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

Merge branch 'master' into vulkan

This commit is contained in:
Triang3l 2020-10-30 22:58:55 +03:00
parent 738cb0b847 ae3b68c7b6
commit cacf702948
5 changed files with 188 additions and 245 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

218
src/xenia/gpu/dxbc_shader_translator_alu.cc
View File

@ -68,32 +68,34 @@ void DxbcShaderTranslator::ProcessVectorAluOperation(
break; break;
case AluVectorOpcode::kMul: case AluVectorOpcode::kMul:
case AluVectorOpcode::kMad: { case AluVectorOpcode::kMad: {
bool is_mad = instr.vector_opcode == AluVectorOpcode::kMad; // Not using DXBC mad to prevent fused multiply-add (mul followed by add
if (is_mad) { // may be optimized into non-fused mad by the driver in the identical
DxbcOpMAd(per_component_dest, operands[0], operands[1], operands[2]); // operands case also).
} else { DxbcOpMul(per_component_dest, operands[0], operands[1]);
DxbcOpMul(per_component_dest, operands[0], operands[1]); uint32_t multiplicands_different =
}
// Shader Model 3: 0 or denormal * anything = 0.
// FIXME(Triang3l): Signed zero needs research and handling.
uint32_t absolute_different =
used_result_components & used_result_components &
~instr.vector_operands[0].GetAbsoluteIdenticalComponents( ~instr.vector_operands[0].GetIdenticalComponents(
instr.vector_operands[1]); instr.vector_operands[1]);
if (absolute_different) { if (multiplicands_different) {
// Shader Model 3: +-0 or denormal * anything = +0.
uint32_t is_zero_temp = PushSystemTemp(); uint32_t is_zero_temp = PushSystemTemp();
DxbcOpMin(DxbcDest::R(is_zero_temp, absolute_different), DxbcOpMin(DxbcDest::R(is_zero_temp, multiplicands_different),
operands[0].Abs(), operands[1].Abs()); operands[0].Abs(), operands[1].Abs());
// min isn't required to flush denormals, eq is. // min isn't required to flush denormals, eq is.
DxbcOpEq(DxbcDest::R(is_zero_temp, absolute_different), DxbcOpEq(DxbcDest::R(is_zero_temp, multiplicands_different),
DxbcSrc::R(is_zero_temp), DxbcSrc::LF(0.0f)); DxbcSrc::R(is_zero_temp), DxbcSrc::LF(0.0f));
DxbcOpMovC(DxbcDest::R(system_temp_result_, absolute_different), // Not replacing true `0 + term` with movc of the term because +0 + -0
DxbcSrc::R(is_zero_temp), // should result in +0, not -0.
is_mad ? operands[2] : DxbcSrc::LF(0.0f), DxbcOpMovC(DxbcDest::R(system_temp_result_, multiplicands_different),
DxbcSrc::R(is_zero_temp), DxbcSrc::LF(0.0f),
DxbcSrc::R(system_temp_result_)); DxbcSrc::R(system_temp_result_));
// Release is_zero_temp. // Release is_zero_temp.
PopSystemTemp(); PopSystemTemp();
} }
if (instr.vector_opcode == AluVectorOpcode::kMad) {
DxbcOpAdd(per_component_dest, DxbcSrc::R(system_temp_result_),
operands[2]);
}
} break; } break;
case AluVectorOpcode::kMax: case AluVectorOpcode::kMax:
@ -179,69 +181,40 @@ void DxbcShaderTranslator::ProcessVectorAluOperation(
component_count = 4; component_count = 4;
} }
result_swizzle = DxbcSrc::kXXXX; result_swizzle = DxbcSrc::kXXXX;
uint32_t absolute_different = uint32_t different = uint32_t((1 << component_count) - 1) &
uint32_t((1 << component_count) - 1) & ~instr.vector_operands[0].GetIdenticalComponents(
~instr.vector_operands[0].GetAbsoluteIdenticalComponents( instr.vector_operands[1]);
instr.vector_operands[1]); for (uint32_t i = 0; i < component_count; ++i) {
if (absolute_different) { DxbcOpMul(DxbcDest::R(system_temp_result_, i ? 0b0010 : 0b0001),
// Shader Model 3: 0 or denormal * anything = 0. operands[0].SelectFromSwizzled(i),
// FIXME(Triang3l): Signed zero needs research and handling. operands[1].SelectFromSwizzled(i));
// Add component products only if non-zero. For dp4, 16 scalar if ((different & (1 << i)) != 0) {
// operations in the worst case (as opposed to always 20 for // Shader Model 3: +-0 or denormal * anything = +0 (also not replacing
// eq/movc/eq/movc/dp4 or min/eq/movc/movc/dp4 for preparing operands // true `0 + term` with movc of the term because +0 + -0 should result
// for dp4). // in +0, not -0).
