RPCS3/llvm-mirror
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-11-23 03:02:36 +01:00
Code Issues Projects Releases Wiki Activity

[ARM] Added more patterns to generate SSAT/USAT with shift

Browse Source 
Added patterns to generate an SSAT or USAT with shift for
SSAT/USAT instructions that are matched from IR patterns.

Differential Revision: https://reviews.llvm.org/D88145
This commit is contained in:
Meera Nakrani 2020-09-28 14:50:19 +00:00
parent a43296f995
commit 54f9731add
5 changed files with 85 additions and 20 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

16
lib/Target/ARM/ARMInstrInfo.td
View File

@ -167,9 +167,9 @@ def ARMcmov : SDNode<"ARMISD::CMOV", SDT_ARMCMov,
[SDNPInGlue]>;
def ARMsubs : SDNode<"ARMISD::SUBS", SDTIntBinOp, [SDNPOutGlue]>;
def ARMssatnoshift : SDNode<"ARMISD::SSAT", SDTIntSatNoShOp, []>;
def ARMssat : SDNode<"ARMISD::SSAT", SDTIntSatNoShOp, []>;
def ARMusatnoshift : SDNode<"ARMISD::USAT", SDTIntSatNoShOp, []>;
def ARMusat : SDNode<"ARMISD::USAT", SDTIntSatNoShOp, []>;
def ARMbrcond : SDNode<"ARMISD::BRCOND", SDT_ARMBrcond,
[SDNPHasChain, SDNPInGlue, SDNPOutGlue]>;
@ -4082,9 +4082,9 @@ def : ARMV6Pat<(int_arm_ssat GPRnopc:$a, imm1_32:$pos),
(SSAT imm1_32:$pos, GPRnopc:$a, 0)>;
def : ARMV6Pat<(int_arm_usat GPRnopc:$a, imm0_31:$pos),
(USAT imm0_31:$pos, GPRnopc:$a, 0)>;
def : ARMPat<(ARMssatnoshift GPRnopc:$Rn, imm0_31:$imm),
def : ARMPat<(ARMssat GPRnopc:$Rn, imm0_31:$imm),
(SSAT imm0_31:$imm, GPRnopc:$Rn, 0)>;
def : ARMPat<(ARMusatnoshift GPRnopc:$Rn, imm0_31:$imm),
def : ARMPat<(ARMusat GPRnopc:$Rn, imm0_31:$imm),
(USAT imm0_31:$imm, GPRnopc:$Rn, 0)>;
def : ARMV6Pat<(int_arm_ssat16 GPRnopc:$a, imm1_16:$pos),
(SSAT16 imm1_16:$pos, GPRnopc:$a)>;
@ -4098,6 +4098,14 @@ def : ARMV6Pat<(int_arm_usat (shl GPRnopc:$a, imm0_31:$shft), imm0_31:$pos),
(USAT imm0_31:$pos, GPRnopc:$a, imm0_31:$shft)>;
def : ARMV6Pat<(int_arm_usat (sra GPRnopc:$a, asr_imm:$shft), imm0_31:$pos),
(USAT imm0_31:$pos, GPRnopc:$a, asr_imm:$shft)>;
def : ARMPat<(ARMssat (shl GPRnopc:$Rn, imm0_31:$shft), imm0_31:$pos),
(SSAT imm0_31:$pos, GPRnopc:$Rn, imm0_31:$shft)>;
def : ARMPat<(ARMssat (sra GPRnopc:$Rn, asr_imm:$shft), imm0_31:$pos),
(SSAT imm0_31:$pos, GPRnopc:$Rn, asr_imm:$shft)>;
def : ARMPat<(ARMusat (shl GPRnopc:$Rn, imm0_31:$shft), imm0_31:$pos),
(USAT imm0_31:$pos, GPRnopc:$Rn, imm0_31:$shft)>;
def : ARMPat<(ARMusat (sra GPRnopc:$Rn, asr_imm:$shft), imm0_31:$pos),
(USAT imm0_31:$pos, GPRnopc:$Rn, asr_imm:$shft)>;
//===----------------------------------------------------------------------===//

