fdo-mirrors/mesa
1
0
Fork 0
mirror of https://gitlab.freedesktop.org/mesa/mesa.git synced 2025-12-24 04:30:10 +01:00
Code Issues Projects Releases Packages Wiki Activity Actions 1

aco: don't use v_pack_b32_f16 if 16-bit input denormals are flushed

Browse source 
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7111>
This commit is contained in:
Rhys Perry 2020-10-13 13:38:16 +01:00 • committed by Marge Bot
parent d4b3e869ee
commit d75d12f507
2 changed files with 6 additions and 2 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

2
src/amd/compiler/aco_ir.h
View file

@ -192,6 +192,8 @@ enum fp_denorm {
/* Note that v_rcp_f32, v_exp_f32, v_log_f32, v_sqrt_f32, v_rsq_f32 and
* v_mad_f32/v_madak_f32/v_madmk_f32/v_mac_f32 always flush denormals. */
fp_denorm_flush = 0x0,
fp_denorm_keep_in = 0x1,
fp_denorm_keep_out = 0x2,
fp_denorm_keep = 0x3,
};

6
src/amd/compiler/aco_lower_to_hw_instr.cpp
View file

@ -37,6 +37,7 @@ namespace aco {
struct lower_context {
Program *program;
Block *block;
std::vector<aco_ptr<Instruction>> instructions;
};
@ -981,7 +982,7 @@ uint32_t get_intersection_mask(int a_start, int a_size,
void copy_16bit_literal(lower_context *ctx, Builder& bld, Definition def, Operand op)
{
if (ctx->program->chip_class < GFX10) {
if (ctx->program->chip_class < GFX10 || !(ctx->block->fp_mode.denorm16_64 & fp_denorm_keep_in)) {
unsigned offset = def.physReg().byte() * 8u;
def = Definition(PhysReg(def.physReg().reg()), v1);
Operand def_op(def.physReg(), v1);
@ -1182,7 +1183,7 @@ void do_pack_2x16(lower_context *ctx, Builder& bld, Definition def, Operand lo,
return;
}
if (ctx->program->chip_class >= GFX9) {
if (ctx->program->chip_class >= GFX9 && (ctx->block->fp_mode.denorm16_64 & fp_denorm_keep_in)) {
Instruction* instr = bld.vop3(aco_opcode::v_pack_b32_f16, def, lo, hi);
/* opsel: 0 = select low half, 1 = select high half. [0] = src0, [1] = src1 */
static_cast<VOP3A_instruction*>(instr)->opsel = hi.physReg().byte() | (lo.physReg().byte() >> 1);
@ -1622,6 +1623,7 @@ void lower_to_hw_instr(Program* program)
Block *block = &program->blocks[i];
lower_context ctx;
ctx.program = program;
ctx.block = block;
Builder bld(program, &ctx.instructions);
float_mode config_mode;