ShuriZma
/
xenia-canary
mirror of https://github.com/xenia-canary/xenia-canary.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Partial shader instr predication.

This commit is contained in:
Ben Vanik 2015-01-01 22:55:10 -08:00
parent ba0745abe2
commit 8b8d692f06
4 changed files with 133 additions and 64 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

158
src/xenia/gpu/gl4/gl4_shader_translator.cc
View File

@ -99,6 +99,8 @@ std::string GL4ShaderTranslator::TranslateVertexShader(
Append(" vec4 r%d = state.float_consts[%d];\n", n, n); Append(" vec4 r%d = state.float_consts[%d];\n", n, n);
} }
Append(" vec4 t;\n"); Append(" vec4 t;\n");
Append(" float s;\n"); // scalar result (used for RETAIN_PREV)
Append(" bool p = false;\n"); // predicate temp
// Execute blocks. // Execute blocks.
const auto& execs = vertex_shader->execs(); const auto& execs = vertex_shader->execs();
@ -132,7 +134,8 @@ std::string GL4ShaderTranslator::TranslatePixelShader(
Append(" vec4 r%d = state.float_consts[%d];\n", n, n + 256); Append(" vec4 r%d = state.float_consts[%d];\n", n, n + 256);
} }
Append(" vec4 t;\n"); Append(" vec4 t;\n");
Append(" float s;\n"); // scalar result (used for RETAIN_PREV) Append(" float s;\n"); // scalar result (used for RETAIN_PREV)
Append(" bool p = false;\n"); // predicate temp
// Bring registers local. // Bring registers local.
for (uint32_t n = 0; n < kMaxInterpolators; n++) { for (uint32_t n = 0; n < kMaxInterpolators; n++) {
@ -747,8 +750,7 @@ bool GL4ShaderTranslator::TranslateALU_MAX4v(const instr_alu_t& alu) {
// ... // ...
bool GL4ShaderTranslator::TranslateALU_MAXs(const instr_alu_t& alu) { bool GL4ShaderTranslator::TranslateALU_MAXs(const instr_alu_t& alu) {
AppendDestReg(get_alu_scalar_dest(alu), alu.scalar_write_mask, AppendDestReg(alu.scalar_dest, alu.scalar_write_mask, alu.export_data);
alu.export_data);
Append(" = "); Append(" = ");
if (alu.scalar_clamp) { if (alu.scalar_clamp) {
Append("clamp("); Append("clamp(");
@ -770,14 +772,12 @@ bool GL4ShaderTranslator::TranslateALU_MAXs(const instr_alu_t& alu) {
Append(", 0.0, 1.0)"); Append(", 0.0, 1.0)");
} }
Append(";\n"); Append(";\n");
AppendDestRegPost(get_alu_scalar_dest(alu), alu.scalar_write_mask, AppendDestRegPost(alu.scalar_dest, alu.scalar_write_mask, alu.export_data);
alu.export_data);
return true; return true;
} }
bool GL4ShaderTranslator::TranslateALU_MINs(const instr_alu_t& alu) { bool GL4ShaderTranslator::TranslateALU_MINs(const instr_alu_t& alu) {
AppendDestReg(get_alu_scalar_dest(alu), alu.scalar_write_mask, AppendDestReg(alu.scalar_dest, alu.scalar_write_mask, alu.export_data);
alu.export_data);
Append(" = "); Append(" = ");
if (alu.scalar_clamp) { if (alu.scalar_clamp) {
Append("clamp("); Append("clamp(");
@ -793,15 +793,13 @@ bool GL4ShaderTranslator::TranslateALU_MINs(const instr_alu_t& alu) {
Append(", 0.0, 1.0)"); Append(", 0.0, 1.0)");
} }
Append(";\n"); Append(";\n");
AppendDestRegPost(get_alu_scalar_dest(alu), alu.scalar_write_mask, AppendDestRegPost(alu.scalar_dest, alu.scalar_write_mask, alu.export_data);
alu.export_data);
return true; return true;
} }
