[Mesa-dev] [PATCH] amd: remove support for LLVM 3.9
Dylan Baker
dylan at pnwbakers.com
Fri Feb 2 21:28:13 UTC 2018
I went ahead and pushed my patch, the only things you should need to do for the
meson side now is move with_gallium_radeonsi, and remove the stray `s` in the
error message.
Dylan
Quoting Bas Nieuwenhuizen (2018-02-02 11:56:26)
> I'd prefer if Dylan's patch goes first, as we are bumping swr too otherwise.
>
> Though either way:
>
> Reviewed-by: Bas Nieuwenhuizen <bas at basnieuwenhuizen.nl>
>
> On Fri, Feb 2, 2018 at 8:15 PM, Dylan Baker <dylan at pnwbakers.com> wrote:
> > FYI: I sent out a patch a few minutes ago to set the minimum version more like
> > autotools does, but I'm fine with this landing first.
> >
> > Reviewed-by: Dylan Baker <dylan at pnwbakers.com>
> >
> > Quoting Marek Olšák (2018-02-02 11:07:57)
> >> Adding this:
> >>
> >> diff --git a/meson.build b/meson.build
> >> index 80ea60f..22eb702 100644
> >> --- a/meson.build
> >> +++ b/meson.build
> >> @@ -991,12 +991,12 @@ endif
> >> _llvm = get_option('llvm')
> >> if _llvm == 'auto'
> >> dep_llvm = dependency(
> >> - 'llvm', version : '>= 3.9.0', modules : llvm_modules,
> >> + 'llvm', version : '>= 4.0.0', modules : llvm_modules,
> >> required : with_amd_vk or with_gallium_radeonsi or with_gallium_swr or
> >> with_gallium_opencl,
> >> )
> >> with_llvm = dep_llvm.found()
> >> elif _llvm == 'true'
> >> - dep_llvm = dependency('llvm', version : '>= 3.9.0', modules : llvm_modules)
> >> + dep_llvm = dependency('llvm', version : '>= 4.0.0', modules : llvm_modules)
> >> with_llvm = true
> >> else
> >> dep_llvm = []
> >> @@ -1019,7 +1019,7 @@ if with_llvm
> >> '-DMESA_LLVM_VERSION_PATCH=@0@'.format(_llvm_patch),
> >> ]
> >> elif with_amd_vk or with_gallium_radeonsi or with_gallium_swr
> >> - error('The following drivers requires LLVM: Radv, RadeonSI, SWR. One of
> >> these is enabled, but LLVM is disabled.')
> >> + error('The following drivers require LLVM: Radv, RadeonSI, SWR. One of these
> >> is enabled, but LLVM is disabled.')
> >> endif
> >>
> >> dep_glvnd = []
> >>
> >> Marek
> >>
> >> On Fri, Feb 2, 2018 at 8:02 PM, Bas Nieuwenhuizen <bas at basnieuwenhuizen.nl>
> >> wrote:
> >>
> >> Also change meson.build?
> >>
> >> On Fri, Feb 2, 2018 at 7:34 PM, Marek Olšák <maraeo at gmail.com> wrote:
> >> > From: Marek Olšák <marek.olsak at amd.com>
> >> >
> >> > Only these are supported:
> >> > - LLVM 4.0
> >> > - LLVM 5.0
> >> > - LLVM 6.0
> >> > - master (7.0)
> >> > ---
> >> > configure.ac | 4 +-
> >> > src/amd/common/ac_llvm_build.c | 187
> >> ++++++---------------
> >> > src/amd/common/ac_llvm_helper.cpp | 10 --
> >> > src/amd/common/ac_llvm_util.c | 39 +----
> >> > src/amd/common/ac_llvm_util.h | 14 +-
> >> > src/amd/common/ac_nir_to_llvm.c | 32 +---
> >> > src/amd/vulkan/radv_device.c | 4 -
> >> > src/gallium/drivers/radeonsi/si_compute.c | 3 +-
> >> > src/gallium/drivers/radeonsi/si_get.c | 13 +-
> >> > src/gallium/drivers/radeonsi/si_shader_tgsi_mem.c | 3 +-
> >> > .../drivers/radeonsi/si_shader_tgsi_setup.c | 2 -
> >> > 11 files changed, 72 insertions(+), 239 deletions(-)
> >> >
> >> > diff --git a/configure.ac b/configure.ac
> >> > index a54b7cb..8ed606c 100644
> >> > --- a/configure.ac
> >> > +++ b/configure.ac
> >> > @@ -96,22 +96,22 @@ XDAMAGE_REQUIRED=1.1
> >> > XSHMFENCE_REQUIRED=1.1
> >> > XVMC_REQUIRED=1.0.6
> >> > PYTHON_MAKO_REQUIRED=0.8.0
> >> > LIBSENSORS_REQUIRED=4.0.0
> >> > ZLIB_REQUIRED=1.2.3
> >> >
> >> > dnl LLVM versions
> >> > LLVM_REQUIRED_GALLIUM=3.3.0
> >> > LLVM_REQUIRED_OPENCL=3.9.0
> >> > LLVM_REQUIRED_R600=3.9.0
> >> > -LLVM_REQUIRED_RADEONSI=3.9.0
> >> > -LLVM_REQUIRED_RADV=3.9.0
> >> > +LLVM_REQUIRED_RADEONSI=4.0.0
> >> > +LLVM_REQUIRED_RADV=4.0.0
> >> > LLVM_REQUIRED_SWR=3.9.0
> >> >
> >> > dnl Check for progs
> >> > AC_PROG_CPP
> >> > AC_PROG_CC
> >> > AC_PROG_CXX
> >> > dnl add this here, so the help for this environmnet variable is close to
> >> > dnl other CC/CXX flags related help
> >> > AC_ARG_VAR([CXX11_CXXFLAGS], [Compiler flag to enable C++11 support
