Register implied expression reads in OpLoad/OpAccessChain.

This is required to avoid relying on complex sub-expression elimination
in compilers, and generates cleaner code.

The problem case is if a complex expression is used in an access chain,
like:

Composite comp = buffer[texture(...)];
vec4 a = comp.a + comp.b + comp.c;

Before, we did not have common subexpression tracking for
OpLoad/OpAccessChain, so we easily ended up with code like:

vec4 a = buffer[texture(...)].a + buffer[texture(...)].b + buffer[texture(...)].c;

A good compiler will optimize this, but we should not rely on it, and
forcing texture(...) to a temporary also looks better.

The solution is to add a vector "implied_expression_reads", which works
similarly to expression_dependencies. We also need an extra mechanism in
to_expression which lets us skip expression read checking and do it
later. E.g. for expr -> access chain -> load, we should only trigger
a read of expr when using the loaded expression.

1 files changed, 3 insertions, 2 deletions

diff --git a/reference/opt/shaders/comp/generate_height.comp b/reference/opt/shaders/comp/generate_height.comp
index d9b1a7b8..feb8d41c 100644
--- a/reference/opt/shaders/comp/generate_height.comp
+++ b/reference/opt/shaders/comp/generate_height.comp
@@ -41,6 +41,7 @@ void main()
     }
     uint _276 = _264.x;
     uint _280 = (gl_GlobalInvocationID.y * _276) + gl_GlobalInvocationID.x;
+    uint _290 = (_455 * _276) + _454;
     vec2 _297 = vec2(gl_GlobalInvocationID.xy);
     vec2 _299 = vec2(_264);
     float _309 = sqrt(9.81000041961669921875 * length(_166.uModTime.xy * mix(_297, _297 - _299, greaterThan(_297, _299 * 0.5)))) * _166.uModTime.z;
@@ -48,8 +49,8 @@ void main()
     vec2 _387 = _316.xx;
     vec2 _392 = _316.yy;
     vec2 _395 = _392 * _137.distribution[_280].yx;
-    vec2 _421 = _392 * _137.distribution[(_455 * _276) + _454].yx;
-    vec2 _429 = (_137.distribution[(_455 * _276) + _454] * _387) + vec2(-_421.x, _421.y);
+    vec2 _421 = _392 * _137.distribution[_290].yx;
+    vec2 _429 = (_137.distribution[_290] * _387) + vec2(-_421.x, _421.y);
     _225.heights[_280] = packHalf2x16(((_137.distribution[_280] * _387) + vec2(-_395.x, _395.y)) + vec2(_429.x, -_429.y));
 }