bool GL4ShaderTranslator::TranslateALU_SETXXs(const instr_alu_t& alu, bool GL4ShaderTranslator::TranslateALU_SETXXs(const instr_alu_t& alu,
const char* op) { const char* op) {
AppendDestReg(get_alu_scalar_dest(alu), alu.scalar_write_mask, AppendDestReg(alu.scalar_dest, alu.scalar_write_mask, alu.export_data);
alu.export_data);
Append(" = "); Append(" = ");
if (alu.scalar_clamp) { if (alu.scalar_clamp) {
Append("clamp("); Append("clamp(");
@ -814,8 +812,7 @@ bool GL4ShaderTranslator::TranslateALU_SETXXs(const instr_alu_t& alu,
Append(", 0.0, 1.0)"); Append(", 0.0, 1.0)");
} }
Append(";\n"); Append(";\n");
AppendDestRegPost(get_alu_scalar_dest(alu), alu.scalar_write_mask, AppendDestRegPost(alu.scalar_dest, alu.scalar_write_mask, alu.export_data);
alu.export_data);
return true; return true;
} }
bool GL4ShaderTranslator::TranslateALU_SETEs(const instr_alu_t& alu) { bool GL4ShaderTranslator::TranslateALU_SETEs(const instr_alu_t& alu) {
@ -832,8 +829,7 @@ bool GL4ShaderTranslator::TranslateALU_SETNEs(const instr_alu_t& alu) {
} }
bool GL4ShaderTranslator::TranslateALU_EXP_IEEE(const instr_alu_t& alu) { bool GL4ShaderTranslator::TranslateALU_EXP_IEEE(const instr_alu_t& alu) {
AppendDestReg(get_alu_scalar_dest(alu), alu.scalar_write_mask, AppendDestReg(alu.scalar_dest, alu.scalar_write_mask, alu.export_data);
alu.export_data);
Append(" = "); Append(" = ");
if (alu.scalar_clamp) { if (alu.scalar_clamp) {
Append("clamp("); Append("clamp(");
@ -846,14 +842,12 @@ bool GL4ShaderTranslator::TranslateALU_EXP_IEEE(const instr_alu_t& alu) {
Append(", 0.0, 1.0)"); Append(", 0.0, 1.0)");
} }
Append(";\n"); Append(";\n");
AppendDestRegPost(get_alu_scalar_dest(alu), alu.scalar_write_mask, AppendDestRegPost(alu.scalar_dest, alu.scalar_write_mask, alu.export_data);
alu.export_data);
return true; return true;
} }
bool GL4ShaderTranslator::TranslateALU_RECIP_IEEE(const instr_alu_t& alu) { bool GL4ShaderTranslator::TranslateALU_RECIP_IEEE(const instr_alu_t& alu) {
AppendDestReg(get_alu_scalar_dest(alu), alu.scalar_write_mask, AppendDestReg(alu.scalar_dest, alu.scalar_write_mask, alu.export_data);
alu.export_data);
Append(" = "); Append(" = ");
if (alu.scalar_clamp) { if (alu.scalar_clamp) {
Append("clamp("); Append("clamp(");
@ -866,14 +860,47 @@ bool GL4ShaderTranslator::TranslateALU_RECIP_IEEE(const instr_alu_t& alu) {
Append(", 0.0, 1.0)"); Append(", 0.0, 1.0)");
} }
Append(";\n"); Append(";\n");
AppendDestRegPost(get_alu_scalar_dest(alu), alu.scalar_write_mask, AppendDestRegPost(alu.scalar_dest, alu.scalar_write_mask, alu.export_data);
alu.export_data);
return true; return true;
} }
bool GL4ShaderTranslator::TranslateALU_PRED_SETXXs(const instr_alu_t& alu,
const char* op) {
Append("p = ");
AppendSrcReg(alu.src3_reg, alu.src3_sel, alu.src3_swiz, alu.src3_reg_negate,
alu.abs_constants);
Append(".x %s 0.0;\n", op);
if (!alu.scalar_write_mask) {
return true;
}
AppendDestReg(alu.scalar_dest, alu.scalar_write_mask, alu.export_data);
Append(" = ");
if (alu.scalar_clamp) {
Append("clamp(");
}
Append("(p ? 0.0 : 1.0).xxxx");
if (alu.scalar_clamp) {
Append(", 0.0, 1.0)");
}
Append(";\n");
