llvm-6502/test/CodeGen/R600/store.ll
Tom Stellard 8eaed0f63d R600/SI: Change all instruction assembly names to lowercase.
This matches the format produced by the AMD proprietary driver.

//==================================================================//
// Shell script for converting .ll test cases: (Pass the .ll files
   you want to convert to this script as arguments).
//==================================================================//

; This was necessary on my system so that A-Z in sed would match only
; upper case.  I'm not sure why.
export LC_ALL='C'

TEST_FILES="$*"

MATCHES=`grep -v Patterns SIInstructions.td | grep -o '"[A-Z0-9_]\+["e]' | grep -o '[A-Z0-9_]\+' | sort -r`

for f in $TEST_FILES; do
  # Check that there are SI tests:
  grep -q -e 'verde' -e 'bonaire' -e 'SI' -e 'tahiti' $f
  if [ $? -eq 0 ]; then
    for match in $MATCHES; do
      sed -i -e "s/\([ :]$match\)/\L\1/" $f
    done

    # Try to get check lines with partial instruction names
    sed -i 's/\(;[ ]*SI[A-Z\\-]*: \)\([A-Z_0-9]\+\)/\1\L\2/' $f
  fi
done

sed -i -e 's/bb0_1/BB0_1/g' ../../../test/CodeGen/R600/infinite-loop.ll
sed -i -e 's/SI-NOT: bfe/SI-NOT: {{[^@]}}bfe/g'../../../test/CodeGen/R600/llvm.AMDGPU.bfe.*32.ll ../../../test/CodeGen/R600/sext-in-reg.ll
sed -i -e 's/exp_IEEE/EXP_IEEE/g' ../../../test/CodeGen/R600/llvm.exp2.ll
sed -i -e 's/numVgprs/NumVgprs/g' ../../../test/CodeGen/R600/register-count-comments.ll
sed -i 's/\(; CHECK[-NOT]*: \)\([A-Z_0-9]\+\)/\1\L\2/' ../../../test/CodeGen/R600/select64.ll ../../../test/CodeGen/R600/sgpr-copy.ll

//==================================================================//
// Shell script for converting .td files (run this last)
//==================================================================//

export LC_ALL='C'
sed -i -e '/Patterns/!s/\("[A-Z0-9_]\+[ "e]\)/\L\1/g' SIInstructions.td
sed -i -e 's/"EXP/"exp/g' SIInstrInfo.td

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@221350 91177308-0d34-0410-b5e6-96231b3b80d8
2014-11-05 14:50:53 +00:00