> >> (only needed if not
> >> > enabled by default and different from
> >> -std=c++11)])
> >> > diff --git a/src/amd/common/ac_llvm_build.c b/src/amd/common/ac_llvm_
> >> build.c
> >> > index 6ede60a..3efcaa1 100644
> >> > --- a/src/amd/common/ac_llvm_build.c
> >> > +++ b/src/amd/common/ac_llvm_build.c
> >> > @@ -213,22 +213,21 @@ ac_to_float(struct ac_llvm_context *ctx,
> >> LLVMValueRef v)
> >> > return LLVMBuildBitCast(ctx->builder, v, ac_to_float_type(ctx,
> >> type), "");
> >> > }
> >> >
> >> >
> >> > LLVMValueRef
> >> > ac_build_intrinsic(struct ac_llvm_context *ctx, const char *name,
> >> > LLVMTypeRef return_type, LLVMValueRef *params,
> >> > unsigned param_count, unsigned attrib_mask)
> >> > {
> >> > LLVMValueRef function, call;
> >> > - bool set_callsite_attrs = HAVE_LLVM >= 0x0400 &&
> >> > - !(attrib_mask & AC_FUNC_ATTR_LEGACY);
> >> > + bool set_callsite_attrs = !(attrib_mask & AC_FUNC_ATTR_LEGACY);
> >> >
> >> > function = LLVMGetNamedFunction(ctx->module, name);
> >> > if (!function) {
> >> > LLVMTypeRef param_types[32], function_type;
> >> > unsigned i;
> >> >
> >> > assert(param_count <= 32);
> >> >
> >> > for (i = 0; i < param_count; ++i) {
> >> > assert(params[i]);
> >> > @@ -714,34 +713,20 @@ ac_prepare_cube_coords(struct ac_llvm_context *ctx,
> >> > LLVMValueRef
> >> > ac_build_fs_interp(struct ac_llvm_context *ctx,
> >> > LLVMValueRef llvm_chan,
> >> > LLVMValueRef attr_number,
> >> > LLVMValueRef params,
> >> > LLVMValueRef i,
> >> > LLVMValueRef j)
> >> > {
> >> > LLVMValueRef args[5];
> >> > LLVMValueRef p1;
> >> > -
> >> > - if (HAVE_LLVM < 0x0400) {
> >> > - LLVMValueRef ij[2];
> >> > - ij[0] = LLVMBuildBitCast(ctx->builder, i, ctx->i32, "");
> >> > - ij[1] = LLVMBuildBitCast(ctx->builder, j, ctx->i32, "");
> >> > -
> >> > - args[0] = llvm_chan;
> >> > - args[1] = attr_number;
> >> > - args[2] = params;
> >> > - args[3] = ac_build_gather_values(ctx, ij, 2);
> >> > - return ac_build_intrinsic(ctx, "llvm.SI.fs.interp",
> >> > - ctx->f32, args, 4,
> >> > - AC_FUNC_ATTR_READNONE);
> >> > - }
> >> >
> >> > args[0] = i;
> >> > args[1] = llvm_chan;
> >> > args[2] = attr_number;
> >> > args[3] = params;
> >> >
> >> > p1 = ac_build_intrinsic(ctx, "llvm.amdgcn.interp.p1",
> >> > ctx->f32, args, 4,
> >> AC_FUNC_ATTR_READNONE);
> >> >
> >> > args[0] = p1;
> >> > @@ -755,30 +740,20 @@ ac_build_fs_interp(struct ac_llvm_context *ctx,
> >> > }
> >> >
> >> > LLVMValueRef
> >> > ac_build_fs_interp_mov(struct ac_llvm_context *ctx,
> >> > LLVMValueRef parameter,
> >> > LLVMValueRef llvm_chan,
> >> > LLVMValueRef attr_number,
> >> > LLVMValueRef params)
> >> > {
> >> > LLVMValueRef args[4];
> >> > - if (HAVE_LLVM < 0x0400) {
> >> > - args[0] = llvm_chan;
> >> > - args[1] = attr_number;
> >> > - args[2] = params;
> >> > -
> >> > - return ac_build_intrinsic(ctx,
> >> > - "llvm.SI.fs.constant",
> >> > - ctx->f32, args, 3,
> >> > - AC_FUNC_ATTR_READNONE);
> >> > - }
> >> >
> >> > args[0] = parameter;
> >> > args[1] = llvm_chan;
> >> > args[2] = attr_number;
> >> > args[3] = params;
> >> >
> >> > return ac_build_intrinsic(ctx, "llvm.amdgcn.interp.mov",
> >> > ctx->f32, args, 4,
> >> AC_FUNC_ATTR_READNONE);
> >> > }
> >> >
> >> > @@ -1202,34 +1177,31 @@ ac_build_ddxy(struct ac_llvm_context *ctx,
> >> > result = LLVMBuildFSub(ctx->builder, trbl, tl, "");
> >> > return result;
> >> > }
> >> >
> >> > void
> >> > ac_build_sendmsg(struct ac_llvm_context *ctx,
> >> > uint32_t msg,
> >> > LLVMValueRef wave_id)
> >> > {
> >> > LLVMValueRef args[2];
> >> > - const char *intr_name = (HAVE_LLVM < 0x0400) ? "llvm.SI.sendmsg"
> >> : "llvm.amdgcn.s.sendmsg";
> >> > args[0] = LLVMConstInt(ctx->i32, msg, false);
> >> > args[1] = wave_id;
> >> > - ac_build_intrinsic(ctx, intr_name, ctx->voidt, args, 2, 0);
> >> > + ac_build_intrinsic(ctx, "llvm.amdgcn.s.sendmsg", ctx->voidt,
> >> args, 2, 0);
> >> > }
> >> >
> >> > LLVMValueRef
> >> > ac_build_imsb(struct ac_llvm_context *ctx,
> >> > LLVMValueRef arg,
> >> > LLVMTypeRef dst_type)
> >> > {
> >> > - const char *intr_name = (HAVE_LLVM < 0x0400) ?
