Eliminate G_TRUNC to S1 SGPR

Author: gandhi56

llvm/lib/Target/AMDGPU/AMDGPURegBankLegalizeHelper.cpp

@@ -1132,7 +1132,8 @@ void RegBankLegalizeHelper::applyMappingDst(
       assert(RB == SgprRB);
       Register NewDst = MRI.createVirtualRegister(SgprRB_S32);
       Op.setReg(NewDst);
-      B.buildTrunc(Reg, NewDst);
+      if (!MRI.use_empty(Reg))
+        B.buildTrunc(Reg, NewDst);
       break;
     }
     case InvalidMapping: {

llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-add.mir

@@ -448,3 +448,29 @@ body: |
     %4:_(s32), %5:_(s1) = G_UADDE %0, %1, %3
     S_ENDPGM 0, implicit %4
 ...
+
+---
+name: uadde_s32_ss_scc_use
+legalized: true
+
+body: |
+  bb.0:
+    liveins: $sgpr0, $sgpr1, $sgpr2
+    ; CHECK-LABEL: name: uadde_s32_ss_scc_use
+    ; CHECK: liveins: $sgpr0, $sgpr1, $sgpr2
+    ; CHECK-NEXT: {{  $}}
+    ; CHECK-NEXT: [[COPY:%[0-9]+]]:sgpr(s32) = COPY $sgpr0
+    ; CHECK-NEXT: [[COPY1:%[0-9]+]]:sgpr(s32) = COPY $sgpr1
+    ; CHECK-NEXT: [[COPY2:%[0-9]+]]:sgpr(s32) = COPY $sgpr2
+    ; CHECK-NEXT: [[C:%[0-9]+]]:sgpr(s32) = G_CONSTANT i32 1
+    ; CHECK-NEXT: [[AND:%[0-9]+]]:sgpr(s32) = G_AND [[COPY2]], [[C]]
+    ; CHECK-NEXT: [[UADDE:%[0-9]+]]:sgpr(s32), [[UADDE1:%[0-9]+]]:sgpr(s32) = G_UADDE [[COPY]], [[COPY1]], [[AND]]
+    ; CHECK-NEXT: S_ENDPGM 0, implicit [[UADDE]](s32)
+    %0:_(s32) = COPY $sgpr0
+    %1:_(s32) = COPY $sgpr1
+    %2:_(s32) = COPY $sgpr2
+    %3:_(s1) = G_TRUNC %2
+    %4:_(s32), %5:_(s1) = G_UADDE %0, %1, %3
+    %6:_(s32) = G_ANYEXT %5:_(s1)
+    S_ENDPGM 0, implicit %4
+...

llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-sext.mir

@@ -77,10 +77,14 @@ body: |
     ; CHECK-NEXT: [[C2:%[0-9]+]]:sgpr(s32) = G_CONSTANT i32 0
     ; CHECK-NEXT: [[SELECT:%[0-9]+]]:sgpr(s32) = G_SELECT [[AND]](s32), [[C1]], [[C2]]
     ; CHECK-NEXT: [[TRUNC:%[0-9]+]]:sgpr(s16) = G_TRUNC [[SELECT]](s32)
+    ; CHECK-NEXT: [[C3:%[0-9]+]]:sgpr(s16) = G_CONSTANT i16 255
+    ; CHECK-NEXT: [[AND1:%[0-9]+]]:sgpr(s16) = G_AND [[TRUNC]], [[C3]]
     %0:_(s32) = COPY $sgpr0
     %1:_(s32) = COPY $sgpr1
     %2:_(s1) = G_ICMP intpred(eq), %0, %1
     %3:_(s16) = G_SEXT %2
+    %4:_(s16) = G_CONSTANT i16 255
+    %5:_(s16) = G_AND %3, %4
 ...
 
 ---
@@ -215,9 +219,13 @@ body: |
     ; CHECK-NEXT: [[C2:%[0-9]+]]:sgpr(s32) = G_CONSTANT i32 0
     ; CHECK-NEXT: [[SELECT:%[0-9]+]]:sgpr(s32) = G_SELECT [[AND]](s32), [[C1]], [[C2]]
     ; CHECK-NEXT: [[TRUNC:%[0-9]+]]:sgpr(s16) = G_TRUNC [[SELECT]](s32)
+    ; CHECK-NEXT: [[C3:%[0-9]+]]:sgpr(s16) = G_CONSTANT i16 255
+    ; CHECK-NEXT: [[AND1:%[0-9]+]]:sgpr(s16) = G_AND [[TRUNC]], [[C3]]
     %0:_(s32) = COPY $sgpr0
     %1:_(s1) = G_TRUNC %0
     %2:_(s16) = G_SEXT %1
+    %3:_(s16) = G_CONSTANT i16 255
+    %4:_(s16) = G_AND %2, %3
 ...
 
 ---

llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-zext.mir

@@ -73,10 +73,14 @@ body: |
     ; CHECK-NEXT: [[C1:%[0-9]+]]:sgpr(s32) = G_CONSTANT i32 0
     ; CHECK-NEXT: [[SELECT:%[0-9]+]]:sgpr(s32) = G_SELECT [[AND]](s32), [[C]], [[C1]]
     ; CHECK-NEXT: [[TRUNC:%[0-9]+]]:sgpr(s16) = G_TRUNC [[SELECT]](s32)
+    ; CHECK-NEXT: [[C2:%[0-9]+]]:sgpr(s16) = G_CONSTANT i16 255
+    ; CHECK-NEXT: [[AND1:%[0-9]+]]:sgpr(s16) = G_AND [[TRUNC]], [[C2]]
     %0:_(s32) = COPY $sgpr0
     %1:_(s32) = COPY $sgpr1
     %2:_(s1) = G_ICMP intpred(eq), %0, %1
     %3:_(s16) = G_ZEXT %2
+    %4:_(s16) = G_CONSTANT i16 255
+    %5:_(s16) = G_AND %3, %4
 ...
 
 ---
@@ -209,9 +213,13 @@ body: |
     ; CHECK-NEXT: [[C1:%[0-9]+]]:sgpr(s32) = G_CONSTANT i32 0
     ; CHECK-NEXT: [[SELECT:%[0-9]+]]:sgpr(s32) = G_SELECT [[AND]](s32), [[C]], [[C1]]
     ; CHECK-NEXT: [[TRUNC:%[0-9]+]]:sgpr(s16) = G_TRUNC [[SELECT]](s32)
+    ; CHECK-NEXT: [[C2:%[0-9]+]]:sgpr(s16) = G_CONSTANT i16 255
+    ; CHECK-NEXT: [[AND1:%[0-9]+]]:sgpr(s16) = G_AND [[TRUNC]], [[C2]]
     %0:_(s32) = COPY $sgpr0
     %1:_(s1) = G_TRUNC %0
     %2:_(s16) = G_ZEXT %1
+    %3:_(s16) = G_CONSTANT i16 255
+    %4:_(s16) = G_AND %2, %3
 ...
 
 ---