diff options
Diffstat (limited to 'src')
-rw-r--r-- | src/gallium/drivers/radeon/AMDGPUInstructions.td | 21 | ||||
-rw-r--r-- | src/gallium/drivers/radeon/R600CodeEmitter.cpp | 24 | ||||
-rw-r--r-- | src/gallium/drivers/radeon/R600ISelLowering.cpp | 5 | ||||
-rw-r--r-- | src/gallium/drivers/radeon/R600Instructions.td | 17 | ||||
-rw-r--r-- | src/gallium/drivers/radeon/R600LowerInstructions.cpp | 41 | ||||
-rw-r--r-- | src/gallium/drivers/radeon/SIInstructions.td | 17 |
6 files changed, 37 insertions, 88 deletions
diff --git a/src/gallium/drivers/radeon/AMDGPUInstructions.td b/src/gallium/drivers/radeon/AMDGPUInstructions.td index 10eceb6ce53..0433c8dcd95 100644 --- a/src/gallium/drivers/radeon/AMDGPUInstructions.td +++ b/src/gallium/drivers/radeon/AMDGPUInstructions.td @@ -84,6 +84,27 @@ class POW_Common <AMDGPUInst log_ieee, AMDGPUInst exp_ieee, AMDGPUInst mul, (exp_ieee (mul rc:$src1, (log_ieee rc:$src0))) >; +/* Other helper patterns */ +/* --------------------- */ + +/* Extract element pattern */ +class Extract_Element <ValueType sub_type, ValueType vec_type, + RegisterClass vec_class, int sub_idx, + SubRegIndex sub_reg>: Pat< + (sub_type (vector_extract (vec_type vec_class:$src), sub_idx)), + (EXTRACT_SUBREG vec_class:$src, sub_reg) +>; + +/* Insert element pattern */ +class Insert_Element <ValueType elem_type, ValueType vec_type, + RegisterClass elem_class, RegisterClass vec_class, + int sub_idx, SubRegIndex sub_reg> : Pat < + + (vec_type (vector_insert (vec_type vec_class:$vec), + (elem_type elem_class:$elem), sub_idx)), + (INSERT_SUBREG vec_class:$vec, elem_class:$elem, sub_reg) +>; + include "R600Instructions.td" include "SIInstrInfo.td" diff --git a/src/gallium/drivers/radeon/R600CodeEmitter.cpp b/src/gallium/drivers/radeon/R600CodeEmitter.cpp index 2367713500c..c951d9f3bad 100644 --- a/src/gallium/drivers/radeon/R600CodeEmitter.cpp +++ b/src/gallium/drivers/radeon/R600CodeEmitter.cpp @@ -297,15 +297,6 @@ void R600CodeEmitter::emitALUInstr(MachineInstr &MI) switch (MI.getOpcode()) { default: break; - /* Custom swizzle instructions, ignore the last two operands */ - case AMDIL::SET_CHAN: - numOperands = 2; - break; - - case AMDIL::VEXTRACT_v4f32: - numOperands = 2; - break; - /* XXX: Temp Hack */ case AMDIL::STORE_OUTPUT: numOperands = 2; @@ -369,13 +360,7 @@ void R600CodeEmitter::emitSrc(const MachineOperand & MO) if (isReduction) { emitByte(reductionElement); } else if (MO.isReg()) { - const MachineInstr * parent = MO.getParent(); - /* The source channel for EXTRACT is stored in operand 2. */ - if (parent->getOpcode() == AMDIL::VEXTRACT_v4f32) { - emitByte(parent->getOperand(2).getImm()); - } else { - emitByte(TRI->getHWRegChan(MO.getReg())); - } + emitByte(TRI->getHWRegChan(MO.getReg())); } else { emitByte(0); } @@ -418,10 +403,6 @@ void R600CodeEmitter::emitDst(const MachineOperand & MO) const MachineInstr * parent = MO.getParent(); if (isReduction) { emitByte(reductionElement); - - /* The destination element for SET_CHAN is stored in the 3rd operand. */ - } else if (parent->getOpcode() == AMDIL::SET_CHAN) { - emitByte(parent->getOperand(2).getImm()); } else if (parent->getOpcode() == AMDIL::VCREATE_v4f32) { emitByte(ELEMENT_X); } else { @@ -651,12 +632,9 @@ unsigned int R600CodeEmitter::getHWInst(const MachineInstr &MI) case AMDIL::STORE_OUTPUT: case AMDIL::VCREATE_v4i32: case AMDIL::VCREATE_v4f32: - case AMDIL::VEXTRACT_v4f32: - case AMDIL::VINSERT_v4f32: case AMDIL::LOADCONST_i32: case AMDIL::LOADCONST_f32: case AMDIL::MOVE_v4i32: - case AMDIL::SET_CHAN: /* Instructons to reinterpret bits as ... */ case AMDIL::IL_ASINT_f32: case AMDIL::IL_ASINT_i32: diff --git a/src/gallium/drivers/radeon/R600ISelLowering.cpp b/src/gallium/drivers/radeon/R600ISelLowering.cpp index 104f4c5aeb8..f92fe2641a5 100644 --- a/src/gallium/drivers/radeon/R600ISelLowering.cpp +++ b/src/gallium/drivers/radeon/R600ISelLowering.cpp @@ -23,6 +23,11 @@ R600TargetLowering::R600TargetLowering(TargetMachine &TM) : { setOperationAction(ISD::MUL, MVT::i64, Expand); // setSchedulingPreference(Sched::VLIW); + addRegisterClass(MVT::v4f32, &AMDIL::R600_Reg128RegClass); + addRegisterClass(MVT::f32, &AMDIL::R600_Reg32RegClass); + + setOperationAction(ISD::EXTRACT_VECTOR_ELT, MVT::v4f32, Legal); + setOperationAction(ISD::INSERT_VECTOR_ELT, MVT::v4f32, Legal); } MachineBasicBlock * R600TargetLowering::EmitInstrWithCustomInserter( diff --git a/src/gallium/drivers/radeon/R600Instructions.td b/src/gallium/drivers/radeon/R600Instructions.td index af6b8be5190..02043fdeea5 100644 --- a/src/gallium/drivers/radeon/R600Instructions.td +++ b/src/gallium/drivers/radeon/R600Instructions.td @@ -863,13 +863,6 @@ let isCodeGenOnly = 1 in { [] >; - def SET_CHAN : AMDGPUShaderInst < - (outs R600_Reg128:$dst), - (ins R600_Reg32:$src0, i32imm:$src1), - "SET_CHAN $dst, $src0, $src1", - [] - >; - def MULLIT : AMDGPUShaderInst < (outs R600_Reg128:$dst), (ins R600_Reg32:$src0, R600_Reg32:$src1, R600_Reg32:$src2), @@ -930,6 +923,16 @@ def LOAD_VTX : AMDGPUShaderInst < } //End isPseudo +def : Extract_Element <f32, v4f32, R600_Reg128, 0, sel_x>; +def : Extract_Element <f32, v4f32, R600_Reg128, 1, sel_y>; +def : Extract_Element <f32, v4f32, R600_Reg128, 2, sel_z>; +def : Extract_Element <f32, v4f32, R600_Reg128, 3, sel_w>; + +def : Insert_Element <f32, v4f32, R600_Reg32, R600_Reg128, 4, sel_x>; +def : Insert_Element <f32, v4f32, R600_Reg32, R600_Reg128, 5, sel_y>; +def : Insert_Element <f32, v4f32, R600_Reg32, R600_Reg128, 6, sel_z>; +def : Insert_Element <f32, v4f32, R600_Reg32, R600_Reg128, 7, sel_w>; + include "R600ShaderPatterns.td" diff --git a/src/gallium/drivers/radeon/R600LowerInstructions.cpp b/src/gallium/drivers/radeon/R600LowerInstructions.cpp index b9f9c7cdb4a..f26d5c87b02 100644 --- a/src/gallium/drivers/radeon/R600LowerInstructions.cpp +++ b/src/gallium/drivers/radeon/R600LowerInstructions.cpp @@ -317,10 +317,6 @@ bool R600LowerInstructionsPass::runOnMachineFunction(MachineFunction &MF) break; } - case AMDIL::VEXTRACT_v4f32: - MI.getOperand(2).setImm(MI.getOperand(2).getImm() - 1); - continue; - case AMDIL::NEGATE_i32: BuildMI(MBB, I, MBB.findDebugLoc(I), TII->get(AMDIL::SUB_INT)) .addOperand(MI.getOperand(0)) @@ -349,43 +345,6 @@ bool R600LowerInstructionsPass::runOnMachineFunction(MachineFunction &MF) break; } - case AMDIL::VINSERT_v4f32: - { - - int64_t swz = MI.getOperand(4).getImm(); - int64_t chan; - switch (swz) { - case (1 << 0): - chan = 0; - break; - case (1 << 8): - chan = 1; - break; - case (1 << 16): - chan = 2; - break; - case (1 << 24): - chan = 3; - break; - default: - chan = 0; - fprintf(stderr, "swizzle: %ld\n", swz); - abort(); - break; - } - BuildMI(MBB, I, MBB.findDebugLoc(I), - TM.getInstrInfo()->get(AMDIL::SET_CHAN)) - .addOperand(MI.getOperand(1)) - .addOperand(MI.getOperand(2)) - .addImm(chan); - - BuildMI(MBB, I, MBB.findDebugLoc(I), - TM.getInstrInfo()->get(AMDIL::COPY)) - .addOperand(MI.getOperand(0)) - .addOperand(MI.getOperand(1)); - break; - } - default: continue; } diff --git a/src/gallium/drivers/radeon/SIInstructions.td b/src/gallium/drivers/radeon/SIInstructions.td index 003d3d0b215..27a8b31757a 100644 --- a/src/gallium/drivers/radeon/SIInstructions.td +++ b/src/gallium/drivers/radeon/SIInstructions.td @@ -907,28 +907,11 @@ def : Pat < >; -/* Extract element pattern */ -class Extract_Element <ValueType sub_type, ValueType vec_type, - RegisterClass vec_class, int sub_idx, - SubRegIndex sub_reg>: Pat< - (sub_type (vector_extract (vec_type vec_class:$src), sub_idx)), - (EXTRACT_SUBREG vec_class:$src, sub_reg) ->; - def : Extract_Element <f32, v4f32, VReg_128, 0, sel_x>; def : Extract_Element <f32, v4f32, VReg_128, 1, sel_y>; def : Extract_Element <f32, v4f32, VReg_128, 2, sel_z>; def : Extract_Element <f32, v4f32, VReg_128, 3, sel_w>; -class Insert_Element <ValueType elem_type, ValueType vec_type, - RegisterClass elem_class, RegisterClass vec_class, - int sub_idx, SubRegIndex sub_reg> : Pat < - - (vec_type (vector_insert (vec_type vec_class:$vec), - (elem_type elem_class:$elem), sub_idx)), - (INSERT_SUBREG vec_class:$vec, elem_class:$elem, sub_reg) ->; - def : Insert_Element <f32, v4f32, VReg_32, VReg_128, 4, sel_x>; def : Insert_Element <f32, v4f32, VReg_32, VReg_128, 5, sel_y>; def : Insert_Element <f32, v4f32, VReg_32, VReg_128, 6, sel_z>; |