> >> "llvm.AMDGPU.flbit.i32" :
> >> > -
> >> "llvm.amdgcn.sffbh.i32";
> >> > - LLVMValueRef msb = ac_build_intrinsic(ctx, intr_name,
> >> > + LLVMValueRef msb = ac_build_intrinsic(ctx,
> >> "llvm.amdgcn.sffbh.i32",
> >> > dst_type, &arg, 1,
> >> > AC_FUNC_ATTR_READNONE);
> >> >
> >> > /* The HW returns the last bit index from MSB, but NIR/TGSI wants
> >> > * the index from LSB. Invert it by doing "31 - msb". */
> >> > msb = LLVMBuildSub(ctx->builder, LLVMConstInt(ctx->i32, 31,
> >> false),
> >> > msb, "");
> >> >
> >> > LLVMValueRef all_ones = LLVMConstInt(ctx->i32, -1, true);
> >> > LLVMValueRef cond = LLVMBuildOr(ctx->builder,
> >> > @@ -1363,151 +1335,90 @@ void ac_build_export(struct ac_llvm_context
> >> *ctx, struct ac_export_args *a)
> >> > args[4] = LLVMConstInt(ctx->i32, a->compr, 0);
> >> > memcpy(args + 5, a->out, sizeof(a->out[0]) * 4);
> >> >
> >> > ac_build_intrinsic(ctx, "llvm.SI.export", ctx->voidt, args, 9,
> >> > AC_FUNC_ATTR_LEGACY);
> >> > }
> >> >
> >> > LLVMValueRef ac_build_image_opcode(struct ac_llvm_context *ctx,
> >> > struct ac_image_args *a)
> >> > {
> >> > - LLVMTypeRef dst_type;
> >> > LLVMValueRef args[11];
> >> > unsigned num_args = 0;
> >> > const char *name = NULL;
> >> > char intr_name[128], type[64];
> >> >
> >> > - if (HAVE_LLVM >= 0x0400) {
> >> > - bool sample = a->opcode == ac_image_sample ||
> >> > - a->opcode == ac_image_gather4 ||
> >> > - a->opcode == ac_image_get_lod;
> >> > -
> >> > - if (sample)
> >> > - args[num_args++] = ac_to_float(ctx, a->addr);
> >> > - else
> >> > - args[num_args++] = a->addr;
> >> > -
> >> > - args[num_args++] = a->resource;
> >> > - if (sample)
> >> > - args[num_args++] = a->sampler;
> >> > - args[num_args++] = LLVMConstInt(ctx->i32, a->dmask, 0);
> >> > - if (sample)
> >> > - args[num_args++] = LLVMConstInt(ctx->i1, a->
> >> unorm, 0);
> >> > - args[num_args++] = ctx->i1false; /* glc */
> >> > - args[num_args++] = ctx->i1false; /* slc */
> >> > - args[num_args++] = ctx->i1false; /* lwe */
> >> > - args[num_args++] = LLVMConstInt(ctx->i1, a->da, 0);
> >> > -
> >> > - switch (a->opcode) {
> >> > - case ac_image_sample:
> >> > - name = "llvm.amdgcn.image.sample";
> >> > - break;
> >> > - case ac_image_gather4:
> >> > - name = "llvm.amdgcn.image.gather4";
> >> > - break;
> >> > - case ac_image_load:
> >> > - name = "llvm.amdgcn.image.load";
> >> > - break;
> >> > - case ac_image_load_mip:
> >> > - name = "llvm.amdgcn.image.load.mip";
> >> > - break;
> >> > - case ac_image_get_lod:
> >> > - name = "llvm.amdgcn.image.getlod";
> >> > - break;
> >> > - case ac_image_get_resinfo:
> >> > - name = "llvm.amdgcn.image.getresinfo";
> >> > - break;
> >> > - default:
> >> > - unreachable("invalid image opcode");
> >> > - }
> >> > + bool sample = a->opcode == ac_image_sample ||
> >> > + a->opcode == ac_image_gather4 ||
> >> > + a->opcode == ac_image_get_lod;
> >> >
> >> > - ac_build_type_name_for_intr(LLVMTypeOf(args[0]), type,
> >> > - sizeof(type));
> >> > -
> >> > - snprintf(intr_name, sizeof(intr_name),
> >> "%s%s%s%s.v4f32.%s.v8i32",
> >> > - name,
> >> > - a->compare ? ".c" : "",
> >> > - a->bias ? ".b" :
> >> > - a->lod ? ".l" :
> >> > - a->deriv ? ".d" :
> >> > - a->level_zero ? ".lz" : "",
> >> > - a->offset ? ".o" : "",
> >> > - type);
> >> > -
> >> > - LLVMValueRef result =
> >> > - ac_build_intrinsic(ctx, intr_name,
> >> > - ctx->v4f32, args, num_args,
> >> > - AC_FUNC_ATTR_READNONE);
> >> > - if (!sample) {
> >> > - result = LLVMBuildBitCast(ctx->builder, result,
> >> > - ctx->v4i32, "");
> >> > - }
> >> > - return result;
> >> > - }
> >> > + if (sample)
> >> > + args[num_args++] = ac_to_float(ctx, a->addr);
> >> > + else
> >> > + args[num_args++] = a->addr;
> >> >
> >> > - args[num_args++] = a->addr;
> >> > args[num_args++] = a->resource;
> >> > -
> >> > - if (a->opcode == ac_image_load ||
> >> > - a->opcode == ac_image_load_mip ||
> >> > - a->opcode == ac_image_get_resinfo) {
> >> > - dst_type = ctx->v4i32;
> >> > - } else {
> >> > - dst_type = ctx->v4f32;
> >> > + if (sample)
> >> > args[num_args++] = a->sampler;
> >> > - }
> >> > -
> >> > args[num_args++] = LLVMConstInt(ctx->i32, a->dmask, 0);
> >> > - args[num_args++] = LLVMConstInt(ctx->i32, a->unorm, 0);
> >> > - args[num_args++] = LLVMConstInt(ctx->i32, 0, 0); /* r128 */
> >> > - args[num_args++] = LLVMConstInt(ctx->i32, a->da, 0);
> >> > - args[num_args++] = LLVMConstInt(ctx->i32, 0, 0); /* glc */
> >> > - args[num_args++] = LLVMConstInt(ctx->i32, 0, 0); /* slc */
> >> > - args[num_args++] = LLVMConstInt(ctx->i32, 0, 0); /* tfe */
> >> > - args[num_args++] = LLVMConstInt(ctx->i32, 0, 0); /* lwe */
> >> > + if (sample)
> >> > + args[num_args++] = LLVMConstInt(ctx->i1, a->unorm, 0);
> >> > + args[num_args++] = ctx->i1false; /* glc */
> >> > + args[num_args++] = ctx->i1false; /* slc */
> >> > + args[num_args++] = ctx->i1false; /* lwe */
> >> > + args[num_args++] = LLVMConstInt(ctx->i1, a->da, 0);
> >> >
> >> > switch (a->opcode) {
> >> > case ac_image_sample:
> >> > - name = "llvm.SI.image.sample";
> >> > + name = "llvm.amdgcn.image.sample";
> >> > break;
> >> > case ac_image_gather4:
> >> > - name = "llvm.SI.gather4";
> >> > + name = "llvm.amdgcn.image.gather4";
> >> > break;
> >> > case ac_image_load:
> >> > - name = "llvm.SI.image.load";
> >> > + name = "llvm.amdgcn.image.load";
> >> > break;
