[SystemZ] Add CodeGen support for integer vector types
This the first of a series of patches to add CodeGen support exploiting
the instructions of the z13 vector facility. This patch adds support
for the native integer vector types (v16i8, v8i16, v4i32, v2i64).
When the vector facility is present, we default to the new vector ABI.
This is characterized by two major differences:
- Vector types are passed/returned in vector registers
(except for unnamed arguments of a variable-argument list function).
- Vector types are at most 8-byte aligned.
The reason for the choice of 8-byte vector alignment is that the hardware
is able to efficiently load vectors at 8-byte alignment, and the ABI only
guarantees 8-byte alignment of the stack pointer, so requiring any higher
alignment for vectors would require dynamic stack re-alignment code.
However, for compatibility with old code that may use vector types, when
*not* using the vector facility, the old alignment rules (vector types
are naturally aligned) remain in use.
These alignment rules are not only implemented at the C language level
(implemented in clang), but also at the LLVM IR level. This is done
by selecting a different DataLayout string depending on whether the
vector ABI is in effect or not.
Based on a patch by Richard Sandiford.
llvm-svn: 236521
2015-05-06 03:25:42 +08:00
|
|
|
; Test insertions of memory values into 0.
|
|
|
|
;
|
|
|
|
; RUN: llc < %s -mtriple=s390x-linux-gnu -mcpu=z13 | FileCheck %s
|
|
|
|
|
|
|
|
; Test VLLEZB.
|
|
|
|
define <16 x i8> @f1(i8 *%ptr) {
|
|
|
|
; CHECK-LABEL: f1:
|
|
|
|
; CHECK: vllezb %v24, 0(%r2)
|
|
|
|
; CHECK: br %r14
|
|
|
|
%val = load i8, i8 *%ptr
|
|
|
|
%ret = insertelement <16 x i8> zeroinitializer, i8 %val, i32 7
|
|
|
|
ret <16 x i8> %ret
|
|
|
|
}
|
|
|
|
|
|
|
|
; Test VLLEZB with the highest in-range offset.
|
|
|
|
define <16 x i8> @f2(i8 *%base) {
|
|
|
|
; CHECK-LABEL: f2:
|
|
|
|
; CHECK: vllezb %v24, 4095(%r2)
|
|
|
|
; CHECK: br %r14
|
|
|
|
%ptr = getelementptr i8, i8 *%base, i64 4095
|
|
|
|
%val = load i8, i8 *%ptr
|
|
|
|
%ret = insertelement <16 x i8> zeroinitializer, i8 %val, i32 7
|
|
|
|
ret <16 x i8> %ret
|
|
|
|
}
|
|
|
|
|
|
|
|
; Test VLLEZB with the next highest offset.
|
|
|
|
define <16 x i8> @f3(i8 *%base) {
|
|
|
|
; CHECK-LABEL: f3:
|
|
|
|
; CHECK-NOT: vllezb %v24, 4096(%r2)
|
|
|
|
; CHECK: br %r14
|
|
|
|
%ptr = getelementptr i8, i8 *%base, i64 4096
|
|
|
|
%val = load i8, i8 *%ptr
|
|
|
|
%ret = insertelement <16 x i8> zeroinitializer, i8 %val, i32 7
|
|
|
|
ret <16 x i8> %ret
|
|
|
|
}
|
|
|
|
|
|
|
|
; Test that VLLEZB allows an index.
|
|
|
|
define <16 x i8> @f4(i8 *%base, i64 %index) {
|
|
|
|
; CHECK-LABEL: f4:
|
|
|
|
; CHECK: vllezb %v24, 0({{%r2,%r3|%r3,%r2}})
|
|
|
|
; CHECK: br %r14
|
|
|
|
%ptr = getelementptr i8, i8 *%base, i64 %index
|
|
|
|
%val = load i8, i8 *%ptr
|
|
|
|
%ret = insertelement <16 x i8> zeroinitializer, i8 %val, i32 7
|
|
|
|
ret <16 x i8> %ret
|
|
|
|
}
|
|
|
|
|
|
|
|
; Test VLLEZH.
|
|
|
|
define <8 x i16> @f5(i16 *%ptr) {
|
|
|
|
; CHECK-LABEL: f5:
|
|
|
|
; CHECK: vllezh %v24, 0(%r2)
|
|
|
|
; CHECK: br %r14
|
|
|
|
%val = load i16, i16 *%ptr
|
|
|
|
%ret = insertelement <8 x i16> zeroinitializer, i16 %val, i32 3
|
|
|
|
ret <8 x i16> %ret
|
|
|
|
}
|
|
|
|
|
|
|
|
; Test VLLEZF.
