1
0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-11-26 12:43:36 +01:00
llvm-mirror/test/CodeGen/AMDGPU/mad_uint24.ll
Nicolai Hähnle 60121ef575 [AMDGPU][SelectionDAG] Don't combine uniform multiplies to MUL_[UI]24
Prefer to keep uniform (non-divergent) multiplies on the scalar ALU when
possible. This significantly improves some game cases by eliminating
v_readfirstlane instructions when the result feeds into a scalar
operation, like the address calculation for a scalar load or store.

Since isDivergent is only an approximation of whether a value is in
SGPRs, it can potentially regress some situations where a uniform value
ends up in a VGPR. These should be rare in real code, although the test
changes do contain a number of examples.

Most of the test changes are just using s_mul instead of v_mul/mad which
is generally better for both register pressure and latency (at least on
GFX10 where sgpr pressure doesn't affect occupancy and vector ALU
instructions have significantly longer latency than scalar ALU). Some
R600 tests now use MULLO_INT instead of MUL_UINT24.

GlobalISel appears to handle more scenarios in the desirable way,
although it can also be thrown off and fails to select the 24-bit
multiplies in some cases.

Alternative solution considered and rejected was to allow selecting
MUL_[UI]24 to S_MUL_I32. I've rejected this because the definition of
those SD operations works is don't-care on the most significant 8 bits,
and this fact is used in some combines via SimplifyDemandedBits.

Based on a patch by Nicolai Hähnle.

Differential Revision: https://reviews.llvm.org/D97063
2021-02-23 15:39:19 +00:00

319 lines
12 KiB
LLVM

; RUN: llc < %s -march=r600 -mcpu=redwood | FileCheck %s --check-prefix=EG --check-prefix=FUNC
; RUN: llc < %s -march=r600 -mcpu=cayman | FileCheck %s --check-prefix=EG --check-prefix=FUNC
; RUN: llc < %s -march=amdgcn -verify-machineinstrs | FileCheck %s --check-prefix=SI --check-prefix=FUNC --check-prefix=GCN
; RUN: llc < %s -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs | FileCheck %s --check-prefix=VI --check-prefix=FUNC --check-prefix=GCN --check-prefix=GCN2
; RUN: llc < %s -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs | FileCheck %s --check-prefix=VI --check-prefix=FUNC --check-prefix=GCN --check-prefix=GCN2
declare i32 @llvm.amdgcn.workitem.id.x() nounwind readnone
; FUNC-LABEL: {{^}}u32_mad24:
; EG: MULLO_INT
; SI: s_mul_i32
; SI: s_add_i32
; VI: s_mul_{{[iu]}}32
; VI: s_add_{{[iu]}}32
define amdgpu_kernel void @u32_mad24(i32 addrspace(1)* %out, i32 %a, i32 %b, i32 %c) {
entry:
%0 = shl i32 %a, 8
%a_24 = lshr i32 %0, 8
%1 = shl i32 %b, 8
%b_24 = lshr i32 %1, 8
%2 = mul i32 %a_24, %b_24
%3 = add i32 %2, %c
store i32 %3, i32 addrspace(1)* %out
ret void
}
; FUNC-LABEL: {{^}}i16_mad24:
; The order of A and B does not matter.
; EG: MULLO_INT {{[* ]*}}T{{[0-9]}}.[[MAD_CHAN:[XYZW]]]
; EG: ADD_INT {{[* ]*}}T{{[0-9]}}.[[MAD_CHAN:[XYZW]]]
; The result must be sign-extended
; EG: BFE_INT {{[* ]*}}T{{[0-9]\.[XYZW]}}, PV.[[MAD_CHAN]], 0.0, literal.x
; EG: 16
; GCN: s_mul_i32 [[MUL:s[0-9]]], {{[s][0-9], [s][0-9]}}
; GCN: s_add_i32 [[MAD:s[0-9]]], [[MUL]], s{{[0-9]}}
; GCN: s_sext_i32_i16 [[EXT:s[0-9]]], [[MAD]]
; GCN: v_mov_b32_e32 v0, [[EXT]]
define amdgpu_kernel void @i16_mad24(i32 addrspace(1)* %out, i16 %a, i16 %b, i16 %c) {
entry:
%0 = mul i16 %a, %b
%1 = add i16 %0, %c
%2 = sext i16 %1 to i32
store i32 %2, i32 addrspace(1)* %out
ret void
}
; FIXME: Need to handle non-uniform case for function below (load without gep).
