aboutsummaryrefslogtreecommitdiff
diff options
context:
space:
mode:
authorChris Lattner <sabre@nondot.org>2007-02-27 17:21:09 +0000
committerChris Lattner <sabre@nondot.org>2007-02-27 17:21:09 +0000
commitdbec2d9dcf6bf942b29db19ecea35e40c5a88506 (patch)
tree6189411ff83f25851ec2575ea5ada18270833ac6
parent84da80d10b472332d079c58d21aa48b82e636274 (diff)
a note
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@34670 91177308-0d34-0410-b5e6-96231b3b80d8
-rw-r--r--lib/Target/X86/README-SSE.txt42
1 files changed, 42 insertions, 0 deletions
diff --git a/lib/Target/X86/README-SSE.txt b/lib/Target/X86/README-SSE.txt
index bab5373d65..2b7f9ae439 100644
--- a/lib/Target/X86/README-SSE.txt
+++ b/lib/Target/X86/README-SSE.txt
@@ -530,3 +530,45 @@ to loads from constant pool.
Floating point max / min are commutable when -enable-unsafe-fp-path is
specified. We should turn int_x86_sse_max_ss and X86ISD::FMIN etc. into other
nodes which are selected to max / min instructions that are marked commutable.
+
+//===---------------------------------------------------------------------===//
+
+We should compile this:
+#include <xmmintrin.h>
+typedef union {
+ int i[4];
+ float f[4];
+ __m128 v;
+} vector4_t;
+void swizzle (const void *a, vector4_t * b, vector4_t * c) {
+ b->v = _mm_loadl_pi (b->v, (__m64 *) a);
+ c->v = _mm_loadl_pi (c->v, ((__m64 *) a) + 1);
+}
+
+to:
+
+_swizzle:
+ movl 4(%esp), %eax
+ movl 8(%esp), %edx
+ movl 12(%esp), %ecx
+ movlps (%eax), %xmm0
+ movlps %xmm0, (%edx)
+ movlps 8(%eax), %xmm0
+ movlps %xmm0, (%ecx)
+ ret
+
+not:
+
+swizzle:
+ movl 8(%esp), %eax
+ movaps (%eax), %xmm0
+ movl 4(%esp), %ecx
+ movlps (%ecx), %xmm0
+ movaps %xmm0, (%eax)
+ movl 12(%esp), %eax
+ movaps (%eax), %xmm0
+ movlps 8(%ecx), %xmm0
+ movaps %xmm0, (%eax)
+ ret
+
+