365 lines
12 KiB
LLVM

; RUN: llc < %s -march=r600 -mcpu=redwood | FileCheck --check-prefix=EG-CHECK --check-prefix=FUNC %s
; RUN: llc < %s -march=r600 -mcpu=cayman | FileCheck --check-prefix=CM-CHECK --check-prefix=FUNC %s
; RUN: llc < %s -march=r600 -mcpu=verde -verify-machineinstrs | FileCheck --check-prefix=SI-CHECK --check-prefix=FUNC %s
;===------------------------------------------------------------------------===;
; Global Address Space
;===------------------------------------------------------------------------===;
; FUNC-LABEL: {{^}}store_i1:
; EG-CHECK: MEM_RAT MSKOR
; SI-CHECK: buffer_store_byte
define void @store_i1(i1 addrspace(1)* %out) {
entry:
store i1 true, i1 addrspace(1)* %out
ret void
}
; i8 store
; EG-CHECK-LABEL: {{^}}store_i8:
; EG-CHECK: MEM_RAT MSKOR T[[RW_GPR:[0-9]]].XW, T{{[0-9]}}.X
; EG-CHECK: VTX_READ_8 [[VAL:T[0-9]\.X]], [[VAL]]
; IG 0: Get the byte index and truncate the value
; EG-CHECK: AND_INT T{{[0-9]}}.[[BI_CHAN:[XYZW]]], KC0[2].Y, literal.x
; EG-CHECK-NEXT: AND_INT * T{{[0-9]}}.[[TRUNC_CHAN:[XYZW]]], [[VAL]], literal.y
; EG-CHECK-NEXT: 3(4.203895e-45), 255(3.573311e-43)
; IG 1: Truncate the calculated the shift amount for the mask
; EG-CHECK: LSHL * T{{[0-9]}}.[[SHIFT_CHAN:[XYZW]]], PV.[[BI_CHAN]], literal.x
; EG-CHECK-NEXT: 3
; IG 2: Shift the value and the mask
; EG-CHECK: LSHL T[[RW_GPR]].X, T{{[0-9]}}.[[TRUNC_CHAN]], PV.[[SHIFT_CHAN]]
; EG-CHECK: LSHL * T[[RW_GPR]].W, literal.x, PV.[[SHIFT_CHAN]]
; EG-CHECK-NEXT: 255
; IG 3: Initialize the Y and Z channels to zero
; XXX: An optimal scheduler should merge this into one of the prevous IGs.
; EG-CHECK: MOV T[[RW_GPR]].Y, 0.0
; EG-CHECK: MOV * T[[RW_GPR]].Z, 0.0
; SI-CHECK-LABEL: {{^}}store_i8:
; SI-CHECK: buffer_store_byte
define void @store_i8(i8 addrspace(1)* %out, i8 %in) {
entry:
store i8 %in, i8 addrspace(1)* %out
ret void
}
; i16 store
; EG-CHECK-LABEL: {{^}}store_i16:
; EG-CHECK: MEM_RAT MSKOR T[[RW_GPR:[0-9]]].XW, T{{[0-9]}}.X
; EG-CHECK: VTX_READ_16 [[VAL:T[0-9]\.X]], [[VAL]]
; IG 0: Get the byte index and truncate the value
; EG-CHECK: AND_INT T{{[0-9]}}.[[BI_CHAN:[XYZW]]], KC0[2].Y, literal.x
; EG-CHECK: AND_INT * T{{[0-9]}}.[[TRUNC_CHAN:[XYZW]]], [[VAL]], literal.y
; EG-CHECK-NEXT: 3(4.203895e-45), 65535(9.183409e-41)
; IG 1: Truncate the calculated the shift amount for the mask
; EG-CHECK: LSHL * T{{[0-9]}}.[[SHIFT_CHAN:[XYZW]]], PV.[[BI_CHAN]], literal.x
; EG-CHECK: 3
; IG 2: Shift the value and the mask
; EG-CHECK: LSHL T[[RW_GPR]].X, T{{[0-9]}}.[[TRUNC_CHAN]], PV.[[SHIFT_CHAN]]
; EG-CHECK: LSHL * T[[RW_GPR]].W, literal.x, PV.[[SHIFT_CHAN]]
; EG-CHECK-NEXT: 65535
; IG 3: Initialize the Y and Z channels to zero
; XXX: An optimal scheduler should merge this into one of the prevous IGs.
; EG-CHECK: MOV T[[RW_GPR]].Y, 0.0
; EG-CHECK: MOV * T[[RW_GPR]].Z, 0.0
; SI-CHECK-LABEL: {{^}}store_i16:
; SI-CHECK: buffer_store_short
define void @store_i16(i16 addrspace(1)* %out, i16 %in) {
