[llvm] r359899 - AMDGPU: Select VOP3 form of sub

Matt Arsenault via llvm-commits llvm-commits at lists.llvm.org
Fri May 3 08:37:07 PDT 2019


Author: arsenm
Date: Fri May  3 08:37:07 2019
New Revision: 359899

URL: http://llvm.org/viewvc/llvm-project?rev=359899&view=rev
Log:
AMDGPU: Select VOP3 form of sub

The VOP3 form should always be the preferred selection form to be
shrunk later.

The r600 sub test needs to be split out because it asserts on the
arguments in the new test during the calling convention lowering.

Added:
    llvm/trunk/test/CodeGen/AMDGPU/r600.sub.ll
      - copied, changed from r359898, llvm/trunk/test/CodeGen/AMDGPU/sub.ll
Modified:
    llvm/trunk/lib/Target/AMDGPU/VOP2Instructions.td
    llvm/trunk/test/CodeGen/AMDGPU/sub.ll

Modified: llvm/trunk/lib/Target/AMDGPU/VOP2Instructions.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/AMDGPU/VOP2Instructions.td?rev=359899&r1=359898&r2=359899&view=diff
==============================================================================
--- llvm/trunk/lib/Target/AMDGPU/VOP2Instructions.td (original)
+++ llvm/trunk/lib/Target/AMDGPU/VOP2Instructions.td Fri May  3 08:37:07 2019
@@ -525,12 +525,12 @@ def : DivergentBinOp<shl, V_LSHLREV_B32_
 
 let SubtargetPredicate = HasAddNoCarryInsts in {
   def : DivergentBinOp<add, V_ADD_U32_e32>;
-  def : DivergentBinOp<sub, V_SUB_U32_e32>;
+  def : DivergentClampingBinOp<sub, V_SUB_U32_e64>;
 }
 
 let SubtargetPredicate = isGFX6GFX7GFX8GFX9, Predicates = [isGFX6GFX7GFX8GFX9] in {
 def : DivergentBinOp<add, V_ADD_I32_e32>;
-def : DivergentBinOp<sub, V_SUB_I32_e32>;
+def : DivergentClampingBinOp<sub, V_SUB_I32_e64>;
 
 def : DivergentBinOp<adde, V_ADDC_U32_e32>;
 def : DivergentBinOp<sube, V_SUBB_U32_e32>;

Copied: llvm/trunk/test/CodeGen/AMDGPU/r600.sub.ll (from r359898, llvm/trunk/test/CodeGen/AMDGPU/sub.ll)
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/AMDGPU/r600.sub.ll?p2=llvm/trunk/test/CodeGen/AMDGPU/r600.sub.ll&p1=llvm/trunk/test/CodeGen/AMDGPU/sub.ll&r1=359898&r2=359899&rev=359899&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/AMDGPU/sub.ll (original)
+++ llvm/trunk/test/CodeGen/AMDGPU/r600.sub.ll Fri May  3 08:37:07 2019
@@ -1,14 +1,8 @@
-; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GCN,SI,FUNC %s
-; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=fiji -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GCN,VI,GFX89,FUNC %s
-; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=gfx900 -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GCN,GFX9,GFX89,FUNC %s
 ; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=redwood -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=EG,FUNC %s
 
 declare i32 @llvm.r600.read.tidig.x() readnone
 
 ; FUNC-LABEL: {{^}}s_sub_i32:
-; GCN: s_load_dwordx2
-; GCN: s_load_dwordx2 s{{\[}}[[A:[0-9]+]]:[[B:[0-9]+]]{{\]}}
-; GCN: s_sub_i32 s{{[0-9]+}}, s[[A]], s[[B]]
 define amdgpu_kernel void @s_sub_i32(i32 addrspace(1)* %out, i32 %a, i32 %b) {
   %result = sub i32 %a, %b
   store i32 %result, i32 addrspace(1)* %out
@@ -16,8 +10,6 @@ define amdgpu_kernel void @s_sub_i32(i32
 }
 
