From 8b3e3beea66d9773adf11ea2e163577d649f3d7c Mon Sep 17 00:00:00 2001 From: Tim Foley Date: Tue, 28 Jan 2020 12:35:13 -0800 Subject: Fix layout for structured buffers of matrices (#1184) When using row-major layout (via command-line or API option), the following sort of declaration: ```hlsl StructuredBuffer gBuffer; ... gBuffer[i] ... ``` Generates unexpected results when compiled to DXBC via fxc or DXIL via dxc, because the fxc/dxc compilers do not respect the matrix layout mode in this specific case (a structured buffer of matrices). Instead, they always use column-major layout, even if row-major was requested by the user. A user can work around this behavior by wrapping the matrix in a `struct`: ```hlsl struct Wrapper { float4x4 wrapped; } SturcturedBuffer gBuffer; ... gBuffer[i].wrapped ... ``` This change simply automates that workaround when compiling for an HLSL-based downstream compiler, so that we get the same behavior across all our backends. The change adds a test case to confirm the behavior across multiple targets, but it turns out we also had a test checked in that confirmed the buggy (or at least surprising) fxc/dxc behavior, so that one had its baselines changed and can work as a regression test for this fix as well. --- tests/compute/matrix-layout-structured-buffer.slang | 17 ++++------------- 1 file changed, 4 insertions(+), 13 deletions(-) (limited to 'tests/compute/matrix-layout-structured-buffer.slang') diff --git a/tests/compute/matrix-layout-structured-buffer.slang b/tests/compute/matrix-layout-structured-buffer.slang index 36596424d..bb7dbb381 100644 --- a/tests/compute/matrix-layout-structured-buffer.slang +++ b/tests/compute/matrix-layout-structured-buffer.slang @@ -1,18 +1,9 @@ // matrix-layout-structured-buffer.slang -// This test is set up to confirm that `StructuredBuffer` types are -// always laid out column-major by fxc/dxc, even when row-major layout has been -// requested globally. -// -// This behavior should be considered a bug in Slang because either: -// -// 1. we should report reflection layout information that acknowledges this behavior, or -// 2. we should alter our HLSL output passed to fxc/dxc to provide consistent -// behavior that matches our reflection data. -// -// For now this test exists to document the situation. It's output can/should -// be updated if we decide to fix the underlying problem by taking option (2). -// +// This test confirms that we apply the matrix layout +// mode requested by the user, even in the case of structured +// buffers of matrices, where fxc/dxc do *not* respect +// the matrix layout mode by default. //TEST(compute):COMPARE_COMPUTE_EX:-slang -compute -xslang -matrix-layout-row-major //TEST(compute):COMPARE_COMPUTE_EX:-slang -compute -xslang -matrix-layout-column-major -- cgit v1.2.3