Mirrors/citra
1
0
Fork 0
mirror of https://github.com/PabloMK7/citra synced 2024-11-15 05:08:23 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

shader_jit_a64: Optimize conditional tests (#229)

Browse source 
These conditional tests are a 1:1 translation from the x64 code but do
not have to be. Reference-values are known at emit-time and can be
embedded as an immediate into an `EOR` instruction rather than moved
into a register. The `TST` instruction can be utilized to more optimally
test and update the `EQ`/`NE` status flags.
This commit is contained in:
Wunk 2024-08-15 01:45:20 -07:00 committed by GitHub
parent e55e619328
commit f248fefe06
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
2 changed files with 19 additions and 18 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

34
src/video_core/shader/shader_jit_a64_compiler.cpp
View file

@ -386,35 +386,33 @@ void JitShader::Compile_SanitizedMul(QReg src1, QReg src2, QReg scratch0) {
} }
void JitShader::Compile_EvaluateCondition(Instruction instr) { void JitShader::Compile_EvaluateCondition(Instruction instr) {
// Note: NXOR is used below to check for equality const u8 refx = instr.flow_control.refx.Value();
const u8 refy = instr.flow_control.refy.Value();
switch (instr.flow_control.op) { switch (instr.flow_control.op) {
// Note: NXOR is used below to check for equality
case Instruction::FlowControlType::Or: case Instruction::FlowControlType::Or:
MOV(XSCRATCH0, (instr.flow_control.refx.Value() ^ 1)); EOR(XSCRATCH0, COND0, refx ^ 1);
MOV(XSCRATCH1, (instr.flow_control.refy.Value() ^ 1)); EOR(XSCRATCH1, COND1, refy ^ 1);
EOR(XSCRATCH0, XSCRATCH0, COND0);
EOR(XSCRATCH1, XSCRATCH1, COND1);
ORR(XSCRATCH0, XSCRATCH0, XSCRATCH1); ORR(XSCRATCH0, XSCRATCH0, XSCRATCH1);
CMP(XSCRATCH0, 0);
break; break;
// Note: TST will AND two registers and set the EQ/NE flags on the result
case Instruction::FlowControlType::And: case Instruction::FlowControlType::And:
MOV(XSCRATCH0, (instr.flow_control.refx.Value() ^ 1)); EOR(XSCRATCH0, COND0, refx ^ 1);
MOV(XSCRATCH1, (instr.flow_control.refy.Value() ^ 1)); EOR(XSCRATCH1, COND1, refy ^ 1);
EOR(XSCRATCH0, XSCRATCH0, COND0); TST(XSCRATCH0, XSCRATCH1);
EOR(XSCRATCH1, XSCRATCH1, COND1);
AND(XSCRATCH0, XSCRATCH0, XSCRATCH1);
break; break;
case Instruction::FlowControlType::JustX: case Instruction::FlowControlType::JustX:
MOV(XSCRATCH0, (instr.flow_control.refx.Value() ^ 1)); CMP(COND0, refx);
EOR(XSCRATCH0, XSCRATCH0, COND0);
break; break;
case Instruction::FlowControlType::JustY: case Instruction::FlowControlType::JustY:
MOV(XSCRATCH0, (instr.flow_control.refy.Value() ^ 1)); CMP(COND1, refy);
EOR(XSCRATCH0, XSCRATCH0, COND1); break;
default:
UNREACHABLE();
break; break;
} }
CMP(XSCRATCH0, 0);
} }
void JitShader::Compile_UniformCondition(Instruction instr) { void JitShader::Compile_UniformCondition(Instruction instr) {

3
src/video_core/shader/shader_jit_a64_compiler.h
View file

@ -94,6 +94,9 @@ private:
*/ */
void Compile_SanitizedMul(oaknut::QReg src1, oaknut::QReg src2, oaknut::QReg scratch0); void Compile_SanitizedMul(oaknut::QReg src1, oaknut::QReg src2, oaknut::QReg scratch0);
/**
* Emits the code to evaluate a conditional instruction and update the host's EQ/NE status-flags
*/
void Compile_EvaluateCondition(Instruction instr); void Compile_EvaluateCondition(Instruction instr);
void Compile_UniformCondition(Instruction instr); void Compile_UniformCondition(Instruction instr);