 ; FUNC-LABEL: {{^}}s_sub_imm_i32:
-; GCN: s_load_dword [[A:s[0-9]+]]
-; GCN: s_sub_i32 s{{[0-9]+}}, 0x4d2, [[A]]
 define amdgpu_kernel void @s_sub_imm_i32(i32 addrspace(1)* %out, i32 %a) {
   %result = sub i32 1234, %a
   store i32 %result, i32 addrspace(1)* %out
@@ -26,9 +18,6 @@ define amdgpu_kernel void @s_sub_imm_i32
 
 ; FUNC-LABEL: {{^}}test_sub_i32:
 ; EG: SUB_INT {{\** *}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
-
-; SI: v_subrev_i32_e32 v{{[0-9]+, vcc, v[0-9]+, v[0-9]+}}
-; GFX9: v_sub_u32_e32 v{{[0-9]+, v[0-9]+, v[0-9]+}}
 define amdgpu_kernel void @test_sub_i32(i32 addrspace(1)* %out, i32 addrspace(1)* %in) {
   %b_ptr = getelementptr i32, i32 addrspace(1)* %in, i32 1
   %a = load i32, i32 addrspace(1)* %in
@@ -40,9 +29,6 @@ define amdgpu_kernel void @test_sub_i32(
 
 ; FUNC-LABEL: {{^}}test_sub_imm_i32:
 ; EG: SUB_INT
-
-; SI: v_sub_i32_e32 v{{[0-9]+}}, vcc, 0x7b, v{{[0-9]+}}
-; GFX9: v_sub_u32_e32 v{{[0-9]+}}, 0x7b, v{{[0-9]+}}
 define amdgpu_kernel void @test_sub_imm_i32(i32 addrspace(1)* %out, i32 addrspace(1)* %in) {
   %a = load i32, i32 addrspace(1)* %in
   %result = sub i32 123, %a
@@ -53,12 +39,6 @@ define amdgpu_kernel void @test_sub_imm_
 ; FUNC-LABEL: {{^}}test_sub_v2i32:
 ; EG: SUB_INT {{\** *}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
 ; EG: SUB_INT {{\** *}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
-
-; SI: v_sub_i32_e32 v{{[0-9]+, vcc, v[0-9]+, v[0-9]+}}
-; SI: v_sub_i32_e32 v{{[0-9]+, vcc, v[0-9]+, v[0-9]+}}
-
-; GFX9: v_sub_u32_e32 v{{[0-9]+, v[0-9]+, v[0-9]+}}
-; GFX9: v_sub_u32_e32 v{{[0-9]+, v[0-9]+, v[0-9]+}}
 define amdgpu_kernel void @test_sub_v2i32(<2 x i32> addrspace(1)* %out, <2 x i32> addrspace(1)* %in) {
   %b_ptr = getelementptr <2 x i32>, <2 x i32> addrspace(1)* %in, i32 1
   %a = load <2 x i32>, <2 x i32> addrspace(1) * %in
@@ -73,16 +53,6 @@ define amdgpu_kernel void @test_sub_v2i3
 ; EG: SUB_INT {{\** *}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
 ; EG: SUB_INT {{\** *}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
 ; EG: SUB_INT {{\** *}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
-
-; SI: v_sub_i32_e32 v{{[0-9]+, vcc, v[0-9]+, v[0-9]+}}
-; SI: v_sub_i32_e32 v{{[0-9]+, vcc, v[0-9]+, v[0-9]+}}
-; SI: v_sub_i32_e32 v{{[0-9]+, vcc, v[0-9]+, v[0-9]+}}
-; SI: v_sub_i32_e32 v{{[0-9]+, vcc, v[0-9]+, v[0-9]+}}
-
-; GFX9: v_sub_u32_e32 v{{[0-9]+, v[0-9]+, v[0-9]+}}
-; GFX9: v_sub_u32_e32 v{{[0-9]+, v[0-9]+, v[0-9]+}}
-; GFX9: v_sub_u32_e32 v{{[0-9]+, v[0-9]+, v[0-9]+}}
-; GFX9: v_sub_u32_e32 v{{[0-9]+, v[0-9]+, v[0-9]+}}
 define amdgpu_kernel void @test_sub_v4i32(<4 x i32> addrspace(1)* %out, <4 x i32> addrspace(1)* %in) {
   %b_ptr = getelementptr <4 x i32>, <4 x i32> addrspace(1)* %in, i32 1
   %a = load <4 x i32>, <4 x i32> addrspace(1) * %in
@@ -93,8 +63,6 @@ define amdgpu_kernel void @test_sub_v4i3
 }
 