> >> > case ac_image_load_mip:
> >> > - name = "llvm.SI.image.load.mip";
> >> > + name = "llvm.amdgcn.image.load.mip";
> >> > break;
> >> > case ac_image_get_lod:
> >> > - name = "llvm.SI.getlod";
> >> > + name = "llvm.amdgcn.image.getlod";
> >> > break;
> >> > case ac_image_get_resinfo:
> >> > - name = "llvm.SI.getresinfo";
> >> > + name = "llvm.amdgcn.image.getresinfo";
> >> > break;
> >> > + default:
> >> > + unreachable("invalid image opcode");
> >> > }
> >> >
> >> > - ac_build_type_name_for_intr(LLVMTypeOf(a->addr), type, sizeof
> >> (type));
> >> > - snprintf(intr_name, sizeof(intr_name), "%s%s%s%s.%s",
> >> > + ac_build_type_name_for_intr(LLVMTypeOf(args[0]), type,
> >> > + sizeof(type));
> >> > +
> >> > + snprintf(intr_name, sizeof(intr_name), "%s%s%s%s.v4f32.%s.v8i32",
> >> > name,
> >> > a->compare ? ".c" : "",
> >> > a->bias ? ".b" :
> >> > a->lod ? ".l" :
> >> > a->deriv ? ".d" :
> >> > a->level_zero ? ".lz" : "",
> >> > a->offset ? ".o" : "",
> >> > type);
> >> >
> >> > - return ac_build_intrinsic(ctx, intr_name,
> >> > - dst_type, args, num_args,
> >> > - AC_FUNC_ATTR_READNONE |
> >> > - AC_FUNC_ATTR_LEGACY);
> >> > + LLVMValueRef result =
> >> > + ac_build_intrinsic(ctx, intr_name,
> >> > + ctx->v4f32, args, num_args,
> >> > + AC_FUNC_ATTR_READNONE);
> >> > + if (!sample) {
> >> > + result = LLVMBuildBitCast(ctx->builder, result,
> >> > + ctx->v4i32, "");
> >> > + }
> >> > + return result;
> >> > }
> >> >
> >> > LLVMValueRef ac_build_cvt_pkrtz_f16(struct ac_llvm_context *ctx,
> >> > LLVMValueRef args[2])
> >> > {
> >> > if (HAVE_LLVM >= 0x0500) {
> >> > LLVMTypeRef v2f16 =
> >> > LLVMVectorType(LLVMHalfTypeInContext(ctx->
> >> context), 2);
> >> > LLVMValueRef res =
> >> > ac_build_intrinsic(ctx, "llvm.amdgcn.cvt.pkrtz",
> >> > @@ -1731,33 +1642,29 @@ void ac_get_image_intr_name(const char
> >> *base_name,
> >> > LLVMTypeRef data_type,
> >> > LLVMTypeRef coords_type,
> >> > LLVMTypeRef rsrc_type,
> >> > char *out_name, unsigned out_len)
> >> > {
> >> > char coords_type_name[8];
> >> >
> >> > ac_build_type_name_for_intr(coords_type, coords_type_name,
> >> > sizeof(coords_type_name));
> >> >
> >> > - if (HAVE_LLVM <= 0x0309) {
> >> > - snprintf(out_name, out_len, "%s.%s", base_name,
> >> coords_type_name);
> >> > - } else {
> >> > - char data_type_name[8];
> >> > - char rsrc_type_name[8];
> >> > -
> >> > - ac_build_type_name_for_intr(data_type, data_type_name,
> >> > - sizeof(data_type_name));
> >> > - ac_build_type_name_for_intr(rsrc_type, rsrc_type_name,
> >> > - sizeof(rsrc_type_name));
> >> > - snprintf(out_name, out_len, "%s.%s.%s.%s", base_name,
> >> > - data_type_name, coords_type_name,
> >> rsrc_type_name);
> >> > - }
> >> > + char data_type_name[8];
> >> > + char rsrc_type_name[8];
> >> > +
> >> > + ac_build_type_name_for_intr(data_type, data_type_name,
> >> > + sizeof(data_type_name));
> >> > + ac_build_type_name_for_intr(rsrc_type, rsrc_type_name,
> >> > + sizeof(rsrc_type_name));
> >> > + snprintf(out_name, out_len, "%s.%s.%s.%s", base_name,
> >> > + data_type_name, coords_type_name, rsrc_type_name);
> >> > }
> >> >
> >> > #define AC_EXP_TARGET (HAVE_LLVM >= 0x0500 ? 0 : 3)
> >> > #define AC_EXP_OUT0 (HAVE_LLVM >= 0x0500 ? 2 : 5)
> >> >
> >> > enum ac_ir_type {
> >> > AC_IR_UNDEF,
> >> > AC_IR_CONST,
> >> > AC_IR_VALUE,
> >> > };
> >> > diff --git a/src/amd/common/ac_llvm_helper.cpp b/src/amd/common/ac_llvm_
> >> helper.cpp
> >> > index 793737c..1d2369b 100644
> >> > --- a/src/amd/common/ac_llvm_helper.cpp
> >> > +++ b/src/amd/common/ac_llvm_helper.cpp
> >> > @@ -58,54 +58,44 @@ void ac_add_attr_dereferenceable(LLVMValueRef val,
> >> uint64_t bytes)
> >> > bool ac_is_sgpr_param(LLVMValueRef arg)
> >> > {
> >> > llvm::Argument *A = llvm::unwrap<llvm::Argument>(arg);
> >> > llvm::AttributeList AS = A->getParent()->getAttributes();
> >> > unsigned ArgNo = A->getArgNo();
> >> > return AS.hasAttribute(ArgNo + 1, llvm::Attribute::InReg);
> >> > }
> >> >
> >> > LLVMValueRef ac_llvm_get_called_value(LLVMValueRef call)
> >> > {
> >> > -#if HAVE_LLVM >= 0x0309
> >> > return LLVMGetCalledValue(call);
> >> > -#else
> >> > - return llvm::wrap(llvm::CallSite(llvm::unwrap<llvm::Instruction>
> >> (call)).getCalledValue());
> >> > -#endif
> >> > }
> >> >
> >> > bool ac_llvm_is_function(LLVMValueRef v)
> >> > {
> >> > -#if HAVE_LLVM >= 0x0309
> >> > return LLVMGetValueKind(v) == LLVMFunctionValueKind;
> >> > -#else
> >> > - return llvm::isa<llvm::Function>(llvm::unwrap(v));
> >> > -#endif
> >> > }
> >> >
> >> > LLVMBuilderRef ac_create_builder(LLVMContextRef ctx,
> >> > enum ac_float_mode float_mode)
> >> > {
> >> > LLVMBuilderRef builder = LLVMCreateBuilderInContext(ctx);
> >> >
> >> > -#if HAVE_LLVM >= 0x0308
> >> > llvm::FastMathFlags flags;
> >> >
> >> > switch (float_mode) {
> >> > case AC_FLOAT_MODE_DEFAULT:
> >> > break;
> >> > case AC_FLOAT_MODE_NO_SIGNED_ZEROS_FP_MATH:
> >> > flags.setNoSignedZeros();
> >> > llvm::unwrap(builder)->setFastMathFlags(flags);
> >> > break;
> >> > case AC_FLOAT_MODE_UNSAFE_FP_MATH:
> >> > #if HAVE_LLVM >= 0x0600
> >> > flags.setFast();
> >> > #else
> >> > flags.setUnsafeAlgebra();
> >> > #endif
> >> > llvm::unwrap(builder)->setFastMathFlags(flags);
> >> > break;
> >> > }
> >> > -#endif
> >> >
> >> > return builder;
> >> > }
> >> > diff --git a/src/amd/common/ac_llvm_util.c b/src/amd/common/ac_llvm_util.