DxbcOpMul(DxbcDest::R(system_temp_result_, 0b0001), DxbcOpMin(DxbcDest::R(system_temp_result_, 0b0100),
operands[0].SelectFromSwizzled(0), operands[0].SelectFromSwizzled(i).Abs(),
operands[1].SelectFromSwizzled(0)); operands[1].SelectFromSwizzled(i).Abs());
if (absolute_different & 0b0001) { DxbcOpEq(DxbcDest::R(system_temp_result_, 0b0100),
DxbcOpMin(DxbcDest::R(system_temp_result_, 0b0010), DxbcSrc::R(system_temp_result_, DxbcSrc::kZZZZ),
operands[0].SelectFromSwizzled(0).Abs(),
operands[1].SelectFromSwizzled(0).Abs());
DxbcOpEq(DxbcDest::R(system_temp_result_, 0b0010),
DxbcSrc::R(system_temp_result_, DxbcSrc::kYYYY),
DxbcSrc::LF(0.0f)); DxbcSrc::LF(0.0f));
DxbcOpMovC(DxbcDest::R(system_temp_result_, 0b0001), DxbcOpMovC(DxbcDest::R(system_temp_result_, i ? 0b0010 : 0b0001),
DxbcSrc::R(system_temp_result_, DxbcSrc::kYYYY),
DxbcSrc::LF(0.0f),
DxbcSrc::R(system_temp_result_, DxbcSrc::kXXXX));
}
for (uint32_t i = 1; i < component_count; ++i) {
bool component_different = (absolute_different & (1 << i)) != 0;
DxbcOpMAd(DxbcDest::R(system_temp_result_,
component_different ? 0b0010 : 0b0001),
operands[0].SelectFromSwizzled(i),
operands[1].SelectFromSwizzled(i),
DxbcSrc::R(system_temp_result_, DxbcSrc::kXXXX));
if (component_different) {
DxbcOpMin(DxbcDest::R(system_temp_result_, 0b0100),
operands[0].SelectFromSwizzled(i).Abs(),
operands[1].SelectFromSwizzled(i).Abs());
DxbcOpEq(DxbcDest::R(system_temp_result_, 0b0100),
DxbcSrc::R(system_temp_result_, DxbcSrc::kZZZZ), DxbcSrc::R(system_temp_result_, DxbcSrc::kZZZZ),
DxbcSrc::LF(0.0f)); DxbcSrc::LF(0.0f),
DxbcOpMovC(DxbcDest::R(system_temp_result_, 0b0001), DxbcSrc::R(system_temp_result_,
DxbcSrc::R(system_temp_result_, DxbcSrc::kZZZZ), i ? DxbcSrc::kYYYY : DxbcSrc::kXXXX));
DxbcSrc::R(system_temp_result_, DxbcSrc::kXXXX),
DxbcSrc::R(system_temp_result_, DxbcSrc::kYYYY));
}
} }
} else { if (i) {
if (component_count == 2) { // Not using DXBC dp# to avoid fused multiply-add, PC GPUs are scalar
DxbcOpDP2(DxbcDest::R(system_temp_result_, 0b0001), operands[0], // as of 2020 anyway, and not using mad for the same reason (mul
operands[1]); // followed by add may be optimized into non-fused mad by the driver
} else if (component_count == 3) { // in the identical operands case also).