 ; FUNC-LABEL: {{^}}test_sub_i16:
-; SI: v_sub_i32_e32 v{{[0-9]+}}, vcc,
-; GFX89: v_sub_u16_e32 v{{[0-9]+, v[0-9]+, v[0-9]+}}
 define amdgpu_kernel void @test_sub_i16(i16 addrspace(1)* %out, i16 addrspace(1)* %in) {
   %tid = call i32 @llvm.r600.read.tidig.x()
   %gep = getelementptr i16, i16 addrspace(1)* %in, i32 %tid
@@ -107,10 +75,6 @@ define amdgpu_kernel void @test_sub_i16(
 }
 
 ; FUNC-LABEL: {{^}}test_sub_v2i16:
-; VI: v_sub_u16_e32 v{{[0-9]+, v[0-9]+, v[0-9]+}}
-; VI: v_sub_u16_sdwa v{{[0-9]+, v[0-9]+, v[0-9]+}}
-
-; GFX9: v_pk_sub_i16
 define amdgpu_kernel void @test_sub_v2i16(<2 x i16> addrspace(1)* %out, <2 x i16> addrspace(1)* %in) {
   %tid = call i32 @llvm.r600.read.tidig.x()
   %gep = getelementptr <2 x i16>, <2 x i16> addrspace(1)* %in, i32 %tid
@@ -123,13 +87,6 @@ define amdgpu_kernel void @test_sub_v2i1
 }
 
 ; FUNC-LABEL: {{^}}test_sub_v4i16:
-; VI: v_sub_u16_e32 v{{[0-9]+, v[0-9]+, v[0-9]+}}
-; VI: v_sub_u16_sdwa v{{[0-9]+, v[0-9]+, v[0-9]+}}
-; VI: v_sub_u16_e32 v{{[0-9]+, v[0-9]+, v[0-9]+}}
-; VI: v_sub_u16_sdwa v{{[0-9]+, v[0-9]+, v[0-9]+}}
-
-; GFX9: v_pk_sub_i16
-; GFX9: v_pk_sub_i16
 define amdgpu_kernel void @test_sub_v4i16(<4 x i16> addrspace(1)* %out, <4 x i16> addrspace(1)* %in) {
   %tid = call i32 @llvm.r600.read.tidig.x()
   %gep = getelementptr <4 x i16>, <4 x i16> addrspace(1)* %in, i32 %tid
@@ -142,9 +99,6 @@ define amdgpu_kernel void @test_sub_v4i1
 }
 
 ; FUNC-LABEL: {{^}}s_sub_i64:
-; GCN: s_sub_u32
-; GCN: s_subb_u32
-
 ; EG: MEM_RAT_CACHELESS STORE_RAW T{{[0-9]+}}.XY
 ; EG-DAG: SUB_INT {{[* ]*}}
 ; EG-DAG: SUBB_UINT
@@ -157,15 +111,6 @@ define amdgpu_kernel void @s_sub_i64(i64
 }
 
