diff options
| author | Qiang Yu <yuq825@gmail.com> | 2022-11-18 16:37:19 +0800 |
|---|---|---|
| committer | Dylan Baker <dylan.c.baker@intel.com> | 2022-11-21 10:05:07 -0800 |
| commit | 9fdfe7751cec5f0f69238d39d56c1f8c2acf3f5b (patch) | |
| tree | b9fc365ceb34b44342a6b54111191a124abbce49 | |
| parent | cba4b07aa0febbf9033d5156952d9a4e211d1773 (diff) | |
| download | mesa-9fdfe7751cec5f0f69238d39d56c1f8c2acf3f5b.tar.gz | |
ac/llvm: fix 16bit varying llvm compile error
Found when 16bit vec3 varying with llvm14 (not found
when llvm15), one 32bit vec4 slot is filled like this:
vec3[0] | undef
vec3[1] | undef
vec3[2] | undef
undef | undef
LLVM error is for the elements with undef:
%287 = insertelement float %280, half %279, i64 0
After this change, we get:
%287 = insertelement <2 x half> %280, half %279, i64 0
Fixes: 279eea5bda2 ("amd/llvm: Transition to LLVM "opaque pointers"")
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Signed-off-by: Qiang Yu <yuq825@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19848>
(cherry picked from commit e3b1f26a2b83538eb175be28b2e785bbe626bf5f)
| -rw-r--r-- | .pick_status.json | 2 | ||||
| -rw-r--r-- | src/amd/llvm/ac_nir_to_llvm.c | 9 |
2 files changed, 10 insertions, 1 deletions
diff --git a/.pick_status.json b/.pick_status.json index 81cf82505e0..7463c3ee44f 100644 --- a/.pick_status.json +++ b/.pick_status.json @@ -265,7 +265,7 @@ "description": "ac/llvm: fix 16bit varying llvm compile error", "nominated": true, "nomination_type": 1, - "resolution": 0, + "resolution": 1, "main_sha": null, "because_sha": "279eea5bda2444fdce21744b972dad5016f0f366" }, diff --git a/src/amd/llvm/ac_nir_to_llvm.c b/src/amd/llvm/ac_nir_to_llvm.c index 8d208f51396..9d3d60603a6 100644 --- a/src/amd/llvm/ac_nir_to_llvm.c +++ b/src/amd/llvm/ac_nir_to_llvm.c @@ -2410,7 +2410,16 @@ static void visit_store_output(struct ac_nir_context *ctx, nir_intrinsic_instr * * using read-modify-write. */ index = LLVMConstInt(ctx->ac.i32, nir_intrinsic_io_semantics(instr).high_16bits, 0); + +#if LLVM_VERSION_MAJOR <= 14 + /* To work around old LLVM bug which won't change the output type to + * LLVMBuildLoad2 type argument. + */ + output = LLVMBuildLoad2(ctx->ac.builder, ctx->ac.f32, output_addr, ""); + output = LLVMBuildBitCast(ctx->ac.builder, output, ctx->ac.v2f16, ""); +#else output = LLVMBuildLoad2(ctx->ac.builder, ctx->ac.v2f16, output_addr, ""); +#endif output = LLVMBuildInsertElement(ctx->ac.builder, output, value, index, ""); value = LLVMBuildBitCast(ctx->ac.builder, output, ctx->ac.f32, ""); } |