|
|
|
|
define <4 x i32> @f6(i32 *%ptr) {
|
|
|
|
; CHECK-LABEL: f6:
|
|
|
|
; CHECK: vllezf %v24, 0(%r2)
|
|
|
|
; CHECK: br %r14
|
|
|
|
%val = load i32, i32 *%ptr
|
|
|
|
%ret = insertelement <4 x i32> zeroinitializer, i32 %val, i32 1
|
|
|
|
ret <4 x i32> %ret
|
|
|
|
}
|
|
|
|
|
|
|
|
; Test VLLEZG.
|
|
|
|
define <2 x i64> @f7(i64 *%ptr) {
|
|
|
|
; CHECK-LABEL: f7:
|
|
|
|
; CHECK: vllezg %v24, 0(%r2)
|
|
|
|
; CHECK: br %r14
|
|
|
|
%val = load i64, i64 *%ptr
|
|
|
|
%ret = insertelement <2 x i64> zeroinitializer, i64 %val, i32 0
|
|
|
|
ret <2 x i64> %ret
|
|
|
|
}
|
2015-05-06 03:26:48 +08:00
|
|
|
|
2015-05-06 03:27:45 +08:00
|
|
|
; Test VLLEZF with a float.
|
|
|
|
define <4 x float> @f8(float *%ptr) {
|
|
|
|
; CHECK-LABEL: f8:
|
|
|
|
; CHECK: vllezf %v24, 0(%r2)
|
|
|
|
; CHECK: br %r14
|
|
|
|
%val = load float, float *%ptr
|
|
|
|
%ret = insertelement <4 x float> zeroinitializer, float %val, i32 1
|
|
|
|
ret <4 x float> %ret
|
|
|
|
}
|
|
|
|
|
2015-05-06 03:26:48 +08:00
|
|
|
; Test VLLEZG with a double.
|
|
|
|
define <2 x double> @f9(double *%ptr) {
|
|
|
|
; CHECK-LABEL: f9:
|
|
|
|
; CHECK: vllezg %v24, 0(%r2)
|
|
|
|
; CHECK: br %r14
|
|
|
|
%val = load double, double *%ptr
|
|
|
|
%ret = insertelement <2 x double> zeroinitializer, double %val, i32 0
|
|
|
|
ret <2 x double> %ret
|
|
|
|
}
|
2018-12-20 21:05:03 +08:00
|
|
|
|
|
|
|
; Test VLLEZF with a float when the result is stored to memory.
|
|
|
|
define void @f10(float *%ptr, <4 x float> *%res) {
|
|
|
|
; CHECK-LABEL: f10:
|
|
|
|
; CHECK: vllezf [[REG:%v[0-9]+]], 0(%r2)
|
|
|
|
; CHECK: vst [[REG]], 0(%r3)
|
|
|
|
; CHECK: br %r14
|
|
|
|
%val = load float, float *%ptr
|
|
|
|
%ret = insertelement <4 x float> zeroinitializer, float %val, i32 1
|
|
|
|
store <4 x float> %ret, <4 x float> *%res
|
|
|
|
ret void
|
|
|
|
}
|
|
|
|
|
|
|
|
; Test VLLEZG with a double when the result is stored to memory.
|
|
|
|
define void @f11(double *%ptr, <2 x double> *%res) {
|
|
|
|
; CHECK-LABEL: f11:
|
|
|
|
; CHECK: vllezg [[REG:%v[0-9]+]], 0(%r2)
|
|
|
|
; CHECK: vst [[REG]], 0(%r3)
|
|
|
|
; CHECK: br %r14
|
|
|
|
%val = load double, double *%ptr
|
|
|
|
%ret = insertelement <2 x double> zeroinitializer, double %val, i32 0
|
|
|
|
store <2 x double> %ret, <2 x double> *%res
|
|
|
|
ret void
|
|
|
|
}
|
|
|
|
|
|
|
|
; Test VLLEZG when the zeroinitializer is shared.
|
|
|
|
define void @f12(i64 *%ptr, <2 x i64> *%res) {
|
|
|
|
; CHECK-LABEL: f12:
|
|
|
|
; CHECK: vllezg [[REG:%v[0-9]+]], 0(%r2)
|
|
|
|
; CHECK: vst [[REG]], 0(%r3)
|
|
|
|
; CHECK: vllezg [[REG1:%v[0-9]+]], 0(%r2)
|
|
|
|
; CHECK: vst [[REG1]], 0(%r3)
|
|
|
|
; CHECK: br %r14
|
|
|
|
%val = load volatile i64, i64 *%ptr
|
|
|
|
%ret = insertelement <2 x i64> zeroinitializer, i64 %val, i32 0
|
|
|
|
store volatile <2 x i64> %ret, <2 x i64> *%res
|
|
|
|
%val1 = load volatile i64, i64 *%ptr
|
|
|
|
%ret1 = insertelement <2 x i64> zeroinitializer, i64 %val1, i32 0
|
|
|
|
store volatile <2 x i64> %ret1, <2 x i64> *%res
|
|
|
|
ret void
|
|
|
|
}
|
|
|
|
|