 ; FUNC-LABEL: {{^}}v_sub_i64:
-; SI: v_sub_i32_e32
-; SI: v_subb_u32_e32
-
-; VI: v_sub_u32_e32
-; VI: v_subb_u32_e32
-
-; GFX9: v_sub_co_u32_e32
-; GFX9: v_subb_co_u32_e32
-
 ; EG: MEM_RAT_CACHELESS STORE_RAW T{{[0-9]+}}.XY
 ; EG-DAG: SUB_INT {{[* ]*}}
 ; EG-DAG: SUBB_UINT
@@ -183,20 +128,6 @@ define amdgpu_kernel void @v_sub_i64(i64
 }
 
 ; FUNC-LABEL: {{^}}v_test_sub_v2i64:
-; SI: v_sub_i32_e32 v{{[0-9]+}}, vcc,
-; SI: v_subb_u32_e32 v{{[0-9]+}}, vcc,
-; SI: v_sub_i32_e32 v{{[0-9]+}}, vcc,
-; SI: v_subb_u32_e32 v{{[0-9]+}}, vcc,
-
-; VI: v_sub_u32_e32 v{{[0-9]+}}, vcc,
-; VI: v_subb_u32_e32 v{{[0-9]+}}, vcc,
-; VI: v_sub_u32_e32 v{{[0-9]+}}, vcc,
-; VI: v_subb_u32_e32 v{{[0-9]+}}, vcc,
-
-; GFX9: v_sub_co_u32_e32 v{{[0-9]+}}, vcc,
-; GFX9: v_subb_co_u32_e32 v{{[0-9]+}}, vcc,
-; GFX9: v_sub_co_u32_e32 v{{[0-9]+}}, vcc,
-; GFX9: v_subb_co_u32_e32 v{{[0-9]+}}, vcc,
 define amdgpu_kernel void @v_test_sub_v2i64(<2 x i64> addrspace(1)* %out, <2 x i64> addrspace(1)* noalias %inA, <2 x i64> addrspace(1)* noalias %inB) {
   %tid = call i32 @llvm.r600.read.tidig.x() readnone
   %a_ptr = getelementptr <2 x i64>, <2 x i64> addrspace(1)* %inA, i32 %tid
@@ -209,32 +140,6 @@ define amdgpu_kernel void @v_test_sub_v2
 }
 
 ; FUNC-LABEL: {{^}}v_test_sub_v4i64:
-; SI: v_sub_i32_e32 v{{[0-9]+}}, vcc,
-; SI: v_subb_u32_e32 v{{[0-9]+}}, vcc,
-; SI: v_sub_i32_e32 v{{[0-9]+}}, vcc,
-; SI: v_subb_u32_e32 v{{[0-9]+}}, vcc,
-; SI: v_sub_i32_e32 v{{[0-9]+}}, vcc,
-; SI: v_subb_u32_e32 v{{[0-9]+}}, vcc,
-; SI: v_sub_i32_e32 v{{[0-9]+}}, vcc,
-; SI: v_subb_u32_e32 v{{[0-9]+}}, vcc,
-
-; VI: v_sub_u32_e32 v{{[0-9]+}}, vcc,
-; VI: v_subb_u32_e32 v{{[0-9]+}}, vcc,
-; VI: v_sub_u32_e32 v{{[0-9]+}}, vcc,
-; VI: v_subb_u32_e32 v{{[0-9]+}}, vcc,
-; VI: v_sub_u32_e32 v{{[0-9]+}}, vcc,
-; VI: v_subb_u32_e32 v{{[0-9]+}}, vcc,
-; VI: v_sub_u32_e32 v{{[0-9]+}}, vcc,
-; VI: v_subb_u32_e32 v{{[0-9]+}}, vcc,
-
-; GFX9: v_sub_co_u32_e32 v{{[0-9]+}}, vcc,
-; GFX9: v_subb_co_u32_e32 v{{[0-9]+}}, vcc,
-; GFX9: v_sub_co_u32_e32 v{{[0-9]+}}, vcc,
-; GFX9: v_subb_co_u32_e32 v{{[0-9]+}}, vcc,
-; GFX9: v_sub_co_u32_e32 v{{[0-9]+}}, vcc,
-; GFX9: v_subb_co_u32_e32 v{{[0-9]+}}, vcc,
-; GFX9: v_sub_co_u32_e32 v{{[0-9]+}}, vcc,
-; GFX9: v_subb_co_u32_e32 v{{[0-9]+}}, vcc,
 define amdgpu_kernel void @v_test_sub_v4i64(<4 x i64> addrspace(1)* %out, <4 x i64> addrspace(1)* noalias %inA, <4 x i64> addrspace(1)* noalias %inB) {
   %tid = call i32 @llvm.r600.read.tidig.x() readnone
   %a_ptr = getelementptr <4 x i64>, <4 x i64> addrspace(1)* %inA, i32 %tid