> >> c
> >> > index 5fd785a..b88c4e4 100644
> >> > --- a/src/amd/common/ac_llvm_util.c
> >> > +++ b/src/amd/common/ac_llvm_util.c
> >> > @@ -39,26 +39,25 @@ static void ac_init_llvm_target()
> >> > LLVMInitializeAMDGPUTarget();
> >> > LLVMInitializeAMDGPUTargetMC();
> >> > LLVMInitializeAMDGPUAsmPrinter();
> >> >
> >> > /* For inline assembly. */
> >> > LLVMInitializeAMDGPUAsmParser();
> >> >
> >> > /* Workaround for bug in llvm 4.0 that causes image intrinsics
> >> > * to disappear.
> >> > * https://reviews.llvm.org/D26348
> >> > + *
> >> > + * "mesa" is the prefix for error messages.
> >> > */
> >> > - if (HAVE_LLVM >= 0x0400) {
> >> > - /* "mesa" is the prefix for error messages */
> >> > - const char *argv[2] = { "mesa", "-simplifycfg-sink-common
> >> =false" };
> >> > - LLVMParseCommandLineOptions(2, argv, NULL);
> >> > - }
> >> > + const char *argv[2] = { "mesa", "-simplifycfg-sink-common=false"
> >> };
> >> > + LLVMParseCommandLineOptions(2, argv, NULL);
> >> > }
> >> >
> >> > static once_flag ac_init_llvm_target_once_flag = ONCE_FLAG_INIT;
> >> >
> >> > LLVMTargetRef ac_get_llvm_target(const char *triple)
> >> > {
> >> > LLVMTargetRef target = NULL;
> >> > char *err_message = NULL;
> >> >
> >> > call_once(&ac_init_llvm_target_once_flag, ac_init_llvm_target);
> >> > @@ -139,81 +138,51 @@ LLVMTargetMachineRef ac_create_target_machine(enum
> >> radeon_family family, enum ac
> >> > triple,
> >> > ac_get_llvm_processor_name(family),
> >> > features,
> >> > LLVMCodeGenLevelDefault,
> >> > LLVMRelocDefault,
> >> > LLVMCodeModelDefault);
> >> >
> >> > return tm;
> >> > }
> >> >
> >> > -
> >> > -#if HAVE_LLVM < 0x0400
> >> > -static LLVMAttribute ac_attr_to_llvm_attr(enum ac_func_attr attr)
> >> > -{
> >> > - switch (attr) {
> >> > - case AC_FUNC_ATTR_ALWAYSINLINE: return LLVMAlwaysInlineAttribute;
> >> > - case AC_FUNC_ATTR_INREG: return LLVMInRegAttribute;
> >> > - case AC_FUNC_ATTR_NOALIAS: return LLVMNoAliasAttribute;
> >> > - case AC_FUNC_ATTR_NOUNWIND: return LLVMNoUnwindAttribute;
> >> > - case AC_FUNC_ATTR_READNONE: return LLVMReadNoneAttribute;
> >> > - case AC_FUNC_ATTR_READONLY: return LLVMReadOnlyAttribute;
> >> > - default:
> >> > - fprintf(stderr, "Unhandled function attribute: %x\n", attr);
> >> > - return 0;
> >> > - }
> >> > -}
> >> > -
> >> > -#else
> >> > -
> >> > static const char *attr_to_str(enum ac_func_attr attr)
> >> > {
> >> > switch (attr) {
> >> > case AC_FUNC_ATTR_ALWAYSINLINE: return "alwaysinline";
> >> > case AC_FUNC_ATTR_INREG: return "inreg";
> >> > case AC_FUNC_ATTR_NOALIAS: return "noalias";
> >> > case AC_FUNC_ATTR_NOUNWIND: return "nounwind";
> >> > case AC_FUNC_ATTR_READNONE: return "readnone";
> >> > case AC_FUNC_ATTR_READONLY: return "readonly";
> >> > case AC_FUNC_ATTR_WRITEONLY: return "writeonly";
> >> > case AC_FUNC_ATTR_INACCESSIBLE_MEM_ONLY: return
> >> "inaccessiblememonly";
> >> > case AC_FUNC_ATTR_CONVERGENT: return "convergent";
> >> > default:
> >> > fprintf(stderr, "Unhandled function attribute: %x\n", attr);
> >> > return 0;
> >> > }
> >> > }
> >> >
> >> > -#endif
> >> > -
> >> > void
> >> > ac_add_function_attr(LLVMContextRef ctx, LLVMValueRef function,
> >> > int attr_idx, enum ac_func_attr attr)
> >> > {
> >> > -#if HAVE_LLVM < 0x0400
> >> > - LLVMAttribute llvm_attr = ac_attr_to_llvm_attr(attr);
> >> > - if (attr_idx == -1) {
> >> > - LLVMAddFunctionAttr(function, llvm_attr);
> >> > - } else {
> >> > - LLVMAddAttribute(LLVMGetParam(function, attr_idx - 1), llvm_attr);
> >> > - }
> >> > -#else
> >> > const char *attr_name = attr_to_str(attr);
> >> > unsigned kind_id = LLVMGetEnumAttributeKindForName(attr_name,
> >> > strlen
> >> (attr_name));
> >> > LLVMAttributeRef llvm_attr = LLVMCreateEnumAttribute(ctx, kind_id,
> >> 0);
> >> >
> >> > if (LLVMIsAFunction(function))
> >> > LLVMAddAttributeAtIndex(function, attr_idx, llvm_attr);
> >> > else
> >> > LLVMAddCallSiteAttribute(function, attr_idx, llvm_attr);
> >> > -#endif
> >> > }
> >> >
> >> > void ac_add_func_attributes(LLVMContextRef ctx, LLVMValueRef function,
> >> > unsigned attrib_mask)
> >> > {
> >> > attrib_mask |= AC_FUNC_ATTR_NOUNWIND;
> >> > attrib_mask &= ~AC_FUNC_ATTR_LEGACY;
> >> >
> >> > while (attrib_mask) {
> >> > enum ac_func_attr attr = 1u << u_bit_scan(&attrib_mask);
> >> > diff --git a/src/amd/common/ac_llvm_util.h b/src/amd/common/ac_llvm_util.
