1
0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-11-01 16:33:37 +01:00
llvm-mirror/test/CodeGen/CellSPU/mul_ops.ll
Evan Cheng 25dcf9b830 Teach dag combine to fold the following transformation more aggressively:
(OP (trunc x), (trunc y)) -> (trunc (OP x, y))

Unfortunately this simple change causes dag combine to infinite looping. The problem is the shrink demanded ops optimization tend to canonicalize expressions in the opposite manner. That is badness. This patch disable those optimizations in dag combine but instead it is done as a late pass in sdisel.

This also exposes some deficiencies in dag combine and x86 setcc / brcond lowering. Teach them to look pass ISD::TRUNCATE in various places.

llvm-svn: 92849
2010-01-06 19:38:29 +00:00

89 lines
2.1 KiB
LLVM

; RUN: llc < %s -march=cellspu > %t1.s
; RUN: grep mpy %t1.s | count 44
; RUN: grep mpyu %t1.s | count 4
; RUN: grep mpyh %t1.s | count 10
; RUN: grep mpyhh %t1.s | count 2
; RUN: grep rotma %t1.s | count 12
; RUN: grep rotmahi %t1.s | count 4
; RUN: grep and %t1.s | count 2
; RUN: grep selb %t1.s | count 6
; RUN: grep fsmbi %t1.s | count 4
; RUN: grep shli %t1.s | count 4
; RUN: grep shlhi %t1.s | count 4
; RUN: grep ila %t1.s | count 2
target datalayout = "E-p:32:32:128-f64:64:128-f32:32:128-i64:32:128-i32:32:128-i16:16:128-i8:8:128-i1:8:128-a0:0:128-v128:128:128-s0:128:128"
target triple = "spu"
; 32-bit multiply instruction generation:
define <4 x i32> @mpy_v4i32_1(<4 x i32> %arg1, <4 x i32> %arg2) {
entry:
%A = mul <4 x i32> %arg1, %arg2
ret <4 x i32> %A
}
define <4 x i32> @mpy_v4i32_2(<4 x i32> %arg1, <4 x i32> %arg2) {
entry:
%A = mul <4 x i32> %arg2, %arg1
ret <4 x i32> %A
}
define <8 x i16> @mpy_v8i16_1(<8 x i16> %arg1, <8 x i16> %arg2) {
entry:
%A = mul <8 x i16> %arg1, %arg2
ret <8 x i16> %A
}
define <8 x i16> @mpy_v8i16_2(<8 x i16> %arg1, <8 x i16> %arg2) {
entry:
%A = mul <8 x i16> %arg2, %arg1
ret <8 x i16> %A
}
define <16 x i8> @mul_v16i8_1(<16 x i8> %arg1, <16 x i8> %arg2) {
entry:
%A = mul <16 x i8> %arg2, %arg1
ret <16 x i8> %A
}
define <16 x i8> @mul_v16i8_2(<16 x i8> %arg1, <16 x i8> %arg2) {
entry:
%A = mul <16 x i8> %arg1, %arg2
ret <16 x i8> %A
}
define i32 @mul_i32_1(i32 %arg1, i32 %arg2) {
entry:
%A = mul i32 %arg2, %arg1
ret i32 %A
}
define i32 @mul_i32_2(i32 %arg1, i32 %arg2) {
entry:
%A = mul i32 %arg1, %arg2
ret i32 %A
}
define i16 @mul_i16_1(i16 %arg1, i16 %arg2) {
entry:
%A = mul i16 %arg2, %arg1
ret i16 %A
}
define i16 @mul_i16_2(i16 %arg1, i16 %arg2) {
entry:
%A = mul i16 %arg1, %arg2
ret i16 %A
}
define i8 @mul_i8_1(i8 %arg1, i8 %arg2) {
entry:
%A = mul i8 %arg2, %arg1
ret i8 %A
}
define i8 @mul_i8_2(i8 %arg1, i8 %arg2) {
entry:
%A = mul i8 %arg1, %arg2
ret i8 %A
}