Modified: llvm/trunk/test/CodeGen/AMDGPU/sub.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/AMDGPU/sub.ll?rev=359899&r1=359898&r2=359899&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/AMDGPU/sub.ll (original)
+++ llvm/trunk/test/CodeGen/AMDGPU/sub.ll Fri May  3 08:37:07 2019
@@ -1,11 +1,10 @@
-; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GCN,SI,FUNC %s
-; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=fiji -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GCN,VI,GFX89,FUNC %s
-; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=gfx900 -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GCN,GFX9,GFX89,FUNC %s
-; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=redwood -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=EG,FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GCN,SI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=fiji -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GCN,VI,GFX89 %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=gfx900 -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GCN,GFX9,GFX89 %s
 
-declare i32 @llvm.r600.read.tidig.x() readnone
+declare i32 @llvm.amdgcn.workitem.id.x() nounwind readnone speculatable
 
-; FUNC-LABEL: {{^}}s_sub_i32:
+; GCN-LABEL: {{^}}s_sub_i32:
 ; GCN: s_load_dwordx2
 ; GCN: s_load_dwordx2 s{{\[}}[[A:[0-9]+]]:[[B:[0-9]+]]{{\]}}
 ; GCN: s_sub_i32 s{{[0-9]+}}, s[[A]], s[[B]]
@@ -15,7 +14,7 @@ define amdgpu_kernel void @s_sub_i32(i32
   ret void
 }
 
-; FUNC-LABEL: {{^}}s_sub_imm_i32:
+; GCN-LABEL: {{^}}s_sub_imm_i32:
 ; GCN: s_load_dword [[A:s[0-9]+]]
 ; GCN: s_sub_i32 s{{[0-9]+}}, 0x4d2, [[A]]
 define amdgpu_kernel void @s_sub_imm_i32(i32 addrspace(1)* %out, i32 %a) {
@@ -24,9 +23,7 @@ define amdgpu_kernel void @s_sub_imm_i32
   ret void
 }
 
-; FUNC-LABEL: {{^}}test_sub_i32:
-; EG: SUB_INT {{\** *}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
-
+; GCN-LABEL: {{^}}test_sub_i32:
 ; SI: v_subrev_i32_e32 v{{[0-9]+, vcc, v[0-9]+, v[0-9]+}}
 ; GFX9: v_sub_u32_e32 v{{[0-9]+, v[0-9]+, v[0-9]+}}
 define amdgpu_kernel void @test_sub_i32(i32 addrspace(1)* %out, i32 addrspace(1)* %in) {
@@ -38,9 +35,7 @@ define amdgpu_kernel void @test_sub_i32(
   ret void
 }
 