> >> h
> >> > index 29dc0c1..3cf385a 100644
> >> > --- a/src/amd/common/ac_llvm_util.h
> >> > +++ b/src/amd/common/ac_llvm_util.h
> >> > @@ -35,23 +35,23 @@
> >> > extern "C" {
> >> > #endif
> >> >
> >> > enum ac_func_attr {
> >> > AC_FUNC_ATTR_ALWAYSINLINE = (1 << 0),
> >> > AC_FUNC_ATTR_INREG = (1 << 2),
> >> > AC_FUNC_ATTR_NOALIAS = (1 << 3),
> >> > AC_FUNC_ATTR_NOUNWIND = (1 << 4),
> >> > AC_FUNC_ATTR_READNONE = (1 << 5),
> >> > AC_FUNC_ATTR_READONLY = (1 << 6),
> >> > - AC_FUNC_ATTR_WRITEONLY = HAVE_LLVM >= 0x0400 ? (1 << 7) : 0,
> >> > - AC_FUNC_ATTR_INACCESSIBLE_MEM_ONLY = HAVE_LLVM >= 0x0400 ? (1 <<
> >> 8) : 0,
> >> > - AC_FUNC_ATTR_CONVERGENT = HAVE_LLVM >= 0x0400 ? (1 << 9) : 0,
> >> > + AC_FUNC_ATTR_WRITEONLY = (1 << 7),
> >> > + AC_FUNC_ATTR_INACCESSIBLE_MEM_ONLY = (1 << 8),
> >> > + AC_FUNC_ATTR_CONVERGENT = (1 << 9),
> >> >
> >> > /* Legacy intrinsic that needs attributes on function
> >> declarations
> >> > * and they must match the internal LLVM definition exactly,
> >> otherwise
> >> > * intrinsic selection fails.
> >> > */
> >> > AC_FUNC_ATTR_LEGACY = (1u << 31),
> >> > };
> >> >
> >> > enum ac_target_machine_options {
> >> > AC_TM_SUPPORTS_SPILL = (1 << 0),
> >> > @@ -87,28 +87,26 @@ LLVMBuilderRef ac_create_builder(LLVMContextRef ctx,
> >> >
> >> > void
> >> > ac_llvm_add_target_dep_function_attr(LLVMValueRef F,
> >> > const char *name, int value);
> >> >
> >> > static inline unsigned
> >> > ac_get_load_intr_attribs(bool can_speculate)
> >> > {
> >> > /* READNONE means writes can't affect it, while READONLY means
> >> that
> >> > * writes can affect it. */
> >> > - return can_speculate && HAVE_LLVM >= 0x0400 ?
> >> > - AC_FUNC_ATTR_READNONE :
> >> > - AC_FUNC_ATTR_READONLY;
> >> > + return can_speculate ? AC_FUNC_ATTR_READNONE :
> >> > + AC_FUNC_ATTR_READONLY;
> >> > }
> >> >
> >> > static inline unsigned
> >> > ac_get_store_intr_attribs(bool writeonly_memory)
> >> > {
> >> > - return writeonly_memory && HAVE_LLVM >= 0x0400 ?
> >> > - AC_FUNC_ATTR_INACCESSIBLE_MEM_ONLY :
> >> > + return writeonly_memory ? AC_FUNC_ATTR_INACCESSIBLE_MEM_ONLY :
> >> > AC_FUNC_ATTR_WRITEONLY;
> >> > }
> >> >
> >> > #ifdef __cplusplus
> >> > }
> >> > #endif
> >> >
> >> > #endif /* AC_LLVM_UTIL_H */
> >> > diff --git a/src/amd/common/ac_nir_to_llvm.c b/src/amd/common/ac_nir_to_
> >> llvm.c
> >> > index 0f7d625..cfcd217 100644
> >> > --- a/src/amd/common/ac_nir_to_llvm.c
> >> > +++ b/src/amd/common/ac_nir_to_llvm.c
> >> > @@ -3628,32 +3628,24 @@ static LLVMValueRef visit_image_load(struct
> >> ac_nir_context *ctx,
> >> > dim == GLSL_SAMPLER_DIM_3D ||
> >> > dim == GLSL_SAMPLER_DIM_SUBPASS ||
> >> > dim == GLSL_SAMPLER_DIM_SUBPASS_MS;
> >> > LLVMValueRef da = is_da ? ctx->ac.i1true : ctx->
> >> ac.i1false;
> >> > LLVMValueRef glc = ctx->ac.i1false;
> >> > LLVMValueRef slc = ctx->ac.i1false;
> >> >
> >> > params[0] = get_image_coords(ctx, instr);
> >> > params[1] = get_sampler_desc(ctx, instr->variables[0],
> >> AC_DESC_IMAGE, NULL, true, false);
> >> > params[2] = LLVMConstInt(ctx->ac.i32, 15, false); /*
> >> dmask */
> >> > - if (HAVE_LLVM <= 0x0309) {
> >> > - params[3] = ctx->ac.i1false; /* r128 */
> >> > - params[4] = da;
> >> > - params[5] = glc;
> >> > - params[6] = slc;
> >> > - } else {
> >> > - LLVMValueRef lwe = ctx->ac.i1false;
> >> > - params[3] = glc;
> >> > - params[4] = slc;
> >> > - params[5] = lwe;
> >> > - params[6] = da;
> >> > - }
> >> > + params[3] = glc;
> >> > + params[4] = slc;
> >> > + params[5] = ctx->ac.i1false;
> >> > + params[6] = da;
> >> >
> >> > ac_get_image_intr_name("llvm.amdgcn.image.load",
> >> > ctx->ac.v4f32, /* vdata */
> >> > LLVMTypeOf(params[0]), /* coords *
> >> /
> >> > LLVMTypeOf(params[1]), /* rsrc */
> >> > intrinsic_name, sizeof
> >> (intrinsic_name));
> >> >
> >> > res = ac_build_intrinsic(&ctx->ac, intrinsic_name, ctx->
> >> ac.v4f32,
> >> > params, 7,
> >> AC_FUNC_ATTR_READONLY);
> >> > }
> >> > @@ -3687,32 +3679,24 @@ static void visit_image_store(struct
> >> ac_nir_context *ctx,
> >> > bool is_da = glsl_sampler_type_is_array(type) ||
> >> > dim == GLSL_SAMPLER_DIM_CUBE ||
> >> > dim == GLSL_SAMPLER_DIM_3D;
> >> > LLVMValueRef da = is_da ? ctx->ac.i1true : ctx->
> >> ac.i1false;
> >> > LLVMValueRef slc = ctx->ac.i1false;
> >> >
> >> > params[0] = ac_to_float(&ctx->ac, get_src(ctx, instr->src
> >> [2]));
> >> > params[1] = get_image_coords(ctx, instr); /* coords */
> >> > params[2] = get_sampler_desc(ctx, instr->variables[0],
> >> AC_DESC_IMAGE, NULL, true, true);
> >> > params[3] = LLVMConstInt(ctx->ac.i32, 15, false); /*
> >> dmask */
