AlexIndustrial/mesa
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

aco: implement 16-bit fsat on GFX8

Browse Source 
GFX8 doesn't have v_med3_f16.

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/26445>
This commit is contained in:
Rhys Perry
2023-12-01 16:20:38 +00:00
committed by Marge Bot
parent de51a21e26
commit 468ee8b80c
2 changed files with 16 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/amd/compiler/aco_instruction_selection.cpp
+5 -1
View File
@@ -2631,9 +2631,13 @@ visit_alu_instr(isel_context* ctx, nir_alu_instr* instr)
break;
}
Temp src = get_alu_src(ctx, instr->src[0]);
if (dst.regClass() == v2b) {
if (dst.regClass() == v2b && ctx->program->gfx_level >= GFX9) {
bld.vop3(aco_opcode::v_med3_f16, Definition(dst), Operand::c16(0u), Operand::c16(0x3c00),
src);
} else if (dst.regClass() == v2b) {
bld.vop2_e64(aco_opcode::v_mul_f16, Definition(dst), Operand::c16(0x3c00), src)
->valu()
.clamp = true;
} else if (dst.regClass() == v1) {
bld.vop3(aco_opcode::v_med3_f32, Definition(dst), Operand::zero(),
Operand::c32(0x3f800000u), src);