
; GFX9-DL-NEXT: s_lshr_b32 s4, s2, 24
|
|
; GFX9-DL-NEXT: v_and_b32_sdwa v4, v0, s5 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
|
|
; GFX9-DL-NEXT: s_lshr_b32 s6, s3, 24
|
|
; GFX9-DL-NEXT: v_and_b32_sdwa v3, v0, s7 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
|
|
; GFX9-DL-NEXT: v_lshl_or_b32 v3, s6, 16, v3
|
|
; GFX9-DL-NEXT: v_lshl_or_b32 v4, s4, 16, v4
|
|
; GFX9-DL-NEXT: v_pk_mul_lo_u16 v3, v4, v3
|
|
; GFX9-DL-NEXT: v_and_b32_sdwa v4, v0, s3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
|
|
; GFX9-DL-NEXT: v_lshrrev_b16_e64 v2, 8, s3
|
|
; GFX9-DL-NEXT: v_lshrrev_b16_e64 v1, 8, s2
|
|
; GFX9-DL-NEXT: v_and_b32_sdwa v0, v0, s2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
|
|
; GFX9-DL-NEXT: v_lshl_or_b32 v2, v2, 16, v4
|
|
; GFX9-DL-NEXT: v_lshl_or_b32 v0, v1, 16, v0
|
|
; GFX9-DL-NEXT: v_pk_mul_lo_u16 v2, v0, v2
|
|
; GFX9-DL-NEXT: v_mov_b32_e32 v0, s0
|
|
; GFX9-DL-NEXT: v_mov_b32_e32 v1, s1
|
|
; GFX9-DL-NEXT: global_load_ushort v4, v[0:1], off
|
|
; GFX9-DL-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX9-DL-NEXT: v_add_u32_e32 v4, v2, v4
|
|
; GFX9-DL-NEXT: v_add_u32_sdwa v2, v4, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
|
|
; GFX9-DL-NEXT: v_add_u32_e32 v2, v2, v3
|
|
; GFX9-DL-NEXT: v_add_u32_sdwa v2, v2, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
|
|
; GFX9-DL-NEXT: global_store_short v[0:1], v2, off
|
|
; GFX9-DL-NEXT: s_endpgm
|
|
;
|
|
; GFX10-DL-LABEL: udot4_acc16_vecMul:
|
|
; GFX10-DL: ; %bb.0: ; %entry
|
|
; GFX10-DL-NEXT: s_load_dwordx2 s[2:3], s[0:1], 0x34
|
|
; GFX10-DL-NEXT: v_mov_b32_e32 v3, 0xffff
|
|
; GFX10-DL-NEXT: ; implicit-def: $vcc_hi
|
|
; GFX10-DL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX10-DL-NEXT: v_mov_b32_e32 v0, s2
|
|
; GFX10-DL-NEXT: v_mov_b32_e32 v1, s3
|
|
; GFX10-DL-NEXT: s_load_dwordx4 s[0:3], s[0:1], 0x24
|
|
; GFX10-DL-NEXT: global_load_ushort v2, v[0:1], off
|
|
; GFX10-DL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX10-DL-NEXT: s_load_dword s0, s[0:1], 0x0
|
|
; GFX10-DL-NEXT: s_load_dword s1, s[2:3], 0x0
|
|
; GFX10-DL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX10-DL-NEXT: v_lshrrev_b16_e64 v4, 8, s0
|
|
