RPCS3/llvm
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm.git synced 2024-12-28 07:05:03 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

[FastISel][AArch64] Fix shift lowering for i8 and i16 value types.

Browse Source 
This fix changes the parameters #r and #s that are passed to the UBFM/SBFM
instruction to get the zero/sign-extension for free.

The original problem was that the shift left would use the 32-bit shift even for
i8/i16 value types, which could leave the upper bits set with "garbage" values.

The arithmetic shift right on the other side would use the wrong MSB as sign-bit
to determine what bits to shift into the value.

This fixes <rdar://problem/17907720>.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@214788 91177308-0d34-0410-b5e6-96231b3b80d8
This commit is contained in:
Juergen Ributzka 2014-08-04 21:49:51 +00:00
parent 3bc3e03f74
commit 2c68cde701
2 changed files with 29 additions and 21 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

28
lib/Target/AArch64/AArch64FastISel.cpp
View File

@ -2202,14 +2202,16 @@ unsigned AArch64FastISel::Emit_LSL_ri(MVT RetVT, unsigned Op0, bool Op0IsKill,
switch (RetVT.SimpleTy) {
default: return 0;
case MVT::i8:
Opc = AArch64::UBFMWri; ImmR = -Shift % 32; ImmS = 7 - Shift; break;
case MVT::i16:
Opc = AArch64::UBFMWri; ImmR = -Shift % 32; ImmS = 15 - Shift; break;
case MVT::i32:
RetVT = MVT::i32;
Opc = AArch64::UBFMWri; ImmR = -Shift % 32; ImmS = 31 - Shift; break;
case MVT::i64:
Opc = AArch64::UBFMXri; ImmR = -Shift % 64; ImmS = 63 - Shift; break;
}
RetVT.SimpleTy = std::max(MVT::i32, RetVT.SimpleTy);
return FastEmitInst_rii(Opc, TLI.getRegClassFor(RetVT), Op0, Op0IsKill, ImmR,
ImmS);
}
@ -2219,15 +2221,13 @@ unsigned AArch64FastISel::Emit_LSR_ri(MVT RetVT, unsigned Op0, bool Op0IsKill,
unsigned Opc, ImmS;
switch (RetVT.SimpleTy) {
default: return 0;
case MVT::i8:
case MVT::i16:
case MVT::i32:
RetVT = MVT::i32;
Opc = AArch64::UBFMWri; ImmS = 31; break;
case MVT::i64:
Opc = AArch64::UBFMXri; ImmS = 63; break;
case MVT::i8: Opc = AArch64::UBFMWri; ImmS = 7; break;
case MVT::i16: Opc = AArch64::UBFMWri; ImmS = 15; break;
case MVT::i32: Opc = AArch64::UBFMWri; ImmS = 31; break;
case MVT::i64: Opc = AArch64::UBFMXri; ImmS = 63; break;
}
RetVT.SimpleTy = std::max(MVT::i32, RetVT.SimpleTy);
return FastEmitInst_rii(Opc, TLI.getRegClassFor(RetVT), Op0, Op0IsKill, Shift,
ImmS);
}
@ -2237,15 +2237,13 @@ unsigned AArch64FastISel::Emit_ASR_ri(MVT RetVT, unsigned Op0, bool Op0IsKill,
unsigned Opc, ImmS;
switch (RetVT.SimpleTy) {
default: return 0;
case MVT::i8:
case MVT::i16:
case MVT::i32:
RetVT = MVT::i32;
Opc = AArch64::SBFMWri; ImmS = 31; break;
case MVT::i64:
Opc = AArch64::SBFMXri; ImmS = 63; break;
case MVT::i8: Opc = AArch64::SBFMWri; ImmS = 7; break;
case MVT::i16: Opc = AArch64::SBFMWri; ImmS = 15; break;
case MVT::i32: Opc = AArch64::SBFMWri; ImmS = 31; break;
case MVT::i64: Opc = AArch64::SBFMXri; ImmS = 63; break;
}
RetVT.SimpleTy = std::max(MVT::i32, RetVT.SimpleTy);
return FastEmitInst_rii(Opc, TLI.getRegClassFor(RetVT), Op0, Op0IsKill, Shift,
ImmS);
}

22
test/CodeGen/AArch64/fast-isel-shift.ll
View File

@ -1,14 +1,14 @@
; RUN: llc -fast-isel -fast-isel-abort -mtriple=arm64-apple-darwin < %s | FileCheck %s
; CHECK-LABEL: lsl_i8
; CHECK: lsl {{w[0-9]*}}, {{w[0-9]*}}, #4
; CHECK: ubfiz {{w[0-9]*}}, {{w[0-9]*}}, #4, #4
define zeroext i8 @lsl_i8(i8 %a) {
%1 = shl i8 %a, 4
ret i8 %1
}
; CHECK-LABEL: lsl_i16
; CHECK: lsl {{w[0-9]*}}, {{w[0-9]*}}, #8
; CHECK: ubfiz {{w[0-9]*}}, {{w[0-9]*}}, #8, #8
define zeroext i16 @lsl_i16(i16 %a) {
%1 = shl i16 %a, 8
ret i16 %1
@ -30,14 +30,14 @@ define i64 @lsl_i64(i64 %a) {
}
; CHECK-LABEL: lsr_i8
; CHECK: lsr {{w[0-9]*}}, {{w[0-9]*}}, #4
; CHECK: ubfx {{w[0-9]*}}, {{w[0-9]*}}, #4, #4
define zeroext i8 @lsr_i8(i8 %a) {
%1 = lshr i8 %a, 4
ret i8 %1
}
; CHECK-LABEL: lsr_i16
; CHECK: lsr {{w[0-9]*}}, {{w[0-9]*}}, #8
; CHECK: ubfx {{w[0-9]*}}, {{w[0-9]*}}, #8, #8
define zeroext i16 @lsr_i16(i16 %a) {
%1 = lshr i16 %a, 8
ret i16 %1
@ -59,14 +59,14 @@ define i64 @lsr_i64(i64 %a) {
}
; CHECK-LABEL: asr_i8
; CHECK: asr {{w[0-9]*}}, {{w[0-9]*}}, #4
; CHECK: sbfx {{w[0-9]*}}, {{w[0-9]*}}, #4, #4
define zeroext i8 @asr_i8(i8 %a) {
%1 = ashr i8 %a, 4
ret i8 %1
}
; CHECK-LABEL: asr_i16
; CHECK: asr {{w[0-9]*}}, {{w[0-9]*}}, #8
; CHECK: sbfx {{w[0-9]*}}, {{w[0-9]*}}, #8, #8
define zeroext i16 @asr_i16(i16 %a) {
%1 = ashr i16 %a, 8
ret i16 %1
@ -87,3 +87,13 @@ define i64 @asr_i64(i64 %a) {
ret i64 %1
}
; CHECK-LABEL: shift_test1
; CHECK: ubfiz {{w[0-9]*}}, {{w[0-9]*}}, #4, #4
; CHECK-NEXT: sbfx {{w[0-9]*}}, {{w[0-9]*}}, #4, #4
define i32 @shift_test1(i8 %a) {
%1 = shl i8 %a, 4
%2 = ashr i8 %1, 4
%3 = sext i8 %2 to i32
ret i32 %3
}