Feedback overhaul

.clang-format

@@ -104,19 +104,32 @@ IfMacros:
 IncludeBlocks:   Regroup
 IncludeCategories:
   - Regex:           '^<ext/.*\.h>'
-    Priority:        2
+    Priority:        4
     SortPriority:    0
     CaseSensitive:   false
-  - Regex:           '^<.*\.h>'
+  # it's critical that <hip/hip_runtime.h> is included before the
+  # <hip/hip_cooperative_groups.h> header
+  # -> I spent a LOT of time trying to get SortPriority to work properly, and
+  #    I've concluded that SortPriority is probably buggy or the documentation isn't
+  #    accurate (or I'm missing something obvious)
+  - Regex:           '^<hip/hip_runtime.h>$'
     Priority:        1
     SortPriority:    0
+    CaseSensitive:   true
+  - Regex:           '^<hip/.*>$'
+    Priority:        2
+    SortPriority:    0
+    CaseSensitive:   true
+  - Regex:           '^<.*\.h>'
+    Priority:        3
+    SortPriority:    0
     CaseSensitive:   false
   - Regex:           '^<.*'
-    Priority:        2
+    Priority:        4
     SortPriority:    0
     CaseSensitive:   false
   - Regex:           '.*'
-    Priority:        3
+    Priority:        5
     SortPriority:    0
     CaseSensitive:   false
 IncludeIsMainRegex: '([-_](test|unittest))?$'

.github/workflows/compilation_checks.yml

@@ -26,7 +26,7 @@ jobs:
         make-type: [hydro, gravity, disk, particles, cosmology, mhd, dust, cooling]
         container:
           #- {name: "CUDA", link: "docker://chollahydro/cholla:cuda_github"}
-          - {name: "HIP", link: "docker://chollahydro/cholla:rocm_github"}
+          - {name: "HIP", link: "ghcr.io/cholla-hydro/cholla-rocm:sha-a08bd2c"}
     env:  # Set environment variables
       CHOLLA_MAKE_TYPE: ${{ matrix.make-type }}
 

builds/make.type.disk

@@ -22,7 +22,7 @@ DFLAGS += -DGRAVITY_5_POINTS_GRADIENT
 
 DFLAGS    += -DMPI_CHOLLA
 DFLAGS    += -DPRECISION=2
-DFLAGS    += -DPLMC
+DFLAGS    += -DPLMP
 DFLAGS    += -DHLLC
 DFLAGS    += -DVL
 

builds/make.type.starblast

@@ -25,7 +25,7 @@ DFLAGS += -DANALYSIS
 DFLAGS    += -DCUDA
 DFLAGS    += -DMPI_CHOLLA 
 DFLAGS    += -DPRECISION=2
-DFLAGS    += -DPLMC
+DFLAGS    += -DPLMP
 DFLAGS    += -DHLLC
 DFLAGS    += -DVL
 

docs/sphinx/Physics/Dual-Energy-Formalism.md

@@ -12,7 +12,7 @@ Problems arise in simulations with large Mach numbers (such as cosmological simu
 
 The solution to this numerical problem is to use the "dual-energy formalism" (more details are provided in [Bryan+ 2014](https://ui.adsabs.harvard.edu/abs/2014ApJS..211...19B)). The core idea is to track an extra separately-advected "thermal energy" field at each cell-location, in addition to the total energy field and use this "thermal energy" field in cases where {math}`(E - E_{\rm thermal})` provides insufficient precision. 
 
-The dual-energy formalism is parameterized by two parameters, {math}`\eta_1` and {math}`\eta_2`. It's easiest to understand their meaning by discussing how they are used. The [Bryan+2014](https://ui.adsabs.harvard.edu/abs/2014ApJS..211...19B) paper describes two main steps:
+The dual-energy formalism is parameterized by two parameters, {math}`\eta_1` and {math}`\eta_2`. It's easiest to understand their meaning by discussing how they are used. The [Bryan+ 2014](https://ui.adsabs.harvard.edu/abs/2014ApJS..211...19B) paper describes two main steps:
 1. During a given timestep, when we want to compute thermal pressure, we compare quotient of the "thermal energy" field divided by {math}`E` to {math}`\eta_1`.
     - When the ratio is smaller than {math}`\eta_1` we use the "thermal energy" field. When it exceeds {math}`\eta_1`, we use {math}`(E-E_{\rm kinetic})`.
     - In effect, {math}`\eta_1` directly parameterizes the precision where the dual-energy formalism kicks in.
@@ -21,7 +21,7 @@ The dual-energy formalism is parameterized by two parameters, {math}`\eta_1` and
     - To motivate this step, it's important to understand that when we separately advect the "thermal energy" and add the {math}`-p(\nabla \cdot {\bf v})\Delta t/ \rho` source term, we are effectively ignoring the effects of shock heating.
     - Consequently, we might want to overwrite the "thermal energy" to capture the effects of shock heating. 
     - The precise condition that dictates when we overwrite the "thermal energy" field involves a comparison of {math}`\eta_2` and the values in neighboring cells. When {math}`\eta_2` is too high, we would effectively exclude shock-heating from weaker shocks. When {math}`\eta_2` is too low we may include spurious heating that is introduced by the truncation error of {math}`(E-E_{\rm kinetic})`.
-    - **NOTE:** [Bryan+2014](https://ui.adsabs.harvard.edu/abs/2014ApJS..211...19B) call this step "synchronization" - we find that name somewhat confusing since it may imply a bidirectional update (updating both "thermal energy" and {math}`E`).
+    - **NOTE:** [Bryan+ 2014](https://ui.adsabs.harvard.edu/abs/2014ApJS..211...19B) call this step "synchronization" - we find that name somewhat confusing since it may imply a bidirectional update (updating both "thermal energy" and {math}`E`).
 
 In practice, Cholla does something slightly different:
 1. It implements step 1 exactly as described above.