-; FUNC-LABEL: {{^}}test_sub_imm_i32:
-; EG: SUB_INT
-
+; GCN-LABEL: {{^}}test_sub_imm_i32:
 ; SI: v_sub_i32_e32 v{{[0-9]+}}, vcc, 0x7b, v{{[0-9]+}}
 ; GFX9: v_sub_u32_e32 v{{[0-9]+}}, 0x7b, v{{[0-9]+}}
 define amdgpu_kernel void @test_sub_imm_i32(i32 addrspace(1)* %out, i32 addrspace(1)* %in) {
@@ -50,10 +45,7 @@ define amdgpu_kernel void @test_sub_imm_
   ret void
 }
 
-; FUNC-LABEL: {{^}}test_sub_v2i32:
-; EG: SUB_INT {{\** *}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
-; EG: SUB_INT {{\** *}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
-
+; GCN-LABEL: {{^}}test_sub_v2i32:
 ; SI: v_sub_i32_e32 v{{[0-9]+, vcc, v[0-9]+, v[0-9]+}}
 ; SI: v_sub_i32_e32 v{{[0-9]+, vcc, v[0-9]+, v[0-9]+}}
 
@@ -68,12 +60,7 @@ define amdgpu_kernel void @test_sub_v2i3
   ret void
 }
 
-; FUNC-LABEL: {{^}}test_sub_v4i32:
-; EG: SUB_INT {{\** *}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
-; EG: SUB_INT {{\** *}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
-; EG: SUB_INT {{\** *}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
-; EG: SUB_INT {{\** *}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
-
+; GCN-LABEL: {{^}}test_sub_v4i32:
 ; SI: v_sub_i32_e32 v{{[0-9]+, vcc, v[0-9]+, v[0-9]+}}
 ; SI: v_sub_i32_e32 v{{[0-9]+, vcc, v[0-9]+, v[0-9]+}}
 ; SI: v_sub_i32_e32 v{{[0-9]+, vcc, v[0-9]+, v[0-9]+}}
@@ -92,11 +79,11 @@ define amdgpu_kernel void @test_sub_v4i3
   ret void
 }
 
-; FUNC-LABEL: {{^}}test_sub_i16:
+; GCN-LABEL: {{^}}test_sub_i16:
 ; SI: v_sub_i32_e32 v{{[0-9]+}}, vcc,
 ; GFX89: v_sub_u16_e32 v{{[0-9]+, v[0-9]+, v[0-9]+}}
 define amdgpu_kernel void @test_sub_i16(i16 addrspace(1)* %out, i16 addrspace(1)* %in) {
-  %tid = call i32 @llvm.r600.read.tidig.x()
+  %tid = call i32 @llvm.amdgcn.workitem.id.x()
   %gep = getelementptr i16, i16 addrspace(1)* %in, i32 %tid
   %b_ptr = getelementptr i16, i16 addrspace(1)* %gep, i32 1
   %a = load volatile i16, i16 addrspace(1)* %gep
@@ -106,13 +93,13 @@ define amdgpu_kernel void @test_sub_i16(
   ret void
 }
 
-; FUNC-LABEL: {{^}}test_sub_v2i16:
+; GCN-LABEL: {{^}}test_sub_v2i16:
 ; VI: v_sub_u16_e32 v{{[0-9]+, v[0-9]+, v[0-9]+}}
 ; VI: v_sub_u16_sdwa v{{[0-9]+, v[0-9]+, v[0-9]+}}
 
 ; GFX9: v_pk_sub_i16
 define amdgpu_kernel void @test_sub_v2i16(<2 x i16> addrspace(1)* %out, <2 x i16> addrspace(1)* %in) {
-  %tid = call i32 @llvm.r600.read.tidig.x()
+  %tid = call i32 @llvm.amdgcn.workitem.id.x()
   %gep = getelementptr <2 x i16>, <2 x i16> addrspace(1)* %in, i32 %tid
   %b_ptr = getelementptr <2 x i16>, <2 x i16> addrspace(1)* %gep, i16 1
   %a = load <2 x i16>, <2 x i16> addrspace(1)* %gep
@@ -122,7 +109,7 @@ define amdgpu_kernel void @test_sub_v2i1
   ret void
 }
 