DxbcOpDP3(DxbcDest::R(system_temp_result_, 0b0001), operands[0], DxbcOpAdd(DxbcDest::R(system_temp_result_, 0b0001),
operands[1]); DxbcSrc::R(system_temp_result_, DxbcSrc::kXXXX),
} else { DxbcSrc::R(system_temp_result_, DxbcSrc::kYYYY));
assert_true(component_count == 4);
DxbcOpDP4(DxbcDest::R(system_temp_result_, 0b0001), operands[0],
operands[1]);
} }
} }
if (component_count == 2) { if (component_count == 2) {
// Add the third operand. Since floating-point addition isn't
// associative, even though adding this in multiply-add for the first
// component would be faster, it's safer to add here, in the end.
DxbcOpAdd(DxbcDest::R(system_temp_result_, 0b0001), DxbcOpAdd(DxbcDest::R(system_temp_result_, 0b0001),
DxbcSrc::R(system_temp_result_, DxbcSrc::kXXXX), DxbcSrc::R(system_temp_result_, DxbcSrc::kXXXX),
operands[2].SelectFromSwizzled(0)); operands[2].SelectFromSwizzled(0));
@ -592,14 +565,13 @@ void DxbcShaderTranslator::ProcessVectorAluOperation(
DxbcOpMov(DxbcDest::R(system_temp_result_, 0b0001), DxbcSrc::LF(1.0f)); DxbcOpMov(DxbcDest::R(system_temp_result_, 0b0001), DxbcSrc::LF(1.0f));
} }
if (used_result_components & 0b0010) { if (used_result_components & 0b0010) {
// Shader Model 3: 0 or denormal * anything = 0.
// FIXME(Triang3l): Signed zero needs research and handling.
DxbcOpMul(DxbcDest::R(system_temp_result_, 0b0010), DxbcOpMul(DxbcDest::R(system_temp_result_, 0b0010),
operands[0].SelectFromSwizzled(1), operands[0].SelectFromSwizzled(1),
operands[1].SelectFromSwizzled(1)); operands[1].SelectFromSwizzled(1));
if (!(instr.vector_operands[0].GetAbsoluteIdenticalComponents( if (!(instr.vector_operands[0].GetIdenticalComponents(
instr.vector_operands[1]) & instr.vector_operands[1]) &
0b0010)) { 0b0010)) {
// Shader Model 3: +-0 or denormal * anything = +0.
DxbcOpMin(DxbcDest::R(system_temp_result_, 0b0100), DxbcOpMin(DxbcDest::R(system_temp_result_, 0b0100),
operands[0].SelectFromSwizzled(1).Abs(), operands[0].SelectFromSwizzled(1).Abs(),
operands[1].SelectFromSwizzled(1).Abs()); operands[1].SelectFromSwizzled(1).Abs());
@ -700,8 +672,7 @@ void DxbcShaderTranslator::ProcessScalarAluOperation(
DxbcOpMul(ps_dest, operand_0_a, operand_0_b); DxbcOpMul(ps_dest, operand_0_a, operand_0_b);
if (instr.scalar_operands[0].components[0] != if (instr.scalar_operands[0].components[0] !=
instr.scalar_operands[0].components[1]) { instr.scalar_operands[0].components[1]) {
// Shader Model 3: 0 or denormal * anything = 0. // Shader Model 3: +-0 or denormal * anything = +0.
// FIXME(Triang3l): Signed zero needs research and handling.
uint32_t is_zero_temp = PushSystemTemp(); uint32_t is_zero_temp = PushSystemTemp();
DxbcOpMin(DxbcDest::R(is_zero_temp, 0b0001), operand_0_a.Abs(), DxbcOpMin(DxbcDest::R(is_zero_temp, 0b0001), operand_0_a.Abs(),
operand_0_b.Abs()); operand_0_b.Abs());
@ -714,58 +685,50 @@ void DxbcShaderTranslator::ProcessScalarAluOperation(
PopSystemTemp(); PopSystemTemp();
} }
break; break;
case AluScalarOpcode::kMulsPrev: { case AluScalarOpcode::kMulsPrev:
// Shader Model 3: 0 or denormal * anything = 0.