; GFX10-DL-NEXT: v_and_b32_sdwa v7, v3, s0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
|
|
; GFX10-DL-NEXT: v_lshrrev_b16_e64 v5, 8, s1
|
|
; GFX10-DL-NEXT: v_and_b32_sdwa v6, v3, s1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
|
|
; GFX10-DL-NEXT: s_lshr_b32 s2, s1, 16
|
|
; GFX10-DL-NEXT: s_lshr_b32 s3, s0, 16
|
|
; GFX10-DL-NEXT: v_lshl_or_b32 v4, v4, 16, v7
|
|
; GFX10-DL-NEXT: s_lshr_b32 s1, s1, 24
|
|
; GFX10-DL-NEXT: v_lshl_or_b32 v5, v5, 16, v6
|
|
; GFX10-DL-NEXT: v_and_b32_sdwa v6, v3, s2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
|
|
; GFX10-DL-NEXT: v_and_b32_sdwa v3, v3, s3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
|
|
; GFX10-DL-NEXT: s_lshr_b32 s0, s0, 24
|
|
; GFX10-DL-NEXT: v_pk_mul_lo_u16 v4, v4, v5
|
|
; GFX10-DL-NEXT: v_lshl_or_b32 v5, s1, 16, v6
|
|
; GFX10-DL-NEXT: v_lshl_or_b32 v3, s0, 16, v3
|
|
; GFX10-DL-NEXT: v_pk_mul_lo_u16 v3, v3, v5
|
|
; GFX10-DL-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX10-DL-NEXT: v_add_nc_u32_e32 v2, v4, v2
|
|
; GFX10-DL-NEXT: v_add_nc_u32_sdwa v2, v2, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
|
|
; GFX10-DL-NEXT: v_add_nc_u32_e32 v2, v2, v3
|
|
; GFX10-DL-NEXT: v_add_nc_u32_sdwa v2, v2, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
|
|
; GFX10-DL-NEXT: global_store_short v[0:1], v2, off
|
|
; GFX10-DL-NEXT: s_endpgm
|
|
<4 x i8> addrspace(1)* %src2,
|
|
i16 addrspace(1)* nocapture %dst) {
|
|
entry:
|
|
%vec1 = load <4 x i8>, <4 x i8> addrspace(1)* %src1
|
|
%vec2 = load <4 x i8>, <4 x i8> addrspace(1)* %src2
|
|
|
|
%cvec1 = zext <4 x i8> %vec1 to <4 x i16>
|
|
%cvec2 = zext <4 x i8> %vec2 to <4 x i16>
|
|
|
|
%mul = mul <4 x i16> %cvec1, %cvec2
|
|
%mul0 = extractelement <4 x i16> %mul, i64 0
|
|
%mul1 = extractelement <4 x i16> %mul, i64 1
|
|
%mul2 = extractelement <4 x i16> %mul, i64 2
|
|
%mul3 = extractelement <4 x i16> %mul, i64 3
|
|
|
|
%acc = load i16, i16 addrspace(1)* %dst, align 4
|
|
%add1 = add i16 %mul0, %acc
|
|
%add2 = add i16 %add1, %mul1
|
|
%add3 = add i16 %add2, %mul2
|
|
%add4 = add i16 %add3, %mul3
|
|
|
|
store i16 %add4, i16 addrspace(1)* %dst, align 4
|
|
ret void
|
|
}
|
|
|
|
; TODO: Support this pattern.