; FUNC-LABEL: {{^}}i8_mad24:
; EG: MULLO_INT {{[* ]*}}T{{[0-9]}}.[[MAD_CHAN:[XYZW]]]
; EG: ADD_INT {{[* ]*}}T{{[0-9]}}.[[MAD_CHAN:[XYZW]]]
; The result must be sign-extended
; EG: BFE_INT {{[* ]*}}T{{[0-9]\.[XYZW]}}, PV.[[MAD_CHAN]], 0.0, literal.x
; EG: 8
; GCN: s_mul_i32 [[MUL:s[0-9]]], {{[s][0-9], [s][0-9]}}
; GCN: s_add_i32 [[MAD:s[0-9]]], [[MUL]], s{{[0-9]}}
; GCN: s_sext_i32_i8 [[EXT:s[0-9]]], [[MAD]]
; GCN: v_mov_b32_e32 v0, [[EXT]]
define amdgpu_kernel void @i8_mad24(i32 addrspace(1)* %out, i8 %a, i8 %b, i8 %c) {
entry:
%0 = mul i8 %a, %b
%1 = add i8 %0, %c
%2 = sext i8 %1 to i32
store i32 %2, i32 addrspace(1)* %out
ret void
}
; This tests for a bug where the mad_u24 pattern matcher would call
; SimplifyDemandedBits on the first operand of the mul instruction
; assuming that the pattern would be matched to a 24-bit mad. This
; led to some instructions being incorrectly erased when the entire
; 24-bit mad pattern wasn't being matched.
; Check that the select instruction is not deleted.
; FUNC-LABEL: {{^}}i24_i32_i32_mad:
; EG: CNDE_INT
; SI: v_cndmask
; GCN2: s_cselect
define amdgpu_kernel void @i24_i32_i32_mad(i32 addrspace(1)* %out, i32 %a, i32 %b, i32 %c, i32 %d) {
entry:
%0 = ashr i32 %a, 8
%1 = icmp ne i32 %c, 0
%2 = select i1 %1, i32 %0, i32 34
%3 = mul i32 %2, %c
%4 = add i32 %3, %d
store i32 %4, i32 addrspace(1)* %out
ret void
}
; FUNC-LABEL: {{^}}extra_and:
; SI-NOT: v_and
; SI: s_mul_i32
; SI: s_mul_i32
; SI: s_add_i32
; SI: s_add_i32
define amdgpu_kernel void @extra_and(i32 addrspace(1)* %arg, i32 %arg2, i32 %arg3) {
bb:
br label %bb4
bb4: ; preds = %bb4, %bb
%tmp = phi i32 [ 0, %bb ], [ %tmp13, %bb4 ]
%tmp5 = phi i32 [ 0, %bb ], [ %tmp13, %bb4 ]
%tmp6 = phi i32 [ 0, %bb ], [ %tmp15, %bb4 ]
%tmp7 = phi i32 [ 0, %bb ], [ %tmp15, %bb4 ]
%tmp8 = and i32 %tmp7, 16777215
%tmp9 = and i32 %tmp6, 16777215
%tmp10 = and i32 %tmp5, 16777215
%tmp11 = and i32 %tmp, 16777215
%tmp12 = mul i32 %tmp8, %tmp11
%tmp13 = add i32 %arg2, %tmp12
%tmp14 = mul i32 %tmp9, %tmp11
%tmp15 = add i32 %arg3, %tmp14
%tmp16 = add nuw nsw i32 %tmp13, %tmp15
%tmp17 = icmp eq i32 %tmp16, 8
br i1 %tmp17, label %bb18, label %bb4
bb18: ; preds = %bb4
store i32 %tmp16, i32 addrspace(1)* %arg
ret void
}
; FUNC-LABEL: {{^}}dont_remove_shift
; SI: s_lshr
; SI: s_mul_i32
; SI: s_mul_i32
; SI: s_add_i32
; SI: s_add_i32
define amdgpu_kernel void @dont_remove_shift(i32 addrspace(1)* %arg, i32 %arg2, i32 %arg3) {
bb:
br label %bb4
bb4: ; preds = %bb4, %bb
