Building and Running GEOSgcm with ifx 2025.2

[mathomp4]

Due to a preprocessor bug (and other issues) with ifx 2025.2, there are many things that will be needed for support.

GEOSgcm v11

• Update GFE in Baselibs with ifx support (new yafyaml, pflogger, pfunit) → ESMA_env
• Update to ESMA_cmake v3.65.0
• Update MAPL to have ifx support → MAPL 2.62
• Update ifx flags to match ifort → WIP: v3: Update ifx flags to match ifort ESMA_cmake#481 (This seems to avoid an ICE in ifx in RRTMGP)
• Fix ifx crash in umwm_cmake (due to -fp-model consistent)
• Update needed in GMAO_gfio/CMakeLists.txt
• Update needed in GMI CMake
• Update needed in Stratchem CMake
• Update needed in ACHEM CMake
• Update needed in FVdycoreCubed_GridComp → v11: Initialize q_con FVdycoreCubed_GridComp#318

GEOSgcm v12

• Update GFE in Baselibs with ifx support (new yafyaml, pflogger, pfunit) → ESMA_env
• Update to ESMA_cmake v4.24.0
• Update MAPL to have ifx support → MAPL 2.62
• Update ifx flags to match ifort → v4: Update ifx flags to match ifort ESMA_cmake#482 (This seems to avoid an ICE in ifx in RRTMGP)
• Fix ifx crash in umwm_cmake (due to -fp-model consistent)
• Update needed in GMAO_gfio/CMakeLists.txt
• Update needed in GMI CMake
• Update needed in Stratchem CMake
• Update needed in ACHEM CMake

[mathomp4]

Build Types

GEOSgcm v11

• Release → Builds, runs, fails layout regress
• Debug → Builds, runs with fix in fv3 (v11: Initialize q_con FVdycoreCubed_GridComp#318)
• Aggressive → Builds, runs

GEOSgcm v12

• Release → Builds, runs, fails layout and openmp regress
• Debug → Builds, runs
  • The above fv3 fix isn't needed here because the line it failed on is removed in v12
• Aggressive →

Throughput Results (ifort v ifx)

GEOSgcm v11 - Release

Resolution	ifort	ifx
c24	501.173	468.458
c90	132.841	144.291
c360	95.995	98.776
c720	74.659	78.435

GEOSgcm v12 - Release

Resolution	ifort	ifx
c24	489.779	495.695
c90	206.609	223.042
c360	120.481	126.018
c720	39.693	44.818