[GPU] Update v_mad_legacy_f32 comment
This commit is contained in:
parent
bc0c2040e2
commit
8bcfcf6452
|
@ -816,10 +816,11 @@ static_assert_size(TextureFetchInstruction, 12);
|
||||||
// move of the third operand in case of zero multiplicands, because the term
|
// 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.
|
// 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
|
// 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
|
// 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
|
// GCN is called v_mad_legacy_f32, not v_fma_legacy_f32 (in 2012-2020, before
|
||||||
// should not use instructions that may be interpreted by the host GPU as
|
// RDNA 2, which removed v_mad_f32 as well) - shader translators should not
|
||||||
// fused multiply-add.
|
// 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
|
||||||
|
|
Loading…
Reference in New Issue