%tmp = phi i32 [ 0, %bb ], [ %tmp13, %bb4 ]
%tmp5 = phi i32 [ 0, %bb ], [ %tmp13, %bb4 ]
%tmp6 = phi i32 [ 0, %bb ], [ %tmp15, %bb4 ]
%tmp7 = phi i32 [ 0, %bb ], [ %tmp15, %bb4 ]
%tmp8 = lshr i32 %tmp7, 8
%tmp9 = lshr i32 %tmp6, 8
%tmp10 = lshr i32 %tmp5, 8
%tmp11 = lshr i32 %tmp, 8
%tmp12 = mul i32 %tmp8, %tmp11
%tmp13 = add i32 %arg2, %tmp12
%tmp14 = mul i32 %tmp9, %tmp11
%tmp15 = add i32 %arg3, %tmp14
%tmp16 = add nuw nsw i32 %tmp13, %tmp15
%tmp17 = icmp eq i32 %tmp16, 8
br i1 %tmp17, label %bb18, label %bb4
bb18: ; preds = %bb4
store i32 %tmp16, i32 addrspace(1)* %arg
ret void
}
; FUNC-LABEL: {{^}}i8_mad_sat_16:
; EG: MULLO_INT {{[* ]*}}T{{[0-9]}}.[[MAD_CHAN:[XYZW]]]
; EG: ADD_INT {{[* ]*}}T{{[0-9]}}.[[MAD_CHAN:[XYZW]]]
; The result must be sign-extended
; EG: BFE_INT {{[* ]*}}T{{[0-9]\.[XYZW]}}, PV.[[MAD_CHAN]], 0.0, literal.x
; EG: 8
; SI: v_mad_u32_u24 [[MAD:v[0-9]]], {{[sv][0-9], [sv][0-9]}}
; VI: v_mad_u16 [[MAD:v[0-9]]], {{[sv][0-9], [sv][0-9]}}
; GCN: v_bfe_i32 [[EXT:v[0-9]]], [[MAD]], 0, 16
; GCN: v_med3_i32 v{{[0-9]}}, [[EXT]],
define amdgpu_kernel void @i8_mad_sat_16(i8 addrspace(1)* %out, i8 addrspace(1)* %in0, i8 addrspace(1)* %in1, i8 addrspace(1)* %in2, i64 addrspace(5)* %idx) {
entry:
%retval.0.i = load i64, i64 addrspace(5)* %idx
%arrayidx = getelementptr inbounds i8, i8 addrspace(1)* %in0, i64 %retval.0.i
%arrayidx2 = getelementptr inbounds i8, i8 addrspace(1)* %in1, i64 %retval.0.i
%arrayidx4 = getelementptr inbounds i8, i8 addrspace(1)* %in2, i64 %retval.0.i
%l1 = load i8, i8 addrspace(1)* %arrayidx, align 1
%l2 = load i8, i8 addrspace(1)* %arrayidx2, align 1
%l3 = load i8, i8 addrspace(1)* %arrayidx4, align 1
%conv1.i = sext i8 %l1 to i16
%conv3.i = sext i8 %l2 to i16
%conv5.i = sext i8 %l3 to i16
%mul.i.i.i = mul nsw i16 %conv3.i, %conv1.i
%add.i.i = add i16 %mul.i.i.i, %conv5.i
%c4 = icmp sgt i16 %add.i.i, -128
%cond.i.i = select i1 %c4, i16 %add.i.i, i16 -128
%c5 = icmp slt i16 %cond.i.i, 127
%cond13.i.i = select i1 %c5, i16 %cond.i.i, i16 127
%conv8.i = trunc i16 %cond13.i.i to i8
%arrayidx7 = getelementptr inbounds i8, i8 addrspace(1)* %out, i64 %retval.0.i
store i8 %conv8.i, i8 addrspace(1)* %arrayidx7, align 1
ret void
}
; FUNC-LABEL: {{^}}i8_mad_32:
; EG: MULLO_INT {{[* ]*}}T{{[0-9]}}.[[MAD_CHAN:[XYZW]]]
; EG: ADD_INT {{[* ]*}}T{{[0-9]}}.[[MAD_CHAN:[XYZW]]]
; The result must be sign-extended
; EG: BFE_INT {{[* ]*}}T{{[0-9]\.[XYZW]}}, PV.[[MAD_CHAN]], 0.0, literal.x
; EG: 8
; SI: v_mad_u32_u24 [[MAD:v[0-9]]], {{[sv][0-9], [sv][0-9]}}