AppendDestRegPost(alu.scalar_dest, alu.scalar_write_mask, alu.export_data);
return true;
}
bool GL4ShaderTranslator::TranslateALU_PRED_SETEs(const instr_alu_t& alu) {
return TranslateALU_PRED_SETXXs(alu, "==");
}
bool GL4ShaderTranslator::TranslateALU_PRED_SETGTs(const instr_alu_t& alu) {
return TranslateALU_PRED_SETXXs(alu, ">");
}
bool GL4ShaderTranslator::TranslateALU_PRED_SETGTEs(const instr_alu_t& alu) {
return TranslateALU_PRED_SETXXs(alu, ">=");
}
bool GL4ShaderTranslator::TranslateALU_PRED_SETNEs(const instr_alu_t& alu) {
return TranslateALU_PRED_SETXXs(alu, "!=");
}
bool GL4ShaderTranslator::TranslateALU_SQRT_IEEE(const instr_alu_t& alu) { bool GL4ShaderTranslator::TranslateALU_SQRT_IEEE(const instr_alu_t& alu) {
AppendDestReg(get_alu_scalar_dest(alu), alu.scalar_write_mask, AppendDestReg(alu.scalar_dest, alu.scalar_write_mask, alu.export_data);
alu.export_data);
Append(" = "); Append(" = ");
if (alu.scalar_clamp) { if (alu.scalar_clamp) {
Append("clamp("); Append("clamp(");
@ -886,14 +913,12 @@ bool GL4ShaderTranslator::TranslateALU_SQRT_IEEE(const instr_alu_t& alu) {
Append(", 0.0, 1.0)"); Append(", 0.0, 1.0)");
} }
Append(";\n"); Append(";\n");
AppendDestRegPost(get_alu_scalar_dest(alu), alu.scalar_write_mask, AppendDestRegPost(alu.scalar_dest, alu.scalar_write_mask, alu.export_data);
alu.export_data);
return true; return true;
} }
bool GL4ShaderTranslator::TranslateALU_MUL_CONST_0(const instr_alu_t& alu) { bool GL4ShaderTranslator::TranslateALU_MUL_CONST_0(const instr_alu_t& alu) {
AppendDestReg(get_alu_scalar_dest(alu), alu.scalar_write_mask, AppendDestReg(alu.scalar_dest, alu.scalar_write_mask, alu.export_data);
alu.export_data);
Append(" = "); Append(" = ");
if (alu.scalar_clamp) { if (alu.scalar_clamp) {
Append("clamp("); Append("clamp(");
@ -913,8 +938,7 @@ bool GL4ShaderTranslator::TranslateALU_MUL_CONST_0(const instr_alu_t& alu) {
Append(", 0.0, 1.0)"); Append(", 0.0, 1.0)");
} }
Append(";\n"); Append(";\n");
AppendDestRegPost(get_alu_scalar_dest(alu), alu.scalar_write_mask, AppendDestRegPost(alu.scalar_dest, alu.scalar_write_mask, alu.export_data);
alu.export_data);
return true; return true;
} }
bool GL4ShaderTranslator::TranslateALU_MUL_CONST_1(const instr_alu_t& alu) { bool GL4ShaderTranslator::TranslateALU_MUL_CONST_1(const instr_alu_t& alu) {
@ -922,8 +946,7 @@ bool GL4ShaderTranslator::TranslateALU_MUL_CONST_1(const instr_alu_t& alu) {
} }
bool GL4ShaderTranslator::TranslateALU_ADD_CONST_0(const instr_alu_t& alu) { bool GL4ShaderTranslator::TranslateALU_ADD_CONST_0(const instr_alu_t& alu) {
AppendDestReg(get_alu_scalar_dest(alu), alu.scalar_write_mask, AppendDestReg(alu.scalar_dest, alu.scalar_write_mask, alu.export_data);
alu.export_data);
Append(" = "); Append(" = ");
if (alu.scalar_clamp) { if (alu.scalar_clamp) {
Append("clamp("); Append("clamp(");
@ -943,8 +966,7 @@ bool GL4ShaderTranslator::TranslateALU_ADD_CONST_0(const instr_alu_t& alu) {
Append(", 0.0, 1.0)"); Append(", 0.0, 1.0)");
} }
Append(";\n"); Append(";\n");
AppendDestRegPost(get_alu_scalar_dest(alu), alu.scalar_write_mask, AppendDestRegPost(alu.scalar_dest, alu.scalar_write_mask, alu.export_data);