entry:
store i16 %in, i16 addrspace(1)* %out
ret void
}
; EG-CHECK-LABEL: {{^}}store_v2i8:
; EG-CHECK: MEM_RAT MSKOR
; EG-CHECK-NOT: MEM_RAT MSKOR
; SI-CHECK-LABEL: {{^}}store_v2i8:
; SI-CHECK: buffer_store_byte
; SI-CHECK: buffer_store_byte
define void @store_v2i8(<2 x i8> addrspace(1)* %out, <2 x i32> %in) {
entry:
%0 = trunc <2 x i32> %in to <2 x i8>
store <2 x i8> %0, <2 x i8> addrspace(1)* %out
ret void
}
; EG-CHECK-LABEL: {{^}}store_v2i16:
; EG-CHECK: MEM_RAT_CACHELESS STORE_RAW
; CM-CHECK-LABEL: {{^}}store_v2i16:
; CM-CHECK: MEM_RAT_CACHELESS STORE_DWORD
; SI-CHECK-LABEL: {{^}}store_v2i16:
; SI-CHECK: buffer_store_short
; SI-CHECK: buffer_store_short
define void @store_v2i16(<2 x i16> addrspace(1)* %out, <2 x i32> %in) {
entry:
%0 = trunc <2 x i32> %in to <2 x i16>
store <2 x i16> %0, <2 x i16> addrspace(1)* %out
ret void
}
; EG-CHECK-LABEL: {{^}}store_v4i8:
; EG-CHECK: MEM_RAT_CACHELESS STORE_RAW
; CM-CHECK-LABEL: {{^}}store_v4i8:
; CM-CHECK: MEM_RAT_CACHELESS STORE_DWORD
; SI-CHECK-LABEL: {{^}}store_v4i8:
; SI-CHECK: buffer_store_byte
; SI-CHECK: buffer_store_byte
; SI-CHECK: buffer_store_byte
; SI-CHECK: buffer_store_byte
define void @store_v4i8(<4 x i8> addrspace(1)* %out, <4 x i32> %in) {
entry:
%0 = trunc <4 x i32> %in to <4 x i8>
store <4 x i8> %0, <4 x i8> addrspace(1)* %out
ret void
}
; floating-point store
; EG-CHECK-LABEL: {{^}}store_f32:
; EG-CHECK: MEM_RAT_CACHELESS STORE_RAW T{{[0-9]+\.X, T[0-9]+\.X}}, 1
; CM-CHECK-LABEL: {{^}}store_f32:
; CM-CHECK: MEM_RAT_CACHELESS STORE_DWORD T{{[0-9]+\.X, T[0-9]+\.X}}
; SI-CHECK-LABEL: {{^}}store_f32:
; SI-CHECK: buffer_store_dword
define void @store_f32(float addrspace(1)* %out, float %in) {
store float %in, float addrspace(1)* %out
ret void
}
; EG-CHECK-LABEL: {{^}}store_v4i16:
; EG-CHECK: MEM_RAT MSKOR
; EG-CHECK: MEM_RAT MSKOR
; EG-CHECK: MEM_RAT MSKOR
; EG-CHECK: MEM_RAT MSKOR
; EG-CHECK-NOT: MEM_RAT MSKOR
; SI-CHECK-LABEL: {{^}}store_v4i16:
; SI-CHECK: buffer_store_short
; SI-CHECK: buffer_store_short
; SI-CHECK: buffer_store_short
; SI-CHECK: buffer_store_short
; SI-CHECK-NOT: buffer_store_byte
define void @store_v4i16(<4 x i16> addrspace(1)* %out, <4 x i32> %in) {
entry:
%0 = trunc <4 x i32> %in to <4 x i16>
store <4 x i16> %0, <4 x i16> addrspace(1)* %out
ret void
}
; vec2 floating-point stores
; EG-CHECK-LABEL: {{^}}store_v2f32:
; EG-CHECK: MEM_RAT_CACHELESS STORE_RAW
; CM-CHECK-LABEL: {{^}}store_v2f32:
; CM-CHECK: MEM_RAT_CACHELESS STORE_DWORD
; SI-CHECK-LABEL: {{^}}store_v2f32:
; SI-CHECK: buffer_store_dwordx2
define void @store_v2f32(<2 x float> addrspace(1)* %out, float %a, float %b) {
entry:
%0 = insertelement <2 x float> <float 0.0, float 0.0>, float %a, i32 0
%1 = insertelement <2 x float> %0, float %b, i32 1
store <2 x float> %1, <2 x float> addrspace(1)* %out
ret void
}
; EG-CHECK-LABEL: {{^}}store_v4i32:
; EG-CHECK: MEM_RAT_CACHELESS STORE_RAW
; EG-CHECK-NOT: MEM_RAT_CACHELESS STORE_RAW