-; FUNC-LABEL: {{^}}test_sub_v4i16:
+; GCN-LABEL: {{^}}test_sub_v4i16:
 ; VI: v_sub_u16_e32 v{{[0-9]+, v[0-9]+, v[0-9]+}}
 ; VI: v_sub_u16_sdwa v{{[0-9]+, v[0-9]+, v[0-9]+}}
 ; VI: v_sub_u16_e32 v{{[0-9]+, v[0-9]+, v[0-9]+}}
@@ -131,7 +118,7 @@ define amdgpu_kernel void @test_sub_v2i1
 ; GFX9: v_pk_sub_i16
 ; GFX9: v_pk_sub_i16
 define amdgpu_kernel void @test_sub_v4i16(<4 x i16> addrspace(1)* %out, <4 x i16> addrspace(1)* %in) {
-  %tid = call i32 @llvm.r600.read.tidig.x()
+  %tid = call i32 @llvm.amdgcn.workitem.id.x()
   %gep = getelementptr <4 x i16>, <4 x i16> addrspace(1)* %in, i32 %tid
   %b_ptr = getelementptr <4 x i16>, <4 x i16> addrspace(1)* %gep, i16 1
   %a = load <4 x i16>, <4 x i16> addrspace(1) * %gep
@@ -141,22 +128,16 @@ define amdgpu_kernel void @test_sub_v4i1
   ret void
 }
 
-; FUNC-LABEL: {{^}}s_sub_i64:
+; GCN-LABEL: {{^}}s_sub_i64:
 ; GCN: s_sub_u32
 ; GCN: s_subb_u32
-
-; EG: MEM_RAT_CACHELESS STORE_RAW T{{[0-9]+}}.XY
-; EG-DAG: SUB_INT {{[* ]*}}
-; EG-DAG: SUBB_UINT
-; EG-DAG: SUB_INT
-; EG-DAG: SUB_INT {{[* ]*}}
 define amdgpu_kernel void @s_sub_i64(i64 addrspace(1)* noalias %out, i64 %a, i64 %b) nounwind {
   %result = sub i64 %a, %b
   store i64 %result, i64 addrspace(1)* %out, align 8
   ret void
 }
 
-; FUNC-LABEL: {{^}}v_sub_i64:
+; GCN-LABEL: {{^}}v_sub_i64:
 ; SI: v_sub_i32_e32
 ; SI: v_subb_u32_e32
 
@@ -165,14 +146,8 @@ define amdgpu_kernel void @s_sub_i64(i64
 
 ; GFX9: v_sub_co_u32_e32
 ; GFX9: v_subb_co_u32_e32
-
-; EG: MEM_RAT_CACHELESS STORE_RAW T{{[0-9]+}}.XY
-; EG-DAG: SUB_INT {{[* ]*}}
-; EG-DAG: SUBB_UINT
-; EG-DAG: SUB_INT
-; EG-DAG: SUB_INT {{[* ]*}}
 define amdgpu_kernel void @v_sub_i64(i64 addrspace(1)* noalias %out, i64 addrspace(1)* noalias %inA, i64 addrspace(1)* noalias %inB) nounwind {
-  %tid = call i32 @llvm.r600.read.tidig.x() readnone
+  %tid = call i32 @llvm.amdgcn.workitem.id.x() readnone
   %a_ptr = getelementptr i64, i64 addrspace(1)* %inA, i32 %tid
   %b_ptr = getelementptr i64, i64 addrspace(1)* %inB, i32 %tid
   %a = load i64, i64 addrspace(1)* %a_ptr
@@ -182,7 +157,7 @@ define amdgpu_kernel void @v_sub_i64(i64
   ret void
 }
 