; VI: v_mad_u16 [[MAD:v[0-9]]], {{[sv][0-9], [sv][0-9]}}
; GCN: v_bfe_i32 [[EXT:v[0-9]]], [[MAD]], 0, 16
define amdgpu_kernel void @i8_mad_32(i32 addrspace(1)* %out, i8 addrspace(1)* %a, i8 addrspace(1)* %b, i8 addrspace(1)* %c, i64 addrspace(5)* %idx) {
entry:
%retval.0.i = load i64, i64 addrspace(5)* %idx
%arrayidx = getelementptr inbounds i8, i8 addrspace(1)* %a, i64 %retval.0.i
%arrayidx2 = getelementptr inbounds i8, i8 addrspace(1)* %b, i64 %retval.0.i
%arrayidx4 = getelementptr inbounds i8, i8 addrspace(1)* %c, i64 %retval.0.i
%la = load i8, i8 addrspace(1)* %arrayidx, align 1
%lb = load i8, i8 addrspace(1)* %arrayidx2, align 1
%lc = load i8, i8 addrspace(1)* %arrayidx4, align 1
%exta = sext i8 %la to i16
%extb = sext i8 %lb to i16
%extc = sext i8 %lc to i16
%mul = mul i16 %exta, %extb
%mad = add i16 %mul, %extc
%mad_ext = sext i16 %mad to i32
store i32 %mad_ext, i32 addrspace(1)* %out
ret void
}
; FUNC-LABEL: {{^}}i8_mad_64:
; EG: MULLO_INT {{[* ]*}}T{{[0-9]}}.[[MAD_CHAN:[XYZW]]]
; EG: ADD_INT {{[* ]*}}T{{[0-9]}}.[[MAD_CHAN:[XYZW]]]
; The result must be sign-extended
; EG: BFE_INT {{[* ]*}}T{{[0-9]\.[XYZW]}}, PV.[[MAD_CHAN]], 0.0, literal.x
; EG: 8
; SI: v_mad_u32_u24 [[MAD:v[0-9]]], {{[sv][0-9], [sv][0-9]}}
; VI: v_mad_u16 [[MAD:v[0-9]]], {{[sv][0-9], [sv][0-9]}}
; GCN: v_bfe_i32 [[EXT:v[0-9]]], [[MAD]], 0, 16
define amdgpu_kernel void @i8_mad_64(i64 addrspace(1)* %out, i8 addrspace(1)* %a, i8 addrspace(1)* %b, i8 addrspace(1)* %c, i64 addrspace(5)* %idx) {
entry:
%retval.0.i = load i64, i64 addrspace(5)* %idx
%arrayidx = getelementptr inbounds i8, i8 addrspace(1)* %a, i64 %retval.0.i
%arrayidx2 = getelementptr inbounds i8, i8 addrspace(1)* %b, i64 %retval.0.i
%arrayidx4 = getelementptr inbounds i8, i8 addrspace(1)* %c, i64 %retval.0.i
%la = load i8, i8 addrspace(1)* %arrayidx, align 1
%lb = load i8, i8 addrspace(1)* %arrayidx2, align 1
%lc = load i8, i8 addrspace(1)* %arrayidx4, align 1
%exta = sext i8 %la to i16
%extb = sext i8 %lb to i16
%extc = sext i8 %lc to i16
%mul = mul i16 %exta, %extb
%mad = add i16 %mul, %extc
%mad_ext = sext i16 %mad to i64
store i64 %mad_ext, i64 addrspace(1)* %out
ret void
}
; The ands are asserting the high bits are 0. SimplifyDemandedBits on
; the adds would remove the ands before the target combine on the mul
; had a chance to form mul24. The mul combine would then see
; extractelement with no known bits and fail. All of the mul/add
; combos in this loop should form v_mad_u32_u24.