alu.export_data);
return true; return true;
} }
bool GL4ShaderTranslator::TranslateALU_ADD_CONST_1(const instr_alu_t& alu) { bool GL4ShaderTranslator::TranslateALU_ADD_CONST_1(const instr_alu_t& alu) {
@ -952,8 +974,7 @@ bool GL4ShaderTranslator::TranslateALU_ADD_CONST_1(const instr_alu_t& alu) {
} }
bool GL4ShaderTranslator::TranslateALU_SUB_CONST_0(const instr_alu_t& alu) { bool GL4ShaderTranslator::TranslateALU_SUB_CONST_0(const instr_alu_t& alu) {
AppendDestReg(get_alu_scalar_dest(alu), alu.scalar_write_mask, AppendDestReg(alu.scalar_dest, alu.scalar_write_mask, alu.export_data);
alu.export_data);
Append(" = "); Append(" = ");
if (alu.scalar_clamp) { if (alu.scalar_clamp) {
Append("clamp("); Append("clamp(");
@ -973,8 +994,7 @@ bool GL4ShaderTranslator::TranslateALU_SUB_CONST_0(const instr_alu_t& alu) {
Append(", 0.0, 1.0)"); Append(", 0.0, 1.0)");
} }
Append(";\n"); Append(";\n");
AppendDestRegPost(get_alu_scalar_dest(alu), alu.scalar_write_mask, AppendDestRegPost(alu.scalar_dest, alu.scalar_write_mask, alu.export_data);
alu.export_data);
return true; return true;
} }
bool GL4ShaderTranslator::TranslateALU_SUB_CONST_1(const instr_alu_t& alu) { bool GL4ShaderTranslator::TranslateALU_SUB_CONST_1(const instr_alu_t& alu) {
@ -985,6 +1005,9 @@ bool GL4ShaderTranslator::TranslateALU_RETAIN_PREV(const instr_alu_t& alu) {
// TODO(benvanik): figure out how this is used. // TODO(benvanik): figure out how this is used.
// It seems like vector writes to export regs will use this to write 1's to // It seems like vector writes to export regs will use this to write 1's to
// components (like w in position). // components (like w in position).
if (!alu.scalar_write_mask) {
return true;
}
assert_true(alu.export_data == 1); assert_true(alu.export_data == 1);
AppendDestReg(alu.vector_dest, alu.scalar_write_mask, alu.export_data); AppendDestReg(alu.vector_dest, alu.scalar_write_mask, alu.export_data);
Append(" = "); Append(" = ");
@ -1065,10 +1088,10 @@ bool GL4ShaderTranslator::TranslateALU(const instr_alu_t* alu, int sync) {
ALU_INSTR(MOVA_FLOORs, 1), // 24 ALU_INSTR(MOVA_FLOORs, 1), // 24
ALU_INSTR(SUBs, 1), // 25 ALU_INSTR(SUBs, 1), // 25
ALU_INSTR(SUB_PREVs, 1), // 26 ALU_INSTR(SUB_PREVs, 1), // 26
ALU_INSTR(PRED_SETEs, 1), // 27 ALU_INSTR_IMPL(PRED_SETEs, 1), // 27
ALU_INSTR(PRED_SETNEs, 1), // 28 ALU_INSTR_IMPL(PRED_SETNEs, 1), // 28
ALU_INSTR(PRED_SETGTs, 1), // 29 ALU_INSTR_IMPL(PRED_SETGTs, 1), // 29
ALU_INSTR(PRED_SETGTEs, 1), // 30 ALU_INSTR_IMPL(PRED_SETGTEs, 1), // 30
ALU_INSTR(PRED_SET_INVs, 1), // 31 ALU_INSTR(PRED_SET_INVs, 1), // 31
ALU_INSTR(PRED_SET_POPs, 1), // 32 ALU_INSTR(PRED_SET_POPs, 1), // 32
ALU_INSTR(PRED_SET_CLRs, 1), // 33 ALU_INSTR(PRED_SET_CLRs, 1), // 33
@ -1093,11 +1116,6 @@ bool GL4ShaderTranslator::TranslateALU(const instr_alu_t* alu, int sync) {
#undef ALU_INSTR #undef ALU_INSTR
#undef ALU_INSTR_IMPL #undef ALU_INSTR_IMPL
if (!alu->scalar_write_mask && !alu->vector_write_mask) {
Append(" // <nop>\n");
return true;
}
if (alu->vector_write_mask) { if (alu->vector_write_mask) {
// Disassemble vector op. // Disassemble vector op.
