From: James Y Knight <jyknight@google.com>
Date: Fri, 9 Oct 2015 21:36:19 +0000 (+0000)
Subject: Fix assert when emitting llvm.pow.f86.
X-Git-Url: http://plrg.eecs.uci.edu/git/?a=commitdiff_plain;h=84477d139097de85c5d33053da2f8772ed0ef80b;p=oota-llvm.git

Fix assert when emitting llvm.pow.f86.

This occurred due to introducing the invalid i64 type after type
legalization had already finished, in an attempt to workaround bitcast
f64 -> v2i32 not doing constant folding.

The *right* thing is to actually fix bitcast, but that has other
complications. So, for now, just get rid of the broken workaround, and
check in a test-case showing that it doesn't crash, with TODOs for
emitting proper code.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@249908 91177308-0d34-0410-b5e6-96231b3b80d8
---

diff --git a/lib/Target/Sparc/SparcISelLowering.cpp b/lib/Target/Sparc/SparcISelLowering.cpp
index 9c8a0a64525..ef2806474a0 100644
--- a/lib/Target/Sparc/SparcISelLowering.cpp
+++ b/lib/Target/Sparc/SparcISelLowering.cpp
@@ -854,11 +854,10 @@ SparcTargetLowering::LowerCall_32(TargetLowering::CallLoweringInfo &CLI,
         // Move from the float value from float registers into the
         // integer registers.
 
-        // TODO: this conversion is done in two steps, because
-        // f64->i64 conversion is done efficiently, and i64->v2i32 is
-        // basically a no-op. But f64->v2i32 is NOT done efficiently
-        // for some reason.
-        Arg = DAG.getNode(ISD::BITCAST, dl, MVT::i64, Arg);
+        // TODO: The f64 -> v2i32 conversion is super-inefficient for
+        // constants: it sticks them in the constant pool, then loads
+        // to a fp register, then stores to temp memory, then loads to
+        // integer registers.
         Arg = DAG.getNode(ISD::BITCAST, dl, MVT::v2i32, Arg);
       }
 
diff --git a/test/CodeGen/SPARC/float-constants.ll b/test/CodeGen/SPARC/float-constants.ll
new file mode 100644
index 00000000000..b3686ebdf44
--- /dev/null
+++ b/test/CodeGen/SPARC/float-constants.ll
@@ -0,0 +1,41 @@
+; RUN: llc < %s -march=sparc | FileCheck %s
+
+; TODO: actually fix the codegen to be optimal. At least we don't
+; crash for now, though...
+
+;; Bitcast should not do a runtime conversion, but rather emit a
+;; constant into integer registers directly.
+
+; CHECK-LABEL: bitcast:
+; TODO-CHECK: sethi 1049856, %o0
+; TODO-CHECK: sethi 0, %o1
+define <2 x i32> @bitcast() {
+  %1 = bitcast double 5.0 to <2 x i32>
+  ret <2 x i32> %1
+}
+
+;; Same thing for a call using a double (which gets passed in integer
+;; registers)
+
+; CHECK-LABEL: test_call
+; TODO-CHECK: sethi 1049856, %o0
+; TODO-CHECK: sethi 0, %o1
+declare void @a(double)
+define void @test_call() {
+  call void @a(double 5.0)
+  ret void
+}
+
+;; And for a libcall emitted from the pow intrinsic.  (libcall
+;; emission happens after SelectionDAG type legalization, so is a bit
+;; different than a normal function call. This was crashing before,
+;; due to an earlier broken workaround for this issue.)
+
+; CHECK-LABEL: test_intrins_call
+; TODO-CHECK: sethi 1049856, %o0
+; TODO-CHECK: sethi 0, %o1
+declare double @llvm.pow.f64(double, double)
+define double @test_intrins_call() {
+  %1 = call double @llvm.pow.f64(double 2.0, double 2.0)
+  ret double %1
+}