12
lib/Target/ARM/ARMInstrThumb2.td
View File

@ -2624,9 +2624,9 @@ def t2USAT16: T2SatI<(ins imm0_15:$sat_imm, rGPR:$Rn),
let Inst{4} = 0;
}
def : T2Pat<(ARMssatnoshift GPRnopc:$Rn, imm0_31:$imm),
def : T2Pat<(ARMssat GPRnopc:$Rn, imm0_31:$imm),
(t2SSAT imm0_31:$imm, GPRnopc:$Rn, 0)>;
def : T2Pat<(ARMusatnoshift GPRnopc:$Rn, imm0_31:$imm),
def : T2Pat<(ARMusat GPRnopc:$Rn, imm0_31:$imm),
(t2USAT imm0_31:$imm, GPRnopc:$Rn, 0)>;
def : T2Pat<(int_arm_ssat GPR:$a, imm1_32:$pos),
(t2SSAT imm1_32:$pos, GPR:$a, 0)>;
@ -2644,6 +2644,14 @@ def : T2Pat<(int_arm_usat (shl GPRnopc:$a, imm0_31:$shft), imm0_31:$pos),
(t2USAT imm0_31:$pos, GPRnopc:$a, imm0_31:$shft)>;
def : T2Pat<(int_arm_usat (sra GPRnopc:$a, asr_imm:$shft), imm0_31:$pos),
(t2USAT imm0_31:$pos, GPRnopc:$a, asr_imm:$shft)>;
def : T2Pat<(ARMssat (shl GPRnopc:$a, imm0_31:$shft), imm0_31:$pos),
(t2SSAT imm0_31:$pos, GPRnopc:$a, imm0_31:$shft)>;
def : T2Pat<(ARMssat (sra GPRnopc:$Rn, asr_imm:$shft), imm0_31:$pos),
(t2SSAT imm0_31:$pos, GPRnopc:$Rn, asr_imm:$shft)>;
def : T2Pat<(ARMusat (shl GPRnopc:$a, imm0_31:$shft), imm0_31:$pos),
(t2USAT imm0_31:$pos, GPRnopc:$a, imm0_31:$shft)>;
def : T2Pat<(ARMusat (sra GPRnopc:$Rn, asr_imm:$shft), imm0_31:$pos),
(t2USAT imm0_31:$pos, GPRnopc:$Rn, asr_imm:$shft)>;
//===----------------------------------------------------------------------===//

28
test/CodeGen/ARM/ssat-with-shift.ll
View File

@ -24,4 +24,32 @@ entry:
ret i32 %0
}
define arm_aapcs_vfpcc i32 @ssat_lsl2(i32 %num){
; CHECK-LABEL: ssat_lsl2:
; CHECK: @ %bb.0: @ %entry
; CHECK-NEXT: ssat r0, #16, r0, lsl #15
; CHECK-NEXT: bx lr
entry:
%shl = shl nsw i32 %num, 15
%0 = icmp sgt i32 %shl, -32768
%1 = select i1 %0, i32 %shl, i32 -32768
%2 = icmp slt i32 %1, 32767
%3 = select i1 %2, i32 %1, i32 32767
ret i32 %3
}
define arm_aapcs_vfpcc i32 @ssat_asr2(i32 %num){
; CHECK-LABEL: ssat_asr2:
; CHECK: @ %bb.0: @ %entry
; CHECK-NEXT: ssat r0, #16, r0, asr #15
; CHECK-NEXT: bx lr
entry:
%shr = ashr i32 %num, 15
%0 = icmp sgt i32 %shr, -32768
%1 = select i1 %0, i32 %shr, i32 -32768
%2 = icmp slt i32 %1, 32767
%3 = select i1 %2, i32 %1, i32 32767
ret i32 %3
}
declare i32 @llvm.arm.ssat(i32, i32)

28
test/CodeGen/ARM/usat-with-shift.ll
View File

@ -24,4 +24,32 @@ entry:
ret i32 %0
}
define arm_aapcs_vfpcc i32 @usat_lsl2(i32 %num){
; CHECK-LABEL: usat_lsl2:
; CHECK: @ %bb.0: @ %entry
; CHECK-NEXT: usat r0, #15, r0, lsl #15
; CHECK-NEXT: bx lr
entry:
%shl = shl nsw i32 %num, 15
%0 = icmp sgt i32 %shl, 0
%1 = select i1 %0, i32 %shl, i32 0
%2 = icmp slt i32 %1, 32767
%3 = select i1 %2, i32 %1, i32 32767
ret i32 %3
}
define arm_aapcs_vfpcc i32 @usat_asr2(i32 %num){
; CHECK-LABEL: usat_asr2:
; CHECK: @ %bb.0: @ %entry
; CHECK-NEXT: usat r0, #15, r0, asr #15
; CHECK-NEXT: bx lr
entry:
%shr = ashr i32 %num, 15
%0 = icmp sgt i32 %shr, 0
%1 = select i1 %0, i32 %shr, i32 0
%2 = icmp slt i32 %1, 32767
%3 = select i1 %2, i32 %1, i32 32767
ret i32 %3
}
declare i32 @llvm.arm.usat(i32, i32)

