[AArch64][GISel] Legalize non-power-of-two G_CTLZ

[lenary]

Summary

This fixes a crash found in PR61549, and adds test coverage for other
sizes that cannot be selected.

[paquette]

Can you add a test for, say, s8 to show that small power-of-2s should still be widened to s32?

[lenary]

In D147516#4243892, @paquette wrote:

Can you add a test for, say, s8 to show that small power-of-2s should still be widened to s32?

Ack, will do. I will also reformat the code to remove the spurious formatting changes

[paquette]

LGTM

[lenary]

I will fix the legalizer-info-validation.mir failure when I land this (imminently).

llvm/lib/Target/AArch64/GISel/AArch64LegalizerInfo.cpp

getActionDefinitionsBuilder(G_BUILD_VECTOR)
       .minScalarOrElt(0, s8)
       .minScalarSameAs(1, 0);
 
   getActionDefinitionsBuilder(G_BUILD_VECTOR_TRUNC).lower();
 
   getActionDefinitionsBuilder(G_CTLZ)
       .legalForCartesianProduct(
           {s32, s64, v8s8, v16s8, v4s16, v8s16, v2s32, v4s32})
-      .scalarize(1);
+      .scalarize(1)
+      .widenScalarToNextPow2(1, /*Min=*/32)
+      .clampScalar(1, s32, s64)
+      .scalarSameSizeAs(0, 1);
   getActionDefinitionsBuilder(G_CTLZ_ZERO_UNDEF).lower();
 
   // TODO: Custom lowering for v2s32, v4s32, v2s64.
   getActionDefinitionsBuilder(G_BITREVERSE)
       .legalFor({s32, s64, v8s8, v16s8})
       .widenScalarToNextPow2(0, /*Min = */ 32)
       .clampScalar(0, s32, s64);
 

llvm/test/CodeGen/AArch64/GlobalISel/legalize-ctlz.mir

bb.0:
     ; CHECK-NEXT: RET_ReallyLR implicit $d0
     %0:_(s64) = COPY $d0
     %2:_(s64) = COPY %0(s64)
     %1:_(s64) = G_CTLZ_ZERO_UNDEF %2(s64)
     $d0 = COPY %1(s64)
     RET_ReallyLR implicit $d0
 
