diff options
author | Tom Stellard <thomas.stellard@amd.com> | 2013-06-25 02:39:20 +0000 |
---|---|---|
committer | Tom Stellard <thomas.stellard@amd.com> | 2013-06-25 02:39:20 +0000 |
commit | 1cb1107c660bdade8b033bae10bf223d977691e5 (patch) | |
tree | 845973531768d7e392b4fe4708589efe9641aa30 /test/CodeGen/R600/packetizer.ll | |
parent | 746f7cafb2d1362de62024a6e62664c3eb3999d2 (diff) | |
download | external_llvm-1cb1107c660bdade8b033bae10bf223d977691e5.zip external_llvm-1cb1107c660bdade8b033bae10bf223d977691e5.tar.gz external_llvm-1cb1107c660bdade8b033bae10bf223d977691e5.tar.bz2 |
R600: Fix typo in R600Schedule.td
This should only make a difference in programs that use a lot of the
vector ALU instructions like BFI_INT and BIT_ALIGN. There is a slight
improvement in the phatk bitcoin mining kernel with this patch on
Evergreen (vector size == 1):
Before:
1173 Instruction Groups / 9520 dwords
After:
1167 Instruction Groups / 9510 dwords
Reviewed-by: Reviewed-by: Vincent Lejeune<vljn at ovi.com>
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@184819 91177308-0d34-0410-b5e6-96231b3b80d8
Diffstat (limited to 'test/CodeGen/R600/packetizer.ll')
-rw-r--r-- | test/CodeGen/R600/packetizer.ll | 34 |
1 files changed, 34 insertions, 0 deletions
diff --git a/test/CodeGen/R600/packetizer.ll b/test/CodeGen/R600/packetizer.ll new file mode 100644 index 0000000..0a405c5 --- /dev/null +++ b/test/CodeGen/R600/packetizer.ll @@ -0,0 +1,34 @@ +; RUN: llc < %s -march=r600 -mcpu=redwood | FileCheck %s +; RUN: llc < %s -march=r600 -mcpu=cayman | FileCheck %s + +; CHECK: @test +; CHECK: BIT_ALIGN_INT T{{[0-9]}}.X +; CHECK: BIT_ALIGN_INT T{{[0-9]}}.Y +; CHECK: BIT_ALIGN_INT T{{[0-9]}}.Z +; CHECK: BIT_ALIGN_INT * T{{[0-9]}}.W + +define void @test(i32 addrspace(1)* %out, i32 %x_arg, i32 %y_arg, i32 %z_arg, i32 %w_arg, i32 %e) { +entry: + %shl = sub i32 32, %e + %x = add i32 %x_arg, 1 + %x.0 = shl i32 %x, %shl + %x.1 = lshr i32 %x, %e + %x.2 = or i32 %x.0, %x.1 + %y = add i32 %y_arg, 1 + %y.0 = shl i32 %y, %shl + %y.1 = lshr i32 %y, %e + %y.2 = or i32 %y.0, %y.1 + %z = add i32 %z_arg, 1 + %z.0 = shl i32 %z, %shl + %z.1 = lshr i32 %z, %e + %z.2 = or i32 %z.0, %z.1 + %w = add i32 %w_arg, 1 + %w.0 = shl i32 %w, %shl + %w.1 = lshr i32 %w, %e + %w.2 = or i32 %w.0, %w.1 + %xy = or i32 %x.2, %y.2 + %zw = or i32 %z.2, %w.2 + %xyzw = or i32 %xy, %zw + store i32 %xyzw, i32 addrspace(1)* %out + ret void +} |