AMDGPU/GlobalISel: Fix import of s_abs_i32 pattern

llvm · Jan 7, 2020 · e699c03 · e699c03
1 parent 9150d6b
commit e699c03
Show file tree

Hide file tree

Showing 2 changed files with 106 additions and 1 deletion.
diff --git a/llvm/lib/Target/AMDGPU/SOPInstructions.td b/llvm/lib/Target/AMDGPU/SOPInstructions.td
@@ -1213,7 +1213,7 @@ def : GCNPat <
 
 def : GCNPat <
   (i32 (smax i32:$x, (i32 (ineg i32:$x)))),
-  (S_ABS_I32 $x)
+  (S_ABS_I32 SReg_32:$x)
 >;
 
 def : GCNPat <

diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/inst-select-abs.mir b/llvm/test/CodeGen/AMDGPU/GlobalISel/inst-select-abs.mir
@@ -0,0 +1,105 @@
+# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
+# RUN: llc -march=amdgcn -mcpu=tahiti -run-pass=instruction-select -verify-machineinstrs %s -o - | FileCheck -check-prefix=GFX6 %s
+# RUN: llc -march=amdgcn -mcpu=gfx900 -run-pass=instruction-select -verify-machineinstrs %s -o - | FileCheck -check-prefix=GFX9 %s
+
+---
+name: smax_neg_abs_pattern_s32_ss
+legalized: true
+regBankSelected: true
+tracksRegLiveness: true
+
+body: |
+  bb.0:
+    liveins: $sgpr0
+
+    ; GFX6-LABEL: name: smax_neg_abs_pattern_s32_ss
+    ; GFX6: liveins: $sgpr0
+    ; GFX6: %src0:sreg_32 = COPY $sgpr0
+    ; GFX6: %smax:sreg_32 = S_ABS_I32 %src0, implicit-def $scc
+    ; GFX6: S_ENDPGM 0, implicit %smax
+    ; GFX9-LABEL: name: smax_neg_abs_pattern_s32_ss
+    ; GFX9: liveins: $sgpr0
+    ; GFX9: %src0:sreg_32 = COPY $sgpr0
+    ; GFX9: %smax:sreg_32 = S_ABS_I32 %src0, implicit-def $scc
+    ; GFX9: S_ENDPGM 0, implicit %smax
+    %src0:sgpr(s32) = COPY $sgpr0
+    %zero:sgpr(s32) = G_CONSTANT i32 0
+    %ineg:sgpr(s32) = G_SUB %zero, %src0
+    %smax:sgpr(s32) = G_SMAX %src0, %ineg
+    S_ENDPGM 0, implicit %smax
+...
+
+---
+name: smax_neg_abs_pattern_s32_ss_commute
+legalized: true
+regBankSelected: true
+tracksRegLiveness: true
+
+body: |
+  bb.0:
+    liveins: $sgpr0
+
+    ; GFX6-LABEL: name: smax_neg_abs_pattern_s32_ss_commute
+    ; GFX6: liveins: $sgpr0
+    ; GFX6: %src0:sreg_32 = COPY $sgpr0
+    ; GFX6: %smax:sreg_32 = S_ABS_I32 %src0, implicit-def $scc
+    ; GFX6: S_ENDPGM 0, implicit %smax
+    ; GFX9-LABEL: name: smax_neg_abs_pattern_s32_ss_commute
+    ; GFX9: liveins: $sgpr0
+    ; GFX9: %src0:sreg_32 = COPY $sgpr0
+    ; GFX9: %smax:sreg_32 = S_ABS_I32 %src0, implicit-def $scc
+    ; GFX9: S_ENDPGM 0, implicit %smax
+    %src0:sgpr(s32) = COPY $sgpr0
+    %zero:sgpr(s32) = G_CONSTANT i32 0
+    %ineg:sgpr(s32) = G_SUB %zero, %src0
+    %smax:sgpr(s32) = G_SMAX %ineg, %src0
+    S_ENDPGM 0, implicit %smax
+...
+
+---
+name: smax_neg_abs_pattern_s32_vv
+legalized: true
+regBankSelected: true
+tracksRegLiveness: true
+
+body: |
+  bb.0:
+    liveins: $vgpr0
+
+    ; GFX6-LABEL: name: smax_neg_abs_pattern_s32_vv
+    ; GFX6: liveins: $vgpr0
+    ; GFX6: %src0:vgpr_32 = COPY $vgpr0
+    ; GFX6: %zero:vgpr_32 = V_MOV_B32_e32 0, implicit $exec
+    ; GFX6: %ineg:vgpr_32, dead %4:sreg_64_xexec = V_SUB_I32_e64 %zero, %src0, 0, implicit $exec
+    ; GFX6: %smax:vgpr_32 = V_MAX_I32_e64 %src0, %ineg, implicit $exec
+    ; GFX6: S_ENDPGM 0, implicit %smax
+    ; GFX9-LABEL: name: smax_neg_abs_pattern_s32_vv
+    ; GFX9: liveins: $vgpr0
+    ; GFX9: %src0:vgpr_32 = COPY $vgpr0
+    ; GFX9: %zero:vgpr_32 = V_MOV_B32_e32 0, implicit $exec
+    ; GFX9: %ineg:vgpr_32 = V_SUB_U32_e64 %zero, %src0, 0, implicit $exec
+    ; GFX9: %smax:vgpr_32 = V_MAX_I32_e64 %src0, %ineg, implicit $exec
+    ; GFX9: S_ENDPGM 0, implicit %smax
+    %src0:vgpr(s32) = COPY $vgpr0
+    %zero:vgpr(s32) = G_CONSTANT i32 0
+    %ineg:vgpr(s32) = G_SUB %zero, %src0
+    %smax:vgpr(s32) = G_SMAX %src0, %ineg
+    S_ENDPGM 0, implicit %smax
+...
+
+# FIXME: Violates constant bus restriction
+# ---
+# name: smax_neg_abs_pattern_s32_vs
+# legalized: true
+# regBankSelected: true
+
+# body: |
+#   bb.0:
+#     liveins: $sgpr0
+
+#     %src0:sgpr(s32) = COPY $sgpr0
+#     %zero:sgpr(s32) = G_CONSTANT i32 0
+#     %ineg:sgpr(s32) = G_SUB %zero, %src0
+#     %smax:vgpr(s32) = G_SMAX %src0, %ineg
+#     S_ENDPGM 0, implicit %smax
+# ...