// FIXME(Triang3l): Signed zero needs research and handling.
uint32_t is_zero_temp = PushSystemTemp();
DxbcOpMin(DxbcDest::R(is_zero_temp, 0b0001), operand_0_a.Abs(),
ps_src.Abs());
// min isn't required to flush denormals, eq is.
DxbcOpEq(DxbcDest::R(is_zero_temp, 0b0001),
DxbcSrc::R(is_zero_temp, DxbcSrc::kXXXX), DxbcSrc::LF(0.0f));
DxbcOpMul(ps_dest, operand_0_a, ps_src);
DxbcOpMovC(ps_dest, DxbcSrc::R(is_zero_temp, DxbcSrc::kXXXX),
DxbcSrc::LF(0.0f), ps_src);
// Release is_zero_temp.
PopSystemTemp();
} break;
case AluScalarOpcode::kMulsPrev2: { case AluScalarOpcode::kMulsPrev2: {
uint32_t test_temp = PushSystemTemp(); uint32_t test_temp = PushSystemTemp();
// Check if need to select the src0.a * ps case. if (instr.scalar_opcode == AluScalarOpcode::kMulsPrev2) {
// ps != -FLT_MAX. // Check if need to select the src0.a * ps case.
DxbcOpNE(DxbcDest::R(test_temp, 0b0001), ps_src, DxbcSrc::LF(-FLT_MAX)); // ps != -FLT_MAX.
// isfinite(ps), or |ps| <= FLT_MAX, or -|ps| >= -FLT_MAX, since -FLT_MAX DxbcOpNE(DxbcDest::R(test_temp, 0b0001), ps_src, DxbcSrc::LF(-FLT_MAX));
// is already loaded to an SGPR, this is also false if it's NaN. // isfinite(ps), or |ps| <= FLT_MAX, or -|ps| >= -FLT_MAX, since
DxbcOpGE(DxbcDest::R(test_temp, 0b0010), -ps_src.Abs(), // -FLT_MAX is already loaded to an SGPR, this is also false if it's
DxbcSrc::LF(-FLT_MAX)); // NaN.
DxbcOpAnd(DxbcDest::R(test_temp, 0b0001), DxbcOpGE(DxbcDest::R(test_temp, 0b0010), -ps_src.Abs(),
DxbcSrc::R(test_temp, DxbcSrc::kXXXX), DxbcSrc::LF(-FLT_MAX));
DxbcSrc::R(test_temp, DxbcSrc::kYYYY)); DxbcOpAnd(DxbcDest::R(test_temp, 0b0001),
// isfinite(src0.b). DxbcSrc::R(test_temp, DxbcSrc::kXXXX),
DxbcOpGE(DxbcDest::R(test_temp, 0b0010), -operand_0_b.Abs(), DxbcSrc::R(test_temp, DxbcSrc::kYYYY));
DxbcSrc::LF(-FLT_MAX)); // isfinite(src0.b).
DxbcOpAnd(DxbcDest::R(test_temp, 0b0001), DxbcOpGE(DxbcDest::R(test_temp, 0b0010), -operand_0_b.Abs(),
DxbcSrc::R(test_temp, DxbcSrc::kXXXX), DxbcSrc::LF(-FLT_MAX));
DxbcSrc::R(test_temp, DxbcSrc::kYYYY)); DxbcOpAnd(DxbcDest::R(test_temp, 0b0001),
// src0.b > 0 (need !(src0.b <= 0), but src0.b has already been checked DxbcSrc::R(test_temp, DxbcSrc::kXXXX),
// for NaN). DxbcSrc::R(test_temp, DxbcSrc::kYYYY));
DxbcOpLT(DxbcDest::R(test_temp, 0b0010), DxbcSrc::LF(0.0f), operand_0_b); // src0.b > 0 (need !(src0.b <= 0), but src0.b has already been checked
DxbcOpAnd(DxbcDest::R(test_temp, 0b0001), // for NaN).