; FUNC-LABEL: {{^}}mad24_known_bits_destroyed:
; GCN: v_mad_u32_u24
; GCN: v_mad_u32_u24
; GCN: v_mad_u32_u24
; GCN: v_mad_u32_u24
; GCN: v_mad_u32_u24
; GCN: v_mad_u32_u24
; GCN: v_mad_u32_u24
; GCN: v_mad_u32_u24
define void @mad24_known_bits_destroyed(i32 %arg, <4 x i32> %arg1, <4 x i32> %arg2, <4 x i32> %arg3, i32 %arg4, i32 %arg5, i32 %arg6, i32 addrspace(1)* %arg7, <4 x i32> addrspace(1)* %arg8) #0 {
bb:
%tmp = and i32 %arg4, 16777215
%tmp9 = extractelement <4 x i32> %arg1, i64 1
%tmp10 = extractelement <4 x i32> %arg3, i64 1
%tmp11 = and i32 %tmp9, 16777215
%tmp12 = extractelement <4 x i32> %arg1, i64 2
%tmp13 = extractelement <4 x i32> %arg3, i64 2
%tmp14 = and i32 %tmp12, 16777215
%tmp15 = extractelement <4 x i32> %arg1, i64 3
%tmp16 = extractelement <4 x i32> %arg3, i64 3
%tmp17 = and i32 %tmp15, 16777215
br label %bb19
bb18: ; preds = %bb19
ret void
bb19: ; preds = %bb19, %bb
%tmp20 = phi i32 [ %arg, %bb ], [ %tmp40, %bb19 ]
%tmp21 = phi i32 [ 0, %bb ], [ %tmp54, %bb19 ]
%tmp22 = phi <4 x i32> [ %arg2, %bb ], [ %tmp53, %bb19 ]
%tmp23 = and i32 %tmp20, 16777215
%tmp24 = mul i32 %tmp23, %tmp
%tmp25 = add i32 %tmp24, %arg5
%tmp26 = extractelement <4 x i32> %tmp22, i64 1
%tmp27 = and i32 %tmp26, 16777215
%tmp28 = mul i32 %tmp27, %tmp11
%tmp29 = add i32 %tmp28, %tmp10
%tmp30 = extractelement <4 x i32> %tmp22, i64 2
%tmp31 = and i32 %tmp30, 16777215
%tmp32 = mul i32 %tmp31, %tmp14
%tmp33 = add i32 %tmp32, %tmp13
%tmp34 = extractelement <4 x i32> %tmp22, i64 3
%tmp35 = and i32 %tmp34, 16777215
%tmp36 = mul i32 %tmp35, %tmp17
%tmp37 = add i32 %tmp36, %tmp16
%tmp38 = and i32 %tmp25, 16777215
%tmp39 = mul i32 %tmp38, %tmp
%tmp40 = add i32 %tmp39, %arg5
store i32 %tmp40, i32 addrspace(1)* %arg7
%tmp41 = insertelement <4 x i32> undef, i32 %tmp40, i32 0
%tmp42 = and i32 %tmp29, 16777215
%tmp43 = mul i32 %tmp42, %tmp11
%tmp44 = add i32 %tmp43, %tmp10
%tmp45 = insertelement <4 x i32> %tmp41, i32 %tmp44, i32 1
%tmp46 = and i32 %tmp33, 16777215
%tmp47 = mul i32 %tmp46, %tmp14
%tmp48 = add i32 %tmp47, %tmp13
%tmp49 = insertelement <4 x i32> %tmp45, i32 %tmp48, i32 2
%tmp50 = and i32 %tmp37, 16777215
%tmp51 = mul i32 %tmp50, %tmp17
%tmp52 = add i32 %tmp51, %tmp16
%tmp53 = insertelement <4 x i32> %tmp49, i32 %tmp52, i32 3
store <4 x i32> %tmp53, <4 x i32> addrspace(1)* %arg8
%tmp54 = add nuw nsw i32 %tmp21, 1
%tmp55 = icmp eq i32 %tmp54, %arg6
br i1 %tmp55, label %bb18, label %bb19
}
attributes #0 = { norecurse nounwind }