diff options
Diffstat (limited to 'src/gallium/drivers/r600/sb/sb_expr.cpp')
| -rw-r--r-- | src/gallium/drivers/r600/sb/sb_expr.cpp | 611 |
1 files changed, 611 insertions, 0 deletions
diff --git a/src/gallium/drivers/r600/sb/sb_expr.cpp b/src/gallium/drivers/r600/sb/sb_expr.cpp new file mode 100644 index 00000000000..e3c7858c3e9 --- /dev/null +++ b/src/gallium/drivers/r600/sb/sb_expr.cpp @@ -0,0 +1,611 @@ +/* + * Copyright 2013 Vadim Girlin <[email protected]> + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: + * Vadim Girlin + */ + +#include <cmath> + +#include "sb_shader.h" + +namespace r600_sb { + +value* get_select_value_for_em(shader& sh, value* em) { + if (!em->def) + return NULL; + + node *predset = em->def; + if (!predset->is_pred_set()) + return NULL; + + alu_node *s = sh.clone(static_cast<alu_node*>(predset)); + convert_predset_to_set(sh, s); + + predset->insert_after(s); + + value* &d0 = s->dst[0]; + d0 = sh.create_temp_value(); + d0->def = s; + return d0; +} + +expr_handler::expr_handler(shader& sh) : sh(sh), vt(sh.vt) {} + +value * expr_handler::get_const(const literal &l) { + value *v = sh.get_const_value(l); + if (!v->gvn_source) + vt.add_value(v); + return v; +} + +void expr_handler::assign_source(value *dst, value *src) { + dst->gvn_source = src->gvn_source; +} + +bool expr_handler::equal(value *l, value *r) { + + assert(l != r); + + if (l->gvalue() == r->gvalue()) + return true; + + if (l->def && r->def) + return defs_equal(l, r); + + if (l->is_rel() && r->is_rel()) + return ivars_equal(l, r); + + return false; +} + +bool expr_handler::ivars_equal(value* l, value* r) { + if (l->rel->gvalue() == r->rel->gvalue() + && l->select == r->select) { + + vvec &lv = l->mdef.empty() ? l->muse : l->mdef; + vvec &rv = r->mdef.empty() ? r->muse : r->mdef; + + // FIXME: replace this with more precise aliasing test + return lv == rv; + } + return false; +} + +bool expr_handler::defs_equal(value* l, value* r) { + + node *d1 = l->def; + node *d2 = r->def; + + if (d1->type != d2->type || d1->subtype != d2->subtype) + return false; + + if (d1->is_pred_set() || d2->is_pred_set()) + return false; + + if (d1->type == NT_OP) { + switch (d1->subtype) { + case NST_ALU_INST: + return ops_equal( + static_cast<alu_node*>(d1), + static_cast<alu_node*>(d2)); +// case NST_FETCH_INST: return ops_equal(static_cast<fetch_node*>(d1), +// static_cast<fetch_node*>(d2); +// case NST_CF_INST: return ops_equal(static_cast<cf_node*>(d1), +// static_cast<cf_node*>(d2); + default: + break; + } + } + return false; +} + +bool expr_handler::try_fold(value* v) { + assert(!v->gvn_source); + + if (v->def) + try_fold(v->def); + + if (v->gvn_source) + return true; + + return false; +} + +bool expr_handler::try_fold(node* n) { + return n->fold_dispatch(this); +} + +bool expr_handler::fold(node& n) { + if (n.subtype == NST_PHI) { + + value *s = n.src[0]; + + // FIXME disabling phi folding for registers for now, otherwise we lose + // control flow information in some cases + // (GCM fails on tests/shaders/glsl-fs-if-nested-loop.shader_test) + // probably control flow transformation is required to enable it + if (s->is_sgpr()) + return false; + + for(vvec::iterator I = n.src.begin() + 1, E = n.src.end(); I != E; ++I) { + value *v = *I; + if (!s->v_equal(v)) + return false; + } + + assign_source(n.dst[0], s); + } else { + assert(n.subtype == NST_PSI); + assert(n.src.size() >= 6); + + value *s = n.src[2]; + assert(s->gvn_source); + + for(vvec::iterator I = n.src.begin() + 3, E = n.src.end(); I != E; I += 3) { + value *v = *(I+2); + if (!s->v_equal(v)) + return false; + } + assign_source(n.dst[0], s); + } + return true; +} + +bool expr_handler::fold(container_node& n) { + return false; +} + +bool expr_handler::fold_setcc(alu_node &n) { + + // TODO + + return false; +} + +bool expr_handler::fold(alu_node& n) { + + if (n.bc.op_ptr->flags & (AF_PRED | AF_KILL)) { + fold_setcc(n); + return false; + } + + switch (n.bc.op_ptr->src_count) { + case 1: return fold_alu_op1(n); + case 2: return fold_alu_op2(n); + case 3: return fold_alu_op3(n); + default: + assert(0); + } + return false; +} + +bool expr_handler::fold(fetch_node& n) { + + unsigned chan = 0; + for (vvec::iterator I = n.dst.begin(), E = n.dst.end(); I != E; ++I) { + value* &v = *I; + if (v) { + if (n.bc.dst_sel[chan] == SEL_0) + assign_source(*I, get_const(0.0f)); + else if (n.bc.dst_sel[chan] == SEL_1) + assign_source(*I, get_const(1.0f)); + } + ++chan; + } + return false; +} + +bool expr_handler::fold(cf_node& n) { + return false; +} + +void expr_handler::apply_alu_src_mod(const bc_alu &bc, unsigned src, + literal &v) { + const bc_alu_src &s = bc.src[src]; + + if (s.abs) + v = fabs(v.f); + if (s.neg) + v = -v.f; +} + +void expr_handler::apply_alu_dst_mod(const bc_alu &bc, literal &v) { + float omod_coeff[] = {2.0f, 4.0, 0.5f}; + + if (bc.omod) + v = v.f * omod_coeff[bc.omod - 1]; + if (bc.clamp) + v = float_clamp(v.f); +} + +bool expr_handler::args_equal(const vvec &l, const vvec &r) { + + assert(l.size() == r.size()); + + int s = l.size(); + + for (int k = 0; k < s; ++k) { + if (!l[k]->v_equal(r[k])) + return false; + } + + return true; +} + +bool expr_handler::ops_equal(const alu_node *l, const alu_node* r) { + const bc_alu &b0 = l->bc; + const bc_alu &b1 = r->bc; + + if (b0.op != b1.op) + return false; + + unsigned src_count = b0.op_ptr->src_count; + + if (b0.index_mode != b1.index_mode) + return false; + + if (b0.clamp != b1.clamp || b0.omod != b1.omod) + return false; + + for (unsigned s = 0; s < src_count; ++s) { + const bc_alu_src &s0 = b0.src[s]; + const bc_alu_src &s1 = b1.src[s]; + + if (s0.abs != s1.abs || s0.neg != s1.neg) + return