-; FUNC-LABEL: {{^}}v_test_sub_v2i64:
+; GCN-LABEL: {{^}}v_test_sub_v2i64:
 ; SI: v_sub_i32_e32 v{{[0-9]+}}, vcc,
 ; SI: v_subb_u32_e32 v{{[0-9]+}}, vcc,
 ; SI: v_sub_i32_e32 v{{[0-9]+}}, vcc,
@@ -198,7 +173,7 @@ define amdgpu_kernel void @v_sub_i64(i64
 ; GFX9: v_sub_co_u32_e32 v{{[0-9]+}}, vcc,
 ; GFX9: v_subb_co_u32_e32 v{{[0-9]+}}, vcc,
 define amdgpu_kernel void @v_test_sub_v2i64(<2 x i64> addrspace(1)* %out, <2 x i64> addrspace(1)* noalias %inA, <2 x i64> addrspace(1)* noalias %inB) {
-  %tid = call i32 @llvm.r600.read.tidig.x() readnone
+  %tid = call i32 @llvm.amdgcn.workitem.id.x() readnone
   %a_ptr = getelementptr <2 x i64>, <2 x i64> addrspace(1)* %inA, i32 %tid
   %b_ptr = getelementptr <2 x i64>, <2 x i64> addrspace(1)* %inB, i32 %tid
   %a = load <2 x i64>, <2 x i64> addrspace(1)* %a_ptr
@@ -208,7 +183,7 @@ define amdgpu_kernel void @v_test_sub_v2
   ret void
 }
 
-; FUNC-LABEL: {{^}}v_test_sub_v4i64:
+; GCN-LABEL: {{^}}v_test_sub_v4i64:
 ; SI: v_sub_i32_e32 v{{[0-9]+}}, vcc,
 ; SI: v_subb_u32_e32 v{{[0-9]+}}, vcc,
 ; SI: v_sub_i32_e32 v{{[0-9]+}}, vcc,
@@ -236,7 +211,7 @@ define amdgpu_kernel void @v_test_sub_v2
 ; GFX9: v_sub_co_u32_e32 v{{[0-9]+}}, vcc,
 ; GFX9: v_subb_co_u32_e32 v{{[0-9]+}}, vcc,
 define amdgpu_kernel void @v_test_sub_v4i64(<4 x i64> addrspace(1)* %out, <4 x i64> addrspace(1)* noalias %inA, <4 x i64> addrspace(1)* noalias %inB) {
-  %tid = call i32 @llvm.r600.read.tidig.x() readnone
+  %tid = call i32 @llvm.amdgcn.workitem.id.x() readnone
   %a_ptr = getelementptr <4 x i64>, <4 x i64> addrspace(1)* %inA, i32 %tid
   %b_ptr = getelementptr <4 x i64>, <4 x i64> addrspace(1)* %inB, i32 %tid
   %a = load <4 x i64>, <4 x i64> addrspace(1)* %a_ptr
@@ -245,3 +220,22 @@ define amdgpu_kernel void @v_test_sub_v4
   store <4 x i64> %result, <4 x i64> addrspace(1)* %out
   ret void
 }
+
+; Make sure the VOP3 form of sub is initially selected. Otherwise pair
+; of opies from/to VCC would be necessary
+
+; GCN-LABEL: {{^}}sub_select_vop3:
+; SI: v_subrev_i32_e64 v0, s[0:1], s0, v0
+; VI: v_subrev_u32_e64 v0, s[0:1], s0, v0
+; GFX9: v_subrev_u32_e32 v0, s0, v0
+
+; GCN: ; def vcc
+; GCN: ds_write_b32
+; GCN: ; use vcc
+define amdgpu_ps void @sub_select_vop3(i32 inreg %s, i32 %v) {
+  %vcc = call i64 asm sideeffect "; def vcc", "={vcc}"()
+  %sub = sub i32 %v, %s
+  store i32 %sub, i32 addrspace(3)* undef
+  call void asm sideeffect "; use vcc", "{vcc}"(i64 %vcc)
+  ret void
+}




More information about the llvm-commits mailing list