diff options
| author | Timur Kristóf <[email protected]> | 2020-01-10 12:06:53 +0100 |
|---|---|---|
| committer | Timur Kristóf <[email protected]> | 2020-01-10 12:30:44 +0100 |
| commit | eccac46cdc6086b7a10adff2ea95de986d70a85d (patch) | |
| tree | e8649066abdae325e004c561c916f7cfd4909e65 /src/amd | |
| parent | a5fe84aefb6858bee123f60b330db1e0287f9cc0 (diff) | |
ac/llvm: Fix ac_build_reduce in wave32 mode.
Previously, when cluster_size was set to 0, it always worked as if
the cluster size was 64. This commit fixes it in wave32 mode by
changing to work as if the cluster size was set to 32.
Signed-off-by: Timur Kristóf <[email protected]>
Reviewed-by: Samuel Pitoiset <[email protected]>
Diffstat (limited to 'src/amd')
| -rw-r--r-- | src/amd/llvm/ac_llvm_build.c | 15 |
1 files changed, 9 insertions, 6 deletions
diff --git a/src/amd/llvm/ac_llvm_build.c b/src/amd/llvm/ac_llvm_build.c index 79831260066..e97eb919c6a 100644 --- a/src/amd/llvm/ac_llvm_build.c +++ b/src/amd/llvm/ac_llvm_build.c @@ -4344,12 +4344,15 @@ ac_build_reduce(struct ac_llvm_context *ctx, LLVMValueRef src, nir_op op, unsign if (cluster_size == 32) return ac_build_wwm(ctx, result); if (ctx->chip_class >= GFX8) { - if (ctx->chip_class >= GFX10) - swap = ac_build_readlane(ctx, result, LLVMConstInt(ctx->i32, 31, false)); - else - swap = ac_build_dpp(ctx, identity, result, dpp_row_bcast31, 0xc, 0xf, false); - result = ac_build_alu_op(ctx, result, swap, op); - result = ac_build_readlane(ctx, result, LLVMConstInt(ctx->i32, 63, 0)); + if (ctx->wave_size == 64) { + if (ctx->chip_class >= GFX10) + swap = ac_build_readlane(ctx, result, LLVMConstInt(ctx->i32, 31, false)); + else + swap = ac_build_dpp(ctx, identity, result, dpp_row_bcast31, 0xc, 0xf, false); + result = ac_build_alu_op(ctx, result, swap, op); + result = ac_build_readlane(ctx, result, LLVMConstInt(ctx->i32, 63, 0)); + } + return ac_build_wwm(ctx, result); } else { swap = ac_build_readlane(ctx, result, ctx->i32_0); |