const auto& iv = vector_alu_instrs[alu->vector_opc]; const auto& iv = vector_alu_instrs[alu->vector_opc];
@ -1154,7 +1172,7 @@ bool GL4ShaderTranslator::TranslateALU(const instr_alu_t* alu, int sync) {
} else { } else {
Append("\t \tOP(%u)\t", alu->scalar_opc); Append("\t \tOP(%u)\t", alu->scalar_opc);
} }
PrintDstReg(get_alu_scalar_dest(*alu), alu->scalar_write_mask, PrintDstReg(alu->scalar_dest, alu->scalar_write_mask,
alu->export_data); alu->export_data);
Append(" = "); Append(" = ");
if (is.num_srcs == 2) { if (is.num_srcs == 2) {
@ -1178,7 +1196,7 @@ bool GL4ShaderTranslator::TranslateALU(const instr_alu_t* alu, int sync) {
Append(" CLAMP"); Append(" CLAMP");
} }
if (alu->export_data) { if (alu->export_data) {
PrintExportComment(get_alu_scalar_dest(*alu)); PrintExportComment(alu->scalar_dest);
} }
Append("\n"); Append("\n");
@ -1196,7 +1214,7 @@ bool GL4ShaderTranslator::TranslateALU(const instr_alu_t* alu, int sync) {
return true; return true;
} }
void GL4ShaderTranslator::PrintDestFecth(uint32_t dst_reg, uint32_t dst_swiz) { void GL4ShaderTranslator::PrintDestFetch(uint32_t dst_reg, uint32_t dst_swiz) {
Append("\tR%u.", dst_reg); Append("\tR%u.", dst_reg);
for (int i = 0; i < 4; i++) { for (int i = 0; i < 4; i++) {
Append("%c", chan_names[dst_swiz & 0x7]); Append("%c", chan_names[dst_swiz & 0x7]);
@ -1247,10 +1265,14 @@ bool GL4ShaderTranslator::TranslateExec(const instr_cf_exec_t& cf) {
Append(" ABSOLUTE_ADDR"); Append(" ABSOLUTE_ADDR");
} }
if (cf.is_cond_exec()) { if (cf.is_cond_exec()) {
Append(" COND(%d)", cf.condition); Append(" COND(%d)", cf.pred_condition);
} }
Append("\n"); Append("\n");
if (cf.is_cond_exec()) {
Append(" if(%cp) {\n", cf.pred_condition ? ' ' : '!');
}
uint32_t sequence = cf.serialize; uint32_t sequence = cf.serialize;
for (uint32_t i = 0; i < cf.count; i++) { for (uint32_t i = 0; i < cf.count; i++) {
uint32_t alu_off = (cf.address + i); uint32_t alu_off = (cf.address + i);
@ -1260,14 +1282,22 @@ bool GL4ShaderTranslator::TranslateExec(const instr_cf_exec_t& cf) {
(const instr_fetch_t*)(dwords_ + alu_off * 3); (const instr_fetch_t*)(dwords_ + alu_off * 3);
switch (fetch->opc) { switch (fetch->opc) {
case VTX_FETCH: case VTX_FETCH:
AppendPredPre(cf.is_cond_exec(), cf.pred_condition,
fetch->vtx.pred_select, fetch->vtx.pred_condition);
if (!TranslateVertexFetch(&fetch->vtx, sync)) { if (!TranslateVertexFetch(&fetch->vtx, sync)) {
return false; return false;
} }
AppendPredPost(cf.is_cond_exec(), cf.pred_condition,
fetch->vtx.pred_select, fetch->vtx.pred_condition);
break; break;
case TEX_FETCH: case TEX_FETCH:
AppendPredPre(cf.is_cond_exec(), cf.pred_condition,
fetch->tex.pred_select, fetch->tex.pred_condition);
if (!TranslateTextureFetch(&fetch->tex, sync)) { if (!TranslateTextureFetch(&fetch->tex, sync)) {
return false; return false;
} }
AppendPredPost(cf.is_cond_exec(), cf.pred_condition,
fetch->tex.pred_select, fetch->tex.pred_condition);
break; break;
case TEX_GET_BORDER_COLOR_FRAC: case TEX_GET_BORDER_COLOR_FRAC:
case TEX_GET_COMP_TEX_LOD: case TEX_GET_COMP_TEX_LOD:
@ -1282,16 +1312,40 @@ bool GL4ShaderTranslator::TranslateExec(const instr_cf_exec_t& cf) {
} }
} else { } else {
const instr_alu_t* alu = (const instr_alu_t*)(dwords_ + alu_off * 3); const instr_alu_t* alu = (const instr_alu_t*)(dwords_ + alu_off * 3);
AppendPredPre(cf.is_cond_exec(), cf.pred_condition, alu->pred_select,
alu->pred_condition);
if (!TranslateALU(alu, sync)) { if (!TranslateALU(alu, sync)) {
return false; return false;
} }
AppendPredPost(cf.is_cond_exec(), cf.pred_condition, alu->pred_select,
alu->pred_condition);
} }
sequence >>= 2; sequence >>= 2;
} }
if (cf.is_cond_exec()) {
Append(" }\n");
}
return true; return true;
} }
void GL4ShaderTranslator::AppendPredPre(bool is_cond_cf, uint32_t cf_condition,
uint32_t pred_select,
uint32_t condition) {
if (pred_select && (!is_cond_cf || cf_condition != condition)) {
Append(" if (%cp) {\n", condition ? ' ' : '!');
}
}
void GL4ShaderTranslator::AppendPredPost(bool is_cond_cf, uint32_t cf_condition,
uint32_t pred_select,
uint32_t condition) {
if (pred_select && (!is_cond_cf || cf_condition != condition)) {
Append(" }\n");
}
}
bool GL4ShaderTranslator::TranslateVertexFetch(const instr_fetch_vtx_t* vtx, bool GL4ShaderTranslator::TranslateVertexFetch(const instr_fetch_vtx_t* vtx,
int sync) { int sync) {
static const struct { static const struct {
@ -1365,7 +1419,7 @@ bool GL4ShaderTranslator::TranslateVertexFetch(const instr_fetch_vtx_t* vtx,
if (vtx->pred_select) { if (vtx->pred_select) {
Append(vtx->pred_condition ? "EQ" : "NE"); Append(vtx->pred_condition ? "EQ" : "NE");
} }
PrintDestFecth(vtx->dst_reg, vtx->dst_swiz); PrintDestFetch(vtx->dst_reg, vtx->dst_swiz);
Append(" = R%u.", vtx->src_reg); Append(" = R%u.", vtx->src_reg);
Append("%c", chan_names[vtx->src_swiz & 0x3]); Append("%c", chan_names[vtx->src_swiz & 0x3]);
if (fetch_types[vtx->format].name) { if (fetch_types[vtx->format].name) {
@ -1481,7 +1535,7 @@ bool GL4ShaderTranslator::TranslateTextureFetch(const instr_fetch_tex_t* tex,
if (tex->pred_select) { if (tex->pred_select) {
Append(tex->pred_condition ? "EQ" : "NE"); Append(tex->pred_condition ? "EQ" : "NE");
} }
PrintDestFecth(tex->dst_reg, tex->dst_swiz); PrintDestFetch(tex->dst_reg, tex->dst_swiz);
Append(" = R%u.", tex->src_reg); Append(" = R%u.", tex->src_reg);
for (int i = 0; i < src_component_count; i++) { for (int i = 0; i < src_component_count; i++) {
Append("%c", chan_names[src_swiz & 0x3]); Append("%c", chan_names[src_swiz & 0x3]);

12
src/xenia/gpu/gl4/gl4_shader_translator.h
View File

@ -98,6 +98,11 @@ class GL4ShaderTranslator {
bool TranslateALU_SETNEs(const ucode::instr_alu_t& alu); bool TranslateALU_SETNEs(const ucode::instr_alu_t& alu);
bool TranslateALU_EXP_IEEE(const ucode::instr_alu_t& alu); bool TranslateALU_EXP_IEEE(const ucode::instr_alu_t& alu);
bool TranslateALU_RECIP_IEEE(const ucode::instr_alu_t& alu); bool TranslateALU_RECIP_IEEE(const ucode::instr_alu_t& alu);