|
|
define amdgpu_kernel void @udot4_acc8_vecMul(<4 x i8> addrspace(1)* %src1,
|
|
; GFX7-LABEL: udot4_acc8_vecMul:
|
|
; GFX7: ; %bb.0: ; %entry
|
|
; GFX7-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x9
|
|
; GFX7-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0xd
|
|
; GFX7-NEXT: s_mov_b32 s3, 0xf000
|
|
; GFX7-NEXT: s_mov_b32 s2, -1
|
|
; GFX7-NEXT: s_movk_i32 s8, 0xff
|
|
; GFX7-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX7-NEXT: buffer_load_ubyte v0, off, s[0:3], 0
|
|
; GFX7-NEXT: s_load_dword s4, s[4:5], 0x0
|
|
; GFX7-NEXT: s_load_dword s5, s[6:7], 0x0
|
|
; GFX7-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX7-NEXT: s_bfe_u32 s6, s4, 0x80008
|
|
; GFX7-NEXT: s_bfe_u32 s10, s5, 0x80008
|
|
; GFX7-NEXT: s_lshr_b32 s11, s5, 16
|
|
; GFX7-NEXT: s_lshr_b32 s12, s5, 24
|
|
; GFX7-NEXT: v_mov_b32_e32 v3, s10
|
|
; GFX7-NEXT: s_lshr_b32 s7, s4, 16
|
|
; GFX7-NEXT: v_mov_b32_e32 v2, s11
|
|
; GFX7-NEXT: s_lshr_b32 s9, s4, 24
|
|
; GFX7-NEXT: v_mov_b32_e32 v1, s12
|
|
; GFX7-NEXT: s_mul_i32 s4, s4, s5
|
|
; GFX7-NEXT: v_mul_u32_u24_e32 v1, s9, v1
|
|
; GFX7-NEXT: v_mul_u32_u24_e32 v2, s7, v2
|
|
; GFX7-NEXT: v_mul_u32_u24_e32 v3, s6, v3
|
|
; GFX7-NEXT: s_and_b32 s5, s4, s8
|
|
; GFX7-NEXT: v_lshlrev_b32_e32 v1, 8, v1
|
|
; GFX7-NEXT: v_and_b32_e32 v2, s8, v2
|
|
; GFX7-NEXT: v_lshlrev_b32_e32 v3, 8, v3
|
|
; GFX7-NEXT: v_or_b32_e32 v1, v2, v1
|
|
; GFX7-NEXT: v_or_b32_e32 v2, s5, v3
|
|
; GFX7-NEXT: v_lshlrev_b32_e32 v1, 16, v1
|
|
; GFX7-NEXT: v_and_b32_e32 v2, 0xffff, v2
|
|
; GFX7-NEXT: v_or_b32_e32 v1, v2, v1
|
|
; GFX7-NEXT: v_lshrrev_b32_e32 v2, 8, v1
|
|
; GFX7-NEXT: v_lshrrev_b32_e32 v3, 16, v1
|
|
; GFX7-NEXT: v_lshrrev_b32_e32 v1, 24, v1
|
|
; GFX7-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX7-NEXT: v_add_i32_e32 v0, vcc, s4, v0
|
|
; GFX7-NEXT: v_add_i32_e32 v0, vcc, v0, v2
|
|
; GFX7-NEXT: v_add_i32_e32 v0, vcc, v0, v3
|
|
; GFX7-NEXT: v_add_i32_e32 v0, vcc, v0, v1
|
|
; GFX7-NEXT: buffer_store_byte v0, off, s[0:3], 0
|
|
; GFX7-NEXT: s_endpgm
|
|
;
|
|
; GFX8-LABEL: udot4_acc8_vecMul:
|
|
; GFX8: ; %bb.0: ; %entry
|
|
; GFX8-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x24
|
|
; GFX8-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x34
|
|
; GFX8-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX8-NEXT: v_mov_b32_e32 v0, s0
|
|
; GFX8-NEXT: v_mov_b32_e32 v1, s1
|
|
; GFX8-NEXT: flat_load_ubyte v2, v[0:1]
|
|
; GFX8-NEXT: s_movk_i32 s0, 0xff
|
|
; GFX8-NEXT: v_mov_b32_e32 v3, s0
|
|
; GFX8-NEXT: s_load_dword s0, s[4:5], 0x0
|
|
; GFX8-NEXT: s_load_dword s1, s[6:7], 0x0
|
|
; GFX8-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX8-NEXT: s_lshr_b32 s2, s0, 24
