mirror of
https://github.com/RPCS3/llvm-mirror.git
synced 2024-11-24 03:33:20 +01:00
ab043ff680
Essentially the same as the GEP change in r230786.
A similar migration script can be used to update test cases, though a few more
test case improvements/changes were required this time around: (r229269-r229278)
import fileinput
import sys
import re
pat = re.compile(r"((?:=|:|^)\s*load (?:atomic )?(?:volatile )?(.*?))(| addrspace\(\d+\) *)\*($| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$)")
for line in sys.stdin:
sys.stdout.write(re.sub(pat, r"\1, \2\3*\4", line))
Reviewers: rafael, dexonsmith, grosser
Differential Revision: http://reviews.llvm.org/D7649
llvm-svn: 230794
67 lines
1.9 KiB
LLVM
67 lines
1.9 KiB
LLVM
; RUN: llc < %s -march=nvptx -mcpu=sm_20 | FileCheck %s
|
|
|
|
; Even though general vector types are not supported in PTX, we can still
|
|
; optimize loads/stores with pseudo-vector instructions of the form:
|
|
;
|
|
; ld.v2.f32 {%f0, %f1}, [%r0]
|
|
;
|
|
; which will load two floats at once into scalar registers.
|
|
|
|
define void @foo(<2 x float>* %a) {
|
|
; CHECK: .func foo
|
|
; CHECK: ld.v2.f32 {%f{{[0-9]+}}, %f{{[0-9]+}}}
|
|
%t1 = load <2 x float>, <2 x float>* %a
|
|
%t2 = fmul <2 x float> %t1, %t1
|
|
store <2 x float> %t2, <2 x float>* %a
|
|
ret void
|
|
}
|
|
|
|
define void @foo2(<4 x float>* %a) {
|
|
; CHECK: .func foo2
|
|
; CHECK: ld.v4.f32 {%f{{[0-9]+}}, %f{{[0-9]+}}, %f{{[0-9]+}}, %f{{[0-9]+}}}
|
|
%t1 = load <4 x float>, <4 x float>* %a
|
|
%t2 = fmul <4 x float> %t1, %t1
|
|
store <4 x float> %t2, <4 x float>* %a
|
|
ret void
|
|
}
|
|
|
|
define void @foo3(<8 x float>* %a) {
|
|
; CHECK: .func foo3
|
|
; CHECK: ld.v4.f32 {%f{{[0-9]+}}, %f{{[0-9]+}}, %f{{[0-9]+}}, %f{{[0-9]+}}}
|
|
; CHECK-NEXT: ld.v4.f32 {%f{{[0-9]+}}, %f{{[0-9]+}}, %f{{[0-9]+}}, %f{{[0-9]+}}}
|
|
%t1 = load <8 x float>, <8 x float>* %a
|
|
%t2 = fmul <8 x float> %t1, %t1
|
|
store <8 x float> %t2, <8 x float>* %a
|
|
ret void
|
|
}
|
|
|
|
|
|
|
|
define void @foo4(<2 x i32>* %a) {
|
|
; CHECK: .func foo4
|
|
; CHECK: ld.v2.u32 {%r{{[0-9]+}}, %r{{[0-9]+}}}
|
|
%t1 = load <2 x i32>, <2 x i32>* %a
|
|
%t2 = mul <2 x i32> %t1, %t1
|
|
store <2 x i32> %t2, <2 x i32>* %a
|
|
ret void
|
|
}
|
|
|
|
define void @foo5(<4 x i32>* %a) {
|
|
; CHECK: .func foo5
|
|
; CHECK: ld.v4.u32 {%r{{[0-9]+}}, %r{{[0-9]+}}, %r{{[0-9]+}}, %r{{[0-9]+}}}
|
|
%t1 = load <4 x i32>, <4 x i32>* %a
|
|
%t2 = mul <4 x i32> %t1, %t1
|
|
store <4 x i32> %t2, <4 x i32>* %a
|
|
ret void
|
|
}
|
|
|
|
define void @foo6(<8 x i32>* %a) {
|
|
; CHECK: .func foo6
|
|
; CHECK: ld.v4.u32 {%r{{[0-9]+}}, %r{{[0-9]+}}, %r{{[0-9]+}}, %r{{[0-9]+}}}
|
|
; CHECK-NEXT: ld.v4.u32 {%r{{[0-9]+}}, %r{{[0-9]+}}, %r{{[0-9]+}}, %r{{[0-9]+}}}
|
|
%t1 = load <8 x i32>, <8 x i32>* %a
|
|
%t2 = mul <8 x i32> %t1, %t1
|
|
store <8 x i32> %t2, <8 x i32>* %a
|
|
ret void
|
|
}
|