bool TranslateALU_PRED_SETXXs(const ucode::instr_alu_t& alu, const char* op);
bool TranslateALU_PRED_SETEs(const ucode::instr_alu_t& alu);
bool TranslateALU_PRED_SETGTs(const ucode::instr_alu_t& alu);
bool TranslateALU_PRED_SETGTEs(const ucode::instr_alu_t& alu);
bool TranslateALU_PRED_SETNEs(const ucode::instr_alu_t& alu);
bool TranslateALU_SQRT_IEEE(const ucode::instr_alu_t& alu); bool TranslateALU_SQRT_IEEE(const ucode::instr_alu_t& alu);
bool TranslateALU_MUL_CONST_0(const ucode::instr_alu_t& alu); bool TranslateALU_MUL_CONST_0(const ucode::instr_alu_t& alu);
bool TranslateALU_MUL_CONST_1(const ucode::instr_alu_t& alu); bool TranslateALU_MUL_CONST_1(const ucode::instr_alu_t& alu);
@ -107,9 +112,14 @@ class GL4ShaderTranslator {
bool TranslateALU_SUB_CONST_1(const ucode::instr_alu_t& alu); bool TranslateALU_SUB_CONST_1(const ucode::instr_alu_t& alu);
bool TranslateALU_RETAIN_PREV(const ucode::instr_alu_t& alu); bool TranslateALU_RETAIN_PREV(const ucode::instr_alu_t& alu);
void PrintDestFecth(uint32_t dst_reg, uint32_t dst_swiz); void PrintDestFetch(uint32_t dst_reg, uint32_t dst_swiz);
void AppendFetchDest(uint32_t dst_reg, uint32_t dst_swiz); void AppendFetchDest(uint32_t dst_reg, uint32_t dst_swiz);
void AppendPredPre(bool is_cond_cf, uint32_t cf_condition,
uint32_t pred_select, uint32_t condition);
void AppendPredPost(bool is_cond_cf, uint32_t cf_condition,
uint32_t pred_select, uint32_t condition);
bool TranslateExec(const ucode::instr_cf_exec_t& cf); bool TranslateExec(const ucode::instr_cf_exec_t& cf);
bool TranslateVertexFetch(const ucode::instr_fetch_vtx_t* vtx, int sync); bool TranslateVertexFetch(const ucode::instr_fetch_vtx_t* vtx, int sync);
bool TranslateTextureFetch(const ucode::instr_fetch_tex_t* tex, int sync); bool TranslateTextureFetch(const ucode::instr_fetch_tex_t* tex, int sync);

9
src/xenia/gpu/ucode.h
View File

@ -234,7 +234,8 @@ XEPACKEDSTRUCT(instr_alu_t, {
uint32_t src3_reg_negate : 1; uint32_t src3_reg_negate : 1;
uint32_t src2_reg_negate : 1; uint32_t src2_reg_negate : 1;
uint32_t src1_reg_negate : 1; uint32_t src1_reg_negate : 1;
uint32_t pred_select : 2; uint32_t pred_condition : 1;
uint32_t pred_select : 1;
uint32_t relative_addr : 1; uint32_t relative_addr : 1;
uint32_t const_1_rel_abs : 1; uint32_t const_1_rel_abs : 1;
uint32_t const_0_rel_abs : 1; uint32_t const_0_rel_abs : 1;
@ -251,10 +252,6 @@ XEPACKEDSTRUCT(instr_alu_t, {
}); });
}); });
inline uint32_t get_alu_scalar_dest(const instr_alu_t& alu) {
return alu.vector_write_mask ? alu.scalar_dest : alu.vector_dest;
}
/* /*
* CF instructions: * CF instructions:
*/ */
@ -301,7 +298,7 @@ XEPACKEDSTRUCT(instr_cf_exec_t, {
XEPACKEDSTRUCTANONYMOUS({ XEPACKEDSTRUCTANONYMOUS({
uint32_t vc_lo : 2; /* vertex cache? */ uint32_t vc_lo : 2; /* vertex cache? */
uint32_t bool_addr : 8; uint32_t bool_addr : 8;
uint32_t condition : 1; uint32_t pred_condition : 1;
uint32_t address_mode : 1; // instr_addr_mode_t uint32_t address_mode : 1; // instr_addr_mode_t