|
|
; GFX8-NEXT: s_lshr_b32 s4, s1, 24
|
|
; GFX8-NEXT: s_lshr_b32 s3, s0, 16
|
|
; GFX8-NEXT: v_mov_b32_e32 v4, s0
|
|
; GFX8-NEXT: v_mov_b32_e32 v5, s1
|
|
; GFX8-NEXT: s_mul_i32 s0, s0, s1
|
|
; GFX8-NEXT: s_lshr_b32 s5, s1, 16
|
|
; GFX8-NEXT: v_mul_u32_u24_sdwa v4, v4, v5 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:BYTE_1 src1_sel:BYTE_1
|
|
; GFX8-NEXT: v_mov_b32_e32 v5, s5
|
|
; GFX8-NEXT: v_and_b32_e32 v3, s0, v3
|
|
; GFX8-NEXT: v_mov_b32_e32 v6, s4
|
|
; GFX8-NEXT: v_mov_b32_e32 v7, s2
|
|
; GFX8-NEXT: v_or_b32_e32 v3, v3, v4
|
|
; GFX8-NEXT: v_mul_u32_u24_e32 v5, s3, v5
|
|
; GFX8-NEXT: v_mul_u32_u24_sdwa v6, v7, v6 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
|
|
; GFX8-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
|
; GFX8-NEXT: v_or_b32_sdwa v4, v5, v6 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
|
|
; GFX8-NEXT: v_or_b32_e32 v4, v3, v4
|
|
; GFX8-NEXT: v_lshrrev_b32_e32 v5, 8, v4
|
|
; GFX8-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX8-NEXT: v_add_u32_e32 v2, vcc, v2, v3
|
|
; GFX8-NEXT: v_add_u32_e32 v2, vcc, v2, v5
|
|
; GFX8-NEXT: v_add_u32_sdwa v2, vcc, v2, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
|
|
; GFX8-NEXT: v_add_u32_sdwa v2, vcc, v2, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_3
|
|
; GFX8-NEXT: flat_store_byte v[0:1], v2
|
|
; GFX8-NEXT: s_endpgm
|
|
;
|
|
; GFX9-NODL-LABEL: udot4_acc8_vecMul:
|
|
; GFX9-NODL: ; %bb.0: ; %entry
|
|
; GFX9-NODL-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x24
|
|
; GFX9-NODL-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x34
|
|
; GFX9-NODL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX9-NODL-NEXT: s_load_dword s2, s[4:5], 0x0
|
|
; GFX9-NODL-NEXT: s_load_dword s3, s[6:7], 0x0
|
|
; GFX9-NODL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX9-NODL-NEXT: s_lshr_b32 s4, s2, 16
|
|
; GFX9-NODL-NEXT: v_mov_b32_e32 v0, s3
|
|
; GFX9-NODL-NEXT: v_mov_b32_e32 v1, s3
|
|
; GFX9-NODL-NEXT: v_mul_lo_u16_e32 v0, s2, v0
|
|
; GFX9-NODL-NEXT: v_mul_lo_u16_sdwa v1, s2, v1 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:BYTE_1 src1_sel:BYTE_1
|
|
; GFX9-NODL-NEXT: v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
|
|
; GFX9-NODL-NEXT: s_lshr_b32 s6, s3, 16
|
|
; GFX9-NODL-NEXT: s_lshr_b32 s7, s3, 24
|
|
; GFX9-NODL-NEXT: v_mov_b32_e32 v1, s6
|
|
; GFX9-NODL-NEXT: v_and_b32_e32 v2, 0xffff, v0
|
|
; GFX9-NODL-NEXT: s_lshr_b32 s5, s2, 24
|
|
; GFX9-NODL-NEXT: v_mov_b32_e32 v0, s7
|
|
; GFX9-NODL-NEXT: v_mul_lo_u16_sdwa v0, s5, v0 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
|
|
; GFX9-NODL-NEXT: v_mul_lo_u16_e32 v1, s4, v1
|
|
; GFX9-NODL-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