 ...
+---
+name:            test_s35
+alignment:       4
+tracksRegLiveness: true
+body:             |
+  bb.1:
+    liveins: $x0
+
+    ; CHECK-LABEL: name: test_s35
+    ; CHECK: liveins: $x0
+    ; CHECK-NEXT: {{  $}}
+    ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s64) = COPY $x0
+    ; CHECK-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 34359738367
+    ; CHECK-NEXT: [[AND:%[0-9]+]]:_(s64) = G_AND [[COPY]], [[C]]
+    ; CHECK-NEXT: [[CTLZ:%[0-9]+]]:_(s64) = G_CTLZ [[AND]](s64)
+    ; CHECK-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 29
+    ; CHECK-NEXT: [[SUB:%[0-9]+]]:_(s64) = G_SUB [[CTLZ]], [[C1]]
+    ; CHECK-NEXT: $x0 = COPY [[SUB]](s64)
+    ; CHECK-NEXT: RET_ReallyLR implicit $x0
+    %1:_(s64) = COPY $x0
+    %0:_(s35) = G_TRUNC %1(s64)
+    %2:_(s35) = G_CTLZ %0(s35)
+    %3:_(s64) = G_ANYEXT %2(s35)
+    $x0 = COPY %3(s64)
+    RET_ReallyLR implicit $x0
+...
+---
+name:            test_s8
+alignment:       4
+tracksRegLiveness: true
+body:             |
+  bb.1:
+    liveins: $w0
+
+    ; CHECK-LABEL: name: test_s8
+    ; CHECK: liveins: $w0
+    ; CHECK-NEXT: {{  $}}
+    ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $w0
+    ; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 255
+    ; CHECK-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[COPY]], [[C]]
+    ; CHECK-NEXT: [[CTLZ:%[0-9]+]]:_(s32) = G_CTLZ [[AND]](s32)
+    ; CHECK-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
+    ; CHECK-NEXT: [[SUB:%[0-9]+]]:_(s32) = G_SUB [[CTLZ]], [[C1]]
+    ; CHECK-NEXT: $w0 = COPY [[SUB]](s32)
+    ; CHECK-NEXT: RET_ReallyLR implicit $w0
+    %1:_(s32) = COPY $w0
+    %0:_(s8) = G_TRUNC %1(s32)
+    %2:_(s8) = G_CTLZ %0(s8)
+    %3:_(s32) = G_ANYEXT %2(s8)
+    $w0 = COPY %3(s32)
+    RET_ReallyLR implicit $w0
+...
+---
+name:            test_s65
+alignment:       4
+tracksRegLiveness: true
+body:             |
+  bb.1:
+    liveins: $x0, $x1
+
+    ; CHECK-LABEL: name: test_s65
+    ; CHECK: liveins: $x0, $x1
+    ; CHECK-NEXT: {{  $}}
+    ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s64) = COPY $x0
+    ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s64) = COPY $x1
+    ; CHECK-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 -1
+    ; CHECK-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+    ; CHECK-NEXT: [[AND:%[0-9]+]]:_(s64) = G_AND [[COPY]], [[C]]
+    ; CHECK-NEXT: [[AND1:%[0-9]+]]:_(s64) = G_AND [[COPY1]], [[C1]]
+    ; CHECK-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
+    ; CHECK-NEXT: [[ICMP:%[0-9]+]]:_(s32) = G_ICMP intpred(eq), [[AND1]](s64), [[C2]]
+    ; CHECK-NEXT: [[CTLZ:%[0-9]+]]:_(s64) = G_CTLZ [[AND]](s64)
+    ; CHECK-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 64
+    ; CHECK-NEXT: [[UADDO:%[0-9]+]]:_(s64), [[UADDO1:%[0-9]+]]:_(s32) = G_UADDO [[CTLZ]], [[C3]]
+    ; CHECK-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
+    ; CHECK-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[UADDO1]], [[C4]]
+    ; CHECK-NEXT: [[UADDE:%[0-9]+]]:_(s64), [[UADDE1:%[0-9]+]]:_(s32) = G_UADDE [[C2]], [[C2]], [[AND2]]
+    ; CHECK-NEXT: [[CTLZ1:%[0-9]+]]:_(s64) = G_CTLZ [[AND1]](s64)
+    ; CHECK-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[ICMP]], [[C4]]
+    ; CHECK-NEXT: [[SELECT:%[0-9]+]]:_(s64) = G_SELECT [[AND3]](s32), [[UADDO]], [[CTLZ1]]
+    ; CHECK-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[ICMP]], [[C4]]
+    ; CHECK-NEXT: [[SELECT1:%[0-9]+]]:_(s64) = G_SELECT [[AND4]](s32), [[UADDE]], [[C2]]
+    ; CHECK-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 63
+    ; CHECK-NEXT: [[USUBO:%[0-9]+]]:_(s64), [[USUBO1:%[0-9]+]]:_(s32) = G_USUBO [[SELECT]], [[C5]]
+    ; CHECK-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[USUBO1]], [[C4]]
+    ; CHECK-NEXT: [[USUBE:%[0-9]+]]:_(s64), [[USUBE1:%[0-9]+]]:_(s32) = G_USUBE [[SELECT1]], [[C2]], [[AND5]]
+    ; CHECK-NEXT: $x0 = COPY [[USUBO]](s64)
+    ; CHECK-NEXT: $x1 = COPY [[USUBE]](s64)
+    ; CHECK-NEXT: RET_ReallyLR implicit $x0, implicit $x1
+    %1:_(s64) = COPY $x0
+    %2:_(s64) = COPY $x1
+    %3:_(s128) = G_MERGE_VALUES %1(s64), %2(s64)
+    %0:_(s65) = G_TRUNC %3(s128)
+    %4:_(s65) = G_CTLZ %0(s65)
+    %7:_(s128) = G_ANYEXT %4(s65)
+    %5:_(s64), %6:_(s64) = G_UNMERGE_VALUES %7(s128)
+    $x0 = COPY %5(s64)
+    $x1 = COPY %6(s64)
+    RET_ReallyLR implicit $x0, implicit $x1
+...

llvm/test/CodeGen/AArch64/GlobalISel/legalizer-info-validation.mir

 # DEBUG-NEXT: .. imm index coverage check SKIPPED: user-defined predicate detected
 # DEBUG-NEXT: G_CTTZ (opcode {{[0-9]+}}): 2 type indices, 0 imm indices
 # DEBUG-NEXT: .. type index coverage check SKIPPED: user-defined predicate detected
 # DEBUG-NEXT: .. imm index coverage check SKIPPED: user-defined predicate detected
 # DEBUG-NEXT: G_CTTZ_ZERO_UNDEF (opcode {{[0-9]+}}): 2 type indices, 0 imm indices
 # DEBUG-NEXT: .. type index coverage check SKIPPED: user-defined predicate detected
 # DEBUG-NEXT: .. imm index coverage check SKIPPED: user-defined predicate detected
 # DEBUG-NEXT: G_CTLZ (opcode {{[0-9]+}}): 2 type indices, 0 imm indices
-# DEBUG-NEXT: .. the first uncovered type index: 2, OK
-# DEBUG-NEXT: .. the first uncovered imm index: 0, OK
+# DEBUG-NEXT: .. type index coverage check SKIPPED: user-defined predicate detected
+# DEBUG-NEXT: .. imm index coverage check SKIPPED: user-defined predicate detected
 # DEBUG-NEXT: G_CTLZ_ZERO_UNDEF (opcode {{[0-9]+}}): 2 type indices, 0 imm indices
 # DEBUG-NEXT: .. type index coverage check SKIPPED: user-defined predicate detected
 # DEBUG-NEXT: .. imm index coverage check SKIPPED: user-defined predicate detected
 # DEBUG-NEXT: G_CTPOP (opcode {{[0-9]+}}): 2 type indices, 0 imm indices
 # DEBUG-NEXT: .. type index coverage check SKIPPED: user-defined predicate detected
 # DEBUG-NEXT: .. imm index coverage check SKIPPED: user-defined predicate detected
 # DEBUG-NEXT: G_BSWAP (opcode {{[0-9]+}}): 1 type index, 0 imm indices
 # DEBUG-NEXT: .. the first uncovered type index: 1, OK