DxbcSrc::R(test_temp, DxbcSrc::kXXXX), DxbcOpLT(DxbcDest::R(test_temp, 0b0010), DxbcSrc::LF(0.0f),
DxbcSrc::R(test_temp, DxbcSrc::kYYYY)); operand_0_b);
DxbcOpIf(true, DxbcSrc::R(test_temp, DxbcSrc::kXXXX)); DxbcOpAnd(DxbcDest::R(test_temp, 0b0001),
// Shader Model 3: 0 or denormal * anything = 0. DxbcSrc::R(test_temp, DxbcSrc::kXXXX),
// ps is already known to be not NaN or Infinity, so multiplying it by 0 DxbcSrc::R(test_temp, DxbcSrc::kYYYY));
// will result in 0. However, src0.a can be anything, so the result should DxbcOpIf(true, DxbcSrc::R(test_temp, DxbcSrc::kXXXX));
// be zero if ps is zero. }
// FIXME(Triang3l): Signed zero needs research and handling. // Shader Model 3: +-0 or denormal * anything = +0.
DxbcOpEq(DxbcDest::R(test_temp, 0b0001), ps_src, DxbcSrc::LF(0.0f)); DxbcOpMin(DxbcDest::R(test_temp, 0b0001), operand_0_a.Abs(),
ps_src.Abs());
// min isn't required to flush denormals, eq is.
DxbcOpEq(DxbcDest::R(test_temp, 0b0001),
DxbcSrc::R(test_temp, DxbcSrc::kXXXX), DxbcSrc::LF(0.0f));
DxbcOpMul(ps_dest, operand_0_a, ps_src); DxbcOpMul(ps_dest, operand_0_a, ps_src);
DxbcOpMovC(ps_dest, DxbcSrc::R(test_temp, DxbcSrc::kXXXX), DxbcOpMovC(ps_dest, DxbcSrc::R(test_temp, DxbcSrc::kXXXX),
DxbcSrc::LF(0.0f), ps_src); DxbcSrc::LF(0.0f), ps_src);
DxbcOpElse(); if (instr.scalar_opcode == AluScalarOpcode::kMulsPrev2) {
DxbcOpMov(ps_dest, DxbcSrc::LF(-FLT_MAX)); DxbcOpElse();
DxbcOpEndIf(); DxbcOpMov(ps_dest, DxbcSrc::LF(-FLT_MAX));
DxbcOpEndIf();
}
// Release test_temp. // Release test_temp.
PopSystemTemp(); PopSystemTemp();
} break; } break;
@ -1023,11 +986,10 @@ void DxbcShaderTranslator::ProcessScalarAluOperation(
case AluScalarOpcode::kMulsc0: case AluScalarOpcode::kMulsc0:
case AluScalarOpcode::kMulsc1: case AluScalarOpcode::kMulsc1:
DxbcOpMul(ps_dest, operand_0_a, operand_1); DxbcOpMul(ps_dest, operand_0_a, operand_1);
if (!(instr.scalar_operands[0].GetAbsoluteIdenticalComponents( if (!(instr.scalar_operands[0].GetIdenticalComponents(
instr.scalar_operands[1]) & instr.scalar_operands[1]) &
0b0001)) { 0b0001)) {
// Shader Model 3: 0 or denormal * anything = 0. // Shader Model 3: +-0 or denormal * anything = +0.
