Home
last modified time | relevance | path

Searched refs:ffma (Results 1 – 13 of 13) sorted by relevance

/third_party/mesa3d/src/intel/compiler/
Dbrw_nir_opt_peephole_ffma.c237 nir_alu_instr *ffma = nir_alu_instr_create(b->shader, nir_op_ffma); in brw_nir_opt_peephole_ffma_instr() local
238 ffma->dest.saturate = add->dest.saturate; in brw_nir_opt_peephole_ffma_instr()
239 ffma->dest.write_mask = add->dest.write_mask; in brw_nir_opt_peephole_ffma_instr()
242 ffma->src[i].src = nir_src_for_ssa(mul_src[i]); in brw_nir_opt_peephole_ffma_instr()
244 ffma->src[i].swizzle[j] = mul->src[i].swizzle[swizzle[j]]; in brw_nir_opt_peephole_ffma_instr()
246 nir_alu_src_copy(&ffma->src[2], &add->src[1 - add_mul_src]); in brw_nir_opt_peephole_ffma_instr()
250 nir_ssa_dest_init(&ffma->instr, &ffma->dest.dest, in brw_nir_opt_peephole_ffma_instr()
253 nir_ssa_def_rewrite_uses(&add->dest.dest.ssa, &ffma->dest.dest.ssa); in brw_nir_opt_peephole_ffma_instr()
255 nir_builder_instr_insert(b, &ffma->instr); in brw_nir_opt_peephole_ffma_instr()
/third_party/mesa3d/src/nouveau/codegen/lib/
Dgm107.asm185 ffma $r5 $r4 $r5 $r0
186 ffma $r0 $r5 neg $r4 $r4
/third_party/mesa3d/src/gallium/drivers/etnaviv/
Detnaviv_compiler_nir_emit.c66 OP(fmul, MUL, 0_1_X), OP(fadd, ADD, 0_X_1), OP(ffma, MAD, 0_1_2),
/third_party/mesa3d/docs/relnotes/
D17.2.3.rst66 - radv: lower ffma in nir.
D21.3.0.rst3540 - nir/algebraic: add various ffma optimizations
3543 - nir: add ffma creation helpers
3544 - nir: create ffma from builders more often
3545 - nir: lower fdot to ffma if lower_ffma=false
3546 - spirv: create ffma more often
D20.3.0.rst3369 - nir: get ffma support from NIR options for nir_lower_flrp
3432 - nir,radeonsi: move ffma fusing to late optimizations for better codegen
3433 - radeonsi: clean up ffma handling
3457 - radeonsi: fuse or lower ffma optimally on all chips
D22.0.0.rst797 - nir: Reorder ffma and fsub combining
3708 - radv,aco: don't lower some ffma instructions
3710 - aco: fix check_vop3_operands() for f16vec2 ffma fneg combine
D19.3.0.rst276 - pan/midgard: Explain ffma
3023 - i965/fs: set rounding mode when emitting fadd, fmul and ffma
D19.1.0.rst419 - panfrost/midgard: Add fround(_even), ftrunc, ffma
2002 - intel/compiler: Lower ffma on Gen4 and Gen5
D19.0.0.rst1877 - glsl: Add "built-in" functions to do ffma(fp64)
D21.0.0.rst1361 - microsoft/compiler: do not lower away 64-bit ffma
D21.1.0.rst5331 - nir: handle negatives in ffma reassociation optimisation
/third_party/mesa3d/src/asahi/compiler/
Dagx_compile.c658 TRIOP(ffma, fma); in agx_emit_alu()