false; + } + return args_equal(l->src, r->src); +} + +bool expr_handler::fold_alu_op1(alu_node& n) { + + assert(!n.src.empty()); + if (n.src.empty()) + return false; + + value* v0 = n.src[0]; + + assert(v0 && n.dst[0]); + + if (!v0->is_const()) { + if ((n.bc.op == ALU_OP1_MOV || n.bc.op == ALU_OP1_MOVA_INT || + n.bc.op == ALU_OP1_MOVA_GPR_INT) + && n.bc.clamp == 0 && n.bc.omod == 0 + && n.bc.src[0].abs == 0 && n.bc.src[0].neg == 0) { + assign_source(n.dst[0], v0); + return true; + } + return false; + } + + literal dv, cv = v0->get_const_value(); + apply_alu_src_mod(n.bc, 0, cv); + + switch (n.bc.op) { + case ALU_OP1_CEIL: dv = ceil(cv.f); break; + case ALU_OP1_COS: dv = cos(cv.f * 2.0f * M_PI); break; + case ALU_OP1_EXP_IEEE: dv = exp2(cv.f); break; + case ALU_OP1_FLOOR: dv = floor(cv.f); break; + case ALU_OP1_FLT_TO_INT: dv = (int)cv.f; break; // FIXME: round modes ???? + case ALU_OP1_FLT_TO_INT_FLOOR: dv = (int32_t)floor(cv.f); break; + case ALU_OP1_FLT_TO_INT_RPI: dv = (int32_t)floor(cv.f + 0.5f); break; + case ALU_OP1_FLT_TO_INT_TRUNC: dv = (int32_t)trunc(cv.f); break; + case ALU_OP1_FLT_TO_UINT: dv = (uint32_t)cv.f; break; + case ALU_OP1_FRACT: dv = cv.f - floor(cv.f); break; + case ALU_OP1_INT_TO_FLT: dv = (float)cv.i; break; + case ALU_OP1_LOG_CLAMPED: + case ALU_OP1_LOG_IEEE: + if (cv.f != 0.0f) + dv = log2(cv.f); + else + // don't fold to NAN, let the GPU handle it for now + // (prevents degenerate LIT tests from failing) + return false; + break; + case ALU_OP1_MOV: dv = cv; break; + case ALU_OP1_MOVA_INT: dv = cv; break; // FIXME ??? +// case ALU_OP1_MOVA_FLOOR: dv = (int32_t)floor(cv.f); break; +// case ALU_OP1_MOVA_GPR_INT: + case ALU_OP1_NOT_INT: dv = ~cv.i; break; + case ALU_OP1_PRED_SET_INV: + dv = cv.f == 0.0f ? 1.0f : (cv.f == 1.0f ? 0.0f : cv.f); break; + case ALU_OP1_PRED_SET_RESTORE: dv = cv; break; + case ALU_OP1_RECIPSQRT_CLAMPED: + case ALU_OP1_RECIPSQRT_FF: + case ALU_OP1_RECIPSQRT_IEEE: dv = 1.0f / sqrt(cv.f); break; + case ALU_OP1_RECIP_CLAMPED: + case ALU_OP1_RECIP_FF: + case ALU_OP1_RECIP_IEEE: dv = 1.0f / cv.f; break; +// case ALU_OP1_RECIP_INT: +// case ALU_OP1_RECIP_UINT: +// case ALU_OP1_RNDNE: dv = floor(cv.f + 0.5f); break; + case ALU_OP1_SIN: dv = sin(cv.f * 2.0f * M_PI); break; + case ALU_OP1_SQRT_IEEE: dv = sqrt(cv.f); break; + case ALU_OP1_TRUNC: dv = trunc(cv.f); break; + + default: + return false; + } + + apply_alu_dst_mod(n.bc, dv); + assign_source(n.dst[0], get_const(dv)); + return true; +} + +bool expr_handler::fold_alu_op2(alu_node& n) { + + if (n.src.size() < 2) + return false; + + value* v0 = n.src[0]; + value* v1 = n.src[1]; + + assert(v0 && v1 && n.dst[0]); + + bool isc0 = v0->is_const(); + bool isc1 = v1->is_const(); + + if (!isc0 && !isc1) + return false; + + literal dv, cv0, cv1; + + if (isc0) { + cv0 = v0->get_const_value(); + apply_alu_src_mod(n.bc, 