; CM-CHECK-LABEL: {{^}}store_v4i32:
; CM-CHECK: MEM_RAT_CACHELESS STORE_DWORD
; CM-CHECK-NOT: MEM_RAT_CACHELESS STORE_DWORD
; SI-CHECK-LABEL: {{^}}store_v4i32:
; SI-CHECK: buffer_store_dwordx4
define void @store_v4i32(<4 x i32> addrspace(1)* %out, <4 x i32> %in) {
entry:
store <4 x i32> %in, <4 x i32> addrspace(1)* %out
ret void
}
; FUNC-LABEL: {{^}}store_i64_i8:
; EG-CHECK: MEM_RAT MSKOR
; SI-CHECK: buffer_store_byte
define void @store_i64_i8(i8 addrspace(1)* %out, i64 %in) {
entry:
%0 = trunc i64 %in to i8
store i8 %0, i8 addrspace(1)* %out
ret void
}
; FUNC-LABEL: {{^}}store_i64_i16:
; EG-CHECK: MEM_RAT MSKOR
; SI-CHECK: buffer_store_short
define void @store_i64_i16(i16 addrspace(1)* %out, i64 %in) {
entry:
%0 = trunc i64 %in to i16
store i16 %0, i16 addrspace(1)* %out
ret void
}
;===------------------------------------------------------------------------===;
; Local Address Space
;===------------------------------------------------------------------------===;
; FUNC-LABEL: {{^}}store_local_i1:
; EG-CHECK: LDS_BYTE_WRITE
; SI-CHECK: ds_write_b8
define void @store_local_i1(i1 addrspace(3)* %out) {
entry:
store i1 true, i1 addrspace(3)* %out
ret void
}
; EG-CHECK-LABEL: {{^}}store_local_i8:
; EG-CHECK: LDS_BYTE_WRITE
; SI-CHECK-LABEL: {{^}}store_local_i8:
; SI-CHECK: ds_write_b8
define void @store_local_i8(i8 addrspace(3)* %out, i8 %in) {
store i8 %in, i8 addrspace(3)* %out
ret void
}
; EG-CHECK-LABEL: {{^}}store_local_i16:
; EG-CHECK: LDS_SHORT_WRITE
; SI-CHECK-LABEL: {{^}}store_local_i16:
; SI-CHECK: ds_write_b16
define void @store_local_i16(i16 addrspace(3)* %out, i16 %in) {
store i16 %in, i16 addrspace(3)* %out
ret void
}
; EG-CHECK-LABEL: {{^}}store_local_v2i16:
; EG-CHECK: LDS_WRITE
; CM-CHECK-LABEL: {{^}}store_local_v2i16:
; CM-CHECK: LDS_WRITE
; SI-CHECK-LABEL: {{^}}store_local_v2i16:
; SI-CHECK: ds_write_b16
; SI-CHECK: ds_write_b16
define void @store_local_v2i16(<2 x i16> addrspace(3)* %out, <2 x i16> %in) {
entry:
store <2 x i16> %in, <2 x i16> addrspace(3)* %out
ret void
}
; EG-CHECK-LABEL: {{^}}store_local_v4i8:
; EG-CHECK: LDS_WRITE
; CM-CHECK-LABEL: {{^}}store_local_v4i8:
; CM-CHECK: LDS_WRITE
; SI-CHECK-LABEL: {{^}}store_local_v4i8:
; SI-CHECK: ds_write_b8
; SI-CHECK: ds_write_b8
; SI-CHECK: ds_write_b8
; SI-CHECK: ds_write_b8
define void @store_local_v4i8(<4 x i8> addrspace(3)* %out, <4 x i8> %in) {
entry:
store <4 x i8> %in, <4 x i8> addrspace(3)* %out
ret void
}
; EG-CHECK-LABEL: {{^}}store_local_v2i32:
; EG-CHECK: LDS_WRITE
; EG-CHECK: LDS_WRITE
; CM-CHECK-LABEL: {{^}}store_local_v2i32:
; CM-CHECK: LDS_WRITE
; CM-CHECK: LDS_WRITE
; SI-CHECK-LABEL: {{^}}store_local_v2i32:
; SI-CHECK: ds_write_b64
define void @store_local_v2i32(<2 x i32> addrspace(3)* %out, <2 x i32> %in) {
entry:
store <2 x i32> %in, <2 x i32> addrspace(3)* %out
ret void
}
; EG-CHECK-LABEL: {{^}}store_local_v4i32:
; EG-CHECK: LDS_WRITE
; EG-CHECK: LDS_WRITE
; EG-CHECK: LDS_WRITE
; EG-CHECK: LDS_WRITE
