Recognize code for doing vector gather/scatter index calculations with

32-bit indices. Instead of shuffling each element out of the index vector,
when all indices are needed, just store the input vector to the stack and
load the elements out.

llvm-svn: 98588

1 files changed, 39 insertions, 0 deletions

diff --git a/llvm/test/CodeGen/X86/gather-addresses.ll b/llvm/test/CodeGen/X86/gather-addresses.ll
new file mode 100644
index 00000000000..07198386b8e
--- /dev/null
+++ b/llvm/test/CodeGen/X86/gather-addresses.ll
@@ -0,0 +1,39 @@
+; RUN: llc -march=x86-64 < %s | FileCheck %s
+
+; When doing vector gather-scatter index calculation with 32-bit indices,
+; bounce the vector off of cache rather than shuffling each individual
+; element out of the index vector.
+
+; CHECK: pand     (%rdx), %xmm0
+; CHECK: movaps   %xmm0, -24(%rsp)
+; CHECK: movslq   -24(%rsp), %rax
+; CHECK: movsd    (%rdi,%rax,8), %xmm0
+; CHECK: movslq   -20(%rsp), %rax
+; CHECK: movhpd   (%rdi,%rax,8), %xmm0
+; CHECK: movslq   -16(%rsp), %rax
+; CHECK: movsd    (%rdi,%rax,8), %xmm1
+; CHECK: movslq   -12(%rsp), %rax
+; CHECK: movhpd   (%rdi,%rax,8), %xmm1
+
+define <4 x double> @foo(double* %p, <4 x i32>* %i, <4 x i32>* %h) nounwind {
+  %a = load <4 x i32>* %i
+  %b = load <4 x i32>* %h
+  %j = and <4 x i32> %a, %b
+  %d0 = extractelement <4 x i32> %j, i32 0
+  %d1 = extractelement <4 x i32> %j, i32 1
+  %d2 = extractelement <4 x i32> %j, i32 2
+  %d3 = extractelement <4 x i32> %j, i32 3
+  %q0 = getelementptr double* %p, i32 %d0
+  %q1 = getelementptr double* %p, i32 %d1
+  %q2 = getelementptr double* %p, i32 %d2
+  %q3 = getelementptr double* %p, i32 %d3
+  %r0 = load double* %q0
+  %r1 = load double* %q1
+  %r2 = load double* %q2
+  %r3 = load double* %q3
+  %v0 = insertelement <4 x double> undef, double %r0, i32 0
+  %v1 = insertelement <4 x double> %v0, double %r1, i32 1
+  %v2 = insertelement <4 x double> %v1, double %r2, i32 2
+  %v3 = insertelement <4 x double> %v2, double %r3, i32 3
+  ret <4 x double> %v3
+}