> >> > - if (HAVE_LLVM <= 0x0309) {
> >> > - params[4] = ctx->ac.i1false; /* r128 */
> >> > - params[5] = da;
> >> > - params[6] = glc;
> >> > - params[7] = slc;
> >> > - } else {
> >> > - LLVMValueRef lwe = ctx->ac.i1false;
> >> > - params[4] = glc;
> >> > - params[5] = slc;
> >> > - params[6] = lwe;
> >> > - params[7] = da;
> >> > - }
> >> > + params[4] = glc;
> >> > + params[5] = slc;
> >> > + params[6] = ctx->ac.i1false;
> >> > + params[7] = da;
> >> >
> >> > ac_get_image_intr_name("llvm.amdgcn.image.store",
> >> > LLVMTypeOf(params[0]), /* vdata */
> >> > LLVMTypeOf(params[1]), /* coords *
> >> /
> >> > LLVMTypeOf(params[2]), /* rsrc */
> >> > intrinsic_name, sizeof
> >> (intrinsic_name));
> >> >
> >> > ac_build_intrinsic(&ctx->ac, intrinsic_name, ctx->
> >> ac.voidt,
> >> > params, 8, 0);
> >> > }
> >> > diff --git a/src/amd/vulkan/radv_device.c b/src/amd/vulkan/radv_device.c
> >> > index aea723c..9fda419 100644
> >> > --- a/src/amd/vulkan/radv_device.c
> >> > +++ b/src/amd/vulkan/radv_device.c
> >> > @@ -1106,25 +1106,21 @@ VkResult radv_CreateDevice(
> >> > device->pbb_allowed = device->physical_device->rad_
> >> info.chip_class >= GFX9 &&
> >> > (device->instance->perftest_flags &
> >> RADV_PERFTEST_BINNING);
> >> >
> >> > /* Disabled and not implemented for now. */
> >> > device->dfsm_allowed = device->pbb_allowed && false;
> >> >
> >> > #ifdef ANDROID
> >> > device->always_use_syncobj = device->physical_device->rad_
> >> info.has_syncobj_wait_for_submit;
> >> > #endif
> >> >
> >> > -#if HAVE_LLVM < 0x0400
> >> > - device->llvm_supports_spill = false;
> >> > -#else
> >> > device->llvm_supports_spill = true;
> >> > -#endif
> >> >
> >> > /* The maximum number of scratch waves. Scratch space isn't
> >> divided
> >> > * evenly between CUs. The number is only a function of the
> >> number of CUs.
> >> > * We can decrease the constant to decrease the scratch buffer
> >> size.
> >> > *
> >> > * sctx->scratch_waves must be >= the maximum posible size of
> >> > * 1 threadgroup, so that the hw doesn't hang from being unable
> >> > * to start any.
> >> > *
> >> > * The recommended value is 4 per CU at most. Higher numbers
> >> don't
> >> > diff --git a/src/gallium/drivers/radeonsi/si_compute.c b/src/gallium/
> >> drivers/radeonsi/si_compute.c
> >> > index ac4fab3..91d6810 100644
> >> > --- a/src/gallium/drivers/radeonsi/si_compute.c
> >> > +++ b/src/gallium/drivers/radeonsi/si_compute.c
> >> > @@ -152,22 +152,21 @@ static void *si_create_compute_state(
> >> > struct si_context *sctx = (struct si_context *)ctx;
> >> > struct si_screen *sscreen = (struct si_screen *)ctx->screen;
> >> > struct si_compute *program = CALLOC_STRUCT(si_compute);
> >> >
> >> > pipe_reference_init(&program->reference, 1);
> >> > program->screen = (struct si_screen *)ctx->screen;
> >> > program->ir_type = cso->ir_type;
> >> > program->local_size = cso->req_local_mem;
> >> > program->private_size = cso->req_private_mem;
> >> > program->input_size = cso->req_input_mem;
> >> > - program->use_code_object_v2 = HAVE_LLVM >= 0x0400 &&
> >> > - cso->ir_type ==
> >> PIPE_SHADER_IR_NATIVE;
> >> > + program->use_code_object_v2 = cso->ir_type ==
> >> PIPE_SHADER_IR_NATIVE;
> >> >
> >> > if (cso->ir_type == PIPE_SHADER_IR_TGSI) {
> >> > program->tokens = tgsi_dup_tokens(cso->prog);
> >> > if (!program->tokens) {
> >> > FREE(program);
> >> > return NULL;
> >> > }
> >> >
> >> > program->compiler_ctx_state.debug = sctx->debug;
> >> > program->compiler_ctx_state.is_debug_context = sctx->
> >> is_debug;
> >> > diff --git a/src/gallium/drivers/radeonsi/si_get.c b/src/gallium/drivers/
> >> radeonsi/si_get.c
> >> > index 36e6018..46d5eee 100644
> >> > --- a/src/gallium/drivers/radeonsi/si_get.c
> >> > +++ b/src/gallium/drivers/radeonsi/si_get.c
> >> > @@ -182,24 +182,22 @@ static int si_get_param(struct pipe_screen
> >> *pscreen, enum pipe_cap param)
> >> > case PIPE_CAP_MEMOBJ:
> >> > case PIPE_CAP_LOAD_CONSTBUF:
> >> > case PIPE_CAP_INT64:
> >> > case PIPE_CAP_INT64_DIVMOD:
> >> > case PIPE_CAP_TGSI_CLOCK:
> >> > case PIPE_CAP_CAN_BIND_CONST_BUFFER_AS_VERTEX:
> >> > case PIPE_CAP_ALLOW_MAPPED_BUFFERS_DURING_EXECUTION:
> >> > case PIPE_CAP_TGSI_ANY_REG_AS_ADDRESS:
> >> > case PIPE_CAP_SIGNED_VERTEX_BUFFER_OFFSET:
> >> > case PIPE_CAP_TGSI_FS_FBFETCH:
> >> > - return 1;
> >> > -
> >> > case PIPE_CAP_TGSI_VOTE:
> >> > - return HAVE_LLVM >= 0x0400;
> >> > + return 1;
> >> >
> >> > case PIPE_CAP_TGSI_BALLOT:
> >> > return HAVE_LLVM >= 0x0500;
> >> >
> >> > case PIPE_CAP_RESOURCE_FROM_USER_MEMORY:
> >> > return !SI_BIG_ENDIAN && sscreen->info.has_userptr;