|
|
; GFX9-NODL-NEXT: v_or_b32_e32 v3, v2, v0
|
|
; GFX9-NODL-NEXT: v_mov_b32_e32 v0, s0
|
|
; GFX9-NODL-NEXT: v_mov_b32_e32 v1, s1
|
|
; GFX9-NODL-NEXT: global_load_ubyte v5, v[0:1], off
|
|
; GFX9-NODL-NEXT: v_lshrrev_b32_e32 v4, 8, v3
|
|
; GFX9-NODL-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX9-NODL-NEXT: v_add_u32_e32 v2, v2, v5
|
|
; GFX9-NODL-NEXT: v_add_u32_e32 v2, v2, v4
|
|
; GFX9-NODL-NEXT: v_add_u32_sdwa v2, v2, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
|
|
; GFX9-NODL-NEXT: v_add_u32_sdwa v2, v2, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_3
|
|
; GFX9-NODL-NEXT: global_store_byte v[0:1], v2, off
|
|
; GFX9-NODL-NEXT: s_endpgm
|
|
;
|
|
; GFX9-DL-LABEL: udot4_acc8_vecMul:
|
|
; GFX9-DL: ; %bb.0: ; %entry
|
|
; GFX9-DL-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x24
|
|
; GFX9-DL-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x34
|
|
; GFX9-DL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX9-DL-NEXT: s_load_dword s2, s[4:5], 0x0
|
|
; GFX9-DL-NEXT: s_load_dword s3, s[6:7], 0x0
|
|
; GFX9-DL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX9-DL-NEXT: s_lshr_b32 s4, s2, 16
|
|
; GFX9-DL-NEXT: v_mov_b32_e32 v0, s3
|
|
; GFX9-DL-NEXT: v_mov_b32_e32 v1, s3
|
|
; GFX9-DL-NEXT: v_mul_lo_u16_e32 v0, s2, v0
|
|
; GFX9-DL-NEXT: v_mul_lo_u16_sdwa v1, s2, v1 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:BYTE_1 src1_sel:BYTE_1
|
|
; GFX9-DL-NEXT: v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
|
|
; GFX9-DL-NEXT: s_lshr_b32 s6, s3, 16
|
|
; GFX9-DL-NEXT: s_lshr_b32 s7, s3, 24
|
|
; GFX9-DL-NEXT: v_mov_b32_e32 v1, s6
|
|
; GFX9-DL-NEXT: v_and_b32_e32 v2, 0xffff, v0
|
|
; GFX9-DL-NEXT: s_lshr_b32 s5, s2, 24
|
|
; GFX9-DL-NEXT: v_mov_b32_e32 v0, s7
|
|
; GFX9-DL-NEXT: v_mul_lo_u16_sdwa v0, s5, v0 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
|
|
; GFX9-DL-NEXT: v_mul_lo_u16_e32 v1, s4, v1
|
|
; GFX9-DL-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
|
|
; GFX9-DL-NEXT: v_or_b32_e32 v3, v2, v0
|
|
; GFX9-DL-NEXT: v_mov_b32_e32 v0, s0
|
|
; GFX9-DL-NEXT: v_mov_b32_e32 v1, s1
|
|
; GFX9-DL-NEXT: global_load_ubyte v5, v[0:1], off
|
|
; GFX9-DL-NEXT: v_lshrrev_b32_e32 v4, 8, v3
|
|
; GFX9-DL-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX9-DL-NEXT: v_add_u32_e32 v2, v2, v5
|
|
; GFX9-DL-NEXT: v_add_u32_e32 v2, v2, v4
|
|
; GFX9-DL-NEXT: v_add_u32_sdwa v2, v2, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
|
|
; GFX9-DL-NEXT: v_add_u32_sdwa v2, v2, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_3
|
|
; GFX9-DL-NEXT: global_store_byte v[0:1], v2, off
|
|
; GFX9-DL-NEXT: s_endpgm
|
|
;
|
|
; GFX10-DL-LABEL: udot4_acc8_vecMul:
|
|
; GFX10-DL: ; %bb.0: ; %entry
|
|
; GFX10-DL-NEXT: s_load_dwordx2 s[2:3], s[0:1], 0x34
|
|
; GFX10-DL-NEXT: ; implicit-def: $vcc_hi
|
|
; GFX10-DL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX10-DL-NEXT: v_mov_b32_e32 v0, s2
|
|
; GFX10-DL-NEXT: v_mov_b32_e32 v1, s3
|
|
; GFX10-DL-NEXT: s_load_dwordx4 s[0:3], s[0:1], 0x24
|
|
; GFX10-DL-NEXT: global_load_ubyte v2, v[0:1], off
|
|
; GFX10-DL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX10-DL-NEXT: s_load_dword s0, s[0:1], 0x0
|
|
; GFX10-DL-NEXT: s_load_dword s1, s[2:3], 0x0
|
|
; GFX10-DL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX10-DL-NEXT: v_lshrrev_b16_e64 v3, 8, s0
|
|
; GFX10-DL-NEXT: v_lshrrev_b16_e64 v4, 8, s1
|
|
; GFX10-DL-NEXT: s_lshr_b32 s2, s0, 24
|
|
; GFX10-DL-NEXT: s_lshr_b32 s3, s1, 24
|
|
; GFX10-DL-NEXT: v_mul_lo_u16_e64 v5, s2, s3
|
|
; GFX10-DL-NEXT: v_mul_lo_u16_e64 v3, v3, v4
|
|
; GFX10-DL-NEXT: v_mul_lo_u16_e64 v4, s0, s1
|
|
; GFX10-DL-NEXT: s_lshr_b32 s0, s0, 16
|
|
; GFX10-DL-NEXT: s_lshr_b32 s1, s1, 16
|
|
; GFX10-DL-NEXT: v_lshlrev_b16_e64 v3, 8, v3
|
|
; GFX10-DL-NEXT: v_or_b32_sdwa v3, v4, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
|
|
; GFX10-DL-NEXT: v_lshlrev_b16_e64 v4, 8, v5
|
|
; GFX10-DL-NEXT: v_mul_lo_u16_e64 v5, s0, s1
|
|
; GFX10-DL-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
|
; GFX10-DL-NEXT: v_or_b32_sdwa v4, v5, v4 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
|
|
; GFX10-DL-NEXT: v_or_b32_e32 v4, v3, v4
|
|
; GFX10-DL-NEXT: v_lshrrev_b32_e32 v5, 8, v4
|
|
; GFX10-DL-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX10-DL-NEXT: v_add_nc_u32_e32 v2, v3, v2
|
|
; GFX10-DL-NEXT: v_add_nc_u32_e32 v2, v2, v5
|
|
; GFX10-DL-NEXT: v_add_nc_u32_sdwa v2, v2, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
|
|
; GFX10-DL-NEXT: v_add_nc_u32_sdwa v2, v2, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_3
|
|
; GFX10-DL-NEXT: global_store_byte v[0:1], v2, off
|
|
; GFX10-DL-NEXT: s_endpgm
|
|
<4 x i8> addrspace(1)* %src2,
|
|
i8 addrspace(1)* nocapture %dst) {
|
|
entry:
|
|
%vec1 = load <4 x i8>, <4 x i8> addrspace(1)* %src1
|
|
%vec2 = load <4 x i8>, <4 x i8> addrspace(1)* %src2
|
|
|
|
%mul = mul <4 x i8> %vec1, %vec2
|
|
%mul0 = extractelement <4 x i8> %mul, i64 0
|
|
%mul1 = extractelement <4 x i8> %mul, i64 1
|
|
%mul2 = extractelement <4 x i8> %mul, i64 2
|
|
%mul3 = extractelement <4 x i8> %mul, i64 3
|
|
|
|
%acc = load i8, i8 addrspace(1)* %dst, align 4
|
|
%add1 = add i8 %mul0, %acc
|
|
%add2 = add i8 %add1, %mul1
|
|
%add3 = add i8 %add2, %mul2
|
|
%add4 = add i8 %add3, %mul3
|
|
|
|
store i8 %add4, i8 addrspace(1)* %dst, align 4
|
|
ret void
|
|
}
|