<div dir="ltr"><div dir="ltr"><div>I personally don't see a point in reimplementing existing LLVM functions. There is also LLVMConstNull for zeros and LLVMConstAllOnes for integer ones, though I think those are mainly good for vectors and people are most used to seeing LLVMConstInt and LLVMConstReal for scalars. </div><div> </div><div>Marek </div> <div class="gmail_quote"><div dir="ltr">On Wed, Dec 19, 2018 at 5:03 PM Rhys Perry <<a href="mailto:pendingchaos02@gmail.com">pendingchaos02@gmail.com</a>> wrote: </div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">I would expect these helpers to be much more efficient than the functions you suggested. They are also (in my opinion) more readable than the suggested functions. I don't think it matters much though, so I'm fine either way. On Tue, 18 Dec 2018 at 02:48, Marek Olšák <<a href="mailto:maraeo@gmail.com" target="_blank">maraeo@gmail.com</a>> wrote: > > On Fri, Dec 7, 2018 at 12:22 PM Rhys Perry <<a href="mailto:pendingchaos02@gmail.com" target="_blank">pendingchaos02@gmail.com</a>> wrote: >> >> Signed-off-by: Rhys Perry <<a href="mailto:pendingchaos02@gmail.com" target="_blank">pendingchaos02@gmail.com</a>> >> --- >> src/amd/common/ac_llvm_build.c | 123 ++++++++++++++++++++++++++++++-- >> src/amd/common/ac_llvm_build.h | 22 +++++- >> src/amd/common/ac_nir_to_llvm.c | 30 ++++---- >> 3 files changed, 154 insertions(+), 21 deletions(-) >> >> diff --git a/src/amd/common/ac_llvm_build.c b/src/amd/common/ac_llvm_build.c >> index 154cc696a2..cc7c6da5a4 100644 >> --- a/src/amd/common/ac_llvm_build.c >> +++ b/src/amd/common/ac_llvm_build.c >> @@ -87,12 +87,16 @@ ac_llvm_context_init(struct ac_llvm_context *ctx, >> ctx->v4f32 = LLVMVectorType(ctx->f32, 4); >> ctx->v8i32 = LLVMVectorType(ctx->i32, 8); >> >> + ctx->i8_0 = LLVMConstInt(ctx->i8, 0, false); >> + ctx->i8_1 = LLVMConstInt(ctx->i8, 1, false); >> ctx->i16_0 = LLVMConstInt(ctx->i16, 0, false); >> ctx->i16_1 = LLVMConstInt(ctx->i16, 1, false); >> ctx->i32_0 = LLVMConstInt(ctx->i32, 0, false); >> ctx->i32_1 = LLVMConstInt(ctx->i32, 1, false); >> ctx->i64_0 = LLVMConstInt(ctx->i64, 0, false); >> ctx->i64_1 = LLVMConstInt(ctx->i64, 1, false); >> + ctx->f16_0 = LLVMConstReal(ctx->f16, 0.0); >> + ctx->f16_1 = LLVMConstReal(ctx->f16, 1.0); >> ctx->f32_0 = LLVMConstReal(ctx->f32, 0.0); >> ctx->f32_1 = LLVMConstReal(ctx->f32, 1.0); >> ctx->f64_0 = LLVMConstReal(ctx->f64, 0.0); >> @@ -201,7 +205,9 @@ ac_get_type_size(LLVMTypeRef type) >> >> static LLVMTypeRef to_integer_type_scalar(struct ac_llvm_context *ctx, LLVMTypeRef t) >> { >> - if (t == ctx->f16 || t == ctx->i16) >> + if (t == ctx->i8) >> + return ctx->i8; >> + else if (t == ctx->f16 || t == ctx->i16) >> return ctx->i16; >> else if (t == ctx->f32 || t == ctx->i32) >> return ctx->i32; >> @@ -268,6 +274,110 @@ ac_to_float(struct ac_llvm_context *ctx, LLVMValueRef v) >> return LLVMBuildBitCast(ctx->builder, v, ac_to_float_type(ctx, type), ""); >> } >> >> +LLVMValueRef ac_get_zerof(struct ac_llvm_context *ctx, LLVMTypeRef t) >> +{ >> + if (t == ctx->f16) >> + return ctx->f16_0; >> + else if (t == ctx->f32) >> + return ctx->f32_0; >> + else if (t == ctx->f64) >> + return ctx->f64_0; >> + else >> + unreachable("Unhandled float size"); >> +} >> + >> +LLVMValueRef ac_get_onef(struct ac_llvm_context *ctx, LLVMTypeRef t) >> +{ >> + if (t == ctx->f16) >> + return ctx->f16_1; >> + else if (t == ctx->f32) >> + return ctx->f32_1; >> + else if (t == ctx->f64) >> + return ctx->f64_1; >> + else >> + unreachable("Unhandled float size"); >> +} >> + >> +LLVMValueRef ac_get_zero(struct ac_llvm_context *ctx, LLVMTypeRef t) >> +{ >> + if (t == ctx->i8) >> + return ctx->i8_0; >> + else if (t == ctx->i16) >> + return ctx->i16_0; >> + else if (t == ctx->i32) >> + return ctx->i32_0; >> + else if (t == ctx->i64) >> + return ctx->i64_0; >> + else >> + unreachable("Unhandled bit size"); >> +} >> + >> +LLVMValueRef ac_get_one(struct ac_llvm_context *ctx, LLVMTypeRef t) >> +{ >> + if (t == ctx->i8) >> + return ctx->i8_1; >> + else if (t == ctx->i16) >> + return ctx->i16_1; >> + else if (t == ctx->i32) >> + return ctx->i32_1; >> + else if (t == ctx->i64) >> + return ctx->i64_1; >> + else >> + unreachable("Unhandled bit size"); >> +} > > > You don't need these helpers. You can just use LLVMConstInt and LLVMConstReal. > >> >> + >> +LLVMTypeRef ac_float_of_size(struct ac_llvm_context *ctx, unsigned bit_size) >> +{ >> + switch (bit_size) { >> + case 16: >> + return ctx->f16; >> + case 32: >> + return ctx->f32; >> + case 64: >> + return ctx->f64; >> + default: >> + unreachable("Unhandled bit size"); >> + } >> +} >> + >> +LLVMTypeRef ac_int_of_size(struct ac_llvm_context *ctx, unsigned bit_size) >> +{ >> + switch (bit_size) { >> + case 8: >> + return ctx->i8; >> + case 16: >> + return ctx->i16; >> + case 32: >> + return ctx->i32; >> + case 64: >> + return ctx->i64; >> + default: >> + unreachable("Unhandled bit size"); >> + } >> +} > > > This is the same as LLVMIntTypeInContext. > > Marek > </blockquote></div></div></div>