Convert vertex loader position cache to std::array

This commit is contained in:
Pokechu22 2022-04-13 16:12:53 -07:00
parent f722bdf7f1
commit 97d0ff58c8
6 changed files with 16 additions and 13 deletions

View File

@ -23,7 +23,7 @@ static void PosMtx_ReadDirect_UByte(VertexLoader* loader)
{ {
u32 posmtx = DataRead<u8>() & 0x3f; u32 posmtx = DataRead<u8>() & 0x3f;
if (loader->m_counter < 3) if (loader->m_counter < 3)
VertexLoaderManager::position_matrix_index[loader->m_counter + 1] = posmtx; VertexLoaderManager::position_matrix_index_cache[loader->m_counter + 1] = posmtx;
DataWrite<u32>(posmtx); DataWrite<u32>(posmtx);
PRIM_LOG("posmtx: {}, ", posmtx); PRIM_LOG("posmtx: {}, ", posmtx);
} }

View File

@ -211,10 +211,11 @@ int VertexLoaderARM64::ReadVertex(VertexComponentFormat attribute, ComponentForm
{ {
CMP(count_reg, 3); CMP(count_reg, 3);
FixupBranch dont_store = B(CC_GT); FixupBranch dont_store = B(CC_GT);
MOVP2R(EncodeRegTo64(scratch2_reg), VertexLoaderManager::position_cache); MOVP2R(EncodeRegTo64(scratch2_reg), VertexLoaderManager::position_cache.data());
ADD(EncodeRegTo64(scratch1_reg), EncodeRegTo64(scratch2_reg), EncodeRegTo64(count_reg), ADD(EncodeRegTo64(scratch1_reg), EncodeRegTo64(scratch2_reg), EncodeRegTo64(count_reg),
ArithOption(EncodeRegTo64(count_reg), ShiftType::LSL, 4)); ArithOption(EncodeRegTo64(count_reg), ShiftType::LSL, 4));
m_float_emit.STUR(write_size, coords, EncodeRegTo64(scratch1_reg), -16); m_float_emit.STUR(write_size, coords, EncodeRegTo64(scratch1_reg),
-int(sizeof(decltype(VertexLoaderManager::position_cache[0]))));
SetJumpTarget(dont_store); SetJumpTarget(dont_store);
} }
@ -422,7 +423,7 @@ void VertexLoaderARM64::GenerateVertexLoader()
// Z-Freeze // Z-Freeze
CMP(count_reg, 3); CMP(count_reg, 3);
FixupBranch dont_store = B(CC_GT); FixupBranch dont_store = B(CC_GT);
MOVP2R(EncodeRegTo64(scratch2_reg), VertexLoaderManager::position_matrix_index); MOVP2R(EncodeRegTo64(scratch2_reg), VertexLoaderManager::position_matrix_index_cache.data());
STR(scratch1_reg, EncodeRegTo64(scratch2_reg), ArithOption(count_reg, true)); STR(scratch1_reg, EncodeRegTo64(scratch2_reg), ArithOption(count_reg, true));
SetJumpTarget(dont_store); SetJumpTarget(dont_store);

View File

@ -31,11 +31,11 @@
namespace VertexLoaderManager namespace VertexLoaderManager
{ {
float position_cache[3][4]; // Used by zfreeze
std::array<std::array<float, 4>, 3> position_cache;
// The counter added to the address of the array is 1, 2, or 3, but never zero. // The counter added to the address of the array is 1, 2, or 3, but never zero.
// So only index 1 - 3 are used. // So only index 1 - 3 are used.
u32 position_matrix_index[4]; std::array<u32, 4> position_matrix_index_cache;
static NativeVertexFormatMap s_native_vertex_map; static NativeVertexFormatMap s_native_vertex_map;
static NativeVertexFormat* s_current_vtx_fmt; static NativeVertexFormat* s_current_vtx_fmt;

View File

@ -53,8 +53,8 @@ void UpdateVertexArrayPointers();
// Position cache for zfreeze (3 vertices, 4 floats each to allow SIMD overwrite). // Position cache for zfreeze (3 vertices, 4 floats each to allow SIMD overwrite).
// These arrays are in reverse order. // These arrays are in reverse order.
extern float position_cache[3][4]; extern std::array<std::array<float, 4>, 3> position_cache;
extern u32 position_matrix_index[4]; extern std::array<u32, 4> position_matrix_index_cache;
// VB_HAS_X. Bitmask telling what vertex components are present. // VB_HAS_X. Bitmask telling what vertex components are present.
extern u32 g_current_components; extern u32 g_current_components;

View File

@ -119,8 +119,9 @@ int VertexLoaderX64::ReadVertex(OpArg data, VertexComponentFormat attribute, Com
{ {
CMP(32, R(count_reg), Imm8(3)); CMP(32, R(count_reg), Imm8(3));
FixupBranch dont_store = J_CC(CC_A); FixupBranch dont_store = J_CC(CC_A);
LEA(32, scratch3, MScaled(count_reg, SCALE_4, -4)); LEA(32, scratch3,
MOVUPS(MPIC(VertexLoaderManager::position_cache, scratch3, SCALE_4), coords); MScaled(count_reg, SCALE_4, -int(VertexLoaderManager::position_cache[0].size())));
MOVUPS(MPIC(VertexLoaderManager::position_cache.data(), scratch3, SCALE_4), coords);
SetJumpTarget(dont_store); SetJumpTarget(dont_store);
} }
}; };
@ -408,7 +409,8 @@ void VertexLoaderX64::GenerateVertexLoader()
// zfreeze // zfreeze
CMP(32, R(count_reg), Imm8(3)); CMP(32, R(count_reg), Imm8(3));
FixupBranch dont_store = J_CC(CC_A); FixupBranch dont_store = J_CC(CC_A);
MOV(32, MPIC(VertexLoaderManager::position_matrix_index, count_reg, SCALE_4), R(scratch1)); MOV(32, MPIC(VertexLoaderManager::position_matrix_index_cache.data(), count_reg, SCALE_4),
R(scratch1));
SetJumpTarget(dont_store); SetJumpTarget(dont_store);
m_native_vtx_decl.posmtx.components = 4; m_native_vtx_decl.posmtx.components = 4;

View File

@ -558,7 +558,7 @@ void VertexManagerBase::CalculateZSlope(NativeVertexFormat* format)
{ {
// If this vertex format has per-vertex position matrix IDs, look it up. // If this vertex format has per-vertex position matrix IDs, look it up.
if (vert_decl.posmtx.enable) if (vert_decl.posmtx.enable)
mtxIdx = VertexLoaderManager::position_matrix_index[3 - i]; mtxIdx = VertexLoaderManager::position_matrix_index_cache[3 - i];
if (vert_decl.position.components == 2) if (vert_decl.position.components == 2)
VertexLoaderManager::position_cache[2 - i][2] = 0; VertexLoaderManager::position_cache[2 - i][2] = 0;