// FIXME(Triang3l): Signed zero needs research and handling.
uint32_t is_zero_temp = PushSystemTemp(); uint32_t is_zero_temp = PushSystemTemp();
DxbcOpMin(DxbcDest::R(is_zero_temp, 0b0001), operand_0_a.Abs(), DxbcOpMin(DxbcDest::R(is_zero_temp, 0b0001), operand_0_a.Abs(),
operand_1.Abs()); operand_1.Abs());

26
src/xenia/gpu/shader.h
View File

@ -235,14 +235,18 @@ struct InstructionOperand {
return false; return false;
} }
// Returns which components of two operands are identical, but may have // Returns which components of two operands will always be bitwise equal
// different signs (for simplicity of usage with GetComponent, treating the // (disregarding component_count for simplicity of usage with GetComponent,
// rightmost component as replicated). // treating the rightmost component as replicated). This, strictly with all
uint32_t GetAbsoluteIdenticalComponents( // conditions, must be used when emulating Shader Model 3 +-0 * x = +0
const InstructionOperand& other) const { // multiplication behavior with IEEE-compliant multiplication (because
// -0 * |-0|, or -0 * +0, is -0, while the result must be +0).
uint32_t GetIdenticalComponents(const InstructionOperand& other) const {
if (storage_source != other.storage_source || if (storage_source != other.storage_source ||
storage_index != other.storage_index || storage_index != other.storage_index ||
storage_addressing_mode != other.storage_addressing_mode) { storage_addressing_mode != other.storage_addressing_mode ||
is_negated != other.is_negated ||
is_absolute_value != other.is_absolute_value) {
return 0; return 0;
} }
uint32_t identical_components = 0; uint32_t identical_components = 0;
@ -252,16 +256,6 @@ struct InstructionOperand {
} }
return identical_components; return identical_components;
} }
// Returns which components of two operands will always be bitwise equal, but
// may have different signs (disregarding component_count for simplicity of
// usage with GetComponent, treating the rightmost component as replicated).
uint32_t GetIdenticalComponents(const InstructionOperand& other) const {
if (is_negated != other.is_negated ||
is_absolute_value != other.is_absolute_value) {
return 0;
}
return GetAbsoluteIdenticalComponents(other);
}
}; };
struct ParsedExecInstruction { struct ParsedExecInstruction {

27
src/xenia/gpu/ucode.h
View File

@ -800,13 +800,26 @@ static_assert_size(TextureFetchInstruction, 12);
// Both are valid only within the current ALU clause. They are not modified // Both are valid only within the current ALU clause. They are not modified
// when the instruction that would write them fails its predication check. // when the instruction that would write them fails its predication check.
// - Direct3D 9 rules (like in GCN v_*_legacy_f32 instructions) for // - Direct3D 9 rules (like in GCN v_*_legacy_f32 instructions) for
// multiplication (0 or denormal * anything = 0) wherever it's present (mul, // multiplication (+-0 or denormal * anything = +0) wherever it's present
// mad, dp, etc.) and for NaN in min/max. It's very important to respect this // (mul, mad, dp, etc.) and for NaN in min/max. It's very important to respect
// rule for multiplication, as games often rely on it in vector normalization // this rule for multiplication, as games often rely on it in vector
// (rcp and mul), Infinity * 0 resulting in NaN breaks a lot of things in // normalization (rcp and mul), Infinity * 0 resulting in NaN breaks a lot of
// games - causes white screen in Halo 3, white specular on characters in GTA // things in games - causes white screen in Halo 3, white specular on
// IV. // characters in GTA IV. The result is always positive zero in this case, no
// TODO(Triang3l): Investigate signed zero handling in multiplication. // matter what the signs of the other operands are, according to R5xx
// Acceleration section 8.7.5 "Legacy multiply behavior" and testing on
// Adreno 200. This means that the following need to be taken into account
// (according to 8.7.2 "ALU Non-Transcendental Floating Point"):
// - +0 * -0 is -0 with IEEE conformance, however, with this legacy SM3
// handling, it should result in +0.
// - +0 + -0 is +0, so multiply-add should not be replaced with conditional
// move of the third operand in case of zero multiplicands, because the term
// may be -0, while the result should be +0 in this case.