uint32_t opc : 4; // instr_cf_opc_t uint32_t opc : 4; // instr_cf_opc_t
}); });

18
src/xenia/gpu/ucode_disassembler.cc
View File

@ -286,7 +286,9 @@ int disasm_alu(Output* output, const uint32_t* dwords, uint32_t alu_off,
if (alu->vector_clamp) { if (alu->vector_clamp) {
output->append(" CLAMP"); output->append(" CLAMP");
} }
if (alu->pred_select) {
output->append(" COND(%d)", alu->pred_condition);
}
if (alu->export_data) { if (alu->export_data) {
print_export_comment(output, alu->vector_dest, type); print_export_comment(output, alu->vector_dest, type);
} }
@ -308,7 +310,7 @@ int disasm_alu(Output* output, const uint32_t* dwords, uint32_t alu_off,
output->append("OP(%u)\t", alu->scalar_opc); output->append("OP(%u)\t", alu->scalar_opc);
} }
print_dstreg(output, get_alu_scalar_dest(*alu), alu->scalar_write_mask, print_dstreg(output, alu->scalar_dest, alu->scalar_write_mask,
alu->export_data); alu->export_data);
output->append(" = "); output->append(" = ");
if (scalar_instructions[alu->scalar_opc].num_srcs == 2) { if (scalar_instructions[alu->scalar_opc].num_srcs == 2) {
@ -335,7 +337,7 @@ int disasm_alu(Output* output, const uint32_t* dwords, uint32_t alu_off,
output->append(" CLAMP"); output->append(" CLAMP");
} }
if (alu->export_data) { if (alu->export_data) {
print_export_comment(output, get_alu_scalar_dest(*alu), type); print_export_comment(output, alu->scalar_dest, type);
} }
output->append("\n"); output->append("\n");
} }
@ -443,6 +445,9 @@ void print_fetch_vtx(Output* output, const instr_fetch_t* fetch) {
output->append(" OFFSET(%u)", vtx->offset); output->append(" OFFSET(%u)", vtx->offset);
} }
output->append(" CONST(%u, %u)", vtx->const_index, vtx->const_index_sel); output->append(" CONST(%u, %u)", vtx->const_index, vtx->const_index_sel);
if (vtx->pred_select) {
output->append(" COND(%d)", vtx->pred_condition);
}
if (1) { if (1) {
// XXX // XXX
output->append(" src_reg_am=%u", vtx->src_reg_am); output->append(" src_reg_am=%u", vtx->src_reg_am);
@ -537,6 +542,9 @@ void print_fetch_tex(Output* output, const instr_fetch_t* fetch) {
output->append(" OFFSET(%u,%u,%u)", tex->offset_x, tex->offset_y, output->append(" OFFSET(%u,%u,%u)", tex->offset_x, tex->offset_y,
tex->offset_z); tex->offset_z);
} }
if (tex->pred_select) {
output->append(" COND(%d)", tex->pred_condition);
}
} }
struct { struct {
@ -614,7 +622,7 @@ void print_cf_exec(Output* output, const instr_cf_t* cf) {
output->append(" ABSOLUTE_ADDR"); output->append(" ABSOLUTE_ADDR");
} }
if (cf->is_cond_exec()) { if (cf->is_cond_exec()) {
output->append(" COND(%d)", cf->exec.condition); output->append(" COND(%d)", cf->exec.pred_condition);
} }
} }
@ -713,7 +721,7 @@ void disasm_exec(Output* output, const uint32_t* dwords, size_t dword_count,
} }
std::string DisassembleShader(ShaderType type, const uint32_t* dwords, std::string DisassembleShader(ShaderType type, const uint32_t* dwords,
size_t dword_count) { size_t dword_count) {
Output* output = new Output(); Output* output = new Output();
instr_cf_t cfa; instr_cf_t cfa;