radeon/llvm: add support for cos/sin intrinsic

Reviewed-by: Tom Stellard <thomas.stellard@amd.com>
This commit is contained in:
Vincent Lejeune 2012-10-03 22:28:59 +02:00
parent 876b42663c
commit 23e11ac835
3 changed files with 15 additions and 12 deletions

View file

@ -21,7 +21,6 @@ let TargetPrefix = "AMDGPU", isTarget = 1 in {
def int_AMDGPU_arl : Intrinsic<[llvm_i32_ty], [llvm_float_ty], [IntrNoMem]>;
def int_AMDGPU_cndlt : Intrinsic<[llvm_float_ty], [llvm_float_ty, llvm_float_ty, llvm_float_ty], [IntrNoMem]>;
def int_AMDGPU_cos : Intrinsic<[llvm_float_ty], [llvm_float_ty], [IntrNoMem]>;
def int_AMDGPU_div : Intrinsic<[llvm_float_ty], [llvm_float_ty, llvm_float_ty], [IntrNoMem]>;
def int_AMDGPU_dp4 : Intrinsic<[llvm_float_ty], [llvm_v4f32_ty, llvm_v4f32_ty], [IntrNoMem]>;
def int_AMDGPU_floor : Intrinsic<[llvm_float_ty], [llvm_float_ty], [IntrNoMem]>;
@ -35,7 +34,6 @@ let TargetPrefix = "AMDGPU", isTarget = 1 in {
def int_AMDGPU_seq : Intrinsic<[llvm_float_ty], [llvm_float_ty, llvm_float_ty], [IntrNoMem]>;
def int_AMDGPU_sgt : Intrinsic<[llvm_float_ty], [llvm_float_ty, llvm_float_ty], [IntrNoMem]>;
def int_AMDGPU_sge : Intrinsic<[llvm_float_ty], [llvm_float_ty, llvm_float_ty], [IntrNoMem]>;
def int_AMDGPU_sin : Intrinsic<[llvm_float_ty], [llvm_float_ty], [IntrNoMem]>;
def int_AMDGPU_sle : Intrinsic<[llvm_float_ty], [llvm_float_ty, llvm_float_ty], [IntrNoMem]>;
def int_AMDGPU_sne : Intrinsic<[llvm_float_ty], [llvm_float_ty, llvm_float_ty], [IntrNoMem]>;
def int_AMDGPU_ssg : Intrinsic<[llvm_float_ty], [llvm_float_ty], [IntrNoMem]>;

View file

@ -905,8 +905,13 @@ let Predicates = [isR600] in {
// Helper pattern for normalizing inputs to triginomic instructions for R700+
// cards.
class TRIG_eg <InstR600 trig, Intrinsic intr> : Pat<
(intr R600_Reg32:$src),
class COS_PAT <InstR600 trig> : Pat<
(fcos R600_Reg32:$src),
(trig (MUL (MOV_IMM_I32 (i32 ALU_LITERAL_X), CONST.TWO_PI_INV), R600_Reg32:$src))
>;
class SIN_PAT <InstR600 trig> : Pat<
(fsin R600_Reg32:$src),
(trig (MUL (MOV_IMM_I32 (i32 ALU_LITERAL_X), CONST.TWO_PI_INV), R600_Reg32:$src))
>;
@ -919,8 +924,8 @@ let Predicates = [isR700] in {
def COS_r700 : COS_Common<0x6F>;
// R700 normalizes inputs to SIN/COS the same as EG
def : TRIG_eg <SIN_r700, int_AMDGPU_sin>;
def : TRIG_eg <COS_r700, int_AMDGPU_cos>;
def : SIN_PAT <SIN_r700>;
def : COS_PAT <COS_r700>;
}
//===----------------------------------------------------------------------===//
@ -997,8 +1002,8 @@ let Predicates = [isEGorCayman] in {
def SSG_eg : SSG_Common<CNDGT_eg, CNDGE_eg>;
def TGSI_LIT_Z_eg : TGSI_LIT_Z_Common<MUL_LIT_eg, LOG_CLAMPED_eg, EXP_IEEE_eg>;
def : TRIG_eg <SIN_eg, int_AMDGPU_sin>;
def : TRIG_eg <COS_eg, int_AMDGPU_cos>;
def : SIN_PAT <SIN_eg>;
def : COS_PAT <COS_eg>;
def FLT_TO_INT_eg : FLT_TO_INT_Common<0x50> {
let Pattern = [];

View file

@ -1133,8 +1133,8 @@ void radeon_llvm_context_init(struct radeon_llvm_context * ctx)
bld_base->op_actions[TGSI_OPCODE_CLAMP].intr_name = "llvm.AMDIL.clamp.";
bld_base->op_actions[TGSI_OPCODE_CMP].emit = build_tgsi_intrinsic_nomem;
bld_base->op_actions[TGSI_OPCODE_CMP].intr_name = "llvm.AMDGPU.cndlt";
bld_base->op_actions[TGSI_OPCODE_COS].emit = build_tgsi_intrinsic_nomem;
bld_base->op_actions[TGSI_OPCODE_COS].intr_name = "llvm.AMDGPU.cos";
bld_base->op_actions[TGSI_OPCODE_COS].emit = build_tgsi_intrinsic_readonly;
bld_base->op_actions[TGSI_OPCODE_COS].intr_name = "llvm.cos.f32";
bld_base->op_actions[TGSI_OPCODE_DIV].emit = build_tgsi_intrinsic_nomem;
bld_base->op_actions[TGSI_OPCODE_DIV].intr_name = "llvm.AMDGPU.div";
bld_base->op_actions[TGSI_OPCODE_ELSE].emit = else_emit;
@ -1175,8 +1175,8 @@ void radeon_llvm_context_init(struct radeon_llvm_context * ctx)
bld_base->op_actions[TGSI_OPCODE_SLT].emit = emit_cmp;
bld_base->op_actions[TGSI_OPCODE_SNE].emit = emit_cmp;
bld_base->op_actions[TGSI_OPCODE_SGT].emit = emit_cmp;
bld_base->op_actions[TGSI_OPCODE_SIN].emit = build_tgsi_intrinsic_nomem;
bld_base->op_actions[TGSI_OPCODE_SIN].intr_name = "llvm.AMDGPU.sin";
bld_base->op_actions[TGSI_OPCODE_SIN].emit = build_tgsi_intrinsic_readonly;
bld_base->op_actions[TGSI_OPCODE_SIN].intr_name = "llvm.sin.f32";
bld_base->op_actions[TGSI_OPCODE_TEX].fetch_args = tex_fetch_args;
bld_base->op_actions[TGSI_OPCODE_TEX].intr_name = "llvm.AMDGPU.tex";
bld_base->op_actions[TGSI_OPCODE_TXB].fetch_args = tex_fetch_args;