// http://developer.amd.com/wordpress/media/2013/10/R5xx_Acceleration_v1.5.pdf
// Multiply-add also appears to be not fused (the SM3 behavior instruction on
// GCN is called v_mad_legacy_f32, not v_fma_legacy_f32) - shader translators
// should not use instructions that may be interpreted by the host GPU as
// fused multiply-add.
enum class AluScalarOpcode : uint32_t { enum class AluScalarOpcode : uint32_t {
// Floating-Point Add // Floating-Point Add

49
tools/build/premake
View File

@ -107,13 +107,14 @@ def has_bin(bin):
return None return None
def shell_call(command, throw_on_error=True, stdout_path=None): def shell_call(command, throw_on_error=True, stdout_path=None, stderr_path=None, shell=False):
"""Executes a shell command. """Executes a shell command.
Args: Args:
command: Command to execute, as a list of parameters. command: Command to execute, as a list of parameters.
throw_on_error: Whether to throw an error or return the status code. throw_on_error: Whether to throw an error or return the status code.
stdout_path: File path to write stdout output to. stdout_path: File path to write stdout output to.
stderr_path: File path to write stderr output to.
Returns: Returns:
If throw_on_error is False the status code of the call will be returned. If throw_on_error is False the status code of the call will be returned.
@ -121,17 +122,22 @@ def shell_call(command, throw_on_error=True, stdout_path=None):
stdout_file = None stdout_file = None
if stdout_path: if stdout_path:
stdout_file = open(stdout_path, 'w') stdout_file = open(stdout_path, 'w')
stderr_file = None
if stderr_path:
stderr_file = open(stderr_path, 'w')
result = 0 result = 0
try: try:
if throw_on_error: if throw_on_error:
result = 1 result = 1
subprocess.check_call(command, shell=False, stdout=stdout_file) subprocess.check_call(command, shell=shell, stdout=stdout_file, stderr=stderr_file)
result = 0 result = 0
else: else:
result = subprocess.call(command, shell=False, stdout=stdout_file) result = subprocess.call(command, shell=shell, stdout=stdout_file, stderr=stderr_file)
finally: finally:
if stdout_file: if stdout_file:
stdout_file.close() stdout_file.close()
if stderr_file:
stderr_file.close()
return result return result
@ -196,42 +202,5 @@ def import_subprocess_environment(args):
os.environ[var.upper()] = setting os.environ[var.upper()] = setting
break break
def git_submodule_update():
"""Runs a full recursive git submodule init and update.
Older versions of git do not support 'update --init --recursive'. We could
check and run it on versions that do support it and speed things up a bit.
"""
if True:
shell_call([
'git',
'submodule',
'update',
'--init',
'--recursive',
])
else:
shell_call([
'git',
'submodule',
'init',
])
shell_call([
'git',
'submodule',
'foreach',
'--recursive',
'git',
'submodule',
'init',
])
shell_call([
'git',
'submodule',
'update',
'--recursive',
])
if __name__ == '__main__': if __name__ == '__main__':
main() main()

111
xenia-build
View File

@ -34,8 +34,11 @@ def main():
# Check git exists. # Check git exists.
if not has_bin('git'): if not has_bin('git'):
print('ERROR: git must be installed and on PATH.') print('WARNING: Git should be installed and on PATH. Version info will be omitted from all binaries!')
sys.exit(1) print('')
elif not git_is_repository():
print('WARNING: The source tree is unversioned. Version info will be omitted from all binaries!')
print('')
# Check python version. # Check python version.
if not sys.version_info[:2] >= (3, 6): if not sys.version_info[:2] >= (3, 6):
@ -185,13 +188,14 @@ def get_bin(binary):
return None return None
def shell_call(command, throw_on_error=True, stdout_path=None, shell=False): def shell_call(command, throw_on_error=True, stdout_path=None, stderr_path=None, shell=False):
"""Executes a shell command. """Executes a shell command.
Args: Args:
command: Command to execute, as a list of parameters. command: Command to execute, as a list of parameters.
throw_on_error: Whether to throw an error or return the status code. throw_on_error: Whether to throw an error or return the status code.
stdout_path: File path to write stdout output to. stdout_path: File path to write stdout output to.
stderr_path: File path to write stderr output to.
Returns: Returns:
If throw_on_error is False the status code of the call will be returned. If throw_on_error is False the status code of the call will be returned.