> >> >
> >> > case PIPE_CAP_DEVICE_RESET_STATUS_QUERY:
> >> > return (sscreen->info.drm_major == 2 &&
> >> > sscreen->info.drm_minor >= 43) ||
> >> > @@ -737,28 +735,23 @@ static unsigned get_max_threads_per_block(struct
> >> si_screen *screen,
> >> > static int si_get_compute_param(struct pipe_screen *screen,
> >> > enum pipe_shader_ir ir_type,
> >> > enum pipe_compute_cap param,
> >> > void *ret)
> >> > {
> >> > struct si_screen *sscreen = (struct si_screen *)screen;
> >> >
> >> > //TODO: select these params by asic
> >> > switch (param) {
> >> > case PIPE_COMPUTE_CAP_IR_TARGET: {
> >> > - const char *gpu;
> >> > - const char *triple;
> >> > -
> >> > - if (HAVE_LLVM < 0x0400)
> >> > - triple = "amdgcn--";
> >> > - else
> >> > - triple = "amdgcn-mesa-mesa3d";
> >> > + const char *gpu, *triple;
> >> >
> >> > + triple = "amdgcn-mesa-mesa3d";
> >> > gpu = ac_get_llvm_processor_name(sscreen->info.family);
> >> > if (ret) {
> >> > sprintf(ret, "%s-%s", gpu, triple);
> >> > }
> >> > /* +2 for dash and terminating NIL byte */
> >> > return (strlen(triple) + strlen(gpu) + 2) * sizeof(char);
> >> > }
> >> > case PIPE_COMPUTE_CAP_GRID_DIMENSION:
> >> > if (ret) {
> >> > uint64_t *grid_dimension = ret;
> >> > diff --git a/src/gallium/drivers/radeonsi/si_shader_tgsi_mem.c b/src/
> >> gallium/drivers/radeonsi/si_shader_tgsi_mem.c
> >> > index b5fad34..ec1de40 100644
> >> > --- a/src/gallium/drivers/radeonsi/si_shader_tgsi_mem.c
> >> > +++ b/src/gallium/drivers/radeonsi/si_shader_tgsi_mem.c
> >> > @@ -289,31 +289,30 @@ static void image_append_args(
> >> > LLVMValueRef i1true = LLVMConstInt(ctx->i1, 1, 0);
> >> > LLVMValueRef r128 = i1false;
> >> > LLVMValueRef da = tgsi_is_array_image(target) ? i1true : i1false;
> >> > LLVMValueRef glc =
> >> > force_glc ||
> >> > inst->Memory.Qualifier & (TGSI_MEMORY_COHERENT |
> >> TGSI_MEMORY_VOLATILE) ?
> >> > i1true : i1false;
> >> > LLVMValueRef slc = i1false;
> >> > LLVMValueRef lwe = i1false;
> >> >
> >> > - if (atomic || (HAVE_LLVM <= 0x0309)) {
> >> > + if (atomic) {
> >> > emit_data->args[emit_data->arg_count++] = r128;
> >> > emit_data->args[emit_data->arg_count++] = da;
> >> > if (!atomic) {
> >> > emit_data->args[emit_data->arg_count++] = glc;
> >> > }
> >> > emit_data->args[emit_data->arg_count++] = slc;
> >> > return;
> >> > }
> >> >
> >> > - /* HAVE_LLVM >= 0x0400 */
> >> > emit_data->args[emit_data->arg_count++] = glc;
> >> > emit_data->args[emit_data->arg_count++] = slc;
> >> > emit_data->args[emit_data->arg_count++] = lwe;
> >> > emit_data->args[emit_data->arg_count++] = da;
> >> > }
> >> >
> >> > /**
> >> > * Append the resource and indexing arguments for buffer intrinsics.
> >> > *
> >> > * \param rsrc the v4i32 buffer resource
> >> > diff --git a/src/gallium/drivers/radeonsi/si_shader_tgsi_setup.c b/src/
> >> gallium/drivers/radeonsi/si_shader_tgsi_setup.c
> >> > index f70e2bd..44294f2 100644
> >> > --- a/src/gallium/drivers/radeonsi/si_shader_tgsi_setup.c
> >> > +++ b/src/gallium/drivers/radeonsi/si_shader_tgsi_setup.c
> >> > @@ -1371,24 +1371,22 @@ void si_llvm_optimize_module(struct
> >> si_shader_context *ctx)
> >> > LLVMAddAlwaysInlinerPass(gallivm->passmgr);
> >> >
> >> > /* This pass should eliminate all the load and store instructions
> >> */
> >> > LLVMAddPromoteMemoryToRegisterPass(gallivm->passmgr);
> >> >
> >> > /* Add some optimization passes */
> >> > LLVMAddScalarReplAggregatesPass(gallivm->passmgr);
> >> > LLVMAddLICMPass(gallivm->passmgr);
> >> > LLVMAddAggressiveDCEPass(gallivm->passmgr);
> >> > LLVMAddCFGSimplificationPass(gallivm->passmgr);
> >> > -#if HAVE_LLVM >= 0x0400
> >> > /* This is recommended by the instruction combining pass. */
> >> > LLVMAddEarlyCSEMemSSAPass(gallivm->passmgr);
> >> > -#endif
> >> > LLVMAddInstructionCombiningPass(gallivm->passmgr);
> >> >
> >> > /* Run the pass */
> >> > LLVMRunPassManager(gallivm->passmgr, ctx->gallivm.module);
> >> >
> >> > LLVMDisposeBuilder(ctx->ac.builder);
> >> > LLVMDisposePassManager(gallivm->passmgr);
> >> > gallivm_dispose_target_library_info(target_library_info);
> >> > }
> >> >
> >> > --
> >> > 2.7.4
> >> >
> >> > _______________________________________________
> >> > mesa-dev mailing list
> >> > mesa-dev at lists.freedesktop.org
> >> > https://lists.freedesktop.org/mailman/listinfo/mesa-dev
> >>
> >>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 488 bytes
Desc: signature
URL: <https://lists.freedesktop.org/archives/mesa-dev/attachments/20180202/02e6411b/attachment-0001.sig>
More information about the mesa-dev
mailing list