; CM-CHECK-LABEL: {{^}}store_local_v4i32:
; CM-CHECK: LDS_WRITE
; CM-CHECK: LDS_WRITE
; CM-CHECK: LDS_WRITE
; CM-CHECK: LDS_WRITE
; SI-CHECK-LABEL: {{^}}store_local_v4i32:
; SI-CHECK: ds_write_b32
; SI-CHECK: ds_write_b32
; SI-CHECK: ds_write_b32
; SI-CHECK: ds_write_b32
define void @store_local_v4i32(<4 x i32> addrspace(3)* %out, <4 x i32> %in) {
entry:
store <4 x i32> %in, <4 x i32> addrspace(3)* %out
ret void
}
; FUNC-LABEL: {{^}}store_local_i64_i8:
; EG-CHECK: LDS_BYTE_WRITE
; SI-CHECK: ds_write_b8
define void @store_local_i64_i8(i8 addrspace(3)* %out, i64 %in) {
entry:
%0 = trunc i64 %in to i8
store i8 %0, i8 addrspace(3)* %out
ret void
}
; FUNC-LABEL: {{^}}store_local_i64_i16:
; EG-CHECK: LDS_SHORT_WRITE
; SI-CHECK: ds_write_b16
define void @store_local_i64_i16(i16 addrspace(3)* %out, i64 %in) {
entry:
%0 = trunc i64 %in to i16
store i16 %0, i16 addrspace(3)* %out
ret void
}
; The stores in this function are combined by the optimizer to create a
; 64-bit store with 32-bit alignment. This is legal for SI and the legalizer
; should not try to split the 64-bit store back into 2 32-bit stores.
;
; Evergreen / Northern Islands don't support 64-bit stores yet, so there should
; be two 32-bit stores.
; EG-CHECK-LABEL: {{^}}vecload2:
; EG-CHECK: MEM_RAT_CACHELESS STORE_RAW
; CM-CHECK-LABEL: {{^}}vecload2:
; CM-CHECK: MEM_RAT_CACHELESS STORE_DWORD
; SI-CHECK-LABEL: {{^}}vecload2:
; SI-CHECK: buffer_store_dwordx2
define void @vecload2(i32 addrspace(1)* nocapture %out, i32 addrspace(2)* nocapture %mem) #0 {
entry:
%0 = load i32 addrspace(2)* %mem, align 4
%arrayidx1.i = getelementptr inbounds i32 addrspace(2)* %mem, i64 1
%1 = load i32 addrspace(2)* %arrayidx1.i, align 4
store i32 %0, i32 addrspace(1)* %out, align 4
%arrayidx1 = getelementptr inbounds i32 addrspace(1)* %out, i64 1
store i32 %1, i32 addrspace(1)* %arrayidx1, align 4
ret void
}
attributes #0 = { nounwind "less-precise-fpmad"="false" "no-frame-pointer-elim"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "unsafe-fp-math"="false" "use-soft-float"="false" }
; When i128 was a legal type this program generated cannot select errors:
; FUNC-LABEL: {{^}}"i128-const-store":
; FIXME: We should be able to to this with one store instruction
; EG-CHECK: STORE_RAW
; EG-CHECK: STORE_RAW
; EG-CHECK: STORE_RAW
; EG-CHECK: STORE_RAW
; CM-CHECK: STORE_DWORD
; CM-CHECK: STORE_DWORD
; CM-CHECK: STORE_DWORD
; CM-CHECK: STORE_DWORD
; SI: buffer_store_dwordx2
; SI: buffer_store_dwordx2
define void @i128-const-store(i32 addrspace(1)* %out) {
entry:
store i32 1, i32 addrspace(1)* %out, align 4
%arrayidx2 = getelementptr inbounds i32 addrspace(1)* %out, i64 1
store i32 1, i32 addrspace(1)* %arrayidx2, align 4
%arrayidx4 = getelementptr inbounds i32 addrspace(1)* %out, i64 2
store i32 2, i32 addrspace(1)* %arrayidx4, align 4
%arrayidx6 = getelementptr inbounds i32 addrspace(1)* %out, i64 3
store i32 2, i32 addrspace(1)* %arrayidx6, align 4
ret void
}