@ -199,21 +203,49 @@ def shell_call(command, throw_on_error=True, stdout_path=None, shell=False):
stdout_file = None stdout_file = None
if stdout_path: if stdout_path:
stdout_file = open(stdout_path, 'w') stdout_file = open(stdout_path, 'w')
stderr_file = None
if stderr_path:
stderr_file = open(stderr_path, 'w')
result = 0 result = 0
try: try:
if throw_on_error: if throw_on_error:
result = 1 result = 1
subprocess.check_call(command, shell=shell, stdout=stdout_file) subprocess.check_call(command, shell=shell, stdout=stdout_file, stderr=stderr_file)
result = 0 result = 0
else: else:
result = subprocess.call(command, shell=shell, stdout=stdout_file) result = subprocess.call(command, shell=shell, stdout=stdout_file, stderr=stderr_file)
finally: finally:
if stdout_file: if stdout_file:
stdout_file.close() stdout_file.close()
if stderr_file:
stderr_file.close()
return result return result
def get_git_head_info(): def generate_version_h():
"""Generates a build/version.h file that contains current git info.
"""
if git_is_repository():
(branch_name, commit, commit_short) = git_get_head_info()
else:
branch_name = 'tarball'
commit = ':(-dont-do-this'
commit_short = ':('
contents = '''// Autogenerated by `xb premake`.
#ifndef GENERATED_VERSION_H_
#define GENERATED_VERSION_H_
#define XE_BUILD_BRANCH "%s"
#define XE_BUILD_COMMIT "%s"
#define XE_BUILD_COMMIT_SHORT "%s"
#define XE_BUILD_DATE __DATE__
#endif // GENERATED_VERSION_H_
''' % (branch_name, commit, commit_short)
with open('build/version.h', 'w') as f:
f.write(contents)
def git_get_head_info():
"""Queries the current branch and commit checksum from git. """Queries the current branch and commit checksum from git.
Returns: Returns:
@ -247,58 +279,28 @@ def get_git_head_info():
return branch_name, commit, commit_short return branch_name, commit, commit_short
def generate_version_h(): def git_is_repository():
"""Generates a build/version.h file that contains current git info. """Checks if git is available and this source tree is versioned.
""" """
(branch_name, commit, commit_short) = get_git_head_info() if not has_bin('git'):
contents = '''// Autogenerated by `xb premake`. return False
#ifndef GENERATED_VERSION_H_ return shell_call([
#define GENERATED_VERSION_H_ 'git',
#define XE_BUILD_BRANCH "%s" 'rev-parse',
#define XE_BUILD_COMMIT "%s" '--is-inside-work-tree',
#define XE_BUILD_COMMIT_SHORT "%s" ], throw_on_error=False, stdout_path=os.devnull, stderr_path=os.devnull) == 0
#define XE_BUILD_DATE __DATE__
#endif // GENERATED_VERSION_H_
''' % (branch_name, commit, commit_short)
with open('build/version.h', 'w') as f:
f.write(contents)
def git_submodule_update(): def git_submodule_update():
"""Runs a full recursive git submodule init and update. """Runs a full recursive git submodule init and update.
Older versions of git do not support 'update --init --recursive'. We could
check and run it on versions that do support it and speed things up a bit.
""" """
if True: shell_call([
shell_call([ 'git',
'git', 'submodule',
'submodule', 'update',
'update', '--init',
'--init', '--recursive',
'--recursive', ])
])
else:
shell_call([
'git',
'submodule',
'init',
])
shell_call([
'git',
'submodule',
'foreach',
'--recursive',
'git',
'submodule',
'init',
])
shell_call([
'git',
'submodule',
'update',
'--recursive',
])
def get_clang_format_binary(): def get_clang_format_binary():
@ -491,7 +493,10 @@ class SetupCommand(Command):
# Setup submodules. # Setup submodules.
print('- git submodule init / update...') print('- git submodule init / update...')
git_submodule_update() if git_is_repository():
git_submodule_update()
else:
print('WARNING: Git not available or not a repository. Dependencies may be missing.')
print('') print('')
print('- running premake...') print('- running premake...')