21
test/CodeGen/Thumb2/mve-satmul-loops.ll
View File

@ -1145,8 +1145,7 @@ define arm_aapcs_vfpcc void @ssatmul_4_q15(i16* nocapture readonly %pSrcA, i16*
; CHECK-NEXT: ldrsh r0, [r12], #2
; CHECK-NEXT: ldrsh r1, [r6], #2
; CHECK-NEXT: muls r0, r1, r0
; CHECK-NEXT: asrs r0, r0, #15
; CHECK-NEXT: ssat r0, #16, r0
; CHECK-NEXT: ssat r0, #16, r0, asr #15
; CHECK-NEXT: strh r0, [r4], #2
; CHECK-NEXT: le lr, .LBB5_7
; CHECK-NEXT: .LBB5_8: @ %for.cond.cleanup
@ -1279,8 +1278,7 @@ define arm_aapcs_vfpcc void @ssatmul_8_q15(i16* nocapture readonly %pSrcA, i16*
; CHECK-NEXT: ldrsh r0, [r12], #2
; CHECK-NEXT: ldrsh r1, [r6], #2
; CHECK-NEXT: muls r0, r1, r0
; CHECK-NEXT: asrs r0, r0, #15
; CHECK-NEXT: ssat r0, #16, r0
; CHECK-NEXT: ssat r0, #16, r0, asr #15
; CHECK-NEXT: strh r0, [r4], #2
; CHECK-NEXT: le lr, .LBB6_7
; CHECK-NEXT: .LBB6_8: @ %for.cond.cleanup
@ -1410,8 +1408,7 @@ define arm_aapcs_vfpcc void @ssatmul_8i_q15(i16* nocapture readonly %pSrcA, i16*
; CHECK-NEXT: ldrsh r0, [r12], #2
; CHECK-NEXT: ldrsh r1, [r6], #2
; CHECK-NEXT: muls r0, r1, r0
; CHECK-NEXT: asrs r0, r0, #15
; CHECK-NEXT: ssat r0, #16, r0
; CHECK-NEXT: ssat r0, #16, r0, asr #15
; CHECK-NEXT: strh r0, [r4], #2
; CHECK-NEXT: le lr, .LBB7_7
; CHECK-NEXT: .LBB7_8: @ %for.cond.cleanup
@ -2218,8 +2215,7 @@ define arm_aapcs_vfpcc void @ssatmul_4_q7(i8* nocapture readonly %pSrcA, i8* noc
; CHECK-NEXT: ldrsb r0, [r12], #1
; CHECK-NEXT: ldrsb r1, [r6], #1
; CHECK-NEXT: muls r0, r1, r0
; CHECK-NEXT: asrs r0, r0, #7
; CHECK-NEXT: ssat r0, #8, r0
; CHECK-NEXT: ssat r0, #8, r0, asr #7
; CHECK-NEXT: strb r0, [r4], #1
; CHECK-NEXT: le lr, .LBB13_7
; CHECK-NEXT: .LBB13_8: @ %for.cond.cleanup
@ -2346,8 +2342,7 @@ define arm_aapcs_vfpcc void @ssatmul_8_q7(i8* nocapture readonly %pSrcA, i8* noc
; CHECK-NEXT: ldrsb r0, [r12], #1
; CHECK-NEXT: ldrsb r1, [r6], #1
; CHECK-NEXT: muls r0, r1, r0
; CHECK-NEXT: asrs r0, r0, #7
; CHECK-NEXT: ssat r0, #8, r0
; CHECK-NEXT: ssat r0, #8, r0, asr #7
; CHECK-NEXT: strb r0, [r4], #1
; CHECK-NEXT: le lr, .LBB14_7
; CHECK-NEXT: .LBB14_8: @ %for.cond.cleanup
@ -2480,8 +2475,7 @@ define arm_aapcs_vfpcc void @ssatmul_16_q7(i8* nocapture readonly %pSrcA, i8* no
; CHECK-NEXT: ldrsb r0, [r12], #1
; CHECK-NEXT: ldrsb r1, [r6], #1
; CHECK-NEXT: muls r0, r1, r0
; CHECK-NEXT: asrs r0, r0, #7
; CHECK-NEXT: ssat r0, #8, r0
; CHECK-NEXT: ssat r0, #8, r0, asr #7
; CHECK-NEXT: strb r0, [r4], #1
; CHECK-NEXT: le lr, .LBB15_7
; CHECK-NEXT: .LBB15_8: @ %for.cond.cleanup
@ -2611,8 +2605,7 @@ define arm_aapcs_vfpcc void @ssatmul_16i_q7(i8* nocapture readonly %pSrcA, i8* n
; CHECK-NEXT: ldrsb r0, [r12], #1
; CHECK-NEXT: ldrsb r1, [r6], #1
; CHECK-NEXT: muls r0, r1, r0
; CHECK-NEXT: asrs r0, r0, #7
; CHECK-NEXT: ssat r0, #8, r0
; CHECK-NEXT: ssat r0, #8, r0, asr #7
; CHECK-NEXT: strb r0, [r4], #1
; CHECK-NEXT: le lr, .LBB16_7
; CHECK-NEXT: .LBB16_8: @ %for.cond.cleanup