0, cv0); + } + + if (isc1) { + cv1 = v1->get_const_value(); + apply_alu_src_mod(n.bc, 1, cv1); + } + + if (isc0 && isc1) { + switch (n.bc.op) { + case ALU_OP2_ADD: dv = cv0.f + cv1.f; break; + case ALU_OP2_ADDC_UINT: + dv = (uint32_t)(((uint64_t)cv0.u + cv1.u)>>32); break; + case ALU_OP2_ADD_INT: dv = cv0.i + cv1.i; break; + case ALU_OP2_AND_INT: dv = cv0.i & cv1.i; break; + case ALU_OP2_ASHR_INT: dv = cv0.i >> (cv1.i & 0x1F); break; + case ALU_OP2_BFM_INT: + dv = (((1 << (cv0.i & 0x1F)) - 1) << (cv1.i & 0x1F)); break; + case ALU_OP2_LSHL_INT: dv = cv0.i << cv1.i; break; + case ALU_OP2_LSHR_INT: dv = cv0.u >> cv1.u; break; + case ALU_OP2_MAX: + case ALU_OP2_MAX_DX10: dv = cv0.f > cv1.f ? cv0.f : cv1.f; break; + case ALU_OP2_MAX_INT: dv = cv0.i > cv1.i ? cv0.i : cv1.i; break; + case ALU_OP2_MAX_UINT: dv = cv0.u > cv1.u ? cv0.u : cv1.u; break; + case ALU_OP2_MIN: + case ALU_OP2_MIN_DX10: dv = cv0.f < cv1.f ? cv0.f : cv1.f; break; + case ALU_OP2_MIN_INT: dv = cv0.i < cv1.i ? cv0.i : cv1.i; break; + case ALU_OP2_MIN_UINT: dv = cv0.u < cv1.u ? cv0.u : cv1.u; break; + case ALU_OP2_MUL: + case ALU_OP2_MUL_IEEE: dv = cv0.f * cv1.f; break; + case ALU_OP2_MULHI_INT: + dv = (int32_t)(((int64_t)cv0.u * cv1.u)>>32); break; + case ALU_OP2_MULHI_UINT: + dv = (uint32_t)(((uint64_t)cv0.u * cv1.u)>>32); break; + case ALU_OP2_MULLO_INT: + dv = (int32_t)(((int64_t)cv0.u * cv1.u) & 0xFFFFFFFF); break; + case ALU_OP2_MULLO_UINT: + dv = (uint32_t)(((uint64_t)cv0.u * cv1.u) & 0xFFFFFFFF); break; + case ALU_OP2_OR_INT: dv = cv0.i | cv1.i; break; + case ALU_OP2_SUB_INT: dv = cv0.i - cv1.i; break; + case ALU_OP2_XOR_INT: dv = cv0.i ^ cv1.i; break; + + case ALU_OP2_SETE: dv = cv0.f == cv1.f ? 1.0f : 0.0f; break; + + default: + return false; + } + + } else { // one source is const + + // TODO handle 1 * anything, 0 * anything, 0 + anything, etc + + return false; + } + + apply_alu_dst_mod(n.bc, dv); + assign_source(n.dst[0], get_const(dv)); + return true; +} + +bool expr_handler::fold_alu_op3(alu_node& n) { + + if (n.src.size() < 3) + return false; + + // TODO handle CNDxx by some common path + + value* v0 = n.src[0]; + value* v1 = n.src[1]; + value* v2 = n.src[2]; + + assert(v0 && v1 && v2 && n.dst[0]); + + bool isc0 = v0->is_const(); + bool isc1 = v1->is_const(); + bool isc2 = v2->is_const(); + + if (!isc0 && !isc1 && !isc2) + return false; + + literal dv, cv0, cv1, cv2; + + if (isc0) { + cv0 = v0->get_const_value(); + apply_alu_src_mod(n.bc, 0, cv0); + } + + if (isc1) { + cv1 = v1->get_const_value(); + apply_alu_src_mod(n.bc, 1, cv1); + } + + if (isc2) { + cv2 = v2->get_const_value(); + apply_alu_src_mod(n.bc, 2, cv2); + } + + if (isc0 && isc1 && isc2) { + switch (n.bc.op) { + case ALU_OP3_MULADD: dv = cv0.f * cv1.f + cv2.f; break; + + // TODO + + default: + return false; + } + + } else { + + // TODO + + return false; + } + + apply_alu_dst_mod(n.bc, dv); + assign_source(n.dst[0], get_const(dv)); + return true; +} + +unsigned invert_setcc_condition(unsigned cc, bool &swap_args) { + unsigned ncc = 0; + + switch (cc) { + case AF_CC_E: ncc = AF_CC_NE; break; + case AF_CC_NE: ncc = AF_CC_E; break; + case AF_CC_GE: ncc = AF_CC_GT; swap_args = true; break; + case AF_CC_GT: ncc = AF_CC_GE; swap_args = true; break; + default: + assert(!"unexpected condition code"); + break; + } + return ncc; +} + +unsigned get_setcc_opcode(unsigned cc, unsigned cmp_type, bool int_dst) { + + if (int_dst && cmp_type == AF_FLOAT_CMP) { + switch (cc) { + case AF_CC_E: return ALU_OP2_SETE_DX10; + case AF_CC_NE: return ALU_OP2_SETNE_DX10; + case AF_CC_GT: return ALU_OP2_SETGT_DX10; + case AF_CC_GE: return ALU_OP2_SETGE_DX10; + } + } else { + + switch(cmp_type) { + case AF_FLOAT_CMP: { + switch (cc) { + case AF_CC_E: return ALU_OP2_SETE; + case AF_CC_NE: return ALU_OP2_SETNE; + case AF_CC_GT: return ALU_OP2_SETGT; + case AF_CC_GE: return ALU_OP2_SETGE; + } + break; + } + case AF_INT_CMP: { + switch (cc) { + case AF_CC_E: return ALU_OP2_SETE_INT; + case AF_CC_NE: return ALU_OP2_SETNE_INT; + case AF_CC_GT: return ALU_OP2_SETGT_INT; + case AF_CC_GE: return ALU_OP2_SETGE_INT; + } + break; + } + case AF_UINT_CMP: { + switch (cc) { + case AF_CC_GT: return ALU_OP2_SETGT_UINT; + case AF_CC_GE: return ALU_OP2_SETGE_UINT; + } + break; + } + } + } + + assert(!"unexpected cc&cmp_type combination"); + return ~0u; +} + +unsigned get_predsetcc_opcode(unsigned cc, unsigned cmp_type) { + + switch(cmp_type) { + case AF_FLOAT_CMP: { + switch (cc) { + case AF_CC_E: return ALU_OP2_PRED_SETE; + case AF_CC_NE: return ALU_OP2_PRED_SETNE; + case AF_CC_GT: return ALU_OP2_PRED_SETGT; + case AF_CC_GE: return ALU_OP2_PRED_SETGE; + } + break; + } + case AF_INT_CMP: { + switch (cc) { + case AF_CC_E: return ALU_OP2_PRED_SETE_INT; + case AF_CC_NE: return ALU_OP2_PRED_SETNE_INT; + case AF_CC_GT: return ALU_OP2_PRED_SETGT_INT; + case AF_CC_GE: return ALU_OP2_PRED_SETGE_INT; + } + break; + } + case AF_UINT_CMP: { + switch (cc) { + case AF_CC_GT: return ALU_OP2_PRED_SETGT_UINT; + case AF_CC_GE: return ALU_OP2_PRED_SETGE_UINT; + } + break; + } + } + + assert(!"unexpected cc&cmp_type combination"); + return ~0u; +} + +void convert_predset_to_set(shader& sh, alu_node* a) { + + unsigned flags = a->bc.op_ptr->flags; + unsigned cc = flags & AF_CC_MASK; + unsigned cmp_type = flags & AF_CMP_TYPE_MASK; + + bool swap_args = false; + + cc = invert_setcc_condition(cc, swap_args); + + unsigned newop = get_setcc_opcode(cc, cmp_type, true); + + a->dst.resize(1); + a->bc.set_op(newop); + + if (swap_args) { + std::swap(a->src[0], a->src[1]); + std::swap(a->bc.src[0], a->bc.src[1]); + } + + a->bc.update_exec_mask = 0; + a->bc.update_pred = 0; +} + +} // namespace r600_sb |